&EPA
United States
Envirofunmlal Protection
Agnncy
Health Risk and Exposure Assessment
for Ozone
First External Review Draft
-------
DISCLAIMER
This preliminary draft document has been prepared by staff from the Risk and Benefits Group,
Office of Air Quality Planning and Standards, U.S. Environmental Protection Agency. Any
opinions, findings, conclusions, or recommendations are those of the authors and do not
necessarily reflect the views of the EPA. This document is being circulated for informational
purposes and to facilitate discussion with the Clean Air Scientific Advisory Committee
(CAS AC) on the overall structure, areas of focus, and level of detail to be included in an external
review draft Policy Assessment, which EPA plans to release for CASAC review and public
comment later this year. Questions related to this preliminary draft document should be
addressed to Karen Wesson, U.S. Environmental Protection Agency, Office of Air Quality
Planning and Standards, C504-02, Research Triangle Park, North Carolina 27711 (email:
wesson.karen@epa.gov).
-------
EPA 452/P-12-001
July 2012
Health Risk and Exposure Assessment for Ozone
First External Review Draft
U.S. Environmental Protection Agency
Office of Air and Radiation
Office of Air Quality Planning and Standards
Health and Environmental Impacts Division
Risk and Benefits Group
Research Triangle Park, North Carolina 27711
-------
This page left intentionally blank.
-------
TABLE OF CONTENTS
Table of Contents i
List of Acronyms/Abbreviations v
1 Introduction 1-1
1.1 History 1-3
1.2 Current Risk and Exposure Assessment: Goals and Planned Approach 1-6
1.3 Organization of Document 1-7
2 Conceptual Model 2-1
2.1 O3 Chemistry 2-1
2.2 Sources of O3 and O3 Precursors 2-2
2.3 Exposure Pathways and Important Microenvironments 2-3
2.4 At-risk Populations 2-5
2.5 Health Endpoints 2-7
2.6 References 2-10
3 Scope 3-1
3.1 Overview of Exposure and Risk Assessments from Last Review 3-2
3.2.1 Overview of Exposure Assessment from Last Review 3-2
3.2.2 Overview of Risk Assessment from Last Review 3-3
3.2 Plan for the Current Exposure and Risk Assessments 3-5
3.2.1 Air Quality Data 3-6
3.2.2 Exposure Assessment 3-9
3.2.3 Lung Function Risk Assessment 3-10
3.2.4 Urban Area Epidemiology Based Risk Assessment 3-14
3.2.5 National-scale Mortality Risk Assessment 3-18
3.2.6 Characterization of Uncertainty and Variability in the Context of the O3 Risk Assessment.... 3-19
3.2.7 Presentation of Risk Estimates to Inform the O3 NAAQS Policy Assessment 3-22
3.3 References 3-23
4 Air Quality Considerations 4-1
4.1 Introduction 4-1
4.2 Overview of Ozone Monitoring and Air Quality 4-1
-------
4.3 Overview of Air Quality Inputs to Risk and Exposure Assessments 4-5
4.3.1 Urban-scale Air Quality Inputs 4-5
4.3.2 National-scale Air Quality Inputs 4-13
4.4 References 4-17
5 Characterization of Population Exposure 5-1
5.1 Introduction 5-1
5.2 Ozone Exposure Studies 5-2
5.3 Exposure Modeling 5-4
5.3.1 The APEX Model 5-4
5.3.2 Key Algorithms 5-6
5.3.3 Model Output 5-16
5.4 Scope of Exposure Assessment 5-18
5.4.1 Selection of Urban Areas 5-18
5.4.2 Time Periods Modeled 5-18
5.4.3 Populations Modeled 5-19
5.4.4 Microenvironments Modeled 5-20
5.4.5 Benchmark Levels Modeled 5-21
5.5 Variability and Uncertainty 5-22
5.5.1 Treatment of Variability 5-22
5.5.2 Charactizati on of Uncertainty 5-23
5.6 Exposure Assessment Results 5-24
5.6.1 Overview 5-24
5.6.2 Exposure Modeling Results 5-24
5.6.3 Characterization of Factors Influencing High Exposures 5-40
5.6.4 Discussions of Exposure Modeling Results 5-51
5.7 References 5-54
6 Characterization of Health Risk Based on Controlled Human Exposure Studies 6-1
6.1 Introduction 6-?
6.2 General Approach 6-?
6.3 Selection of Health Endpoints 6-?
6.4 Approach to Calculating Risk Esimates 6-?
-------
6.4.1 Exposure-Response Functions Used in Prior Reviews 6-?
6.4.2 McDonnell-Stewart-Smith Model 6-?
6.5 Risk Estimates 6-?
6.5.1 Recent Air Quality 6-?
6.5.2 JustMeeting Current Ozone Standard 6-?
6.6 Sensitivity Analyses 6-?
6.6.1 Alterantive Assumptions About the Shape of the Exposure-Response Functions 6-?
6.6.2 Alternative Ventilation Rates Algorithm 6-?
6.7 Uncertainty and Variability 6-1
6.7.1 Exposure-Response Functions Used in Prior Reviews 6-?
6.7.2 McDonnell-Stewart-Smith Model 6-?
6.8 References 6-?
7 Characterization of Health Risk Based on Epidemiological Studies 7-1
7.1 General Approach 7-1
7.1.1 Basic Structure of the Risk Assessment 7-1
7.1.2 Calculating OS-Related Health Effects Incidence 7-10
7.2 Air Quality Considerations 7-12
7.2.1 Characterizing Recent Conditions 7-13
7.2.2 Estimating U.S. Background 7-16
7.2.3 Simulating Air Quality to Just Meet Current and Alternative Standards 7-17
7.3 Selection of Model Inputs 7-19
7.3.1 Selection and Delineation of Urban Study Areas 7-19
7.3.2 Selection of Epidemiological Studies and Concentration-Response Functions 7-22
7.3.3 Defining O3 Concentration Ranges (down to the LML) for Which There Is an Increase
Confidence in Estimating Risk 7-30
7.3.4 Baseline Health Effects Incidence and Prevalence Data 7-32
7.3.5 Population (Demographic) Data 7-34
7.4 Addressing Variability and Uncertainty 7-34
7.4.1 Treatment of Key Sources of Variability 7-37
7.4.2 Qualitative Assessment of Uncertainty 7-40
7.5 Urban Study Area Results 7-46
in
-------
7.5.1 Assessment of Health Risk Associated with Recent Conditions 7-66
7.5.2 Assessment of Health Risk Associated with Simulating Meeting the Current Suite of O3
Standards 7-69
7.6 Key Observations Drawn from the Urban Case Study Analysis of OS-Related Risk 7-72
7.6.1 Overall Confidence in Risk Assessment and Risk Estimates 7-72
7.6.2 Risk Estimates Generated for Both the Recent Conditions and Simulation of Meeting the Current
Standard 7-74
7.7 Potential Refinements for the Second Draft Risk Assessment 7-76
7.7.1 Potential Sensitivity Analyses 7-76
7.7.2 Additional Refinements to the Core Risk Estimates Completed for the First Draft REA 7-77
7.7.3 Treatment of both Long-term Exposure-related Mortality and Morbidity Endpoints 7-79
7.8 References 7-82
8 National-scale Assessment of Short-term Mortality Related to O3 Exposure 8-1
8.1 Introduction 8-1
8.1.1 Methods 8-2
8.1.2 Re suits 8-10
8.1.3 Discussion 8-21
8.2 Evaluating the Representativeness of the Urban Study Areas in the National Context 8-22
8.2.1 Analysis Based on Consideration of National Distributions of Risk-based Attributes 8-24
8.2.2 Analysis Based on Consideration of National Distributions of 03-related Mortality Risk 8-46
8.2.3 Discussion 8-49
8.3 References 8-50
9 Synthesis and Integration of Results 9-1
9.1 Summary of Key Results of Population Exposure Assessment 9-1
9.2 Summary of Key Results of Health Risk Based on Controlled Human Exposure Studies 9-5
9.3 Summary of Key Results of Health Risk Based on Epidemiological Studies 9-5
9.4 Observations 9-9
IV
-------
LIST OF ACRONYMS/ABBREVIATIONS
AER
AHRQ
APEX
AQI
AQS
ATUS
BenMAP
BRFSS
BSA
CAA
CASAC
CDC
CDF
CH4
CHAD
CI
CMAQ
C02
C-R
ED
EGU
EPA
ER
eVNA
EVR
FEM
FEV1
FRM
FVC
air exchange rate
Agency for Healthcare Research and Quality
Air Pollution Exposure Model
Air Quality Index
Air Quality System
American Time Use Survey
Benefits Mapping and Analysis Program
Behavioral Risk Factor Surveillance System
body surface area
Clean Air Act
Clean Air Science Advisory Committee
Center for Disease Control and Prevention
cumulative distribution functions
methane
Consolidated Human Activity Database
confidence interval
Community Multi-scale Air Quality
carbon dioxide
Concentration Response (function)
emergency department
electric generating unit
U.S. Environmental Protection Agency
emergency room
enhanced Voronoi Neighbor Averaging
equivalent ventilation rate
Federal Equivalent Method
one-second forced expiratory volume
Federal Reference Method
forced vital capacity
-------
HA
HDDM
HNO3
HO2
HUCP
IPCC
IRP
ISA
LML
MATS
METs
MSA
MT
NAAQS
NCDC
NEI
NO
NO2
NOX
03
OAQPS
OH
PA
PDI
PI
PM
ppb
ppm
PRB
REA
RR
hospital admissions
Higher-order Decoupled Direct Method
nitric acid
hydro-peroxy radical
Healthcare Cost and Utilization Program
Intergovernmental Panel on Climate Change
Integrated Review Plan
Integrated Science Assessment
lowest measured level
Modeled Attainment Test Software
metabolic equivalents of work
Metropolitan Statistical Area
metric ton
National Ambient Air Quality Standards
National Climatic Data Center
National Emissions Inventory
nitric oxide
nitrite
nitrogen oxides
Ozone
Office of Air Quality Planning and Standards
hydroxyl radical
Policy Assessment
pain on deep inspiration
posterior interval
particulate matter
parts per billion
parts per million
Policy Relevant Background
Risk and Exposure Assessment
relative risk
-------
SAB Science Advisory Board
SEDD State Emergency Department Databases
SES socioeconomic status
SID State Inpatient Databases
SO2 sulfur dioxide
STE stratosphere-troposphere exchange
TRIM Expo Total Risk Integrated Methodology Inhalation Exposure
VE ventilation rate
VNA Voronoi Neighbor Averaging
VOC volatile organic carbon
WHO World Health Organization
-------
1 1 INTRODUCTION
2 The U.S. Environmental Protection Agency (EPA) is presently conducting a review of
3 the national ambient air quality standards (NAAQS) for ozone (Os) and related photochemical
4 oxidants. An overview of the approach to reviewing the Os NAAQS is presented in the
5 Integrated Review Plan for the Ozone National Ambient Air Quality Standards (IRP, US EPA,
6 201 la). The IRP discusses the schedule for the review; the approaches to be taken in developing
7 key scientific, technical, and policy documents; and the key policy-relevant issues that will frame
8 our consideration of whether the current NAAQS for 63 should be retained or revised.
9 Sections 108 and 109 of the Clean Air Act (CAA) govern the establishment and periodic
10 review of the NAAQS. These standards are established for pollutants that may reasonably be
11 anticipated to endanger public health and welfare, and whose presence in the ambient air results
12 from numerous or diverse mobile or stationary sources. The NAAQS are to be based on air
13 quality criteria, which are to accurately reflect the latest scientific knowledge useful in indicating
14 the kind and extent of identifiable effects on public health or welfare that may be expected from
15 the presence of the pollutant in ambient air. The EPA Administrator is to promulgate and
16 periodically review, at five-year intervals, "primary" (health-based) and "secondary" (welfare-
17 based) NAAQS for such pollutants. Based on periodic reviews of the air quality criteria and
18 standards, the Administrator is to make revisions in the criteria and standards, and promulgate
19 any new standards, as may be appropriate. The Act also requires that an independent scientific
20 review committee advise the Administrator as part of this NAAQS review process, a function
21 performed by the Clean Air Scientific Advisory Committee (CASAC).1
22 The current primary NAAQS for Os is set at a level of 0.075 ppm, based on the annual
23 fourth-highest daily maximum 8-hr average concentration, averaged over three years, and the
24 secondary standard is identical to the primary standard (73 FR 16436). The EPA initiated the
1 The Clean Air Scientific Advisory Committee (CAS AC) was established under section 109(d)(2) of the
Clean Air Act (CAA) (42 U.S.C. 7409) as an independent scientific advisory committee. CASAC provides advice,
information and recommendations on the scientific and technical aspects of air quality criteria and NAAQS under
sections 108 and 109 of the CAA. The CASAC is a Federal advisory committee chartered under the Federal
Advisory Committee Act (FACA). See
http://vosemite.epa.gov/sab/sabpeople.nsfAVebComniitteesSubcomniittees/CASAC%20Particulate%20Matter%20R
eview%20Panel for a list of the CASAC PM Panel members and current advisory activities.
1-1
-------
1 current review of the Os NAAQS on September 29, 2008 with an announcement of the
2 development of an 63 Integrated Science Assessment and a public workshop to discuss policy -
3 relevant science to inform EPA's integrated plan for the review of the Oj NAAQS (73 FR
4 56581). The NAAQS review process includes four key phases: planning, science assessment,
5 risk/exposure assessment, and policy assessment/rulemaking.2 A workshop was held on October
6 29-30, 2008 to discuss policy-relevant scientific and technical information to inform EPA's
7 planning for the Os NAAQS review. Following the workshop, EPA developed a planning
8 document, the Integrated Review Plan for the Ozone National Ambient Air Quality Standards
9 (IRP; US EPA, 2011), which outlined the key policy-relevant issues that frame this review, the
10 process and schedule for the review, and descriptions of the purpose, contents, and approach for
11 developing the other key documents for this review.3 In June 2012, EPA completed the third
12 draft of the Os ISA, assessing the latest available policy-relevant scientific information to inform
13 the review of the Os standards. The Integrated Science Assessment for Ozone and Related
14 Photochemical Oxidants - Third External Review Draft (ISA; US EPA, 2012), includes an
15 evaluation of the scientific evidence on the health effects of Os, including information on
16 exposure, physiological mechanisms by which 63 might adversely impact human health, an
17 evaluation of the toxicological and controlled human exposure study evidence, and an evaluation
18 of the epidemiological evidence including information on reported concentration-response (C-R)
19 relationships for (Vrelated morbidity and mortality associations, including consideration of
20 effects on susceptible populations.4
21 The EPA's Office of Air Quality Planning and Standards (OAQPS) has developed this
22 first draft quantitative health risk and exposure assessment (REA) describing preliminary
23 quantitative assessments of exposure to Os and Os-related risks to public health to support the
24 review of the primary Os standards. This draft document presents the conceptual model, scope,
25 methods, key results, observations, and related uncertainties associated with the quantitative
26 analyses performed. The REA builds upon the health effects evidence presented and assessed in
2 For more information on the NAAQS review process see http://www.epa.gov/ttn/naaqs/review.html.
3 On March 30, 2009, EPA held a public consultation with the CASAC Ozone Panel on the draft IRP. The
final IRP took into consideration comments received from CASAC and the public on the draft plan as well as input
from senior Agency managers.
4 The ISA also evaluates scientific evidence for the effects of O3 on public welfare which EPA will consider
in its review of the secondary O3 NAAQS. Building upon the effects evidence presented in the ISA, OAQPS has
also developed a second REA titled Ozone Welfare Effects Risk and Exposure Assessment (US EPA, 2012).
1-2
-------
1 the ISA, as well as CAS AC advice (Samet, 20011) and public comments on a scope and methods
2 planning document for the REA (here after, "Scope and Methods Plan", US EPA, 2011).
3 Revisions to this draft REA will draw upon the final ISA and will reflect consideration of
4 CASAC and public comments on this draft.
5 The ISA and REA will inform the development of a Policy Assessment (PA) and
6 rulemaking steps that will lead to final decisions on the primary Os NAAQS, as described in the
7 IRP. The PA will include staff analysis of the scientific basis for alternative policy options for
8 consideration by senior EPA management prior to rulemaking. The PA integrates and interprets
9 information from the ISA and the REA to frame policy options for consideration by the
10 Administrator. The PA is intended to link the Agency's scientific and technical assessments,
11 presented in the ISA and REA, to judgments required of the Administrator in determining
12 whether it is appropriate to retain or revise the current Os standards. Development of the PA is
13 also intended to facilitate elicitation of CASAC's advice to the Agency and recommendations on
14 any new standards or revisions to existing standards as may be appropriate, as provided for in the
15 Clean Air Act (CAA). The first draft PA is planned for release around the middle of August
16 2012 for review by the CASAC Os Panel and the public concurrently with their review of this
17 first draft REA September 11-13, 2012.
18 1.1 HISTORY
19 As part of the last O3 NAAQS review completed in March 2008, EPA's OAQPS
20 conducted quantitative risk and exposure assessments to estimate exposures above health
21 benchmarks and risks of various health effects associated with exposure to ambient 63 in a
22 number of urban study areas selected to illustrate the public health impacts of this pollutant (U.S.
23 EPA 2007a, U.S. EPA 2007b). The assessment scope and methodology were developed with
24 considerable input from CASAC and the public, with CASAC generally concluding that the
25 exposure assessment reflected generally accepted modeling approaches, and that the risk
26 assessments were well done, balanced and reasonably communicated (Henderson, 2006a). The
27 final quantitative risk and exposure assessments took into consideration CASAC advice
28 (Henderson, 2006a; Henderson, 2006b) and public comments on two drafts of the risk and
29 exposure assessments.
1-3
-------
1 The exposure and health risk assessment conducted in the last review developed exposure
2 and health risk estimates for 12 urban areas across the U.S. based on 2002 to 2004 air quality
3 data. That assessment provided annual or O3 season-specific exposure and risk estimates for
4 these years of air quality and for air quality scenarios simulating just meeting the then-existing 8-
5 hour Os standard set in 1997 at a level of 0.08 ppm and several alternative 8-hour standards. The
6 strengths and limitations in the assessment were characterized, and analyses of key uncertainties
7 were presented.
8 Exposure estimates from the last assessment were used as an input to the risk assessment
9 for lung function responses (a health endpoint for which exposure-response functions were
10 available from controlled human exposure studies). Exposure estimates were developed for the
11 general population and population groups including school age children with asthma as well as
12 all school age children. The exposure estimates also provided information on exposures to
13 ambient Os concentrations at and above specified benchmark levels (referred to as "exposures of
14 concern") to provide some perspective on the public health impacts of health effects associated
15 with Os exposures in controlled human exposure studies that could not be evaluated in the
16 quantitative risk assessment (e.g., lung inflammation, increased airway responsiveness, and
17 decreased resistance to infection).
18 The last human risk assessment included risk estimates based on both controlled human
19 exposure studies and epidemiological and field studies. Ozone-related risk estimates for lung
20 function decrements were generated using probabilistic exposure-response relationships based on
21 data from controlled human exposure studies, together with probabilistic exposure estimates
22 from the exposure analysis. For several other health endpoints, Os-related risk estimates were
23 generated using concentration-response relationships reported in epidemiological or field studies,
24 together with ambient air quality concentrations, baseline health incidence rates, and population
25 data for the various locations included in the assessment. Health endpoints included in the
26 assessment based on epidemiological or field studies included: hospital admissions for
27 respiratory illness in four urban areas, premature mortality in 12 urban areas, and respiratory
28 symptoms in asthmatic children in 1 urban area.
29 The last exposure and risk assessment helped to inform the last review and the final
30 decision to revise the primary Os NAAQS to a level of 0.075 ppm, as discussed in the Final Rule
31 notice (73 FR 16436; March 27, 2008). As an initial matter, in considering the adequacy of the
1-4
-------
1 then-current standard, while the Administrator placed primary consideration on the body of
2 scientific evidence of (Vrelated health effects, he also considered the exposure and risk
3 assessment results and related uncertainties. In so doing, the Administrator considered the
4 estimated percentages of asthmatic and all school age children likely to experience exposures
5 (while at moderate or greater exertion) at and above the benchmark levels of 0.080, 0.070 and
6 0.060 ppm upon simulation of just meeting the then-current standard, as well as the year-to-year
7 and city-to-city variability and the uncertainties is those estimates. He also considered the
8 estimated health risks for lung function decrements, respiratory symptoms, respiratory-related
9 hospital admissions and mortality upon simulation of just meeting the then-current standard, as
10 well as the variability and uncertainties in those estimates. He recognized that these risk
11 estimates were indicative of a much broader array of O3-related health endpoints that could not
12 be included in the quantitative assessment (e.g., school absences, increased medication use,
13 emergency department visits) which primarily affect at-risk populations. In considering this
14 information, the Administrator concluded that the estimated exposures and risks were important
15 from a public health perspective and that they provide additional support to the evidence-based
16 conclusion that the then-current standard needed to be revised.
17 In considering the level at which a revised primary 63 standard should be set, within the
18 proposed range of 0.070 to 0.075 ppm, the Administrator again placed primary consideration on
19 the body of scientific evidence of (Vrelated health effects, while viewing the results of the
20 exposure and risk assessment as providing information in support of his decision. In considering
21 the exposure estimates simulated for meeting alternative standard levels, the Administrator
22 placed greatest weight on estimated exposures at and above the 0.080 ppm benchmark level, less
23 weight on the 0.070 ppm benchmark, and very little weight on the 0.060 ppm benchmark. Given
24 the degree of uncertainty in these estimates, he judged that there was not an appreciable
25 difference, from a public health perspective, in the estimates of exposures associated with just
26 meeting a standard at the upper end (0.075 ppm) versus the lower end (0.070 ppm) of the
27 proposed range of levels. The Administrator placed less weight on the risk estimates for meeting
28 alternative standard levels, and noted that the results suggest a gradual reduction in risks with no
29 clear breakpoint as increasingly lower standard levels are considered. Taken together, the
30 Administrator judged that the exposure and risk information did not provide a clear basis for
31 choosing a specific level within the range of levels being considered. In reaching a final
1-5
-------
1 evidence-based decision to set the standard at a level of 0.075 ppm, the Administrator noted that
2 this level was above the range of levels recommended by CASAC (0.060 to 0.070 ppm). In
3 explaining the basis for this difference with CASAC, the Administrator noted that there is no
4 bright line clearly directing the choice of level, and the choice of an appropriate level is clearly a
5 public health policy judgment. In reaching his final judgment, the Administrator explained in
6 part that CASAC appeared to place greater weight on the results of the risk assessment as a basis
7 for its recommended range, while he more heavily weighed the implications of the uncertainties
8 associated with the exposure and risk assessments.
9 Following promulgation of the revised O3 standard in March 2008, state, public health,
10 environmental, and industry petitioners filed suit against EPA regarding that final decision.
11 At EPA's request the consolidated cases were held in abeyance pending EPA's voluntary
12 reconsideration of the 2008 decision. A notice of proposed rulemaking to reconsider the
13 2008 final decision was issued by the Administrator on January 6, 2010. On September 2,
14 2011, the Office of Management and Budget returned the draft final rule on reconsideration
15 to EPA for further consideration. EPA decided to coordinate further proceedings on its
16 voluntary rulemaking on reconsideration with this ongoing periodic review, by deferring the
17 completion of its voluntary rulemaking on reconsideration until it completes its statutorily-
18 required periodic review. In light of that, the litigation on the 2008 final decision is no
19 longer being held in abeyance and is proceeding. The 2008 Os standards remain in effect.
20 1 2 CURRENT RISK AND EXPOSURE ASSESSMENT: GOALS AND PLANNED
21 APPROACH
22 The goals of the current quantitative exposure and health risk assessments are (1) to
23 provide estimates of the number of people in the general population and in sensitive populations
24 with O3 exposures above benchmark levels while at moderate or greater exertion levels; (2) to
25 provide estimates of the number of people in the general population and in at-risk populations
26 with impaired lung function resulting from exposures to Os; (3) to provide estimates of the
27 potential magnitude of premature mortality and selected morbidity health effects in the
28 population, including at-risk populations, where data are available to assess these groups,
29 associated with recent ambient levels of 63 and with just meeting the current primary 63
30 standard and any alternative standards that might appropriately be considered in selected urban
1-6
-------
1 study areas; (4) to develop a better understanding of the influence of various inputs and
2 assumptions on the exposure and risk estimates to more clearly differentiate alternative standards
3 that might be considered including potential impacts on various at-risk populations; and (5) to
4 gain insights into the distribution of risks and patterns of risk reduction and uncertainties in those
5 risk estimates. In addition, we have conducted an assessment to provide nationwide estimates of
6 the potential magnitude of premature mortality associated with ambient Os exposures to more
7 broadly characterize this risk on a national scale. This assessment includes an evaluation of the
8 distribution of risk across the U. S., to assess the extent to which we have captured the upper end
9 of the risk distribution with our urban study area analyses.
10 This current quantitative risk and exposure assessment builds on the approach used and
11 lessons learned in the last O3 risk and exposure assessment and focuses on improving the
12 characterization of the overall confidence in the exposure and risk estimates, including related
13 uncertainties, by incorporating a number of enhancements, in terms of both the methods and data
14 used in the analyses. This risk assessment considers a variety of health endpoints for which, in
15 staff s judgment, there is adequate information to develop quantitative risk estimates that can
16 meaningfully inform the review of the primary Os NAAQS.
17 The results from this risk and exposure assessment will be considered from a policy
18 perspective in the PA. The PA will also evaluate the entire body of scientific evidence of
19 relationships between 63 and a wide array of health endpoints, including those considered in the
20 risk assessment, from a policy perspective. These evidence-based and exposure/risk-based
21 considerations will inform staffs assessment of various policy options as discussed in the PA.
22 This first draft REA provides an assessment of exposure and risk associated with recent
23 ambient levels of Os and Os air quality simulated to just attain the current primary Os standards.
24 Subsequent drafts of the REA will evaluate potential alternative Os standards based on
25 considerations discussed in the first draft of the Policy Assessment.
26 1.3 ORGANIZATION OF DOCUMENT
27 The remainder of this document, when final, will be organized as follows. Chapter 2
28 provides a conceptual framework for the risk and exposure assessment, including discussions of
29 63 chemistry, sources of 63 precursors, exposure pathways and microenvironments where 63
30 exposure can be high, at-risk populations, and health endpoints associated with 63. This
1-7
-------
1 conceptual framework sets the stage for the scope of the risk and exposure assessments. Chapter
2 3 provides an overview of the scope of the quantitative risk and exposure assessments, including
3 a summary of the previous risk and exposure assessments, and an overview of the current risk
4 and exposure assessments. Chapter 4 discusses air quality considerations relevant to the
5 exposure and risk assessments, including available Os monitoring data, and important inputs to
6 the risk and exposure assessments. Chapter 5 describes the inputs, models, and results for the
7 human exposure assessment, and discusses the literature on exposure to Os, exposure modeling
8 approaches using the Air Pollution Exposure Model (APEX), the scope of the exposure
9 assessment, inputs to the exposure modeling, sensitivity and uncertainty evaluations, and
10 estimation of results. Chapter 6 describes the estimation of health risks based on application of
11 the results of human clinical studies, including discussions of health endpoint selection,
12 approaches to calculating risk, and results. (We note that work is continuing on Chapter 6 and we
13 expect to release a first draft of that chapter in August.) Chapter 7 describes the estimation of
14 health risks in selected urban areas based on application of the results of observational
15 epidemiology studies, including discussions of air quality characterizations, model inputs,
16 variability and uncertainty, and results. Chapter 8 describes the national scale risk
17 characterization and urban area representativeness analysis. Chapter 9 provides an integrative
18 discussion of the exposure and risk estimates generated in the analyses drawing on the results of
19 the analyses based on both clinical and epidemiology studies, and incorporating considerations
20 from the national scale risk characterization.
1-8
-------
2 CONCEPTUAL FRAMEWORK
2 In this chapter, we summarize the conceptual framework for assessing exposures to O3
3 and the associated risks to human populations. This conceptual framework includes elements
4 related to characterization of ambient Os and its relation to population exposures (Section 2.1),
5 important sources of Os precursors including oxides of nitrogen (NOX) and volatile organic
6 compounds (VOC) (Section 2.2), exposure pathways and important microenvironments where
7 63 exposures may be high (Section 2.3), populations that may be at greater risk due to increased
8 exposure or other factors that increase vulnerability and susceptibility (Section 2.4), and health
9 outcomes identified in the literature as associated with ambient 63 (Section 2.5).
10 2.1 OZONE CHEMISTRY
11 63 occurs naturally in the stratosphere where it provides protection against harmful solar
12 ultraviolet radiation, and it is formed closer to the surface in the troposphere by both natural and
13 anthropogenic sources. 63 is not emitted directly into the air, but is created when its two primary
14 precursors, volatile organic compounds (VOC) and oxides of nitrogen (NOX), combine in the
15 presence of sunlight. VOC and NOX are, for the most part, emitted directly into the atmosphere.
16 Carbon monoxide (CO) and methane (CH/t) are also important for Os formation (US EPA, 2012,
17 section 3.2.2).
18 Rather than varying directly with emissions of its precursors, O?, changes in a nonlinear
19 fashion with the concentrations of its precursors. NOX emissions lead to both the formation and
20 destruction of Os, depending on the local quantities of NOX, VOC, and radicals such as the
21 hydroxyl (OH) and hydro-peroxy (HO2) radicals. In areas dominated by fresh emissions of NOX,
22 these radicals are removed via the production of nitric acid (HNO3), which lowers the O3
23 formation rate. In addition, the scavenging of Os by reaction with NO is called "titration," and is
24 often found in downtown metropolitan areas, especially near busy streets and roads, and in
25 power plant plumes. This titration results in local valleys in which ozone concentrations are low
26 compared to surrounding areas. Titration is usually short-lived confined to areas close to strong
27 NOX sources, and the NO2 formed this way leads to Os formation later and further downwind. .
28 Consequently, ozone response to reductions in NOX emissions is complex and may include ozone
29 decreases at some times and locations and increases of ozone to fill in the local valleys of low
2-1
-------
1 ozone. In areas with low NOX concentrations, such as those found in remote continental areas to
2 rural and suburban areas downwind of urban centers, the net production of 63 typically varies
3 directly with NOX concentrations, and increases with increasing NOX emissions.
4 In general, the rate of Os production is limited by either the concentration of VOCs or
5 NOX, and Os formation using these two precursors relies on the relative sources of OH and NOX.
6 When OH radicals are abundant and are not depleted by reaction with NOX and/or other species,
7 O3 production is referred to as being "NOx-limited" (US EPA, 2012, section 3.2.4). In this
8 situation, Os concentrations are most effectively reduced by lowering NOX emissions, rather than
9 lowering emissions of VOCs. When the abundance of OH and other radicals is limited either
10 through low production or reactions with NOX and other species, Os production is sometimes
11 called "VOC-limited" or "radical limted" or "NOX-saturated" (Jaegle et al., 2001), and O3 is most
12 effectively reduced by lowering VOCs. However, even in NOx-saturated conditions, very large
13 decreases in NOX emissions can cause the ozone formation regime to become NOX limited.
14 Consequently, reductions in NOX emissions (when large) can make further emissions reductions
15 more effective at reducing ozone. Between the NOx-limited and NOx-saturated extremes there is
16 a transitional region where Os is relatively insensitive to marginal changes in both NOX and
17 VOCs. In rural areas and downwind of urban areas, Os production is generally NOx-limited.
18 However, across urban areas with high populations, conditions may vary. For contrast, while
19 data from monitors in Nashville, TN suggest NOx-limited conditions exist there, data from
20 monitors in Los Angeles suggest NOx-saturated conditions (US EPA, 2012, Figure 3-3).
21 2.2 SOURCES OF O3 AND O3 PRECURSORS
22 Os precursor emissions can be divided into anthropogenic and natural source categories,
23 with natural sources further divided into biogenic emissions (from vegetation, microbes, and
24 animals) and abiotic emissions (from biomass burning, lightning, and geogenic sources). The
25 anthropogenic precursors of Os originate from a wide variety of stationary and mobile sources.
26 In urban areas, both biogenic and anthropogenic VOCs are important for Os formation.
27 Hundreds of VOCs are emitted by evaporation and combustion processes from a large number of
28 anthropogenic sources. Based on the 2005 national emissions inventory (NEI), solvent use and
29 highway vehicles are the two main sources of VOCs, with roughly equal contributions to total
30 emissions (US EPA, 2012, Figure 3-3). The emissions inventory categories of "miscellaneous"
2-2
-------
1 (which includes agriculture and forestry, wildfires, prescribed burns, and structural fires) and off-
2 highway mobile sources are the next two largest contributing emissions categories with a
3 combined total of over 5.5 million metric tons a year (MT/year).
4 On the U.S. and global scales, emissions of VOCs from vegetation are much larger than
5 those from anthropogenic sources. Emissions of VOCs from anthropogenic sources in the 2005
6 NEI were -17 MT/year (wildfires constitute -1/6 of that total), compared to emissions from
7 biogenic sources of 29 MT/year. Vegetation emits substantial quantities of VOCs, such as
8 isoprene and other terpenoid and sesqui-terpenoid compounds. Most biogenic emissions occur
9 during the summer because of their dependence on temperature and incident sunlight. Biogenic
10 emissions are also higher in southern and eastern states than in northern and western states for
11 these reasons and because of species variations.
12 Anthropogenic NOX emissions are associated with combustion processes. Based on the
13 2005 NEI, the three largest sources of NOX are on-road and off-road mobile sources (e.g.,
14 construction and agricultural equipment) and electric power generation plants (EGUs) (US EPA,
15 2012, Figure 3-3). Emissions of NOX therefore are highest in areas having a high density of
16 power plants and in urban regions having high traffic density. However, it is not possible to
17 make an overall statement about their relative impacts on Os in all local areas because EGUs are
18 sparser than mobile sources, particularly in the west and south and because of the nonlinear
19 chemistry discussed in Section 2.1.
20 Major natural sources of NOX in the U.S. include lightning, soils, and wildfires. Biogenic
21 NOX emissions are generally highest during the summer and occur across the entire country,
22 including areas where anthropogenic emissions are low. It should be noted that uncertainties in
23 estimating natural NOX emissions are much larger than for anthropogenic NOX emissions.
24 Ozone concentrations in a region are affected both by local formation and by transport
25 from surrounding areas. Ozone transport occurs on many spatial scales including local transport
26 between cities, regional transport over large regions of the U.S. and international/long-range
27 transport. In addition, Os is also transfered into the troposphere from the stratosphere, which is
28 rich in O3; through stratosphere-troposphere exchange (STE). These inversions or "foldings" usually
29 occur behind cold fronts, bringing stratospheric air with them (U.S. EPA, 2012, section 3.4.1.1).
30 Contribution to Os concentrations in an area from STE are defined as being part of background Os
31 (U.S. EPA, 2012, section 3.4).
2-3
-------
1 2.3 EXPOSURE PATHWAYS AND IMPORTANT MICROENVIRONMENTS
2 Human exposure to Os involves the contact (via inhalation) between a person and the
3 pollutant in the various locations (or microenvironments) in which people spend their time.
4 Ozone concentrations in some indoor microenvironments, such as within homes or offices, are
5 considerably lower than Os concentrations in similarly located outdoor microenvironments,
6 primarily due to deposition processes and the transformation of Os into other chemical
7 compounds within those indoor microenvironments. Concentrations of Oj may also be quite
8 different in roadway environments, such as might occur while an individual is in a vehicle.
9 Thus, three important classes of microenvironments that should be considered when
10 evaluating population exposures to ambient Os are indoors, outdoors, and in-vehicle. Within
11 each of these broad classes of microenvironments, there are many subcategories, reflecting types
12 of buildings, types of vehicles, etc. The Os ISA evaluated the literature on indoor-outdoor Os
13 concentration relationships and found that studies consistently show that indoor concentrations
14 of Os are often substantially lower than outdoor concentrations unless indoor sources are present.
15 This relationship is greatly affected by the air exchange rate, which can be affected by open
16 windows, use of air conditioning, and other factors. Ratios of indoor to outdoor Os
17 concentrations generally range from about 0.1 to 0.4 (US EPA, 2012, section 4.3.2). In some
18 indoor locations, such as schools, there can be large temporal variability in the indoor-outdoor
19 ratios because of differences in air exchange rates over the day. For example, during the school
20 day, there is an increase in open doors and windows, so the indoor-outdoor ratio is higher during
21 the school day compared with an overall average across all hours and days. In-vehicle
22 concentrations are also likely to be lower than ambient concentrations, although the literature
23 providing quantitative estimates is smaller. Studies of personal exposure to Os have identified
24 that O3 exposures are highest when individuals are in outdoor microenvironments, such as
25 walking outdoors midday, moderate when in vehicle microenvironments, and lowest in
26 residential indoor microenvironments (US EPA, 2012, section 4.3.3). Thus the time spent
27 indoors, outdoors, and in vehicles is likely to be a critical component in estimating Os exposures.
28 Another important issue in characterizing exposure involves consideration of the extent
29 to which people in relevant population groups modify their behavior for the purpose of
30 decreasing their personal exposure to Os based on information about air quality levels made
31 public through the Air Quality Index (AQI). The AQI is the primary tool EPA has used to
2-4
-------
1 provide information on expected occurrences of high levels of 63 and other pollutants. The AQI
2 provides both the expected level of air quality in an area along with a set of actions that
3 individuals and communities can take to reduce exposure to air pollution and thus reduce the risk
4 of health effects associated with breathing ambient air pollution. There are several studies,
5 discussed in the Os ISA, that have evaluated the degree to which populations are aware of the
6 AQI and what actions individuals and communities take in response to AQI values in the
7 unhealthy range. These studies suggest that susceptible populations, such as children, older
8 adults, and asthmatics, modify their behavior in response to days with bad air quality, most
9 commonly by reducing their time spent outdoors or limiting their outdoor activity exertion level.
10 The challenge remains in how to consider averting behaviors as they currently exist within the
11 assessment tools we use and how best to quantitatively estimate the impact on estimated
12 exposures and health risks in response to improved knowledge of participation rates, the varying
13 types of actions performed particularly by potentially susceptible individuals, and the duration of
14 these averting behaviors.
15
16 2.4 AT-RISK POPULATIONS
17 The O3 ISA refers to "at risk" populations as an all-encompassing term used for groups
18 with specific factors that increase the risk of an air pollutant- (e.g., O3) related health effect in a
19 population that group (US EPA, 2012, chapter 8). Populations or lifestages can experience
20 elevated risks from O3 exposure for a number of reasons. These include high levels of exposure
21 due to activity patterns which include a high duration of time in high O3 environments, e.g.
22 outdoor recreation or work, high levels of activity which increase the dose of O3, e.g. high levels
23 of exercise, genetic or other biological factors, e.g. life stage, which predispose an individual to
24 sensitivity to a given dose of O3, pre-existing diseases, e.g. asthma or COPD, and socioeconomic
25 factors which may result in more severe health outcomes, e.g. low access to primary care can
26 lead to increased emergency department visits or hospital admissions. Modeling of exposures to
27 Os should incorporate information on time spent by potentially at-risk populations in key high Os
28 environments. This requires identification of populations with key exposure-related risk factors,
29 e.g. children or adults engaging in activities involving moderate to high levels of outdoor
30 exertion, especially on a repeated basis typical of student athletes or outdoor workers, as well as
2-5
-------
1 identifying populations with high sensitivity to 63, e.g. asthmatic children. It also requires that
2 information on 63 concentrations be carefully mapped to environments where at-risk populations
3 are likely to be exposed, e.g. near roadways where running may occur, or at schools or parks
4 where children are likely to be engaged in outdoor activities.
5 In addition to consideration of factors that lead to increased exposure to Os, modeling of
6 risk from Os exposures should incorporate additional information on factors that can lead to
7 increased dose of Os for a given exposure, e.g. increased breathing rates during periods of
8 exertion. These factors are especially important for risk estimates based on application of the
9 results of controlled human exposure studies which attempt to control for dose-related factors.
10 For risk modeling based on application of observational epidemiology results, it is also important
11 to understand characteristics of study populations that can impact observed relationships between
12 ambient Os and population health responses.
13 The Os ISA identifies a number of factors which have been associated with modifications
14 of the effect of ambient Os on health outcomes. Building on the causal framework used
15 throughout the Os ISA, conclusions are made regarding the strength of evidence for each factor
16 that may contribute to increased risk of an (Vrelated health effect based on the evaluation and
17 synthesis of evidence across scientific disciplines. The 63 ISA categorizes potential risk
18 modifying factors by the degree of available evidence. These categories include "adequate
19 evidence," "suggestive evidence," "inadequate evidence," and "evidence of no effect." See
20 Table 8-1 of the O3 ISA for a discussion of these categories (US EPA, 2012, chapter 8).
21 Factors categorized as having adequate evidence include asthma, lifestage (children <18
22 and older adults >65 are more susceptible than young and middle aged adults), diets with
23 nutritional deficiencies, and working outdoors. For example, children are the group considered
24 to be at greatest risk because they breathe more air per pound of body weight, are more likely to
25 be active outdoors when 63 levels are high, are more likely than adults to have asthma, and their
26 lungs continue to develop until they are fully grown. Factors categorized as having suggestive
27 evidence include genetic markers, sex (some studies have shown that females are at greater risk
28 of mortality from O3 compared to males), low socioeconomic status, and obesity. Factors
29 characterized as having inadequate evidence include influenza and other respiratory infections,
30 COPD, cardiovascular disease, diabetes, hyperthyroidism, race, and smoking (US EPA, 2012,
31 section 8.5, Table 8-4).
2-6
-------
1 Populations with greater proportions of individuals with characteristics associated with
2 higher risk from Os exposure are likely to have a greater risk from any given level of Os. As a
3 result, risk assessments focused on identifying populations with high levels of O3 risk should
4 focus on locations with high proportions of at-risk populations, including children and older
5 adults and people with asthma and low socioeconomic status.
6 2.5 HEALTH ENDPOINTS
7 The Os ISA identifies a wide range of health outcomes associated with short-term
8 exposure to ambient Os, including an array of morbidity effects as well as premature mortality.
9 The ISA also identifies several morbidity effects and some evidence for premature mortality
10 associated with longer-term exposures to Os. In considering health endpoints that are
11 appropriate for a risk assessment, it is useful to focus on endpoints that cover susceptible
12 populations, provide additional information about patterns or magnitude of risk, have public
13 health significance, and have sufficient information available in the literature to provide an
14 appropriate concentration-response function, in the case of epidemiological studies, or an
15 appropriate exposure-response function, in the case of controlled human exposure studies.
16 Generally speaking, epidemiology studies are well suited to risk assessment because they
17 are based on population responses to ambient air pollution exposure, and include responses of
18 populations with a wide range of susceptibility to Os. Further, such studies can evaluate serious
19 health endpoints, including hospital admissions and premature mortality. However,
20 epidemiology studies have not traditionally been based on observations of personal exposure to
21 ambient Os, and instead have used population exposure surrogates, often based on simple
22 averages of Os monitor observations. Controlled human exposure studies are also useful for risk
23 assessment, in combination with population-level assessments of exposure to ambient Os, in that
24 they are based on direct measurement of controlled O3 exposures to individuals. However,
25 controlled human exposure studies are generally focused on small numbers of relatively healthy
26 individuals, and therefore cannot represent the range of susceptibility in the population, and in
27 fact are clearly biased away from highly susceptible individuals. Controlled human exposure
28 studies also can only evaluate less serious indicators of health effects such as one-second forced
29 expiratory volume (FEV1) as an indicator of lung function or respiratory symptoms such as
2-7
-------
1 cough or pain on deep inspiration. Given the strengths and limitations in both types of studies,
2 analyses of risk using the results of both types of studies are appropriate.
3 Estimates of risk based on results of human controlled human exposure studies are
4 valuable because there is clear evidence from these studies that there is a causal relationship
5 between exposures to Os over multiple hours and reductions in lung function at moderate levels
6 of exertion. In addition, results of these studies can be applied to modeled estimates of
7 population exposure to provide additional insights into the types of population exposure
8 characteristics, including activity patterns and microenvironments that are associated with high
9 levels of risk. Estimates of risk based on results of observational epidemiology studies are
10 valuable because they often focus on more serious health endpoints which could not be assessed
11 in controlled human exposure studies. Epidemiological studies generally evaluate health
12 outcomes in an entire population or subpopulation, which includes both more sensitive and less
13 sensitive individuals, and thus may be able to identify more serious health effects in at-risk
14 subpopulations which cannot be evaluated in controlled human exposure studies which generally
15 exclude individuals likely to experience significant adverse health effects from Os exposure.
16 Epidemiological studies of 63 documented in the ISA have evaluated the relationship between
17 63 and various endpoints including respiratory symptoms, respiratory-related hospitalizations
18 and emergency department (ED) visits, and premature mortality.
19 The 63 ISA makes overall causal determinations based on the full range of evidence
20 including epidemiological, controlled human exposure and toxicological studies. Figure 2-1
21 shows the Os health effects which have been categorized by strength of evidence for causality in
22 the Os ISA (US EPA, 2012, chapter 2). These determinations support causal relationships
23 between short-term exposure to Os and respiratory effects, including respiratory-related
24 morbidity and mortality, a likely causal relationship with all-cause total mortality, and are
25 suggestive of a causal relationship for cardiovascular and central nervous system effects. The
26 determinations also support a likely causal relationship between long-term 63 exposures and
27 respiratory effects (including respiratory symptoms, new-onset asthma, and respiratory
28 mortality), and are suggestive of causal relationships between long-term O3 exposures and
29 mortality as well as cardiovascular, reproductive and developmental, and central nervous system
30 effects.
2-8
-------
Short term exposures
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
Not likely
Inadequate
to infer
Cardiovascular effects
Ccn tral nervous system
effects
Suggestive
Mortality
Likely
Respiratory effects
Causal
Cancer
Cardiovascularcffects Kespiratory effects
(morbidity and mortality)
Rep reductive and
developmental effects
Central nervous system
effects
Mortality
Long term exposures
Figure 2-1. Causal Determinations for O3 Health Effects
The ISA identifies several responses to short-term Os exposure that have been evaluated
in controlled human exposure studies (US EPA, 2012, section 6.2.1). These include decreased
inspiratory capacity, decreased forced vital capacity (FVC) and forced expiratory volume in 1
second (FEV1); mild bronchoconstriction; rapid, shallow breathing patterns during exercise;
symptoms of cough and pain on deep inspiration (PDI); and pulmonary inflammation. While
such studies provide direct evidence of relationships between short-term 63 exposure and an
array of respiratory-related effects, there are only sufficient exposure-response data at different
concentrations to develop quantitative risk estimates for Os-related decrements in FEV1.
Within the broad category of respiratory morbidity effects, the epidemiology literature
has provided effect estimates for a wide range of health endpoints associated with short-term Os
exposures which can be used in risk assessment. These health endpoints include lung function,
respiratory symptoms and medication use, respiratory-related hospital admissions and emergency
department visits. In the case of respiratory symptoms, the evidence is most consistently
supportive of the relationship between short-term ambient 63 metrics and respiratory symptoms
2-9
-------
1 and asthma medication use in children with asthma, but not for 63 and these health outcomes in
2 children without asthma. In the case of hospital admissions, there is evidence of associations
3 between shot-term ambient O3 metrics and general respiratory-related hospital admissions as
4 well as more specific asthma-related hospital admissions.
5 With regard to mortality, studies have evaluated associations between short-term ambient
6 O?, metrics and all-cause, non-accidental, and cause-specific (usually respiratory or
7 cardiovascular) mortality. The evidence from respiratory-related morbidity studies provides
8 strong support for respiratory-related mortality for which a causal determination has been made.
9 There are also a number of large studies that have found associations between 63 and all-cause
10 and all non-accidental mortality for which a likely causal determination has been made. Thus, it
11 is appropriate to assess risks for respiratory-related mortality as well as for all-cause total
12 mortality associated with Os exposure.
13 With regard to effects associated with long-term Os exposures, ISA reports a likely causal
14 relationship between Os and respiratory-related effects, including respiratory symptoms, new-
15 onset asthma, and respiratory mortality.. This suggests that for long-term exposures, when
16 comparing the evidence for respiratory-related mortality and total mortality, the evidence is most
17 supportive of risks for respiratory-related mortality, supported by the strong evidence for
18 respiratory morbidity. As a result, it is appropriate to consider including respiratory mortality
19 rather than total mortality in the risk assessment, and to give consideration to additional such
20 respiratory-related health endpoints.
21
22 2.6 REFERENCES
23 US EPA. 2012. Integrated Science Assessment of Ozone and Related Photochemical Oxidants
24 (Third External Review Draft). U.S. Environmental Protection Agency, Washington, DC,
25 EPA/600/R-10/076C, 2012.S EPA.
2-10
-------
1 3 SCOPE
2 This chapter provides an overview of the scope and key design elements of this
3 quantitative exposure and health risk assessment. The design of this assessment began with a
4 review of the exposure and risk assessments completed during the last Os NAAQS review (US
5 EPA, 2007a,b), with an emphasis on considering key limitations and sources of uncertainty
6 recognized in that analysis.
7 As an initial step in the current Os NAAQS review, in October 2009, EPA invited outside
8 experts, representing a broad range of expertise (e.g., epidemiology, human and animal
9 toxicology, statistics, risk/exposure analysis, atmospheric science) to participate in a workshop
10 with EPA staff to help inform EPA's plan for the review. The participants discussed key policy -
11 relevant issues that would frame the review and the most relevant new science that would be
12 available to inform our understanding of these issues. One workshop session focused on
13 planning for quantitative risk and exposure assessments, taking into consideration what new
14 research and/or improved methodologies would be available to inform the design of quantitative
15 exposure and health risk assessment. Based in part on the workshop discussions, EPA developed
16 a draft IRP (US EPA, 2009) outlining the schedule, process, and key policy-relevant questions
17 that would frame this review. On November 13, 2009, EPA held a consultation with CASAC on
18 the draft IRP (74 FR 54562, October 22, 2009), which included opportunity for public comment.
19 The final IRP incorporated comments from CASAC (Samet, 2009) and the public on the draft
20 plan as well as input from senior Agency managers. The final IRP included initial plans for
21 quantitative risk and exposure assessments for both human health and welfare (US EPA, 201 la,
22 chapters 5 and 6).
23 As a next step in the design of these quantitative assessments, OAQPS staff developed
24 more detailed planning documents, O?, National Ambient Air Quality Standards: Scope and
25 Methods Plan for Health Risk and Exposure Assessment (Health Scope and Methods Plan; US
26 EPA, 201 Ib) and Os National Ambient Air Quality Standards: Scope and Methods Plan for
27 Welfare Risk and Exposure Assessment (Welfare Scope and Methods Plan, US EPA, 201 Ic).
28 These Scope and Methods Plans were the subject of a consultation with CASAC on May 19-20,
29 2011 (76 FR 23809, April 28, 2011). Based on consideration of CASAC (Samet, 2011) and
30 public comments on the Scope and Methods Plan and information in the second draft ISA, we
3-1
-------
1 modified the scope and design of the quantitative risk assessment and provided a memo with
2 updates to information presented in the Scope and Methods Plans (Wegman, 2012). The Scope
3 and Methods Plans together with the update memo provide the basis for the discussion of the
4 scope of this exposure and risk assessment provided in this chapter.
5 In presenting the scope and key design elements of the current risk assessment, this
6 chapter first provides a brief overview of the quantitative exposure and risk assessment
7 completed for the previous Os NAAQS review in section 3.1, including key limitations and
8 uncertainties associated with that analysis. Section 3.2 provides a summary of the design of the
9 exposure assessment. Section 3.3 provides a summary of the design of the risk assessment based
10 on application of results of human clinical studies. Section 3.4 provides a summary of the design
11 of the risk assessment based on application of results of epidemiology studies.
12
13 3.1 OVERVIEW OF EXPOSURE AND RISK ASSESSMENTS FROM LAST
14 REVIEW
15 3.1.1 OVER VIEW OF EXPOSURE ASSESSMENT FROM LAST REVIEW
16 The exposure and health risk assessment conducted in the review completed in March
17 2008 developed exposure and health risk estimates for 12 urban areas across the U.S., which
18 were chosen, based on the location of Os epidemiological studies and to represent a range of
19 geographic areas, population demographics, and Os climatology. That analysis was in part based
20 upon the exposure and health risk assessments done as part of the review completed in 1997.l
21 The exposure and risk assessment incorporated air quality data (i.e., 2002 through 2004) and
22 provided annual or Os season-specific exposure and risk estimates for these recent years of air
23 quality and for air quality scenarios simulating just meeting the existing 8-hour Os standard and
24 several alternative 8-hour Os standards. Exposure estimates were used as an input to the risk
25 assessment for lung function responses (a health endpoint for which exposure-response functions
26 were available from controlled human exposure studies). Exposure estimates were developed for
1 In the 1994-1997 Ozone NAAQS review, EPA conducted exposure analyses for the general population,
children who spent more time outdoors, and outdoor workers. Exposure estimates were generated for 9 urban areas
for as is air quality and for just meeting the existing 1-hour standard and several alternative 8-hour standards.
Several reports that describe these analyses can be found at:
http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_pr.html.
3-2
-------
1 the general population and population groups including school age children with asthma as well
2 as all school-age children. The exposure estimates also provided information on population
3 exposures exceeding potential health effect benchmark levels that were identified based on the
4 observed occurrence of health endpoints not explicitly modeled in the health risk assessment
5 (e.g., lung inflammation, increased airway responsiveness, and decreased resistance to infection)
6 associated with 6-8 hour exposures to Os in controlled human exposure studies.
7 The exposure analysis took into account several important factors including the
8 magnitude and duration of exposures, frequency of repeated high exposures, and breathing
9 rate of individuals at the time of exposure. Estimates were developed for several indicators
10 of exposure to various levels of 63 air quality, including counts of people exposed one or
11 more times to a given O3 concentration while at a specified breathing rate, and counts of
12 person-occurrences which accumulate occurrences of specific exposure conditions over all
13 people in the population groups of interest over an Os season.
14 As discussed in the 2007 Staff Paper (US EPA, 2007c) and in Section Ha of the O3
15 Final Rule (73 FR 16440 to 16442, March 27, 2008), the most important uncertainties
16 affecting the exposure estimates were related to modeling human activity patterns over an
17 63 season, modeling of variations in ambient concentrations near roadways, and modeling
18 of air exchange rates that affect the amount of 63 that penetrates indoors. Another important
19 uncertainty, discussed in more detail in the Staff Paper (US EPA, 2007c, section 4.3.4.7),
20 was the uncertainty in energy expenditure values which directly affected the modeled
21 breathing rates. These were important since they were used to classify exposures occurring
22 when children were engaged in moderate or greater exertion and health effects observed in
23 the controlled human exposure studies generally occurred under these exertion levels for 6
24 to 8-hour exposures to O?, concentrations at or near 0.08 ppm. Reports that describe these
25 analyses (U.S. EPA, 2007a,c; Langstaff, 2007) can be found at:
26 http://www.epa.gov/ttn/naaqs/standards/O3/s_O3_index.html.
27
28 3.1.2 OVERVIEW OF RISK ASSESSMENT FROM LAST REVIEW
29 The human health risk assessment presented in the review completed in March 2008 was
30 designed to estimate population risks in a number of urban areas across the U.S., consistent with
31 the scope of the exposure analysis described above (U.S. EPA, 2007b,c). The risk assessment
3-3
-------
1 included risk estimates based on both controlled human exposure studies and epidemiological
2 and field studies. (Vrelated risk estimates for lung function decrements were generated using
3 probabilistic exposure-response relationships based on data from controlled human exposure
4 studies, together with probabilistic exposure estimates from the exposure analysis. For several
5 other health endpoints, Os-related risk estimates were generated using concentration-response
6 relationships reported in epidemiological or field studies, together with ambient air quality
7 concentrations, baseline health incidence rates, and population data for the various locations
8 included in the assessment. Health endpoints included in the assessment based on
9 epidemiological or field studies included: hospital admissions for respiratory illness in four urban
10 areas, premature mortality in 12 urban areas, and respiratory symptoms in asthmatic children in 1
11 urban area.
12 In the health risk assessment conducted in the previous review, EPA recognized that there
13 were many sources of uncertainty and variability in the inputs to the assessment and that there
14 was a high degree of uncertainty in the resulting risk estimates. The statistical uncertainty
15 surrounding the estimated Os coefficients in epidemiology-based concentration-response
16 functions as well as the shape of the exposure-response relationship chosen for the lung function
17 risk assessment were addressed quantitatively. Additional uncertainties were addressed through
18 sensitivity analyses and/or qualitatively. The risk assessment conducted for the previous 63
19 NAAQS review incorporated some of the variability in key inputs to the assessment by using
20 location-specific inputs (e.g., location-specific concentration-response functions, baseline
21 incidence rates and population data, and air quality data for epidemiological-based endpoints,
22 location specific air quality data and exposure estimates for the lung function risk assessment). In
23 that review, several urban areas were included in the health risk assessment to provide some
24 sense of the variability in the risk estimates across the U.S.
25 Key observations and insights from the 63 risk assessment, in addition to important
26 caveats and limitations, were addressed in Section II.B of the Final Rule notice (73 FR 16440 to
27 14 16443, March 27, 2008). In general, estimated risk reductions associated with going from
28 current O3 levels to just meeting the current and alternative 8-hour standards showed patterns of
29 decreasing estimated risk associated with just meeting the lower alternative 8-hour standards
30 considered. Furthermore, the estimated percentage reductions in risk were strongly influenced by
31 the baseline air quality year used in the analysis, which was due to significant year-to-year
3-4
-------
1 variability in 63 concentrations. There was also noticeable city-to-city variability in the
2 estimated (Vrelated incidence of morbidity and mortality across the 12 urban areas.
3 Uncertainties associated with estimated policy-relevant background (PRB) concentrations2 were
4 also addressed and revealed differential impacts on the risk estimates depending on the health
5 effect considered as well as the location. EPA also acknowledged that at the time of the previous
6 review there were considerable uncertainties surrounding estimates of Os C-R coefficients and
7 the shape for concentration-response relationships and whether or not a population threshold or
8 non-linear relationship exists within the range of concentrations examined in the epidemiological
9 studies.
10
11 3.2 PLAN FOR THE CURRENT EXPOSURE AND RISK ASSESSMENTS
12 The Scope and Methods Plan, including updates (U.S. EPA, 201 Ib; Wegman, 2012),
13 outlined a planned approach for conducting the current quantitative Os exposure and risk
14 assessments, including broad design issues as well as more detailed aspects of the analyses. A
15 critical step in designing the quantitative risk and exposure assessments is to clearly identify the
16 policy-relevant questions to be addressed by these assessments. More specifically, we have
17 identified the following goals for the exposure and risk assessment: (1) to provide estimates of
18 the number of people in the general population and in sensitive populations with Os exposures
19 above benchmark levels; (2) to provide estimates of the number of people in the general
20 population and in sensitive populations with impaired lung function resulting from exposures to
21 Os; (3) to provide estimates of the potential magnitude of premature mortality and/or selected
22 morbidity health effects in the population, including sensitive populations, associated with recent
23 ambient levels of Os and with just meeting the current O?, standard and any alternative standards
24 that might be considered in selected urban study areas; (4) to develop a better understanding of
25 the influence of various inputs and assumptions on the risk estimates to more clearly differentiate
26 alternative standards that might be considered including potential impacts on various sensitive
27 populations; (5) to gain insights into the distribution of risks and patterns of risk reduction and
2Policy-relevant background (PRB) ozone has been defined in previous reviews as the distribution of ozone
concentrations that would be observed in the U. S. in the absence of anthropogenic (man-made) emissions of ozone
precursor emissions (e.g., VOC, CO, NOx) in the U.S., Canada, and Mexico.
3-5
-------
1 uncertainties in those risk estimates; and (6) to understand the national mortality burden
2 associated with recent ambient Os, and how well the risk estimates for the set of urban areas
3 modeled reflect the national distribution of mortality risk. In addition, we are evaluating the
4 degree to which current evidence supports estimation of morbidity and mortality associated with
5 longer-term exposures to Os.
6 The planned approaches for conducting the exposure and risk analyses are briefly
7 summarized below. We begin with a discussion of the air quality data that will be used in both
8 the exposure and risk assessments, and then discuss each component of the exposure and risk
9 assessments.
10
11 3.2.1 AIR QUALITY DATA
12 Air quality inputs to the exposure and risk assessments include: (1) recent air quality data for
13 Os from suitable monitors and meteorological data for each selected urban study area; (2) simulated
14 air quality that reflects changes in the distribution of Os air quality estimated to occur when an area
15 just meets the current or alternative Os standards under consideration3, and (3) Os air quality
16 surfaces for recent years covering the entire continental U.S. for use in the national-scale assessment.
17 The urban area exposure and risk analyses are based on the five most recent years of air
18 quality data available at this time, 2006-2010. We are including 5 years to reflect the
19 considerable variability in meteorological conditions and the variation in Os precursor emissions
20 that have occurred in recent years. The analyses mostly focus on the Os season of May to
21 September but also include analysis of additional Os measurements during the rest of the year.
22 The required Os monitoring season varies for the urban areas as described in more detail in
23 Chapter 4.
24 Only Os data collected by Federal reference or equivalent methods (FRMs or FEMs) are
25 used in the urban area risk and exposure assessments, consistent with the use of such data in most
26 of the health effects studies. In developing the Os air quality surfaces for the national-scale
27 analysis, a combination of monitoring data and modeled Os concentrations is used to provide
3 Estimates of U.S. background concentrations (concentrations of ozone estimated to occur if
all U.S. anthropogenic emissions of NOx and VOC are eliminated) were used to set a lower bounds
for simulating air quality for just meeting the current ozone standard.
3-6
-------
1 greater coverage across the U.S. The procedure for fusing O3 monitor data with modeling results
2 is described further in Chapter 4.
3 Several O3 metrics are generated for use in the urban area exposure and risk analyses. The
4 exposure analyses use hourly O3 concentrations, while the risk analyses use several different
5 averaging times. The specific metrics used in each analysis are discussed further in following
6 chapters. In addition to temporal averages of O3 concentrations, spatial averages are also
7 generated for use in the risk analyses based on the specific averaging method applied in the
8 epidemiology studies. Based on the specific approaches used in the source epidemiology studies,
9 we develop a data set for each urban area based on a composite of all monitors according to the
10 method in the epidemiologic study. As in the last review, some monitoring sites may be omitted, if
11 needed, to best match the set of monitors that were used in the epidemiological studies.
12 Simulation of just meeting the current O3 standard is accomplished in this first draft
13 REA using a quadratic rollback method similar to what was implemented in the previous risk
14 and exposure analysis for the 2008 O3 NAAQS review (U.S. EPA, 2007a,b,c). This choice
15 was based on analyses of historical O3 data which found, from comparing the reductions over
16 time in daily ambient O3 levels in two locations with sufficient ambient air quality data, that
17 reductions tended to be roughly quadratic. Based on the current understanding of how O3
18 forms and reacts to changes in emissions, reductions in emissions that would be needed to
19 meet the current standards are likely to lead to reductions in hourly concentrations for most
20 hours of the day, but may have little impact on concentrations for some hours, and in some
21 cases can lead to increases in O3 concentrations particularly during nighttime hours. The
22 quadratic rollback method has difficulty representing these complexities in O3 chemistry and
23 reduces O3 concentrations over all hours. To address this issue in the rollback methodology for
24 the first draft REA, we are planning to impose a lower bound on O3 concentration values
25 based on modeled O3 levels after eliminating all U.S. anthropogenic emissions of O3
26 precursors (NOx and VOC). These estimates will be developed using the GEOS-Chem global
27 chemical transport model. This approach is applied so that O3 concentrations for any particular
28 hour cannot go below the estimated lower bound values.
29 For the second draft REA, we are evaluating approaches for simulating attainment of
30 current and alternative standards that are based on modeling the response of O3 concentrations to
31 reductions in anthropogenic NOx and VOC emissions, using the Higher-order Decoupled Direct
3-7
-------
1 Method (HDDM) capabilities in the Community Multi-scale Air Quality (CMAQ) model. This
2 modeling incorporates all known emissions, including emissions from nonanthropogenic sources
3 and anthropogenic emissions from sources in and outside of the U.S. As a result, the need to
4 specify values for U.S. background is not necessary, as it is incorporated in the modeling
5 directly. In simulations of just meeting the standards used to inform the exposure and risk
6 assessment, HDDM sensitivities can be applied relative to ambient measurements of Os to
7 estimate how ozone concentrations would respond to changes in anthropogenic emissions within
8 the U.S. The evaluation of this new approach is presented in Chapter 4 of this REA and in more
9 detail in Simon et al. (2012).
10 In the previous review, background 63 (referred to in that review as policy relevant
11 background, or PRB) was incorporated into the REA by calculating only risk in excess of PRB.
12 CASAC members recommended that EPA move away from using PRB in calculating risks
13 (Henderson, 2007). EPA is following this advice in the current REA, and as a result, the air
14 quality assessment will not include estimates of background Os, with the exception of providing
15 a floor for O?, concentrations when implementing the quadratic rollback method to simulate
16 attainment of the current standards. The evidence and information on background 63 that is
17 assessed in the Integrated Science Assessment (ISA) will now be considered in the Policy
18 Assessment (PA). With regard to background 63 concentrations, the PA will consider available
19 information on ambient 63 concentrations resulting from natural sources, anthropogenic sources
20 outside the U.S., and anthropogenic sources outside of North America.
21 In providing a broader national characterization of Os air quality in the U.S., this REA
22 draws upon air quality data analyzed in the Os ISA as well as national and regional trends in air
23 quality as evaluated in EPA's Air Quality Status and Trends document (U.S. EPA, 2008a), and
24 EPA's Report on the Environment (U.S. EPA, 2008b). This information along with additional
25 analyses is used to develop a broad characterization of current air quality across the nation. This
26 characterization includes tables of areas and population in the U. S. exceeding current 63
27 standards (and potential alternative standards in the second draft REA). Also included are data
28 on the expected number of days on which the O3 standards are exceeded, adjusting for the
29 number of days monitored. Further, Os levels in locations and time periods relevant to areas
30 assessed in key short-term epidemiological studies used in the risk analysis are characterized.
31 Information on the spatial and temporal characterization of Os across the national monitoring
3-8
-------
1 network is also provided. This information is used in the comparison of the attributes of the
2 selected urban study areas to national distributions of attributes to help place the results of that
3 assessment into a broader national context.
4 3.2.2 EXPOSURE ASSESSMENT
5 The scope of the exposure assessment will ultimately include the full set of 16 urban
6 areas4. For this first draft REA, we have modeled 4 of the 16 urban areas, including Atlanta,
7 Denver, Los Angeles, and Philadelphia. All 16 areas will be modeled in the second draft REA.
8 These areas were selected to be generally representative of a variety of populations, geographic
9 areas, climates, and different 63 and co-pollutant levels, and are areas where epidemiologic
10 studies have been conducted that support the quantitative risk assessment. In addition to
11 providing population exposures for estimation of lung function effects, the exposure modeling
12 will provide a characterization of urban air pollution exposure environments and activities
13 resulting in the highest exposures, differences in which may partially explain the heterogeneity
14 across urban areas seen in the risks associated with Os air pollution.
15 Population exposure to ambient Os levels will be evaluated using version 4.4 of the
16 APEX model. The model and updated documentation are available at
17 http://www.epa.gov/ttn/fera/apex_download.html. APEX is based on the current state of
18 knowledge of inhalation exposure modeling. Exposure estimates are generated for recent 63
19 levels, based on 2006-2010 air quality data, and for O3 levels resulting from simulations of just
20 meeting the current 8-hour Os NAAQS and alternative Os standards, based on adjusting 2006-
21 2010 air quality data. Exposure estimates are generated for 1) the general population, 2) school-
22 age children (ages 5 to 18), 3) asthmatic school-age children, 4) outdoor workers, and 5) the
23 elderly population (aged 65 and older). This choice of population groups includes a strong
24 emphasis on children, which reflects the results of the last review in which children, especially
25 those who are active outdoors, were identified as the most important at-risk group.
26 The exposure estimates will be used as an input to the portion of the health risk
27 assessment that is based on exposure-response relationships derived from controlled human
4 These cities are Atlanta, GA; Baltimore, MD; Boston, MA; Chicago, IL; Cleveland, OH; Dallas, TX;
Denver, CO; Detroit, MI; Houston, TX; Los Angeles, CA; New York, NY; Philadelphia, PA; Seattle, WA;
Sacramento, CA; St. Louis, MO; and Washington, D.C.
3-9
-------
1 exposure studies. The exposure analysis will also provide information on population exposure
2 exceeding levels of concern that are identified based on evaluation of health effects in the ISA.
3 It will also provide a characterization of populations with high exposures in terms of exposure
4 environments and activities. In addition, the exposure analysis will offer key observations based
5 on the results of the APEX modeling, viewed in the context of factors such as averting behavior
6 and key uncertainties and limitations of the model.
7 3.2.3 LUNG FUNCTION RISK ASSESSMENT
8 Prior EPA risk assessments for Os have included risk estimates for lung function
9 decrements and respiratory symptoms based on analysis of individual data from controlled
10 human exposure studies. The current assessment applies probabilistic exposure-response
11 relationships which are based on analyses of individual data that describe the relationship
12 between a measure of personal exposure to Os and the measure(s) of lung function recorded in
13 the study. The current quantitative risk assessment presents only a partial picture of the risks to
14 public health associated with short-term Os exposures, as controlled human exposure studies
15 have only examined markers of short-term reversible lung responses.
16 The major components in the lung function risk assessment are shown in Figure 3-1. The
17 measure of personal exposure to ambient 63 is typically some function of hourly exposures -
18 e.g., 1-hour maximum or 8-hour maximum. Therefore, the lung function risk assessment based
19 on exposure-response relationships derived from controlled human exposure study data requires
20 estimates of personal exposure to Os, typically on a 1-hour or multi-hour basis. Because data on
21 personal hourly Os exposures are not available, estimates of personal exposures to varying
22 ambient concentrations are derived through the exposure modeling described above. Controlled
23 human exposure studies, carried out in laboratory settings, are generally not specific to any particular
24 real world location. A controlled human exposure studies-based risk assessment can therefore
25 appropriately be carried out for any locations for which there are adequate air quality data on which
26 to base the modeling of personal exposures.
27 Modeling of risks of lung function decrements are based on application of results from
28 controlled human exposure studies. These studies involve volunteer subjects who are exposed
29 while engaged in different exercise regimens to specified levels of Os under controlled
30 conditions for specified amounts of time. The responses measured in such studies have included
3-10
-------
1 measures of lung function, such as forced expiratory volume in one second (FEV1), respiratory
2 symptoms, airway hyper-responsiveness, and inflammation.
3-11
-------
"As is" Ambient
Ozone Levels
Modeled
Background
Ozone Levels
Ambient Population-
Oriented Monitoring
for Selected Urban
Areas
Modeled Hour-by-Hour
Exposures Resulting From
(1) "As is" Ambient Ozone
Levels and (2) Background
Ozone Levels
Controlled Human
Exposure Studies
(various lung
function endpoints)
Exposure-Response
Relationships
Using Exposure
Metrics Based on
Hourly Ozone
Exposures
Modeled Hour-by-Hour
Exposures Resulting From
(1) "As is" Ambient Ozone
Levels and (2) When
Standards are Met
Estimates of Lung
Function Risk
Associated with "As
is" Ozone Levels Over
Background
Estimates of Lung
Function Risk
Reduction Associated
with Meeting
Standards
Current or Alternative
Standards
Figure 3- 1 Major Components of Ozone Lung Function Health Risk Assessment Based on
Controlled Human Exposure Studies
3-12
-------
1 The lung function risk assessment includes lung function decrement risk estimates, using
2 forced expiratory volume in one second (FEV1), for the general population, school age children,
3 asthmatic school age children, outdoor workers, and the elderly population (aged 65 and older)
4 living in 16 urban areas (4 of which are included in this first draft REA) in the U.S. These areas,
5 defined earlier, represent a range of geographic areas, population demographics, and O3
6 climatology. These 16 areas also include the 12 urban areas evaluated in the risk analyses based
7 on concentration-response relationships developed from epidemiological or field studies.
8 This lung function risk assessment estimates lung function decrements (> 10, > 24, and
9 >20% changes in FEV1) in children 5-18 years old associated with 8-hour exposures at moderate
10 exertion. These lung function estimates are based on applying data from adult subjects (18-35
11 years old) to children 5-18. This is based on findings from other chamber studies and summer
12 camp field studies documented in the 1996 O3 Staff Paper (US EPA, 1996a) and 1996 O3
13 Criteria Document (US EPA, 1996b), that lung function changes in healthy children are similar
14 to those observed in healthy adults exposed to O3 under controlled chamber conditions.
15 Risk estimates in this first draft REA are based in part on exposure-response relationships
16 estimated from the combined data sets from multiple O3 controlled human exposure studies. Data
17 from the studies by Folinsbee et al. (1988), Horstman et al. (1990), and McDonnell et al. (1991)
18 in addition to more recent data from Adams (2002, 2003, 2006) are used to estimate exposure-
19 response relationships for > 10, 15, and 20% decrements in FEV1. Based on additional studies
20 identified in the ISA, we will update for the second draft REA the exposure response function
21 using results from two additional recent clinical studies, Kim et al, 2011 and Schelegle, et al,
22 2009.
23 Risk measures estimated for lung function risk assessment the numbers of school age
24 children and other groups experiencing one or more occurrences of a lung function decrement
25 >10, > 15, and > 20% in an O3 season, and total number of occurrences of these lung function
26 decrements in school age children and active school age children.
27 We are also investigating the possibility of using for the second draft REA an alternative
28 model that estimates FEV1 responses for individuals associated with short-term exposures to O3
29 (McDonnell, Stewart, and Smith, 2010). This model is based on the controlled human exposure
30 data included in the prior lung function risk assessment as well as additional data sets for
31 different averaging times and breathing rates. These data were from 15 controlled human O3
3-13
-------
1 exposure studies that included exposure of 541 volunteers (ages 12 18-35 years) on a total of
2 864 occasions (see McDonnell et al., 2007, for a description of these data).
3 3.2.4 URBAN AREA EPIDEMIOLOGY BASED RISK ASSESSMENT
4 As discussed in the Os ISA (US EPA, 2012), a significant number of epidemiological and
5 field studies examining a variety of health effects associated with ambient Os concentrations in
6 various locations throughout the U.S., Canada, Europe, and other regions of the world have been
7 published since the last Os NAAQS review. As a result of the availability of these
8 epidemiological and field studies and air quality information, this first draft REA includes an
9 assessment of selected health risks attributable to recent ambient Os concentrations and health
10 risk reductions associated with attainment of the current O3 standard in selected urban locations
11 in the U.S. The second draft REA will also include assessments of the health risk reductions
12 associated with attainment of alternative Os standards.
13 The major components of the portion of the health risk assessment based on data from
14 epidemiological and field studies are illustrated in Figure 3-2. The approaches used by staff to
15 select health endpoint categories, urban areas, and epidemiology and field studies to consider for
16 inclusion in the risk assessment are discussed below. Epidemiological and field studies provide
17 estimated concentration-response relationships based on data collected in real world settings.
18 Ambient Os concentration is typically measured as the average of monitor-specific
19 measurements, using population-oriented monitors. Population health responses for O3 have
20 included population counts of school absences, emergency room visits, hospital admissions for
21 respiratory and cardiac illness, respiratory symptoms, and premature mortality. Risk assessment
22 based on epidemiological studies typically requires baseline incidence rates and population data
23 for the risk assessment locations. To minimize uncertainties introduced by extrapolation, a risk
24 assessment based on epidemiological studies can be performed for the locations in which the
25 studies were carried out.
26
3-14
-------
Air Quality
Ambient Monitoring
for Selected Urban
Study Areas
Air Quality Adjustment
Procedures to Simulate Just
Meeting Currentand
Alternative* NAAQS
Concentration-Response
Selection of
Epidemiological Studies to
Provide Concentration-
Response Functions
Baseline Health Effects Incidence Rates and Demographics
Estimates of City-specific
Baseline Health Effects
Incidence Rates
City-specific Demographic Data
Health
Risk
Model
Risk Estimates:
Recent Air Quality
Simulating Just Meeting
Current NAAQS
- Incremental difference
in risk compared to
recentair quality
Simulating Just Meeting
Alternative NAAQS
Under Consideration*
- Incremental difference
in risk compared to just
meeting current NAAQS
* This portion of the analysis will be
completed for the second draft REA
2 Figure 3- 2 Overview of Risk Assessment Model Based on Results of Epidemiologic
3 Studies
4 The design of this human health risk assessment reflects goals laid out in the Integrated
5 Review Plan (U.S. EPA, 201 la, section 5.5) including: (1) to provide estimates of the potential
6 magnitude of premature mortality and selected morbidity health effects in the populations in
7 selected urban study areas associated with recent ambient 63 levels and with just meeting the
8 current suite of 63 standards and any alternative standards that might be considered; (2) to
9 develop a better understanding of the influence of various inputs and assumptions on the risk
10 estimates; and (3) to gain insights into the distribution of risks and patterns of risk reduction and
11 uncertainties in those risk estimates.
12 As in the risk assessment for the previous Os NAAQS review, the current risk assessment
13 is focused on modeling risk for a set of selected urban study areas, chosen in order to provide
14 population coverage and to capture the observed heterogeneity in (Vrelated risk across selected
15 urban study areas. This assessment also evaluates the risk results for the selected urban areas
3-15
-------
1 within a broader national context to better characterize the nature, magnitude, extent, variability,
2 and uncertainty of the public health impacts associated with 63 exposures. This national-scale
3 assessment is discussed in the next section.
4 This risk assessment is focused on health effect endpoints for which the weight of the
5 evidence as assessed in the Os ISA supports the judgment that the overall health effect category
6 is at least likely caused by exposure to Os either alone and/or in combination with other
7 pollutants. The analysis includes estimates of mortality risk associated with short-term 8-hour O?,
8 concentrations in all 12 urban case study areas, as well as risk of hospitalization for chronic
9 obstructive pulmonary disease and pneumonia. In addition, the analysis includes additional
10 analysis of hospitalizations for additional respiratory diseases in Los Angeles, New York City,
11 and Detroit, due to limited availability of epidemiology studies covering these endpoints across
12 the 12 urban areas. The analysis also evaluates risks of respiratory related emergency
13 department visits in Atlanta and New York City, and risks of respiratory symptoms in Boston,
14 again based on availability of epidemiology studies in these locations.
15 This analysis will also consider the respiratory mortality and morbidity risks associated
16 with longer-term exposures to 63. The third draft ISA classifies respiratory effects, including
17 respiratory mortality and morbidity, as likely causally related to long-term exposures to 63.
18 However, the availability of epidemiology studies that can provide suitable C-R functions for
19 these endpoints for use in this risk assessment is limited. As a result, for this first draft REA, we
20 are providing a discussion of the potential sources of C-R functions for these endpoints, but are
21 not providing quantitative results, as we are still evaluating the appropriateness of applying the
22 results of the available epidemiology studies for this risk assessment.
23 We have identified multiple options for specifying the concentration-response functions
24 for particular health endpoints. This risk assessment provides an array of reasonable estimates
25 for each endpoint based on the available epidemiological evidence. This array of results
26 provides a limited degree of information on the variability and uncertainty in risk due to
27 differences in study designs, model specification, and analysis years, amongst other differences.
28 However, the second draft REA will provide a more comprehensive set of sensitivity analyses,
29 especially for the short-term exposure mortality estimates, for which we only provide two sets of
30 estimates based on the primary model specifications in the published studies.
3-16
-------
1 As part of the risk assessment, we address both uncertainty and variability. In the case of
2 uncertainty, we use a four-tiered approach developed by the World Health Organization (WHO)
3 and used in the risk assessment completed for the last PM NAAQS review. The WHO's four-
4 tiered approach matches the sophistication of the assessment of uncertainty to the overall
5 complexity of the risk assessment, while also considering the potential magnitude of the impact
6 that the risk assessment can have from a regulatory/policy perspective (e.g., risk assessments that
7 are complex and are associated with significant regulatory initiatives would likely be subjected
8 to more sophisticated uncertainty analysis). The WHO framework includes the use of sensitivity
9 analysis both to characterize the potential impact of sources of uncertainty on risk estimates and
10 to generate an array of reasonable risk estimates. We will implement the WHO framework more
11 completely in the second draft REA. In the case of variability, we identify key sources of
12 variability associated with Os risk (for both short-term and long-term exposure-related endpoints
13 included in the risk assessment) and discuss the degree to which these sources of variability are
14 reflected in the design of the risk assessment.
15 As part of the analysis, we also provide a representativeness analysis designed to support
16 the interpretation of risk estimates generated for the set of urban study areas included in the risk
17 assessment. The representativeness analysis focuses on comparing the urban study areas to
18 national-scale distributions for key Os-risk related attributes (e.g., demographics including
19 socioeconomic status, air-conditioning use, baseline incidence rates and ambient Os levels). The
20 goal of these comparisons is to assess the degree to which the urban study areas provide
21 coverage for different regions of the country as well as for areas likely to experience elevated Os-
22 related risk due to their specific mix of attributes related to Os risk.
23 The risk assessment is implemented using the environmental Benefits Mapping and
24 Analysis Program (BenMAP) (Abt Associates, 2008), EPA's GIS-based computer program for
25 the estimation of health impacts associated with air pollution. BenMAP draws upon a database
26 of population, baseline incidence and effect coefficients to automate the calculation of health
27 impacts. EPA has traditionally relied upon the BenMAP program to estimate the health impacts
28 avoided and economic benefits associated with adopting new air quality rules. The following
29 diagram (Figure 3-3) summarizes the data inputs (in black text) and outputs (in blue text) for a
30 typical BenMAP analysis.
31
3-17
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
Census
Population Data
Modeled Baseline
and Post-Control
Ambient PM2.5
2020 Population
Projections
Woods & Poole
Population
Projections
PM2.5 Incremental Air
Quality Change
PM2.5 Health
Functions
Economic
Valuation
Functions
PM2.5-Related Health
Impacts
Background
Incidence and
Prevalence Rates
Monetized PM2.5-
related Benefits
Blue identifies a user-selected input within the BenMAP model
Green identifies a data input generated outside of the BenMAP
Figure 3- 3 Data Inputs and Outputs for the BenMAP Model
3.2.5 NATIONAL-SCALE MORTALITY RISK ASSESSMENT
The national-scale mortality risk assessment serves two primary purposes. First, it serves
as part of the representativeness analysis discussed above, providing an assessment of the degree
to which the urban study areas included in the risk assessment provide coverage for areas of the
country expected to experience elevated mortality rates due to O3-exposure. Second, it provides a
broader perspective on the distribution of risks associated with recent Os concentrations
throughout the U.S., and provides a more complete understanding of the overall public health
3-18
-------
1 burden associated with O35. We note that a national-scale assessment such as this was completed
2 for the risk assessment supporting the latest PM NAAQS review (US EPA, 2010) with the results
3 of the analysis being used to support an assessment of the representativeness of the urban study
4 areas assessed in the PM NAAQS risk assessment, as described here for O3.
5 For short-term exposure-related mortality, the assessment provides several estimates of
6 national mortality risk, including a full national-scale estimate including all counties in the
7 continential U.S., and an analysis of just the set of urban areas included in the time series studies
8 that provide the effect estimates used to generate the risk estimates for short-term in the urban
9 case study areas. We have higher confidence in the analysis based on the large urban areas
10 included in the epidemiology studies, but the information from the full analysis of all counties is
11 useful to gain understanding of the potential magnitude of risk in less urbanized areas.
12 3.2.6 CHARACTERIZATION OF UNCERTAINTY AND VARIABILITY IN THE
13 CONTEXT OF THE O3 RISK ASSESSMENT
14 An important component of this population health risk assessment is the characterization
15 of both uncertainty and variability. Variability refers to the heterogeneity of a variable of interest
16 within a population or across different populations. For example, populations in different regions
17 of the country may have different behavior and activity patterns (e.g., air conditioning use, time
18 spent indoors) that affect their exposure to ambient O3 and thus the population health response.
19 The composition of populations in different regions of the country may vary in ways that can
20 affect the population response to exposure to O3 - e.g., two populations exposed to the same
21 levels of O3 might respond differently if one population is older than the other. Variability is
22 inherent and cannot be reduced through further research. Refinements in the design of a
23 population risk assessment are often focused on more completely characterizing variability in
24 key factors affecting population risk - e.g., factors affecting population exposure or response -in
5 In the previous O3 NAAQS review, CASAC commented that "There is an underestimation of the affected
population when one considers only twelve urban "Metropolitan Statistical Areas" (MSAs). The CASAC
acknowledges that EPA may have intended to illustrate a range of impacts rather than be comprehensive in their
analyses. However, it must be recognized that ozone is a regional pollutant that will affect people living outside
these 12 MSAs, as well as inside and outside other urban areas." Inclusion of the national-scale mortality risk
assessment partially addresses this concern by providing a broader characterization of risk for an important ozone
health endpoint.
3-19
-------
1 order to produce risk estimates whose distribution adequately characterizes the distribution in the
2 underlying population(s).
3 Uncertainty refers to the lack of knowledge regarding the actual values of inputs to an
4 analysis. Models are typically used in analyses, and there is uncertainty about the true values of
5 the parameters of the model (parameter uncertainty) - e.g., the value of the coefficient for Os in a
6 C-R function. There is also uncertainty about the extent to which the model is an accurate
7 representation of the underlying physical systems or relationships being modeled (model
8 uncertainty) - e.g., the shapes of C-R functions. In addition, there may be some uncertainty
9 surrounding other inputs to an analysis due to possible measurement error—e.g., the values of
10 daily 63 concentrations in a risk assessment location, or the value of the baseline incidence rate
11 for a health effect in a population6.
12 In any risk assessment, uncertainty is, ideally, reduced to the maximum extent possible
13 through improved measurement of key variables and ongoing model refinement. However,
14 significant uncertainty often remains, and emphasis is then placed on characterizing the nature of
15 that uncertainty and its impact on risk estimates. The characterization of uncertainty can be both
16 qualitative and, if a sufficient knowledgebase is available, quantitative.
17 The characterization of uncertainty associated with risk assessment is often addressed in
18 the regulatory context using a tiered approach in which progressively more sophisticated
19 methods are used to evaluate and characterize sources of uncertainty depending on the overall
20 complexity of the risk assessment (WHO, 2008). Guidance documents developed by EPA for
21 assessing air toxics-related risk and Superfund Site risks as well as recent guidance from the
22 World Health Organization specify multitier approaches for addressing uncertainty.
23 For the Os risk assessment, we are using a tiered framework developed by WHO to guide
24 the characterization of uncertainty. The WHO guidance presents a four-tiered approach, where
25 the decision to proceed to the next tier is based on the outcome of the previous tier's assessment.
26 The four tiers described in the WHO guidance include:
6 It is also important to point out that failure to characterize variability in an input used in modeling can also
introduce uncertainty into the analysis. This reflects the important link between uncertainty and variability with the
effort to accurately characterize variability in key model inputs actually reflecting an effort to reduce uncertainty.
3-20
-------
1 Tier 0: recommended for routine screening assessments, uses default uncertainty factors
2 (rather than developing site-specific uncertainty characterizations);
3 Tier 1: the lowest level of site-specific uncertainty characterization, involves qualitative
4 characterization of sources of uncertainty (e.g., a qualitative assessment of the general magnitude
5 and direction of the effect on risk results);
6 Tier 2: site-specific deterministic quantitative analysis involving sensitivity analysis,
7 interval-based assessment, and possibly probability bounded (high-and low-end) assessment; and
8 Tier 3: uses probabilistic methods to characterize the effects on risk estimates of sources
9 of uncertainty, individually and combined.
10 With this four-tiered approach, the WHO framework provides a means for systematically
11 linking the characterization of uncertainty to the sophistication of the underlying risk assessment.
12 Ultimately, the decision as to which tier of uncertainty characterization to include in a risk
13 assessment will depend both on the overall sophistication of the risk assessment and the
14 availability of information for characterizing the various sources of uncertainty.
15 This risk assessment for the O?, NAAQS review is relatively complex, thereby warranting
16 consideration of a full probabilistic (WHO Tier 3) uncertainty analysis. However, limitations in
17 available information prevent this level of analysis from being completed for all important
18 elements of uncertainty. In particular, the incorporation of uncertainty related to key elements of
19 C-R functions (e.g., competing lag structures, alternative functional forms, etc.) into a full
20 probabilistic WHO Tier 3 analysis would require that probabilities be assigned to each
21 competing specification of a given model element (with each probability reflecting a subjective
22 assessment of the probability that the given specification is the correct description of reality).
23 However, for most model elements there is insufficient information on which to base these
24 probabilities. One approach that has been taken in such cases is expert elicitation; however, this
25 approach is resource-and time-intensive and consequently, it is not feasible to use this technique
26 in support of this Os risk assessment.7
7 While a full probabilistic uncertainty analysis is not undertaken for this risk assessment, we provide a limited
assessment using the confidence intervals associated with effects estimates (obtained from epidemiological studies)
to incorporate statistical uncertainty associated with sample size considerations in the presentation of risk estimates.
Technically, this type of probabilistic simulation represents a Tier 3 uncertainty analysis, although as noted here, it
will be limited and only address uncertainty related to the fit of the C-R functions.
3-21
-------
For most elements of this risk assessment, rather than conducting a full probabilistic
uncertainty analysis, we include a qualitative discussion of the potential impact of uncertainty on
risk results (WHO Tierl). The second draft REA will include additional sensitivity analyses
4 assessing the potential impact of sources of uncertainty on risk results (WHO Tier 2). For
5
6
7
U.k3k3\^k3k3111g LllV^ L/WL\^11L1CI.1 1111L/U-\^L W.L OWLJ.1 \^\^O W.L LJ.11\^\^1 IClllll V Wll 1101V IV^OLJ-ILO » V V _L _L\_/ _l_l\^l -^ f • -A. *-*A
sensitivity analyses, we will include only those alternative specifications for input parameters or
modeling approaches that are deemed to have scientific support in the literature (and so represent
alternative reasonable input parameter values or modeling options). This means that the array of
risk estimates presented in this assessment are expected to represent reasonable risk estimates
9 that can be used to provide some information regarding the potential impacts of uncertainty in
10 the model elements.
11 3.2.7 PRESENTATION OF RISK ESTIMATES TO INFORM THE O3NAAQS
12 POLICY ASSESSMENT
13 We plan to conduct the risk assessment in two phases. Phase 1, presented in this first
14 draft REA, includes analysis of risk associated with recent air quality and simulating air quality
15 to just meet the current Os NAAQS. Phase 2, which will be included in the second draft REA,
16 will focus on evaluating risk associated with simulating Os air quality that just meets alternative
17 Os NAAQS under consideration.
18 We present risk estimates in two ways: (1) total (absolute) health effects incidence for
19 recent air quality and simulations of air quality just meeting the current and alternative NAAQS
20 under consideration, and (2) risk reduction estimates, reflecting the difference between (a) risks
21 associated with recent air quality compared to risks associated with just meeting the current
22 NAAQS and (b) in Phase 2, reflecting the difference between risks associated with just meeting
23 the current NAAQS compared to risks associated with just meeting alternative NAAQS under
24 consideration.
25 We present an array of risk estimates in order to provide additional context for
26 understanding the potential impact of uncertainty on the risk estimates. We include risk
27 modeled across the full distribution of Os concentrations, as well as core risk estimates ozone
28 concentrations down to zero and down to a surrogate for the lowest measured levels (LML) in
29 the epidemiology studies. According to the Os ISA, the controlled human exposure and
30 epidemiologic studies that examined the shape of the C-R function and the potential presence of
31 a threshold have indicated a generally linear C-R function with no indication of a threshold in
3-22
-------
1 analyses that have examined the 8-hour concentrations used in this risk analysis (US EPA, 2012,
2 section 2.5.4.4). The approach most consistent with the statistical models reported in the
3 epidemiological studies is to apply the concentration-response functions to all ozone
4 concentrations down to zero. However, consistent with the conclusions of the ISA, we also
5 recognize that confidence in the nature of the concentration-response function and the magnitude
6 of the risks associated with very low concentrations of ozone is reduced because there are few
7 ozone measurements at the lowest levels in many of the urban areas included in the studies. As a
8 result, the LML provides a cutoff value above which we have higher confidence in the estimated
9 risks. In our judgment, the two sets of estimates based on estimating risk down to zero and
10 estimating risk down to the LML provide a reasonable bound on estimated total risks, reflecting
11 uncertainties about the C-R function below the lowest ozone levels evaluated in the studies.
12 The results of the representativeness analysis are presented using cumulative probability
13 plots (for the national-level distribution of Os risk-related parameters) with the locations where
14 the individual urban study areas fall within those distributions noted in the plots using vertical
15 lines. Similar types of plots are used to present the distribution of national-scale mortality
16 estimates based on the national-scale risk assessment, showing the location of the urban case
17 study areas within the overall national distribution.
18
19 3.3 REFERENCES
20 Abt Associates Inc. (2008). Environmental Benefits Mapping and Analysis Program (Version
21 3.0). Bethesda, MD. Prepared for Environmental Protection Agency, Office of Air Quality
22 Planning and Standards, Air Benefits and Cost Group. Research Triangle Park, NC.
23 Folinsbee, L.J.; McDonnell, W.F.; Horstman, D.H. (1988). Pulmonary function and symptom
24 responses after 6.6-hour exposure to 0.12 ppm Os with moderate exercise. JAPCA. 38: 28-35.
25 Henderson, R. 2007. Clean Air Scientific Advisory Committee's (CAS AC) Review of the
26 Agency's Final O3 Staff Paper. EPA-CASAC-07-002. March 26.
3-23
-------
1 Horstman, D.H.; Folinsbee, L.J.; Ives, P.J.; Abdul-Salaam, S.; McDonnell, W.F. (1990). O3
2 concentration and pulmonary response relationships for 6.6-hr exposures with five hrs of
3 moderate exercise to 0.08, 0.10, and 0. \2ppm. Am Rev RespirDis. 142:1158-1163.
4 Kim, Chong S., Neil E. Alexis, Ana G. Rappold, Howard Kehrl, Milan J. Hazucha, John C. Lay,
5 Mike T. Schmitt, Martin Case, Robert B. Devlin, David B. Peden, and David Diaz-Sanchez.
6 "Lung Function and Inflammatory Responses in Healthy Young Adults Exposed to 0.06 ppm O3
7 for 6.6 Hours." American Journal of Respiratory and Critical Care Medicine 183, no. 9 (2011):
8 1215-1221.
9 Langstaff, J.E. (2007). OAQPS Staff Memorandum to O3 NAAQS Review Docket (EPA
10 HQ-OAR-2005-0172). Subject: Analysis of Uncertainty in O3 Population Exposure Modeling.
11 [January 31, 2007]. Available at: http://www.epa.gov/ttn/naaqs/standards/O3/s_O3_cr_td.html
12 McDonnell, W.F. et al. (1991). Respiratory response of humans exposed to low levels of O3 for
13 6.6 hours. American Review of Respirtory Disease 147:804-810.
14 McDonnell W.F., Stewart P.W., Smith M. V. (2007). The temporal dynamics of O3 -induced
15 FEV1 changes in humans: an exposure-response model. Inhal Toxicol 19:483-494.
16 McDonnell W.F., Stewart P.W., Smith M.V. (2010). Prediction of O3 -induced lung function
17 responses in humans. Inhal Toxicol. 22(2): 160-8.
18 Samet, J. 2009. Consultation on EPA's Draft Integrated Review Plan for the National Ambient
19 Air Quality Standards for Paniculate Matter. EPA-CASAC-10-004. Decembers.
20 Samet, J. 2011. Consultation on EPA's O3 National Ambient Air Quality Standards: Scope and
21 Methods Plan for Health Risk and Exposure Assessment (April 2011) and O3 National Ambient
22 Air Quality Standards: Scope and Methods Plan for Welfare Risk and Exposure Assessment
23 (April 2011). EPA-CASAC-11-008. June 21.
3-24
-------
1 Schelegle, Edward S., Christopher A. Morales, William F. Walby, Susan Marion, and Roblee P.
2 Allen. "6.6-Hour Inhalation of Os Concentrations from 60 to 87 Parts per Billion in Healthy
3 Humans." American Journal of Respiratory and Critical Care Medicine 180 (2009): 265-272.
4 Simon, H., Baker, K., Possiel, N., Akhtar, F., Napelenok, S., Timin, B., Wells, B. (2012) Model-
5 based rollback using the higher order direct decoupled method (HDDM). Available on the
6 Internet at: http://www.epa.gov/ttn/naaqs/standards/ozone/s o3 2008 rea.html.
7
8 US EPA. 1996a. Review of National Ambient Air Quality Standards for Os : Assessment of
9 Scientific and Technical Information - OAQPS Staff Paper. EPA/452/R-96-007. Office of Air
10 Quality Planning and Standards, Research Triangle Park, NC. Available from: NTIS,
11 Springfield, VA; PB96-203435. Available at:
12 http://www.epa.gov/ttn/naaqs/standards/O3 /s_o3_pr.html
13 US EPA. 1996b. Air Quality Criteria for O3 and Related Photochemical Oxidants. EPA/600/P-
14 93/004aF-cF. Office of Research and Development, National Center for Environmental
15 Assessment, Research Triangle Park, NC. Available at:
16 http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=2831.
17 US EPA. 2007a. Os Population Exposure Analysis for Selected Urban Areas. Office of Air
18 Quality Planning and Standards, RTF, NC. EPA-452/R-07-010. July.
19 US EPA. 2007b. O3 Health Risk Assessment for Selected Urban Areas. Office of Air Quality
20 Planning and Standards, RTF, NC. EPA 452/R-07-009. July.
21 US EPA, 2007c. Review of the National Ambient Air Quality Standards for O3: Policy
22 Assessment of Scientific and Technical Information. OAQPS Staff Paper. Research Triangle
23 Park, NC: Office of Air Quality Planning and Standards; report no. EPA-452/R-07-007a.
24 Available at: http://www.epa.gov/ttn/naaqs/standards/O3/s_O3_cr_sp.html
3-25
-------
1 US EPA. 2008a. National Air Quality: Status and Trends Through 2007. Office of Air Quality
2 Planning and Standards. Research Triangle Park, NC. EPA-454/R-08-006. November.
3 Available at: http://www.epa.gov/airtrends/2008/index.html.
4 US EPA 2008b. EPA 's Report on the Environment. Washington, DC EPA/600/R-07/045F. May
5 2008.
6 Available at: http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=190806
7 US EPA. 2009. Integrated Review Plan for the O3 National Ambient Air Quality
8 Standards Review: External Review Draft. Environmental Media Assessment Group,
9 National Center for Environmental Assessment and Health and Environmental Impacts Division,
10 Office of Air Quality Planning and Standards, RTF, NC. EPA 452/D-09-001. September.
11 US EPA. 2010. Quantitative Health Risk Assessment for Particulate Matter. Office of Air
12 Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-10-005. Available at:
13 http://www.epa.gov/ttn/naaqs/standards/pm/data/PM_RA_FINAL_June_2010.pdf
14 US EPA, 201 la. Integrated Review Plan for the O3 National Ambient Air Quality Standards.
15 National Center for Environmental Assessment, Office of Research and Development and Office
16 of Air Quality Planning and Standards, Office of Air and Radiation. RTF, NC. EPA 452/R-11 -
17 006. April.
18 US EPA, 201 Ib. O3 National Ambient Air Quality Standards: Scope and Methods Plan for
19 Health Risk and Exposure Assessment. Office of Air Quality Planning and Standards. RTF, NC.
20 EPA-452/P-11-001. April.
21 US EPA, 201 Ic. O3 National Ambient Air Quality Standards: Scope and Methods Plan for
22 Welfare Risk and Exposure Assessment. Office of Air Quality Planning and Standards. RTF,
23 NC. EPA-452/P-11-002. April.
3-26
-------
1 Wegman, L. 2012. Updates to information presented in the Scope and Methods Plans for the 63
2 NAAQS Health and Welfare Risk and Exposure Assessments. Memorandum from Lydia
3 Wegman, Division Director, Health and Environmental Impacts Division, Office of Air Quality
4 Planning and Standards, Office of Air and Radiation, US EPA to Holly Stallworth, Designated
5 Federal Officer, Clean Air Scientific Advisory Committee, US EPA Science Advisory Board
6 Staff Office. May 2, 2012.
7 World Health Organization. 2008. Harmonization Project Document No. 6. Part 1: Guidance
8 Document on Characterizing and Communicating Uncertainty in Exposure Assessment.
9 Available at: http://www.who.int/ipcs/methods/harmonization/areas/exposure/en/.
10
11
3-27
-------
1 4 AIR QUALITY CONSIDERATIONS
2 4.1 INTRODUCTION
3 Air quality information is used in the risk and exposure analyses (Chapters 5-7) to assess
4 risk and exposure resulting from recent O^ concentrations, as well as to estimate the relative
5 change in risk and exposure resulting from adjusted Oj concentrations after simulating just
6 meeting the current Os standard of 0.075 ppm. For the population exposure analyses discussed in
7 Chapter 5, 16 urban areas will ultimately be modeled1. Four of these urban areas are modeled
8 for this first draft REA, and as a result, air quality information from those 4 urban areas was
9 analyzed for this first draft. The four urban areas evaluated for this first draft include Atlanta,
10 GA; Denver, CO; Los Angeles, CA and Philadelphia, PA. The lung function risk assessment
11 discussed in Chapter 6 uses the same air quality data as the population exposure assessment and
12 models the same four urban areas for the first draft. For the epidemiology-based risk assessment
13 discussed in Chapter 7, 12 of the 16 areas evaluated for population exposure are included, and air
14 quality data for all 12 of these urban areas were analyzed. These 12 urban areas include the 4
15 cities evaluated in the first draft exposure assessment as well as: Baltimore, MD; Boston, MA;
16 Cleveland, OH; Detroit, MI; Houston, TX; New York, NY; Sacramento, CA; and St. Louis, MO.
17 In addition, Chapter 8 includes an assessment of the national-scale Os mortality risk burden
18 based on national-scale air quality information. This chapter describes the air quality information
19 used in these analyses, providing an overview of monitoring data and air quality (section 4.2) as
20 well as an overview of air quality inputs to the risk and exposure assessments (section 4.3).
21 4.2 OVERVIEW OF OZONE MONITORING AND AIR QUALITY
22 To monitor compliance with the NAAQS, state and local environmental agencies operate
23 Os monitoring sites at various locations, depending on the population of the area and typical peak
24 O3 concentrations (US EPA, 2012a, sections 3.5.6.1, 3.7.4). In 2010, there were 1,250 state and
25 local Os monitors reporting concentrations to EPA (US EPA, 2012a, Figures 3-21 and 3-22).
26 The minimum number of Os monitors required in a Metropolitan Statistical Area (MSA) ranges
27 from zero, for areas with a population under 350,000 and with no recent history of an Os design
28 value greater than 85% of the NAAQS, to four, for areas with a population greater than 10
29 million and an Os design value greater than 85% of the NAAQS.2 In areas for which Os
1 These cities are Atlanta, GA; Baltimore, MD; Boston, MA; Chicago, IL; Cleveland, OH; Dallas, TX; Denver, CO;
Detroit, MI; Houston, TX; Los Angeles, CA; New York, NY; Philadelphia, PA; Seattle, WA; Sacramento, CA; St.
Louis, MO; and Washington, D.C.
2The current monitor and probe siting requirements have an urban focus and do not address siting in non-urban, rural
areas. States may operate ozone monitors in non-urban or rural areas to meet other objectives (e.g., support for
research studies of atmospheric chemistry or ecosystem impacts).
4-1
-------
1
2
3
4
5
6
monitors are required, at least one site must be designed to record the maximum concentration
for that particular metropolitan area. Since 63 concentrations often decrease significantly in the
colder parts of the year in many areas, Os is required to be monitored only during the "ozone
season," which varies by state (US EPA, 2012a, section 3.5.6 and Figure 3-20).3
7
8
9
10
11
12
13
14
15
8-hour Ozone Design Values, 2008-2010
• 40-65 ppb (249 Sites)
O 66-70 ppb (309 Sites)
O 71-75 ppb (303 Sites)
• 76-90 ppb (168 Sites)
• 91-112 ppb (36 Sites)
Figure 4-1 Individual monitor 8-h daily max Os design values displayed for the 2008-
2010 period (U.S. EPA, 2012, Figure 3-52A)
Figure 4-1 shows the location and 8-h Oj, design values (3-year average of the annual 4*
highest daily maximum 8-hour Oi concentration) for all available monitors in the US for the
2008-2010 period. All 12 of the selected urban areas have 2008-2010 8-h O3 design values at or
above the current standard. Figure 4-2 shows how the 4th highest 8-h daily max Os
concentrations vary for each of the 12 urban areas from 2006-2010. In general, all twelve cities
Some States and Territories operate ozone monitors year-round, including Arizona, California, Hawaii, Louisiana,
Nevada, New Mexico, Puerto Rico, Texas, American Samoa, Guam and the Virgin Islands.
4-2
-------
1
2
3
4
5
show a decrease in 03 concentrations between 2006 and 2010, with an average decrease in the 4*
highest 8-h daily max Oj, concentration of 9 ppb. However, there is significant year-to-year
variability, with some locations, such as Sacramento and Houston, showing increases in some
years relative to 2006 even though the 2010 values are somewhat lower.
6
7
8
9
10
11
12
13
14
15
16
17
18
Changes in Ozone Air Quality in Selected 12 Urban
Areas 2006-2010
130
2010
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
Saint Louis
Figure 4-2 Trends in 8-h daily max 63 for the selected 12 urban areas analyzed in the
risk and exposure assessment for 2006-2010 (annual 4th highest 8-h daily
max Os concentrations in ppm)
Table 4-1 gives the number of monitors and the required 63 monitoring season for each
of the 12 selected urban areas. The counties listed as part of each of the 12 urban areas are based
on the counties included in the Zanobetti and Schwartz (2008) study of 63 and mortality in 48
U.S. cities between 1989 and 2000, which is used in the epidemiology-based health risk
assessment4. Also listed in Table 4-1 are the 8-h O3 design values for 2006-2008 and 2008-2010.
All of the cities, except for Sacramento (which showed no change), had a decrease in the 63
design value concentrations between the two 3-year periods with an average change of 7 ppb.
It should be noted that the counties included in Table 4-1 are those analyzed in the epidemiology-based risk
assessment (Chapter 7) but differ from the counties included in the population exposure (Chapter 5) and the lung
function risk assessment (Chapter 6). These differences are explained in Chapters 5-7.
4-3
-------
1 Table 4-1: Information on the 12 Urban Case Study Areas in the Risk Assessment
Study Area
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
Counties5
Cobb County, GA
DeKalb County, GA
Fulton County, GA
Gwinnett County, GA
Baltimore City, MD
Baltimore County, MD
Middlesex County, MA
Norfolk County, MA
Suffolk County, MA
Cuyahoga County, OH
Denver County, CO
Wayne County, MI
Harris County, TX
Los Angeles County, CA
Bronx County, NY
Kings County, NY
New York County, NY
Queens County, NY
Richmond County, NY
Philadelphia County, PA
Sacramento County, CA
St. Louis City, MO
St. Louis County, MO
Population
(2010)
3,105,873
1,425,990
2,895,958
1,280,122
600,158
1,820,584
4,092,459
9,818,605
8,175,133
1,526,006
1,418,788
1,318,248
#ofO3
Monitors
5
3
5
4
3
4
17
17
8
4
8
8
Required Oj
Monitoring
Season
March -
October
April -
October
April -
September
April -
October
M^arch -
September
April -
September
January -
December
January -
December
April -
October
April -
October
January -
December
April -
October
2006-
2008
(PPb)6
95
91
82
84
86
82
91
119
89
92
102
85
2008-
2010
(PPb)6
80
89
76
77
78
75
84
112
84
83
102
77
Counties listed here reflect those included in the Zanobetti and Schwartz (2008) study of ozone and mortality in
48 U.S. cities between 1989 and 2000.
6 These are values of the highest 4th high 8-h max average (ppb) for the counties listed for each urban area. It
should be noted that sometimes monitors with higher values occurred within the urban area but outside of the
counties included in the Zanobetti and Schwartz (2008) study and those values are not included in this table.
4-4
-------
1 4.3 OVERVIEW OF AIR QUALITY INPUTS TO RISK AND EXPOSURE
2 ASSESSMENTS
3 The air quality information input into the risk and exposure assessments includes both
4 recent air quality data from the years 2006-2010, as well as air quality data adjusted to reflect
5 just meeting the current Os standard of 0.075 ppm. In this section, we summarize these air
6 quality inputs and discuss the methodology used to simulate air quality to meet the current
7 standard. Additional information is provided in Wells et al. (2012) and Simon et al. (2012).
8
9 4.3.1 Urban-scale Air Quality Inputs
10 4.3.1.1 Recent Air Quality
11 The air quality monitoring data used to inform the first draft Ozone Risk and Exposure
12 Assessments were hourly Oi concentrations collected between 1/1/2006 and 12/31/2010 from all
13 US monitors meeting EPA's siting, method, and quality assurance criteria in 40 CFR Part 58.
14 These data were extracted from EPA's Air Quality System (AQS) database7 on June 27, 2011.
15 Regionally concurred exceptional event data (i.e. data certified by the monitoring agency to have
16 been affected by natural phenomena such as wildfires or stratospheric intrusions, and concurred
17 upon by the EPA regional office) were not included in the assessments. However, concurred
18 exceptional events were rare, accounting for less than 0.01% of the total observations. All
19 concurred exceptional events in 2006-2010 were related to wildfires in California in 2008. There
20 were no concurrences of exceptional event data for stratospheric intrusions in 2006-2010 in the
21 data extracted on June 27, 2011.
22 In order to compare the monitoring data to the NAAQS, the data were split into two
23 overlapping 3-year periods, 2006-2008 and 2008-2010. The Oi monitors were checked for data
24 completeness within each period, and all monitors lacking sufficient data to calculate a valid 3-
25 year design value were excluded (see 40 CFR Part 50, Appendix P). All subsequent air quality
26 data analyses described in this chapter were performed separately on the monitoring data within
27 each of the two design value periods.
28 The sections below summarize the recent air quality data input into the epidemiological
29 study-based risk assessment, and the exposure and clinical study-based risk assessment. More
30 details on these inputs are also provided in Wells et al. (2012).
31
7 EPA's Air Quality System (AQS) database is a state-of-the-art repository for many types of air quality and related
monitoring data. AQS contains monitoring data for the six criteria pollutants dating back to the 1970's, as well as more recent
additions such as air toxics, meteorology, and quality assurance data. At present, AQS receives ozone monitoring data collected
hourly from over 1,300 monitors, and is quality assured by one of over 100 state, local, or tribal air quality monitoring agencies.
4-5
-------
1 Epidemiology Based Risk Assessment
2 Air quality concentration data for the epidemiology-based risk analyses are input into the
3 environmental Benefits Mapping and Analysis Program (BenMAP; Abt Associates, 2010a) for
4 assessment. Gaps of 1 or 2 hours in the hourly concentration data were interpolated. These short
5 gaps tend to occur at regular intervals in the monitoring data due to a requirement for monitoring
6 agencies to turn off their monitors for brief periods in order to perform quality control checks.
7 Generally, quality control checks are performed during nighttime hours (between 12:00 AM and
8 6:00 AM) when Os concentrations tend to be lowest. Missing intervals of 3 hours or more were
9 infrequent and were not replaced.
10 The air quality monitoring data for the 12 urban areas were area-wide spatial averages of
11 the hourly Os concentrations within each area. The area boundaries were chosen to match the
12 study areas in Zanobetti & Schwartz (2008) which generally covered the urban population
13 centers within the larger metropolitan areas. The ambient data from the monitors within each
14 area were averaged hour-by-hour within EPA's required Os monitoring season. Although some
15 monitoring data were collected outside of the required season, often fewer monitors in an area
16 remained in operation outside of the required season.
17 For input into BenMAP, four daily metrics were calculated from the spatially averaged
18 hourly Os concentrations. These metrics were:
19 1. Daily maximum 1-hour concentration
20 2. Daily maximum 8-hour concentration
21 3. Daytime 8-hour average concentration (10:OOAM to 6:OOPM)
22 4. Daily 24-hour average concentration
23
24 Exposure Modeling and Clinical Study Based Risk Assessment
25 For the exposure modeling and clinical study based risk assessment, the air quality data are input
26 in the Air Pollutants Exposure (APEX) model, also referred to as the Total Risk Integrated
27 Methodology Inhalation Exposure (TREVI.Expo) model (U.S. EPA, 2012b,c). For estimating
28 ambient Os concentrations to use in the exposure model, we use hourly Os concentrations from
29 the AQS. The specific monitors used in the urban areas modeled and the method for estimating
30 and replacing missing data are described in Appendix 4-B.
31
32 4.3.1.2 Air Quality after Simulating "Just Meeting" Current O3 Standard
33 In addition to recent air quality concentrations, the risk and exposure assessments also
34 consider the relative change in risk and exposure when considering the distribution of 63
35 concentrations after simulating "just meeting" the current Os standard of 0.075 ppm. The
4-6
-------
1 sections below summarize the methodology applied for this first draft REA to simulate just
2 meeting the current NAAQS by "rolling back" the baseline distribution of recent 63
3 concentrations and an alternative simulation approach being considered for the 2n draft of the
4 REA. More details on these inputs are also provided in Wells et al. (2012), and a more complete
5 description of the alternative simulation approach is provided in Simon et al. (2012).
6
1 Methods
8 The "quadratic rollback" method was used in the previous Os NAAQS review to adjust
9 ambient 63 concentrations to simulate minimally meeting current and alternative standards (U.S.
10 EPA, 2007). As the name implies, quadratic rollback uses a quadratic equation to reduce high
11 concentrations at a greater rate than low concentrations. The intent is to simulate reductions in
12 63 resulting from unspecified reductions in precursor emissions, without greatly affecting
13 concentrations near ambient background levels (Duff et al., 1998).
14 Two independent analyses (Johnson, 2002; Rizzo, 2005; 2006) were conducted to
15 compare quadratic rollback with other methods such as linear (proportional) rollback and
16 distributional (Weibull) rollback. Both analyses used different rollback methods to reduce
17 concentrations from a high 63 year to simulate levels achieved during a low 63 year, then
18 compared the results to the ambient concentrations observed during the low Os year. Both
19 analyses concluded that the quadratic rollback method resulted in an 8-hour 03 distribution most
20 similar to that of the ambient concentrations.
21 In this review, quadratic rollback was used to simulate reductions in Os concentrations in
22 areas which failed to meet EPA's current Os NAAQS of 0.075 ppm (75 ppb). Hourly Os
23 concentrations were reduced so that the highest design value in each area was exactly 75 ppb, the
24 highest value meeting the NAAQS. Concentrations at the remaining monitors in each area were
25 similarly reduced using the quadratic rollback coefficients calculated at the highest monitor.
26 Quadratic rollback was performed independently within each area for two design value periods,
27 2006-2008 and 2008-2010. In some of the 12 urban areas, the monitor with the highest design
28 value was not within the area boundaries chosen to match the study areas in Zanobetti &
29 Schwartz (2008). In these cases, the high monitor was included in the quadratic rollback, and the
30 ozone concentrations at the monitors within the Zanobetti & Schwartz (2008) study area were
31 similarly reduced. In this way, while the high monitor outside of the study area would have been
32 simulated to have a design value of 75 ppb to just meet the standard, the design value at the
33 monitors within the study area would have been simulated to have design values below 75 ppb.
34 To avoid reducing Os concentrations below background levels, background "floor"
35 values were set defining minimum values beyond which quadratic rollback would not be applied.
4-7
-------
1 Background concentrations were estimated from two GEOS-Chem modeling simulations for the
2 model year of 2006: one with zero U.S. anthropogenic emissions (i.e. U.S. background) but with
3 all other anthropogenic and natural emissions globally, and the other with all anthropogenic and
4 biogenic emissions included (i.e. base case) (Zhang et al., 2011). The monitors in each study
5 area were paired with their appropriate GEOS-Chem grid cells, potentially matching multiple
6 monitors to the same cell. The paired hourly U.S. background and base case concentrations were
7 then spatially averaged in the same way as the 63 monitoring data (as described in 4.3.1.1).
8 Medians by area, month, and hour of the day were calculated from the spatially-averaged U.S.
9 background and base case modeled concentrations, and ratios of the U.S. background to base
10 case concentrations were calculated to provide monthly diurnal profiles of the ratio of U.S.
o
11 background to total ozone for every month for every area . Next, the U.S. background ratios
12 were multiplied by the respective monitored values in each of the 5 years, 2006-2010, to obtain
13 the U.S. background floor values.
14 The U.S. background floor values were compared to the hourly "rolled back" air quality
15 values for each area. If there was an hour for which the 63 concentration had been "rolled back"
16 to below the U.S. background floor value, then that hourly concentration value was set equal to
17 whichever was lower: the U.S. background floor value or to the original monitored O^
18 concentration value for that hour.
19 Figure 4-3 shows diurnal profiles of seasonally averaged U.S. background floor values
20 for the 12 urban case study areas in the risk assessment. The U.S. background floor values show
21 a diurnal pattern similar to that of the observed 63 concentrations, with the highest values
22 occurring in the early afternoon hours and the lowest values occurring around sunrise.
23 Generally, the highest U.S. background values occurred in the spring, while the other three
24 seasons were more difficult to distinguish. Denver was a notable exception to this pattern,
25 having nearly identical U.S. background floor values in the spring and summer months.
26 Figure 4-4 shows box-and-whisker plots of the U.S. background floor values in the 12
27 urban case study areas. The distribution of the U.S. background floor values varied from area to
28 area, but generally ranged from near 0 to between 30 and 40 ppb, with median between 10 and
29 20 ppb.
30
Values were set equal to one, if greater than one.
4-8
-------
1
2
3
4
5
6
1
8
o
-=1-
Atlanta
'*«.
o
-=1-
CD
en
o
CN
Baltimore
Boston
CD
f-4
Cleveland
6 12 18 24
Denver
o
c-4
6 12 18 24 0
Detroit o
6 12 18 24
Houston
o
CN
0 6 12 18 24 0
CD Hew York o
12 18 24
CD
(N
CD
c-t
Philadelphia
CD
-=J-
CD
r-j
6 12 18 24 0
Sacramento CD
1=1
CN
6 12 18 24
Los Angeles
6 12 18 24
Saint Louis
0 6 12 18 24
* * * * Summer
0 6 12 18
* * * • Autumn
24
0 6 12 18 24
* * * * Winter
0 6 12 18 24
• • • * Spring
Figure 4-3 Diurnal Profiles of Seasonally Averaged U.S. Background Floor Values in
the Urban Case Study Areas
Notes: Values shown are 2006-2010 averages, in parts per billion. Seasons were defined as Spring = March - May,
Summer = June - August, Autumn = September - November, Winter = December - February. Winter values are
missing for Cleveland because no monitoring data were available for that period.)
4-9
-------
LTJ
1
2
3
4
5
6
7
8
9
10
11
12
13
14
(
(
T3
—
T3
I
:,
1
1
4
1
5
_l
ri
C
D
n
D
0
1
I
Zi
Tii
u
>
LI
J
C
5
>
5
D
-
'
C
_ i
5
u
D
-
5
fl
D
D
i
i
1
=
yl
u
D
31
i
yl
D
i
(
*-
^
^
<*
5
D
j
i
c
^
5
u
3
TJ
L
"
D
U
^
J
73
-
(
fs
3
D
J
__i
T3
0
Figure 4-4 Distribution of U.S. Background Floor Values in the Urban Case Study
Areas
Table 4-2 contains selected summary statistics generated to evaluate the frequency and
magnitude of the U.S. background adjustments in the quadratic rollback procedure. Overall,
over 20% of the rollback concentrations were adjusted, however, the average magnitude of the
adjustments was very small (< 0.2 ppb), and even the largest adjustment was less than 5 ppb.
Over 95% of the adjustments simply returned the rollback concentrations to their original
monitored values instead of the modeled U.S background value, and again the average
magnitude of the adjustment was very small (< 0.2 ppb). In conclusion, the U.S. background
adjustment procedure had little effect on the rollback concentrations.
Table 4-2 Frequency and Magnitude of the U.S. Background Adjustments, 2006 - 2008
Urban Area
Atlanta
Baltimore
Boston
Cleveland
% Rollback
Values
Adjusted
16.7
19.7
16.4
20.0
% Replaced
with Monitor
Values
97.2
96.8
96.3
96.2
% Replaced
with Floor
Values
2.8
3.2
3.7
3.8
Average
Adjustment
(ppb)
0.10
0.15
0.17
0.18
Maximum
Adjustment
(ppb)
2.3
2.2
1.2
1.6
4-10
-------
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
Saint Louis
OVERALL
14.4
14.9
28.4
24.6
16.4
18.7
24.3
12.8
20.5
96.2
96.8
96.4
93.9
96.7
96.2
92.1
97.1
95.5
3.8
3.2
3.6
6.1
o o
J.J
3.8
7.9
2.9
4.5
0.20
0.13
0.15
0.29
0.09
0.16
0.34
0.11
0.17
2.4
1.3
1.6
4.5
1.4
2.0
3.0
1.1
4.5
1
2
3
4
5
6
7
8
9
10
Figure 4-5 shows seasonal average diurnal profiles of the observed and rollback composite
monitor values in the 12 urban case study areas for 2006-2008. The gray and blue lines are
averages over the required Os monitoring season (see Table 4-1), while the red and green lines
are averages over the "peak" O?, months, June - August. The June - August averages are higher
than the 63 season averages, except in Houston where the highest 63 concentrations are often
observed in April-May and September-October. The diurnal patterns are generally quite similar
from area to area, with most of the variation occurring in the peak concentration heights during
the daytime hours.
4-11
-------
o
CD
CD
CM
Atlanta
o
CO
Baltimore
o
-------
1 For this first draft of the REA, we have evaluated approaches for simulating attainment of
2 current and alternative standards that are based on modeling the response of 63 concentrations to
3 reductions in anthropogenic NOX and VOC emissions, using the Higher-Order Decoupled Direct
4 Method (HDDM) capabilities in the Community Multi-scale Air Quality (CMAQ) model. This
5 modeling incorporates all known emissions, including emissions from non-anthropogenic
6 sources and anthropogenic emissions from sources in and outside of the U.S. As a result, the
7 need to specify values for U.S. background concentrations is not necessary, as it is incorporated
8 in the modeling directly. In simulations of just meeting the standards used to inform the
9 exposure and risk assessment, HDDM sensitivities can be applied relative to ambient
10 measurements of 63 to estimate how ozone concentrations would respond to changes in
11 anthropogenic emissions within the U.S. Application of this approach also addresses the
12 recommendation by the National Research Council of the National Academies (NRC, 2008) to
13 explore how emissions reductions might effect temporal and spatial variations in 63
14 concentrations, and to include information on how NOX versus VOC control strategies might
15 affect risk and exposure to 63. The new approach using HDDM, discussed in detail in Simon et
16 al., 2012, seems promising, and EPA staff propose to use it in simulating just meeting the current
17 and alternative 03 standards for the second draft of the REA.
18
19 4.3.2 National-scale Air Quality Inputs
20 In contrast to the urban study areas analysis, the national-scale analysis employs a data
21 fusion approach that takes advantage of the accuracy of monitor observations and the
22 comprehensive spatial information of the CMAQ modeling system to create a national-scale
23 "fused" spatial surface of seasonal average 03. The spatial surface is created by fusing 2006-
24 2008 measured Oi concentrations with the 2007 CMAQ model simulation, which was run for a
25 12 km gridded domain, using the EPA's Model Attainment Test Software (MATS; Abt
26 Associates, 201 Ob), which employs the Voronoi Neighbor Averaging (VNA) technique (Timin et
27 al., 2010) enhanced with information on the spatial gradient of 63 provided by CMAQ results.
28 More details on the ambient measurements and the 2007 CMAQ model simulation, as well as the
29 spatial fusion technique, can be found in Wells et al. (2012) and Hall et al. (2012). It should also
30 be noted that this same spatial fusion technique was employed for a national-scale risk
31 assessment by Fann et al. (2012) to produce "fused" spatial fields for 03 and PM2.5 and in the PM
32 NAAQS REA to produce a national-scale spatial field for PM2.5 (U.S. EPA, 2010).
33 Two "fused" spatial surfaces were created for: (1) the May-September mean of the 8-hr
34 daily maximum (consistent with the metric used by Bell et al. 2004); and (2) the June-August
35 mean of the 8-hr daily mean from 10am to 6pm (consistent with the metric used by Zanobetti
36 and Schwartz 2008) Os concentrations across the continental U.S. Figure 4-6 and Figure 4-7
4-13
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
show the geographic distribution of these spatial surfaces. Figure 4-8 shows the frequency and
cumulative percent of the seasonal average 63 concentrations by grid cell, using both metrics.
May-September average 8-hr daily maximum concentrations are most frequently in the 40-50
ppb range, while June-August average 8-hr daily mean concentrations are more evenly
distributed across a range of 20-70 ppb. Maximum concentrations for the June-August mean of
the 8-hr daily mean concentrations from 10am to 6pm are generally higher than for the May-
September mean of the 8-hr daily maximum concentrations since the seasonal definition is
limited to the summer months when Oi tends to be highest. The maximum, minimum, mean,
median, and 95* percentile concentrations for both 8-hr daily maximum and 8-hr daily mean are
shown in Table 4-3. These seasonal average metrics are not equivalent to the averaging time for
the current NAAQS, which is based on the 4* highest value rather than seasonal mean, so the
values should not be directly compared against the NAAQS.
15
16
17
18
19
20
£
ppb
Figure 4-6
Seasonal (May-September) average 8-hr, daily maximum baseline Os
concentrations (ppb) at the surface, based on a 2007 CMAQ model
simulation fused with average 2006-2008 observations from the Os monitor
network.
4-14
-------
S?
ppb
1
2
3
4
5
6
7
Figure 4-7
Seasonal (June-August) average 8-hr, daily mean (10am-6pm) baseline Os
concentrations (ppb) at the surface, based on a 2007 CMAQ model
simulation fused with average 2006-2008 observations from the Os monitor
network.
4-15
-------
1
2
3
4
5
6
7
e*
cr
1'
0.4
0.3
0.2
0.1
0
20
100%
0%
20
•June-August average Shr daily mean 10am-6pm
>May-September average 8hr daily maximum
30
40
50
60
70
30
40 50
Concentration (ppb)
60
70
Figure 4-8
Frequency and cumulative percent of May-September average 8-hr daily
maximum and the June-August 8-hr daily mean (10am-6pm) Os
concentration (ppb) by gridcell, based on 2006-2008 monitor observations
fused with 2007 CMAQ-modeled O3 levels.
4-16
-------
1
2
3
4
Table 4-3 Statistical characterization of the May-September average 8-hr daily
maximum and the June-August 8-hr daily mean (10am-6pm) Os
concentration (ppb), based on 2006-2008 monitor observations fused with
2007 CMAQ-modeled O3 levels.
Maximum
Minimum
Mean
Median
95th Percentile
May-September average 8-hr daily
maximum concentration (ppb)
65.0
19.7
41.8
42.6
51.6
June-August average daily 10am -
6pm daily mean concentration
(ppb)
85.5
18.0
40.4
41.3
55.1
5
6
7
8
9
10
1 1
1 2
13
14
15
16
17
18
4.4 REFERENCES
Abt Associates, Inc. (2010a). Environmental Benefits and Mapping Program (Version 4.0).
Bethesda, MD. Prepared for U.S. Environmental Protection Agency Office of Air Quality
Planning and Standards. Research Triangle Park, NC. Available on the Internet at
.
Abt Associates, Inc. (201 Ob). Model Attainment Test Software (Version 2). Bethesda, MD.
Prepared for the U.S. Environmental Protection Agency Office of Air Quality Planning
and Standards. Research Triangle Park, NC. Available on the Internet at:
http://www.epa.gov/scram001/modelingapps.mats.htm.
Bell, M.L., A. McDermott, S.L. Zeger, J.M. Samet, F. Dominici. (2004). Ozone and short-term
mortality in 95 US urban communities, 1987-2000. JAMA, 292:2372-2378.
19 Duff, M., Horst, R. L., Johnson, T. R. (1998). Quadratic Rollback: A Technique to Model
20 Ambient Concentrations Due to Undefined Emission Controls. Presented at the Air and
21 Waste Management Annual Meeting, San Diego, CA. June 14-18, 1998.
22 Fann N, Lamson AD, Anenberg SC, Wesson K, Risley D, Hubbell BJ. (2012). Estimating the
23 national public health burden associated with exposure to ambient PM2.5 and ozone. Risk
24 Analysis, 32:81-95.
4-17
-------
1 Hall, E., Eyth, A., Phillips, S. (2012) Hierarchical Bayesian Model (HBM)-Derived Estimates of
2 Air Quality for 2007: Annual Report. EPA/600/R-12/538. Available on the Internet at:
3 http://www.epa.gov/heasd/sources/proj ects/CDC/AnnualReports/2007_HBM.pdf
4 Johnson, T. (2002). A Guide to Selected Algorithms, Distributions, and Databases Used in
5 Exposure Models Developed by the Office of Air Quality Planning and Standards.
6 Prepared by TRJ Environmental, Inc. for U.S. Environmental Protection Agency, Office
7 of Research and Development, Research Triangle Park, NC.
8 National Research Council of the National Academies (2008). Estimating Mortality Risk
9 Reduction and Economic Benefits from Controlling Ozone Air Pollution. The National
10 Academies Press, Washington, D.C.
11 Rizzo, M. (2005). A Comparison of Different Rollback Methodologies Applied to Ozone
12 Concentrations. November 7, 2005. Available at:
13 http ://www. epa.gov/ttn/naaq s/standards/ozone/s_o3_cr_td.html
14 Rizzo, M. (2006). A Distributional Comparison between Different Rollback Methodologies
15 Applied to Ambient Ozone Concentrations. May 31, 2006. Available on the Internet:
16 http ://www. epa.gov/ttn/naaq s/standards/ozone/s_o3_cr_td.html
17 Simon, H., Baker, K., Possiel, N., Akhtar, F., Napelenok, S., Timin, B., Wells, B. (2012) Model-
18 based rollback using the higher order direct decoupled method (HDDM). Available on
19 the Internet at: http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_2008_rea.html.
20 Timin B, Wesson K, Thurman J. Application of Model and Ambient Data Fusion Techniques to
21 Predict Current and Future Year PM2.s Concentrations in Unmonitored Areas. (2010). Pp.
22 175-179 in Steyn DG, Rao St (eds). Air Pollution Modeling and Its Application XX.
23 Netherlands: Springer.
24 U.S. EPA, 2007. Review of the National Ambient Air Quality Standards for Ozone: Policy
25 Assessment of Scientific and Technical Information. OAQPS Staff Paper. U.S.
26 Environmental Protection Agency
27 Office of Air Quality Planning and Standards. Research Triangle Park, North Carolina. EPA-
28 452/R-07-007.U.S. Environmental Protection Agency. (2012a). Integrated Science
29 Assessment for Ozone and Related Photochemical Oxidants: Third External Review
30 Draft, U. S. Environmental Protection Agency, Research Triangle Park, NC. EPA/600/R-
31 10/076C
4-18
-------
1 U.S. Environmental Protection Agency (2012b). Total Risk Integrated Methodology (TRIM) -
2 Air Pollutants Exposure Model Documentation (TRIM.Expo / APEX, Version 4.4)
3 Volume I: User's Guide. Office of Air Quality Planning and Standards, U.S.
4 Environmental Protection Agency, Research Triangle Park, NC. EPA-452/B-12-00la.
5 Available at: http://www.epa.gov/ttn/fera/human apex.html
6 U.S. Environmental Protection Agency (2012c). Total Risk Integrated Methodology (TRIM) -
7 Air Pollutants Exposure Model Documentation (TRIM.Expo / APEX, Version 4.4)
8 Volume II: Technical Support Document. Office of Air Quality Planning and Standards,
9 U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/B-12-
10 OOlb. Available at: http://www.epa.gov/ttn/fera/human_apex.html
11 Wells, B., Wesson, K., Jenkins, S. (2012). Analysis of Recent U.S. Ozone Air Quality Data to
12 Support the O3 NAAQS Review and Quadratic Rollback Simulations to Support the First
13 Draft of the Risk and Exposure Assessment. Available on the Internet at:
14 http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_2008_rea.html
15 Zhang, L., DJ. Jacob, N.V. Smith-Downey, D.A. Wood, D. Blewitt, C.C. Carouge, A. van
16 Donkelaar, D.B. A. Jones, L.T. Murray, Y. Wang. (2011). Improved estimate of the
17 policy-relevant background ozone in the United States using the GEOS-Chem global
18 model with l/2°x2/3° horizontal resolution over North America. Atmos Environ,
19 45:6769-6776.
20 Zanobetti, A., and J. Schwartz. (2008). Mortality displacement in the association of ozone with
21 mortality: An analysis of 48 cities in the United States. Am J Resp Crit Care Med,
22 177:184-189.
23
24
25
4-19
-------
1
2 5 CHARACTERIZATION OF HUMAN EXPOSURE TO OZONE
3 5.1 INTRODUCTION
4 As part of the last O3 NAAQS review, EPA conducted exposure analyses for the general
5 population, all school-age children (ages 5-18), active school-age children, and asthmatic school-
6 age children (EPA, 2007a,b). Exposure estimates were generated for 12 urban areas for recent
7 years of air quality and for just meeting the existing 8-hr standard and several alternative 8-hr
8 standards. EPA also conducted a health risk assessment that produced risk estimates for the
9 number of children and percent of children experiencing impaired lung function and other
10 respiratory symptoms associated with the exposures estimated for these same 12 urban areas.
11 The exposure analysis conducted for the current review builds upon the methodology and
12 lessons learned from the exposure analyses conducted in previous reviews (U.S. EPA, 1996a,
13 2007a,b), as well as information provided in the third draft ISA (EPA, 2012a). EPA will be
14 conducting exposure modeling for 16 urban areas located across the U.S., listed in Table 5-3). In
15 this first draft REA, results are presented for four of these areas, Atlanta, Denver, Los Angeles,
16 and Philadelphia.
17 Population exposures to ambient 63 levels are modeled using the Air Pollutants Exposure
18 (APEX) model, also referred to as the Total Risk Integrated Methodology Inhalation Exposure
19 (TREVI.Expo) model (U.S. EPA, 2012b,c). Exposure estimates are developed for 63 levels in
20 recent years, based on 2006 to 2010 ambient air quality measurements. Exposures are also
21 estimated for O3 levels associated with just meeting the current 8-hr O3 NAAQS, based on
22 adjusting data derived from the ambient monitoring network as described in Chapter 4 with
23 additional details in Wells et al. (2012). Exposures are modeled for 1) the general population, 2)
24 school-age children (ages 5-18), and 3) asthmatic school-age children. The strong emphasis on
25 children reflects the finding of the last O3 NAAQS review (EPA, 2007a) and the ISA (EPA,
26 2012a, Chapter 8) that children are an important at-risk group.
27 This chapter provides a brief overview of the types of studies that provide data on which
28 this analysis is based, followed by a description of the exposure model used for this analysis, the
29 model input data, and the results of the analysis. The final sections of this chapter summarize the
30 sensitivity analyses and model evaluation that have been conducted for the APEX exposure
31 model, and plans for additional analyses to be included in the second draft REA.
5-1
-------
1 5.2 OZONE EXPOSURE STUDIES
2 Many studies have produced information and data supporting the development of
3 methods for estimating human exposure to ambient Os over the past several decades. These
4 studies have been reviewed in the ISA and previous EPA Ozone Air Quality Criteria Documents
5 (U.S. EPA, 1986, 1996b, 2006, 2012a). The types of studies which provide the basis for
6 modeling human exposure to Os include studies of people's activities, work and exercise
7 patterns, physiology, physics and Os-related chemistry in microenvironments, atmospheric
8 modeling of O3, chamber studies of atmospheric chemistry, and modeling of meteorology.
9 Measurements that have proven to be useful for understanding and estimating exposure obtained
10 from personal exposure assessment studies include fixed-site ambient concentrations,
11 concentrations in specific indoor and outdoor microenvironments, personal exposure levels,
12 personal activity patterns, air exchange rates, infiltration rates, deposition and decay rates, and
13 meteorology.
14 Exposure Concepts and Definitions
15 Human exposure to a contaminant is defined as "contact at a boundary between a human
16 and the environment at a specific contaminant concentration for a specific interval of time," and
17 has units of concentration times duration (National Research Council, 1991). For airborne
18 pollutants the contact boundary is nasal and oral openings in the body, and personal exposure of
19 an individual to a chemical in the air for a discrete time period is quantified as (Lioy, 1990;
20 National Research Council, 1991):
21 Z[tl,t2} = ]t L,(t)at (4_i)
22 where E[rl5r2] is the personal exposure during the time period from t\ to ti, and C(f) is the
23 concentration at time t in the breathing zone. We refer to the exposure concentration to mean the
24 concentration to which one is exposed. The breathing rate (ventilation rate) at the time of
25 exposure is an important determinant of the dose received by the individual. Although we do not
26 estimate dose, we refer to intake as the total amount of Os inhaled (product of exposure
27 concentration, duration, and minute ventilation rate).
28 Personal exposure to Os can be estimated directly by monitoring the concentration of Os
29 in the person's breathing zone (close to the nose/mouth) using a personal exposure monitor.
5-2
-------
1 Exposure can also be estimated indirectly, by estimating or monitoring the concentrations over
2 time in locations in which the individual spends time and estimating the time and duration the
3 individual spends in each location, as well as the level of activity and resulting ventilation rate.
4 In both of these methods, Equation 4-1 is used to calculate an estimate of personal exposure. A
5 key concept in modeling exposure is the microenvironment, a term that refers to the immediate
6 surroundings of an individual. A microenvironment is a location in which pollutant
7 concentrations are relatively homogeneous for short periods of time. Microenvironments can be
8 outdoors or indoors; some examples are outdoors near the home, outdoors near the place of
9 work, bedrooms, kitchens, vehicles, stores, restaurants, street-corner bus stops, schools, and
10 places of work. A bedroom may be treated as a different microenvironment than a kitchen if the
11 concentrations are significantly different in the two rooms. The concentrations in a
12 microenvironment typically change over time; for example, Os concentrations in a kitchen while
13 cooking with a gas stove may be lower than when these activities are not being performed, due to
14 scavenging of Os by nitric oxide (NO) emissions from the gas burned.
15 An important factor affecting the concentrations of Os indoors is the degree to which the
16 ambient outdoor air is transported indoors. This can be modeled using physical factors such as
17 air exchange rates (AERs), deposition and decay rates, and penetration factors. The volumetric
18 exchange rate (m3/hour) is the rate of air exchange between the indoor and outdoor air. The AER
19 between indoors and outdoors is the number of complete air exchanges per hour and is equal to
20 the volumetric exchange rate divided by the volume of the well-mixed indoor air. Indoor
21 concentrations of Os can be decreased by uptake of Os by surfaces and by chemical reactions.
22 The deposition and chemical decay rates are the rates (per hour) at which Os is removed from
23 the air by surface uptake and chemical reactions. Some exposure models employ an infiltration
24 factor, which is conceptually useful if distinguishing between the air exchange processes of air
25 blowing through open doors and windows and the infiltration of air through smaller openings.
26 Since measurements of AERs account for both of these processes (including infiltration), this
27 distinction is not useful in applied modeling of 63 exposures and will not be discussed further
28 here. Simpler exposure models use a "factor model" approach to estimate indoor O3
29 concentrations by multiplying the ambient outdoor concentrations by an indoor/outdoor
30 concentration ratio, referred to as & penetration factor.
5-3
-------
1 5.3 EXPOSURE MODELING
2 Models of human exposure to airborne pollutants are typically driven by estimates of
3 ambient outdoor concentrations of the pollutants, which vary by time of day as well as by
4 location. These outdoor concentration estimates may be provided by measurements, by air
5 quality models, or by a combination of these. Simulations of scenarios where current or
6 alternative ozone standards are just met require some form of modeling. The main purpose of
7 this exposure analysis is to allow comparisons of population exposures to 63 within each urban
8 area, associated with recent air quality levels and with several potential alternative air quality
9 standards or scenarios. Human exposure, regardless of the pollutant, depends on where an
10 individual is located and what they are doing. Inhalation exposure models are useful in
11 realistically estimating personal exposures and intake based on activity-specific ventilation rates,
12 particularly when recognizing that these measurements cannot be performed for a given
13 population. This section provides a brief overview of the model used by EPA to estimate 63
14 population exposure. A more detailed technical description of APEX is provided in Appendix
15 5 A.
16 5.3.1 The APEX Model
17 The EPA has developed the APEX model for estimating human population exposure to
18 criteria and air toxic pollutants. APEX also serves as the human inhalation exposure model
19 within the Total Risk Integrated Methodology (TRIM) framework (Richmond et al., 2002; EPA
20 2012b,c). APEX is conceptually based on the probabilistic NAAQS Exposure Model (pNEM)
21 that was used in the 1996 O3 NAAQS review (Johnson et al., 1996a; 1996b; 1996c). Since that
22 time the model has been restructured, improved, and expanded to reflect conceptual advances in
23 the science of exposure modeling and newer input data available for the model. Key
24 improvements to algorithms include replacement of the cohort approach with a probabilistic
25 sampling approach focused on individuals, accounting for fatigue and oxygen debt after exercise
26 in the calculation of ventilation rates, and new approaches for construction of longitudinal
27 activity patterns for simulated persons. Major improvements to data input to the model include
28 updated AERs, census and commuting data, and the daily time-activities database. These
29 improvements are described later in this chapter.
30 APEX is a probabilistic model designed to account for the numerous sources of
31 variability that affect people's exposures. APEX simulates the movement of individuals through
5-4
-------
1 time and space and estimates their exposure to a given pollutant in indoor, outdoor, and in-
2 vehicle microenvironments. The model stochastically generates simulated individuals using
3 census-derived probability distributions for demographic characteristics. The population
4 demographics are drawn from the year 2000 Census at the tract level, and a national commuting
5 database based on 2000 census data provides home-to-work commuting flows between tracts.1
6 Any number of simulated individuals can be modeled, and collectively they approximate a
7 random sampling of people residing in a particular study area.
8 Daily activity patterns for individuals in a study area, an input to APEX, are obtained
9 from detailed diaries that are compiled in the Consolidated Human Activity Database (CHAD)
10 (McCurdy et al., 2000; EPA, 2002). The diaries are used to construct a sequence of activity
11 events for simulated individuals consistent with their demographic characteristics, day type, and
12 season of the year, as defined by ambient temperature regimes (Graham & McCurdy, 2004). The
13 time-location-activity diaries input to APEX contain information regarding an individuals' age,
14 sex, race, employment status, occupation, day-of-week, daily maximum hourly average
15 temperature, the location, start time, duration, and type of each activity performed. Much of this
16 information is used to best match the activity diary with the generated personal profile, using
17 age, sex, employment status, day of week, and temperature as first-order characteristics. The
18 approach is designed to capture the important attributes contributing to an individuals' behavior,
19 and of particular relevance here, time spent outdoors (Graham and McCurdy, 2004).
20 Furthermore, these diary selection criteria give credence to the use of the variable data that
21 comprise CHAD (e.g., data collected were from different seasons, different states of origin, etc.).
22 Contributing to the uncertainty of the simulated diary sequences is that the approach for creating
23 year-long activity sequences is based on a cross-sectional activity data base of 24-hour records.
24 The typical subject in the time/activity studies in CHAD provided less than 2 days of diary data.
25 APEX calculates the concentration in the microenvironment associated with each event in an
26 individual's activity pattern and sums the event-specific exposures within each hour to obtain a
27 continuous series of hourly exposures spanning the time period of interest.
28 APEX has a flexible approach for modeling microenvironmental concentrations, where
29 the user can define the microenvironments to be modeled and their characteristics. Typical
30 indoor microenvironments include residences, schools, and offices. Outdoor microenvironments
There are approximately 65,400 census tracts in the -3,200 counties in the U.S.
5-5
-------
1 include near roadways, at bus stops, and playgrounds. Inside cars, trucks, and mass transit
2 vehicles are microenvironments which are classified separately from indoors and outdoors.
3 Activity-specific simulated breathing rates of individuals are used in APEX to
4 characterize intake received from an exposure. These breathing, or ventilation, rates are derived
5 from energy expenditure estimates for each activity included in CHAD and are adjusted for age-
6 and sex-specific physiological parameters associated with each simulated individual. Energy
7 expenditure estimates themselves are derived from METS (metabolic equivalents of work)
8 distributions associated with every activity in CHAD (McCurdy et al., 2000), largely based upon
9 the Ainsworth et al. (1993) "Compendium of Physical Activities." METS are a dimensionless
10 ratio of the activity-specific energy expenditure rate to the basal or resting energy expenditure
11 rate, and the metric is used by exercise physiologists and clinical nutritionists to estimate work
12 undertaken by individuals as they go through their daily life (Montoye et al., 1996). This
13 approach is discussed more thoroughly in McCurdy (2000).
14 5.3.2 Key Algorithms
15 Ozone concentrations in each microenvironment are estimated using either a mass-
16 balance or transfer factors approach, selected by the user. The user specifies probability
17 distributions for the parameters that are used in the microenvironment model that reflect the
18 observed variabilities in the parameters. These distributions can depend on the values of other
19 variables calculated in the model or input to APEX. For example, the distribution of AERs in a
20 home, office, or car can depend on the type of heating and air conditioning present, which are
21 also stochastic inputs to the model, as well as the ambient temperature. The user can choose to
22 keep the value of a stochastic parameter constant for the entire simulation (which would be
23 appropriate for the volume of a house), or can specify that a new value shall be drawn hourly,
24 daily, or seasonally from specified distributions. APEX also allows the user to specify diurnal,
25 weekly, or seasonal patterns for various microenvironmental parameters. The distributions of
26 parameters input to APEX characterize the variability of parameter values, and are not intended
27 to reflect uncertainties in the parameter estimates.
28 The mass balance method used within APEX assumes that the air in an enclosed
29 microenvironment is well-mixed and that the air concentration is fairly spatially uniform at a
30 given time within the microenvironment. The following four processes are modeled to predict
31 the concentration of an air pollutant in such a microenvironment:
5-6
-------
1 • Inflow of air into the microenvironment;
2 • Outflow of air from the microenvironment;
3 • Removal of a pollutant from the microenvironment due to deposition, filtration, and
4 chemical degradation; and
5 • Emissions from sources of a pollutant inside the microenvironment.
6 The transfer factors model is simpler than the mass balance model, however, still most
7 parameters are derived from distributions rather than single values, to account for observed
8 variability. It does not calculate concentration in a microenvironment from the concentration in
9 the previous hour and it has only two parameters, a proximity factor, used to account for
10 proximity of the microenvironment to sources or sinks of pollution, or other systematic
11 differences between concentrations just outside the microenvironment and the ambient
12 concentrations (at the measurements site), and a penetration factor, which quantifies the degree
13 to which the outdoor air penetrates into the microenvironment. When there are no indoor
14 sources, the penetration factor is essentially the ratio of the concentration in the
15 microenvironment to the outdoor concentration.
16 Regardless of the method used to estimate the microenvironmental concentrations, APEX
17 calculates a time series of exposure concentrations that a simulated individual experiences during
18 the modeled time period. APEX estimates the exposure using the concentrations calculated for
19 each microenvironment and the time spent in each of a sequence of microenvironments visited
20 according to the "activity diary" of each individual. The hourly average exposures of each
21 simulated individual are time-weighted averages of the within-hour exposures. From hourly
22 exposures, APEX calculates the time series of 8-hr and daily average exposures that simulated
23 individuals experience during the simulation period. APEX then statistically summarizes and
24 tabulates the hourly, 8-hr, and daily exposures.
25 Estimation of Ambient Air Quality
26 For estimating ambient 63 concentrations to use in the exposure model, the urban areas
27 modeled here have several monitors measuring hourly Oj concentrations (ranging from 12 in the
28 Atlanta area to 51 in the Los Angeles area, for 2008). Having multiple monitors in the simulated
29 areas collecting time-resolved data allows for the utilization of APEX spatial and temporal
30 capabilities in estimating exposure. Since APEX uses actual records of where individuals are
5-7
-------
1 located at specific times of the day, more realistic exposure estimates are obtained in simulating
2 the contact of individuals with these spatially and temporally diverse concentrations. Primary
3 uncertainties in the air quality data input to the model result from estimating concentrations at
4 locations which may not be in close proximity to monitoring sites (as estimated by spatial
5 interpolation of actual data points) and from the method used to estimate missing data for some
6 hours or days. In addition, concentrations of Os near roadways are particularly difficult to
7 estimate due to the rapid reaction of Os with nitric oxide emitted from motor vehicles.
8 We have modeled the 63 seasons for 2006 to 2010 to account for year-to-year variability
9 of air quality and meteorology in recent years. Having this wide range of air quality data across
10 multiple years available for use in the exposure simulation has a direct impact on more
11 realistically estimating the range of exposures, rather than using a single year of air quality data.
12 Estimation of Concentrations in Indoor Microenvironments
13 The importance of estimation of concentrations in indoor microenvironments (e.g.,
14 homes, offices, schools, restaurants, vehicles) is underscored by the finding that personal
15 exposure measurements of 63 may not be well-correlated with ambient measurements and indoor
16 concentrations are usually much lower than ambient concentrations (EPA, 2012a, Section 4.3.3).
17 APEX has been designed to better estimate human exposure through use of algorithms
18 that attempt to capture the full range of Os concentrations expected within several important
19 microenvironments. Parameters used to estimate the concentrations in microenvironments can
20 be highly variable, both between microenvironments (e.g., different houses have varying
21 characteristics) and within microenvironments (e.g., the characteristics of a given house can vary
22 over time). Since APEX is a probabilistic model, if data accurately characterizing this variability
23 are provided to the model, then such variabilities would not result in uncertainties in the
24 estimation of the microenvironmental concentrations. Thus, it is the input data used in
25 development of the parameters that are the limiting factor, and to date, APEX uses the most
26 current available data to develop required distributions of parameters for estimation of
27 microenvironmental concentrations.
28 Air Exchange Processes
29 The air exchange rate is the single most important factor in determining the relationship
30 between outdoor and indoor concentrations of O3. AERs are highly variable, both within a
5-8
-------
1 microenvironment over time and between microenvironments of the same type. AERs depend
2 on the physical characteristics of a microenvironment and also on the behavior of the occupants
3 of the microenvironment. There is a strong dependence on temperature, and some dependence
4 on other atmospheric conditions, such as wind. APEX uses probabilistic distributions of AERs
5 which were derived from several measurement studies in a number of locations, and are stratified
6 by both temperature and the presence or absence of air conditioning. These two variables are the
7 most influential variables influencing AER distributions (see Appendix 5B).
8 Removal Processes
9 Concentrations within indoor microenvironments can be reduced due to removal
10 processes such as deposition to surfaces and by reaction with other chemicals in the air.
11 Deposition is modeled probabilistically in APEX by a using a distribution of decay rates.
12 The lack of a better treatment of indoor air chemistry is not considered to be a significant
13 limitation of APEX for modeling Os.
14 Characterization of Population Demographics and Activity Patterns
15 By using actual time-location-activity diaries that capture the duration and frequency of
16 occurrence of visitations/activities performed, APEX can simulate expected variability in human
17 behavior, both within and between individuals. Fundamentals of energy expenditure are then
18 used to estimate relative intensity of activities performed. This, combined with
19 microenvironmental concentrations, allows for the reasonable estimation of the magnitude,
20 frequency, pattern, and duration of exposures an individual experiences.
21 CHAD is the most complete, high quality source of human activity data for use in
22 exposure modeling. The database contains over 38,000 individual daily diaries including time-
23 location-activity patterns for individuals of both sexes across a wide range of ages (<1 to 94).
24 The database is geographically diverse, containing diaries from individuals residing in major
25 cities, suburban and rural areas across the U.S. Time spent performing activities within
26 particular locations can be on a minute-by minute basis, thus avoiding the smoothing of potential
27 peak exposures longer time periods would give. Table 5-1 summarizes the studies in CHAD
28 used in this modeling analysis.
29 There are some limitations to the database, however, many of which are founded in the
30 individual studies from which activity patterns were derived (Graham and McCurdy, 2004). A
5-9
-------
1 few questions remain regarding the representativeness of CHAD diaries to the simulated
2 population, such as the age of diary data (i.e., some data were generated in the 1980s) and diary
3 structure differences (i.e., real-time versus recall method of data collection). Many of the
4 assumptions about use of these activity patterns in exposure modeling are strengthened by the
5 manner in which they are used by APEX, through focusing on the most important individual
6 attributes that contribute to variability in human behavior (e.g., age, sex, time spent outdoors, day
7 of week, ambient temperature, occupation).
8 The extent to which the human activity database provides a balanced representation of
9 the population being modeled is likely to vary across areas. Although the algorithm that
10 constructs activity sequences accounts to some extent for the effects of population demographics
11 and local climate on activity, this adjustment procedure may not account for all inter-city
12 differences in people's activities. A new methodology has been developed to more appropriately
13 assign individual diaries to reflect time-location-activity patterns in simulated individuals
14 (discussed further in section 4.5.3). Input distributions used in the new procedure for
15 constructing multi-day activity patterns are based on longitudinal activity data from children of a
16 specific age range (appropriate for this application where similar aged children are the primary
17 focus), however the data used were limited to one study and may not be appropriate for other
18 simulated individuals. Thus, there are limitations in approximating within-person variance and
19 between-person variance for certain variables (e.g., time spent outdoors). Personal activity
20 patterns are also likely to be affected by many local factors, including topography, land use,
21 traffic patterns, mass transit systems, and recreational opportunities, which are not incorporated
22 in the current exposure analysis approach due to the complexity of scale and lack of data to
23 support the development of a reasonable approach.
5-10
-------
1 Table 5-1. Studies in CHAD used in this analysis
Study name
Baltimore
Retirement Home
Study (EPA)
California Youth
Activity Patterns
Study (CARS)
California Adults
Activity Patterns
Study (CARS)
California Children
Activity Patterns
Study (CARS)
Cincinnati Activity
Patterns Study
(EPRI)
Denver CO
Personal Exposure
Study (EPA)
Los Angeles Ozone
Exposure Study:
Elementary School
Los Angeles Ozone
Exposure Study:
High School
Geographic
coverage
One building
in Baltimore
California
California
California
Cincinnati
metro, area
Denver
metro, area
Los Angeles
Los Angeles
Study time
period
01/1997-02/1997,
07/1998-08/1998
10/1987-09/1988
10/1987-09/1988
04/1989- 02/1990
03/1985-04/1985,
08/1985
11/1982-02/1983
10/1989
09/1990-10/1990
Subject
ages
72-93
12- 17
18-94
<1 - 11
-------
National Human
Activity Pattern
Study (NHAPS):
Air
National Human
Activity Pattern
Study (NHAPS):
Water
National Study of
Avoidance of S
(NSAS)
Population Study of
Income Dynamics
PSID CDS I (Univ.
Michigan I)
Population Study of
Income Dynamics
PSID CDS II (Univ.
Michigan II)
RTI Ozone
Averting Behavior
RTF Panel (EPA)
Seattle
Washington, B.C.
(EPA)
Totals
National
National
7 U.S.
metropolitan
areas
National
National
35 U.S.
metropolitan
areas
RTF, NC
Seattle, WA
Wash., D.C.
metro, area
09/1992-10/1994
09/1992-10/1994
06/2009-09/2009
02/1997-12/1997
01/2002-12/2003
07/2002-08/2003
06/2000-05/2001
10/1999-03/2002
11/1982-02/1983
1982 - 2009
-------
1 Averting Behavior and Exposure
2 A growing area of air pollution research involves evaluating the actions persons might
3 perform in response to high Os concentration days (ISA, section 4.1.1). Most commonly termed
4 averting behaviors, they can be broadly characterized as personal activities that either reduce
5 pollutant emissions or limit personal exposure levels. The latter topic is of particular interest in
6 this REA due to the potential negative impact it could have on 63 concentration-response (C-R)
7 functions used to estimate health risk and on time expenditure and activity exertion levels
8 recorded in the CHAD diaries used by APEX to estimate 63 exposures. To this end, we have
9 performed an additional review of the available literature here beyond that summarized in the
10 ISA to include several recent technical reports that collected and/or evaluated averting behavior
11 data. Our purpose is to generate a few reasonable quantitative approximations that allow us to
12 better understand how averting behavior might affect our current population exposure and risk
13 estimates. We expect that the continued development and communication of air quality
14 information via all levels of environmental, health, and meteorological organizations will only
15 further increase awareness of air pollution, its associated health effects, and the recommended
16 actions to take to avoid exposure, thus making averting behaviors and participation rates an even
17 more important consideration in future O3 exposure and risk assessments. The following is a
18 summary of our current findings, with details provided in Graham (2012).
19 The first element considered in our evaluation is peoples' general perception of air
20 pollution and whether they were aware of alert notification systems. The prevalence of
21 awareness was variable; about 50% to 90% of survey study participants acknowledge or were
22 familiar with air quality systems (e.g., Blanken et al., 1991; KS DOH, 2006; Mansfield et al.,
23 2006; Semenza et al., 2008) and was dependent on several factors. In studies that considered a
24 persons' health status, e.g., asthmatics or parents of asthmatic children, there was a consistently
25 greater degree of awareness (approximately a few to 15 percentage points) when compared to
26 that of non-asthmatics. Residing in an urban area was also an important influential factor raising
27 awareness, as both the number of high air pollution events and their associated alerts are greater
28 when compared to rural areas. Of lesser importance, though remaining a statistically significant
29 influential variable, were several commonly correlated demographic attributes such as age,
30 education-level, and income-level, with each factor positively associated with awareness.
31 The second element considered in our evaluation was the type of averting behaviors
32 performed. For our purposes in this Os REA, the most relevant studies were those evaluating
5-13
-------
1 outdoor time expenditure, more specifically, the duration of outdoor events and the associated
2 exertion level of activities performed while outdoors. This is because both of these variables are
3 necessary to understanding O3 exposure and adverse effects and in accurately estimating human
4 health risk.
5 As stated above regarding air quality awareness, asthmatics consistently indicated a
6 greater likelihood of performing averting behaviors compared to non-asthmatics - estimated to
7 differ by about a factor of two. This difference could be the combined effect of those persons
8 having been advised by health professional to avoid high air pollution events and them being
9 aware of alert notification systems. Based on the survey studies reviewed, we estimate that 30%
10 of asthmatics may reduce their outdoor activity level on alert days (e.g., KS DOH, 2006;
11 McDermott et al., 2006; Wen et al., 2009).2 An estimate of 15%, derived from reductions in
12 public attendance at outdoor events (Zivin and Neidell, 2009) is consistent with the above
13 estimate when considering that it is likely represented by a non-asthmatic population. That said,
14 both attenuation and the re-establishment of averting behavior was apparent when considering a
15 few to several days above high pollution alert levels (either occurring over consecutive days or
16 across an entire year) (McDermott et al., 2006; Zivin and Neidell, 2009), suggesting that
17 participation in averting behavior over a multiday period for an individual is complex and likely
18 best represented by a time and activity-dependent function rather than a simple point estimate.
19 There were only a few studies offering quantitative estimates of durations of averting
20 behavior, either considering outdoor exertion level or outdoor time (Bresnahan et al., 1997;
21 Mansfield et.al, 2006, Neidell, 2010; Sexton, 2011). Each of these studies considered outdoor
22 time expenditure during the afternoon hours. Based on the studies reviewed, we estimate that
23 outdoor time/exertion during afternoon hours may be reduced by about 20-40 minutes in
24 response to an air quality alert notification. Generally requisite factors include: a high alert level
25 for the day (e.g., red or greater on the AQI), high 63 concentrations (above the NAAQS), and
26 persons having a compromised health condition (e.g., asthmatic or elderly).
27 The third element considered in our evaluation is how to further define the impact of
28 averting behavior on modeled exposure estimates.3 As described in section 5.3.2, APEX uses
29 time location activity data (diaries) from CHAD to estimate population exposures. These diaries
2 Many of these studies do not account for one important factor when using a recall questionnaire design: whether
the participant's stated response to air pollution is the same as the action they performed.
3 The discussion of another important effect of averting behavior is on concentration-response functions (more
relevant to the risk assessment in chapter 7). This is presented in the ISA (section 4.1.2).
5-14
-------
1 come from a number of differing studies; some were generated as part of an air pollution
2 research study, some may have been collected during a summer/ozone season, while some diary
3 days may have corresponded with high O3 concentration and air quality alert days. At this time,
4 none of the diary days used by APEX have been identified as representing days where a person
5 did or did not perform an averting behavior to reduce their exposure. In considering the above
6 discussion regarding the potential rate of participation and averting actions performed, it is
7 possible that some of the CHAD diary days express times where that selected individual may
8 have reduced their time spent outdoors or outdoor exertion level. Currently, without having an
9 identifier for averting behavior, the diaries are assigned randomly4 to a simulated persons' day
10 and do not consider ambient 63 concentrations. Therefore, there may be instances where, on a
11 given day, a simulated person does appear to engage in averting behavior (a diary having less
12 time than usual spent outdoors in the afternoon), while for most other persons on the same day
13 (or the same person on a different high concentration day) there is no averting behavior.
14 Therefore, averting behavior may be incorporated into our exposure modeling, albeit to an
15 unknown degree,5 though definitely generating low-biased estimates of exposures that would
16 occur in the complete absence of averting behavior.
17 Modeling Physiological Processes
18 The modeling of physiological processes that are relevant to the exposure and intake of
19 Os is a complicated endeavor, particularly when attempting to capture inter- and intra-personal
20 variability in these rates. APEX has a physiological module capable of estimating ventilation
21 rates (VE) for every activity performed by an individual, which primarily drives 63 intake dose6
22 rate estimates. See Isaacs, et al. (2008) and Chapter 7 of the APEX TSD (EPA, 2012c) for a
23 discussion of this module. Briefly, the module is based on the relationship between energy
24 expenditure and oxygen consumption rate, thus both within- and between-person variability in
25 ventilation can be addressed through utilization of the unique sequence of events individuals go
26 through each simulated day. These activity-specific VE estimates, when normalized by BSA, are
27 then used to characterize an individual's exertion level in compiling the summary exposure
28 tables (Table 5-2). One of the key determinants of estimated VE is the exertion level of an
4 APEX uses maximum temperature in assigning diaries for a select day in an area, capturing some variability in O3
concentrations.
5 Neither the participation rate nor the duration of averting for simulated persons is being strictly controlled for by
the model.
6 Intake dose is a measure related to dose; it is the amount of ozone that enters the lungs.
5-15
-------
1 individual's activity, where exertion levels have units of metabolic equivalents of work (MET),
2 which is the ratio of energy expenditure for an activity to the person's basal, or resting, metabolic
3 rate.
4 There are some limitations in using MET values for this purpose, due mostly to the
5 manner in which the time-location-activity diaries were generated and subsequent estimates of
6 exertion level. An individual (or their caregiver if younger than eight years old) would record
7 the activity performed with a start and end time, with no information on the associated exertion
8 level of the activity. Exertion level (MET) was then inferred by developers of the CHAD
9 database (McCurdy et al., 2000) using standard values and distributions of those values reported
10 by an expert panel of exercise physiologists (Ainsworth et al., 1993). Although this approach
11 allows for an appropriate range of exertion levels to be assigned to the individuals' activities
12 (and to the simulated population), children's activity levels fluctuate widely within a single
13 activity category; their pattern is often characterized as having bursts of high energy expenditure
14 within a longer time frame of less energy expenditure (Freedson, 1989). These fluctuations in
15 energy expenditure that occur within an activity (and thus a simulated event) are not well
16 captured by the MET assignment procedure.
17 5.3.3 Model Output
18 There are several useful indicators of exposure of people to O3 air pollution and resulting
19 intake of O3. In this analysis, exposure indicators include daily maximum 1-hr and 8-hr average
20 O3 exposures, stratified by a measure of the level of exertion at the time of exposure. Factors
21 that are important in calculating these indicators include the magnitude and duration of exposure,
22 frequency of repeated high exposures, and the breathing rate of individuals at the time of
23 exposure. The level of exertion of individuals engaged in particular activities is measured by an
24 equivalent ventilation rate (EVR), ventilation normalized by body surface area (BSA, in m2),
25 which is calculated as VE /BSA, where VE is the ventilation rate (liters/minute). Table 5-2 lists
26 the ranges of EVR corresponding to moderate and heavy levels of exertion.
27 Table 5-2. Exertion levels in terms of equivalent ventilation rates (liters/min-m2 BSA)
Averaging time Moderate exertion Heavy exertion
1 hour 16-30 EVR > 30 EVR
8 hour 13-27 EVR > 27 EVR
28 from Whitfield et al., 1996, page 15.
5-16
-------
1
2 APEX calculates two general types of exposure estimates: counts of the estimated
3 number of people exposed to a specified 63 concentration level and the number of times per 63
4 season that they are so exposed; the latter metric is in terms of person-occurrences or person-
5 days. The former highlights the number of individuals exposed one or more times per Oj season
6 to the exposure indicator of interest. In the case where the exposure indicator is a benchmark
7 concentration level, the model estimates the number of people who are expected to experience
8 exposures to that level of air pollution, or higher, at least once during the modeled period. APEX
9 also reports counts of individuals with multiple exposures. The person-occurrences measure
10 estimates the number of times per season that individuals are exposed to the exposure indicator
11 of interest and then accumulates these estimates for the entire population residing in an area.
12 This metric conflates people and occurrences: one occurrence for each of 10 people is counted
13 the same as 10 occurrences for one person.
14 APEX tabulates and displays the two measures for exposures above levels ranging from
15 0.0 to 0.16 ppm by 0.01 ppm increments, where the exposures are:
16 • Daily maximum 1-hour average exposures
17 • Daily maximum 8-hour average exposures
18 • Daily average exposures.
19 These results are tabulated for the following population groups:
20 • All ages and activity levels
21 • Children at all activity levels
22 • Asthmatic children.
23 Separate output tables are produced for different levels of exertion concomitant with the
24 exposures:
25 • All exertion levels
26 • Moderate and greater exertion levels
27 APEX also produces tables of the time spent in different microenvironments, stratified by
28 exposure levels.
5-17
-------
1 5.4 SCOPE OF EXPOSURE ASSESSMENT
2 5.4.1 Selection of Urban Areas to be Modeled
3 The selection of urban areas to include in the exposure analysis takes into consideration
4 the location of 63 epidemiological studies, the availability of ambient 63 data, and the desire to
5 represent a range of geographic areas, population demographics, and O3 climatology. The criteria
6 and considerations that went into selection of urban areas for the Os risk assessment included the
7 following:
• The overall set of urban locations should represent a range of geographic areas, urban
population demographics, and climatology.
• The locations should be focused on areas that do not meet or are close to not meeting the
current 8-hr Os NAAQS and should include the largest areas with major Os
nonattainment problems.
• There must be sufficient O3 air quality data for the recent 2006-2010 period.
• The areas should include the 12 cities modeled in the epidemiologic-based risk
assessment.
8 Based on these criteria, we chose the 16 urban areas listed in Table 5-3 to develop population
9 exposure estimates.7 As mentioned above, in this first draft REA, results are presented for four
10 of these areas, Atlanta, Denver, Los Angeles, and Philadelphia. The geographic extents of these
11 four modeled areas are illustrated in Appendix 5B.
12 5.4.1 Time Periods Modeled
13 We have modeled the Os seasons for 2006 to 2010. The exposure periods modeled are
14 the Os seasons for which routine hourly Os monitoring data are available. These periods include
15 most of the high-ozone events in each area. The time periods modeled for each area are listed in
16 Table 5-3. The number of ozone monitors in each area varies slightly from year-to-year. The
17 number of monitors in 2008 used in the exposure modeling are 12 for the Atlanta area, 17 for
18 Denver, 51 for Los Angeles, and 19 for Philadelphia.
19
20
21
7 In the remainder of this chapter the city name in bold in Table 4-2 is used to represent the entire urban
area.
5-18
-------
1 Table 5-3. Urban Areas and Time Periods Modeled"
Urban Area (CBSAs or Counties)
Period modeled
Atlanta area, GA (Barrow, Bartow, Bibb, Butts, Carroll Floyd, Cherokee,
Clarke, Clayton, Cobb, Coweta, Dawson, De Kalb, Douglas, Fayette,
Forsyth, Fulton, Gwinnett, Hall, Haralson, Heard, Henry, Jasper, Lamar,
Meriwether, Gilmer, Newton, Paulding, Pickens, Pike, Polk, Rockdale,
Spalding, Troup, Upson, Walton, Chambers (AL))
Baltimore-Towson, MD
Boston area, MA (Barnstable, Bristol, Dukes, Essex, Middlesex, Nantucket,
Norfolk, Plymouth, Suffolk, Worcester)
Chicago-Naperville-Joliet, IL-IN-WI
Cleveland-Akron-Elyria, OH
Dallas-Fort Worth-Arlington, TX
Denver area, CO (Adams, Arapahoe, Boulder, Broomfield, Clear Creek,
Denver, Douglas, Elbert, Gilpin, Jefferson, Park, Larimer, Weld)
Detroit-Warren-Livonia, MI
Houston-Sugar Land-Baytown, TX
Los Angeles-Long Beach-Riverside, CA (Los Angeles, Orange, Riverside
(part), San Bernardino (part), Ventura (part))
New York-Northern New Jersey-Long Island, NY-NJ-PA
Philadelphia-Camden-Wilmington, PA-NJ-DE-MD
Sacramento—Arden-Arcade—Roseville, CA
Seattle-Tacoma-Bellevue, WA
St. Louis, MO-IL
Washington-Arlington-Alexandria, DC-VA-MD-WV
March 1 to Oct. 31
April 1 to Oct. 31
April 1 to Sept. 30
April 1 to Sept. 30
April 1 to Oct. 31
Jan. 1 to Dec. 30
April 1 to Sept. 30
April 1 to Sept. 30
Jan. 1 to Dec. 30
Jan. 1 to Dec. 30
April 1 to Sept. 30
April 1 to Oct. 31
Jan. 1 to Dec. 30
May 1 to Sept. 30
April 1 to Oct. 31
April 1 to Oct. 31
2 In this first draft REA, Atlanta, Denver, Los Angeles, and Philadelphia are modeled.
3 5.4.2 Populations Modeled
4 Exposure modeling was conducted for the general population residing in each area
5 modeled, as well as for school-age children (ages 5 to 18) and asthmatic school-age children.
6 Due to the increased amount of time spent outdoors engaged in relatively high levels of physical
7 activity (which increases intake), school-age children as a group are particularly at risk for
8 experiencing Os-related health effects (EPA, 2012a, Chapter 8). We report results for school-age
9 children down to age five, however, there is a trend for younger children to attend school. Some
10 states allow 4-year-olds to attend kindergarten, and most states have preschool programs for
11 children younger than five. In 2000, six percent of U.S. children ages 3 to 19 who attend school
5-19
-------
1 were younger than five years old (2000 Census Summary File 3, Table QT-P19: School
2 Enrollment). We are not taking these younger children into account in our analysis due to a lack
3 of information which would let us characterize this group of children.
4 The population of asthmatic children is estimated for each city using asthma prevalence
5 data from the National Health Interview Surveys (NHIS) (Dey and Bloom, 2005). Asthma
6 prevalence rates for children aged 0 to 17 years were calculated for each age, sex, and
7 geographic region. For this analysis, asthma prevalence was defined as the probability of a
8 "Yes" response to the question: "do you still have asthma?" among those that responded "Yes"
9 or "No" to this question. A detailed description of this analysis is presented in Appendix 5B.
10 5.4.3 Microenvironments Modeled
11 In APEX, microenvironments provide the exposure locations for modeled individuals.
12 For exposures to be accurately estimated, it is important to have realistic microenvironments that
13 are matched closely to where people are physically located on a daily and hourly basis. As
14 discussed in section 4.3.2 above, the two methods available in APEX for calculating pollutant
15 concentrations within microenvironments are a mass balance model and a transfer factor
16 approach. Table 5-4 lists the 28 microenvironments selected for this analysis and the exposure
17 calculation method for each. The parameters used in this analysis for modeling these
18 microenvironments are described in Appendix 5B.
19
20 Table 5-4. Microenvironments modeled
1
2
O
4
5
6
7
8
9
10
Microenvironment
Indoor - Residence
Indoor - Community Center or Auditorium
Indoor - Restaurant
Indoor - Hotel, Motel
Indoor - Office building, Bank, Post office
Indoor - Bar, Night club, Cafe
Indoor - School
Indoor - Shopping mall, Non-grocery store
Indoor - Grocery store, Convenience store
Indoor - Metro- Sub way-Train station
Calculation
Method
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Parameters1
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
5-20
-------
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Indoor - Hospital, Medical care facility
Indoor - Industrial, factory, warehouse
Indoor - Other indoor
Outdoor - Residential
Outdoor - Park or Golf course
Outdoor - Restaurant or Cafe
Outdoor - School grounds
Outdoor - Boat
Outdoor - Other outdoor non-residential
Near-road - Metro-Subway-Train stop
Near-road - Within 10 yards of street
Near-road - Parking garage (covered or below ground)
Near-road - Parking lot (open), Street parking
Near-road - Service station
Vehicle - Cars and Light Duty Trucks
Vehicle - Heavy Duty Trucks
Vehicle - Bus
Vehicle - Train, Subway
Mass balance
Mass balance
Mass balance
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
AER and DE
AER and DE
AER and DE
None
None
None
None
None
None
PR
PR
PR
PR
PR
PE and PR
PE and PR
PE and PR
PE and PR
1 l AER=air exchange rate, DE=decay-deposition rate, PR=proximity factor, PE=penetration factor
2 5.4.1 Benchmark Levels Modeled
3 Benchmark levels used in this assessment include concentrations of 0.060, 0.070 and
4 0.080 ppm, which are the same benchmark levels used in the exposure assessment conducted in
5 the last review. Estimating exposures to ambient O3 concentrations at and above these
6 benchmark levels is intended to provide some perspective on the public health impacts of O3-
7 related health effects that have been demonstrated in human clinical and toxicological studies,
8 but cannot currently be evaluated in quantitative risk assessments, such as lung inflammation,
9 increased airway responsiveness, and decreased resistance to infection. The 0.080 ppm
10 benchmark represents an exposure level at which there is a substantial amount of clinical
11 evidence demonstrating a range of O3-related effects including lung inflammation and airway
12 responsiveness in healthy individuals. The 0.070 ppm benchmark reflects evidence that
13 asthmatics have larger and more serious effects than healthy people as well as a substantial body
14 of epidemiological evidence of associations with O3 levels that extend will below 0.080 ppm.
5-21
-------
1 The 0.060 ppm benchmark additionally represents the lowest exposure level at which OS-related
2 effects have been observed in clinical studies of healthy individuals.
3 5.5 VARIABILITY AND UNCERTAINTY
4 An important issue associated with any population exposure or risk assessment is the
5 characterization of variability and uncertainty. Variability refers to the inherent heterogeneity in
6 a population or variable of interest (e.g., residential air exchange rates). The degree of variability
7 cannot be reduced through further research, only better characterized with additional
8 measurement. Uncertainty refers to the lack of knowledge regarding the values of model input
9 variables (i.e., parameter uncertainty), the physical systems or relationships used (i.e., use of
10 input variables to estimate exposure or risk or model uncertainty)., and in specifying the scenario
11 that is consistent with purpose of the assessment (i.e., scenario uncertainty). Uncertainty is,
12 ideally, reduced to the maximum extent possible through improved measurement of key
13 parameters and iterative model refinement. The approaches used to assess variability and to
14 characterize uncertainty in this REA are discussed in the following two sections. Each section
15 also contains a concise summary of the identified components contributing to uncertainly and
16 how each source may affect the estimated exposures.
17 5.5.1 Treatment of Variability
18 The purpose for addressing variability in this REA is to ensure that the estimates of
19 exposure and risk reflect the variability of ambient O3 concentrations, population characteristics,
20 associated Os exposure and dose, and potential health risk across the study area and for the
21 simulated at-risk populations. In this REA, there are several algorithms that account for
22 variability of input data when generating the number of estimated benchmark exceedances or
23 health risk outputs. For example, variability may arise from differences in the population
24 residing within census tracts (e.g., age distribution) and the activities that may affect population
25 exposure to 63 (e.g., time spent inside vehicles, performing moderate or greater exertion level
26 activities outdoors). A complete range of potential exposure levels and associated risk estimates
27 can be generated when appropriately addressing variability in exposure and risk assessments;
28 note however that the range of values obtained would be within the constraints of the input
29 parameters, algorithms, or modeling system used, not necessarily the complete range of the true
30 exposure or risk values.
5-22
-------
1 Where possible, staff identified and incorporated the observed variability in input data
2 sets to estimate model parameters within the exposure assessment rather than employing
3 standard default assumptions and/or using point estimates to describe model inputs. The details
4 regarding variability distributions used in data inputs are described in Appendix 5B. To the
5 extent possible given the data available for the assessment, staff accounted for variability within
6 the exposure modeling. APEX has been designed to account for variability in some of the input
7 data, including the physiological variables that are important inputs to determining ventilation
8 rates. As a result, APEX addresses much of the variability in factors that affect human exposure.
9 Important sources of the variability accounted for in this analysis are summarized in Appendix
10 5D.
11 5.5.2 Characterization of Uncertainty
12 While it may be possible to capture a range of exposure or risk values by accounting for
13 variability inherent to influential factors, the true exposure or risk for any given individual within
14 a study area is largely unknown. To characterize health risks, exposure and risk assessors
15 commonly use an iterative process of gathering data, developing models, and estimating
16 exposures and risks, given the goals of the assessment, scale of the assessment performed, and
17 limitations of the input data available. However, significant uncertainty often remains and
18 emphasis is then placed on characterizing the nature of that uncertainty and its impact on
19 exposure and risk estimates.
20 The REA's for the previous 63, NC>2, 862, and CO NAAQS reviews each presented a
21 characterization of uncertainty of exposure modeling (Langstaff, 2007; EPA 2008, 2009, 2010).
22 Details regarding those approaches and a summary of the key findings of those reports that are
23 most relevant to the current ozone exposure assessment are provided in Appendix 5D. The most
24 influential elements of uncertainty are the following:
25 • Activity patterns
26 • Air exchange rates (AERs)
27 • Spatial variability in 63 concentrations
28 • METs distributions
29 • Resting metabolic rate and ventilation rate equations
5-23
-------
1 In the second draft REA, we plan to present the results of sensitivity analyses for each of
2 these five elements. Activity pattern sensitivity analyses will include restricting diaries to more
3 recent years, restricting diaries to be city-specific, and simulating activity patterns for specific
4 cohorts, including school children and outdoor workers. These will include the treatment of
5 activity patterns that can lead to repeated exposures to high ozone. Air exchange rates sensitivity
6 analyses will include restricting AERs to be city-specific. The sensitivity analyses for spatial
7 variability in Os concentrations will include varying the radius of influence of the air quality
8 monitors and using photochemical grid modeling results with the monitored concentrations to
9 improve the spatial interpolation of 63 concentrations. The influence of METs distributions,
10 resting metabolic rate equations, and ventilation rate equations will be ascertained by using
11 updated METs distributions and alternative resting metabolic rate and ventilation rate equations.
12 5.6 EXPOSURE ASSESSMENT RESULTS
13 5.6.1 Overview
14 The results of the exposure analysis are presented as a series of graphs focusing on a
15 range of benchmark levels, described in Chapter 2 and in Section 5.4.1 above, as being of
16 particular health concern. A range of concentrations in the air quality data measured over the five
17 year period (2006-2010) were used in the exposure model, providing a range of estimated
18 exposures output by the model. Exposure results are presented for recent air quality (base years)
19 and for air quality adjusted to just meet the current standards, based on 2006-2008 and 2008-
20 2010 design values, as described in Chapter 3. Estimates of exposures for the year 2008 were
21 developed for both of these sets of design values. This section first addresses the exposures
22 estimated for school children using figures and follows those with tables of estimates of
23 exposures for school-age children (ages 5-18), asthmatic school-age children, and the general
24 population, under moderate or greater exertion.
25 5.6.2 Exposure Modeling Results
26 A series of figures are presented for each of the benchmark levels (0.060, 0.070, and
27 0.080 ppm-8hr), for each of the five years, 2006 - 2010. Exposure estimates are presented for
28 those individuals experiencing moderate or greater levels of exertion averaged over the same 8-
29 hr period that the exposure occurred. The exertion level is characterized by breathing rates, as
30 described in Section 5.3.3. Results for school-age children exposed to 63 while engaged in
31 moderate exertion are presented in each of the subsequent figures. Results for asthmatic school -
5-24
-------
1 age children have similar exposure outcomes and patterns across the urban areas modeled (see
2 the sets of tables following the figures).
3 The next set of figures (Figure 5-1 though Figure 5-15) shows the percent of school-age
4 children who experience at least one 8-hour average exposure above the benchmark levels of
5 0.06, 0.07, and 0.08 ppm-8hr, while at the same time engaged in activities resulting in moderate
6 or greater exertion. On each figure the base case air quality exposure scenario can be compared
7 to exposures with air quality just meeting the current standard. "75 6-8" denotes the current
8 standard of 75 ppb based on 2006-2008 design values, and "75 8-10" denotes the current
9 standard of 75 ppb based on 2008-2010 design values. Note that the year 2008 has results for
10 both of these current standard scenarios, since it occurs in both of the design value periods 2006-
11 2008 and 2008-2010. For example, in Figure 5-7, 18 percent of school-age children in Atlanta
12 are estimated to have experienced one or more 8-hours average exposure of at least 0.06 ppm,
13 while engaged in moderate or greater exertion. When the air quality is adjusted to just meet the
14 current standard based on the 2006-2008 design value for Atlanta, this estimate is reduced to 12
15 percent. When the air quality is adjusted to just meet the current standard based on the 2008-
16 2010 design value for Atlanta, this estimate is 3 percent.
17
5-25
-------
Figure 5-1. Percent of Children in 2006 With 8-hour Exposures > 0.06 ppm Concomitant
With Moderate or Greater Exertion
base
756-,
a\
ATLA
DENY
LA
PHIL
Figure 5-2. Percent of Children in 2006 With 8-hour Exposures > 0.07 ppm Concomitant
With Moderate or Greater Exertion
base
7567
LJ
0%
ATLA
DENY
LA
PHIL
5-26
-------
Figure 5-3. Percent of Children in 2006 With 8-hour Exposures > 0.08 ppm Concomitant
With Moderate or Greater Exertion
base
756-,
ATLA
DENY
LA
PHIL
Figure 5-4. Percent of Children in 2007 With 8-hour Exposures > 0.06 ppm Concomitant
With Moderate or Greater Exertion
base
7567
ATLA
DENY
LA
PHIL
5-27
-------
Figure 5-5. Percent of Children in 2007 With 8-hour Exposures > 0.07 ppm Concomitant
With Moderate or Greater Exertion
base
756-,
ATLA
DENY
LA
PHIL
Figure 5-6. Percent of Children in 2007 With 8-hour Exposures > 0.08 ppm Concomitant
With Moderate or Greater Exertion
base
7567
ATLA
DENY
LA
PHIL
5-28
-------
Figure 5-7. Percent of Children in 2008 With 8-hour Exposures > 0.06 ppm Concomitant
With Moderate or Greater Exertion
base
75 8-10
756-
ATLA DENY
PHIL
Figure 5-8. Percent of Children in 2008 With 8-hour Exposures > 0.07 ppm Concomitant
With Moderate or Greater Exertion
758-10 / /—£
2%
a
1%
o%
4%
756-
0%
0%
o%
a
1%
ATLA DENY LA PHIL
5-29
-------
Figure 5-9. Percent of Children in 2008 With 8-hour Exposures > 0.08 ppm Concomitant
With Moderate or Greater Exertion
base
75 8-10
0%
0%
0%
756-
0%
0%
0%
0%
ATLA
DENY
LA
PHIL
0%
Figure 5-10. Percent of Children in 2009 With 8-hour Exposures > 0.06 ppm Concomitant
With Moderate or Greater Exertion
base
ATLA
DENY
LA
PHIL
5-30
-------
Figure 5-11. Percent of Children in 2009 With 8-hour Exposures > 0.07 ppm Concomitant
With Moderate or Greater Exertion
base
ATLA
DENY
LA
PHIL
Figure 5-12. Percent of Children in 2009 With 8-hour Exposures > 0.08 ppm Concomitant
With Moderate or Greater Exertion
base
ATLA
DENY
LA
PHIL
5-31
-------
Figure 5-13. Percent of Children in 2010 With 8-hour Exposures > 0.06 ppm Concomitant
With Moderate or Greater Exertion
base
75 8-/0
ATLA
DENY
LA
PHIL
Figure 5-14. Percent of Children in 2010 With 8-hour Exposures > 0.07 ppm Concomitant
With Moderate or Greater Exertion
base
ATLA
DENY
LA
PHIL
5-32
-------
Figure 5-15. Percent of Children in 2010 With 8-hour Exposures > 0.08 ppm Concomitant
With Moderate or Greater Exertion
base
ATLA
DENY
LA
PHIL
The following tables present results for school-age children, asthmatic school-age children, and
the general population.
5-33
-------
Table 4-5. Percent of people with 1 or more 8-hour exposures above different levels (ppb-8hr), Children (moderate exertion)
City
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
myear
2006
2006
2006
2006
2007
2007
2007
2007
2008
2008
2008
2008
2009
2009
2009
2009
2010
2010
2010
2010
Mean
Mean
Mean
Mean
Above 0
base
96.8%
96.3%
97.2%
96.4%
96.9%
96.4%
97.2%
96.4%
96.8%
96.4%
97.3%
96.4%
96.9%
96.4%
97.4%
96.4%
97.0%
96.4%
97.4%
96.5%
96.9%
96.4%
97.3%
96.4%
Above 60
75/4
2006-8
11.7%
14.2%
5.2%
8.3%
14.4%
7.6%
3.1%
11.3%
2.8%
5.6%
4.1%
7.5%
9.7%
9.1%
4.1%
9.0%
Above 60
75/4
2008-10
11.5%
15.8%
5.0%
16.2%
5.3%
8.1%
3.6%
2.9%
8.3%
9.0%
1.8%
18.4%
8.4%
11.0%
3.5%
12.5%
Above 60
base
31.9%
29.8%
36.7%
28.6%
35.8%
22.4%
32.5%
33.6%
18.1%
20.8%
35.2%
26.4%
10.1%
11.7%
31.5%
9.3%
14.5%
13.3%
23.1%
30.1%
22.1%
19.6%
31.8%
25.6%
Above 70
75/4
2006-8
1.7%
1.4%
0.5%
0.3%
2.1%
0.3%
0.2%
1.8%
0.2%
0.2%
0.3%
0.6%
1.3%
0.6%
0.3%
0.9%
Above 70
75/4
2008-10
1.8%
1.3%
0.4%
3.8%
0.7%
0.6%
0.2%
0.0%
0.9%
0.4%
0.0%
2.8%
1.1%
0.7%
0.2%
2.2%
Above 70
base
16.1%
10.4%
21.1%
11.8%
19.2%
4.8%
15.5%
16.4%
4.4%
2.9%
18.0%
10.6%
1.9%
1.1%
14.8%
0.6%
2.5%
1.0%
8.1%
10.9%
8.8%
4.0%
15.5%
10.1%
Above 80
75/4
2006-8
0.1%
0.0%
0.0%
0.0%
0.1%
0.0%
0.0%
0.1%
0.0%
0.0%
0.0%
0.0%
0.1%
0.0%
0.0%
0.0%
Above 80
75/4
2008-10
0.2%
0.1%
0.0%
0.3%
0.1%
0.0%
0.0%
0.0%
0.1%
0.0%
0.0%
0.1%
0.1%
0.0%
0.0%
0.1%
Above 80
base
5.7%
1.5%
9.7%
2.4%
7.0%
0.3%
5.8%
5.7%
0.7%
0.2%
7.0%
2.6%
0.2%
0.0%
5.4%
0.0%
0.3%
0.0%
2.3%
1.5%
2.8%
0.4%
6.0%
2.4%
5-34
-------
Table 4-6. Number of people with 1 or more 8-hour exposures above different levels (ppb-8hr), Children (moderate exertion)
City
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
myear
2006
2006
2006
2006
2007
2007
2007
2007
2008
2008
2008
2008
2009
2009
2009
2009
2010
2010
2010
2010
Mean
Mean
Mean
Mean
Above 0
base
829,000
532,000
3,510,000
1,120,000
829,000
540,000
3,500,000
1,120,000
828,000
540,000
3,510,000
1,120,000
828,000
537,000
3,520,000
1,120,000
829,000
537,000
3,520,000
1,120,000
829,000
537,000
3,510,000
1,120,000
Above 60
75/4
2006-8
100,000
78,500
186,000
96,600
123,000
42,500
111,000
130,000
24,300
31,100
147,000
86,600
82,700
50,700
148,000
105,000
Above 60
75/4
2008-10
98,700
88,500
182,000
188,000
44,900
45,100
129,000
33,800
71,100
50,200
66,000
213,000
71,600
61,300
126,000
145,000
Above 60
base
273,000
165,000
1,330,000
332,000
307,000
126,000
1,170,000
389,000
154,000
116,000
1,270,000
306,000
86,000
65,000
1,140,000
108,000
124,000
74,100
836,000
348,000
189,000
109,000
1,150,000
297,000
Above 70
75/4
2006-8
14,300
7,500
16,800
3,480
18,000
1,680
7,730
20,400
1,390
871
9,390
6,860
11,200
3,350
11,300
10,200
Above 70
75/4
2008-10
15,100
7,070
15,400
43,800
5,900
3,140
5,960
338
7,730
2,310
1,190
32,800
9,580
4,170
7,530
25,600
Above 70
base
138,000
57,600
762,000
137,000
165,000
26,700
558,000
190,000
37,500
16,100
651,000
123,000
16,300
6,030
534,000
7,220
21,100
5,640
292,000
127,000
75,400
22,400
559,000
117,000
Above 80
75/4
2006-8
1,110
51
0
235
936
0
0
1,460
76
39
0
0
707
30
0
564
Above 80
75/4
2008-10
1,760
390
224
3,120
439
52
0
0
592
0
0
1,610
929
147
75
1,580
Above 80
base
48,400
8,200
349,000
27,300
60,000
1,920
209,000
65,900
5,920
871
252,000
29,500
2,040
195
195,000
104
2,310
13
83,100
17,300
23,700
2,240
218,000
28,000
5-35
-------
Table 4-11. Percent of people with 1 or more 8-hour exposures above different levels (ppb-8hr), Asthmatic children (moderate exertion)
City
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
myear
2006
2006
2006
2006
2007
2007
2007
2007
2008
2008
2008
2008
2009
2009
2009
2009
2010
2010
2010
Mean
Mean
Mean
Mean
Above 0
base
96.9%
96.3%
97.7%
96.7%
97.0%
96.5%
98.0%
96.9%
97.0%
96.8%
97.2%
97.1%
97.3%
96.2%
97.3%
96.7%
97.3%
96.5%
97.0%
97.1%
96.5%
97.5%
96.9%
Above 60
75/4
2006-8
11.7%
14.9%
5.1%
8.6%
15.0%
7.6%
3.5%
12.4%
3.0%
6.0%
4.0%
7.6%
9.9%
9.4%
4.2%
9.5%
Above 60
75/4
2008-10
11.8%
16.5%
5.0%
17.0%
5.4%
8.3%
3.6%
3.0%
8.9%
8.6%
18.6%
8.7%
11.2%
3.4%
12.9%
Above 60
base
32.8%
30.8%
38.0%
29.4%
36.6%
23.7%
32.5%
35.2%
18.4%
22.2%
36.9%
27.9%
10.1%
11.6%
32.3%
9.5%
15.1%
13.0%
30.5%
22.7%
20.3%
26.4%
Above 70
75/4
2006-8
1.7%
1.3%
0.6%
0.3%
1.7%
0.3%
0.4%
1.8%
0.2%
0.1%
0.4%
0.6%
1.2%
0.6%
0.4%
0.9%
Above 70
75/4
2008-10
2.0%
1.4%
0.5%
4.1%
0.6%
0.6%
0.1%
0.0%
0.8%
0.4%
2.6%
1.1%
0.8%
0.2%
2.2%
Above 70
base
16.0%
11.0%
21.8%
12.5%
19.8%
5.1%
16.7%
17.9%
4.6%
3.3%
18.2%
10.8%
1.7%
1.1%
15.2%
0.7%
2.4%
1.1%
10.8%
9.0%
4.3%
10.5%
Above 80
75/4
2006-8
0.1%
0.1%
0.0%
0.0%
0.1%
0.0%
0.0%
0.1%
0.0%
0.0%
0.0%
0.0%
0.1%
0.0%
0.0%
0.0%
Above 80
75/4
2008-10
0.2%
0.1%
0.0%
0.3%
0.0%
0.0%
0.0%
0.0%
0.1%
0.0%
0.2%
0.1%
0.0%
0.0%
0.1%
Above 80
base
5.6%
1.5%
10.4%
2.6%
7.3%
0.3%
6.6%
6.0%
0.6%
0.2%
6.9%
2.8%
0.1%
0.1%
5.4%
0.0%
0.2%
0.0%
1.4%
2.8%
0.4%
2.5%
5-36
-------
Table 4-12. Number of people with 1 or more 8-hour exposures above different levels (ppb-8hr), Asthmatic children (moderate exertion)
City
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
myear
2006
2006
2006
2006
2007
2007
2007
2007
2008
2008
2008
2008
2009
2009
2009
2009
2010
2010
2010
Mean
Mean
Mean
Mean
Above 0
base
83,900
47,800
311,000
129,000
83,900
48,800
312,000
128,000
83,900
48,900
318,000
131,000
81,200
47,700
319,000
130,000
81,300
47,800
131,000
82,900
48,200
315,000
130,000
Above 60
75/4
2006-8
10,200
7,380
16,300
11,500
13,000
3,840
11,000
16,300
2,580
3,020
13,000
10,200
8,560
4,740
13,400
12,700
Above 60
75/4
2008-10
10,200
8,320
16,300
22,900
4,540
4,100
11,800
4,050
7,460
4,270
25,000
7,390
5,560
11,100
17,300
Above 60
base
28,400
15,300
121,000
39,300
31,700
12,000
103,000
46,500
15,900
11,200
121,000
37,600
8,450
5,760
106,000
12,800
12,600
6,420
41,000
19,400
10,100
35,400
Above 70
75/4
2006-8
1,510
643
1,780
419
1,490
169
1,120
2,340
172
65
1,270
831
1,060
292
1,390
1,200
Above 70
75/4
2008-10
1,760
728
1,790
5,460
496
298
373
52
649
182
3,530
967
403
770
3,010
Above 70
base
13,800
5,440
69,300
16,700
17,100
2,570
53,000
23,700
3,950
1,680
59,800
14,500
1,450
532
49,700
909
2,040
558
14,600
7,680
2,160
14,100
Above 80
75/4
2006-8
76
26
0
26
57
0
0
104
19
0
0
0
51
9
0
43
Above 80
75/4
2008-10
210
26
149
338
0
0
0
0
57
0
234
89
9
50
191
Above 80
base
4,830
720
33,100
3,430
6,300
169
21,100
7,910
554
91
22,700
3,720
114
26
17,700
26
153
0
1,870
2,390
201
3,390
5-37
-------
Table 4-17. Percent of people with 1 or more 8-hour exposures above different levels (ppb-8hr), All people (moderate exertion)
City
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
myear
2006
2006
2006
2006
2007
2007
2007
2007
2008
2008
2008
2008
2009
2009
2009
2009
2010
2010
2010
2010
Mean
Mean
Mean
Mean
Above 0
base
80.5%
79.4%
81.0%
76.5%
80.7%
79.4%
80.9%
76.6%
80.5%
79.5%
80.9%
76.5%
80.7%
79.7%
81.0%
76.3%
80.8%
79.7%
81.0%
76.6%
80.7%
79.6%
81.0%
76.5%
Above 60
75/4
2006-8
7.7%
8.6%
3.2%
4.6%
8.1%
4.6%
2.1%
6.4%
2.1%
3.7%
2.7%
4.3%
6.0%
5.6%
2.7%
5.1%
Above 60
75/4
2008-10
8.0%
10.2%
3.3%
9.5%
3.5%
4.9%
2.4%
1.7%
5.2%
6.1%
1.2%
10.4%
5.6%
7.1%
2.3%
7.2%
Above 60
base
21.6%
18.1%
21.1%
17.0%
22.9%
13.4%
18.7%
20.2%
12.2%
13.4%
21.0%
15.6%
6.5%
7.1%
18.0%
5.3%
9.3%
8.8%
13.3%
17.7%
14.5%
12.1%
18.4%
15.2%
Above 70
75/4
2006-8
1.2%
0.9%
0.4%
0.2%
1.1%
0.2%
0.1%
0.9%
0.1%
0.1%
0.2%
0.4%
0.8%
0.4%
0.2%
0.5%
Above 70
75/4
2008-10
1.4%
0.9%
0.3%
2.2%
0.6%
0.4%
0.1%
0.0%
0.5%
0.3%
0.0%
1.6%
0.8%
0.6%
0.2%
1.3%
Above 70
base
10.4%
6.3%
11.5%
6.5%
10.9%
2.7%
8.8%
9.2%
3.1%
2.0%
10.4%
6.1%
1.3%
0.8%
8.3%
0.5%
1.5%
0.8%
4.8%
6.1%
5.5%
2.5%
8.8%
5.7%
Above 80
75/4
2006-8
0.1%
0.0%
0.0%
0.0%
0.1%
0.0%
0.0%
0.1%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
Above 80
75/4
2008-10
0.2%
0.0%
0.0%
0.2%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.0%
0.1%
0.1%
0.0%
0.0%
0.1%
Above 80
base
3.6%
0.9%
5.4%
1.3%
3.6%
0.2%
3.4%
3.1%
0.6%
0.1%
4.3%
1.5%
0.2%
0.0%
3.2%
0.0%
0.1%
0.0%
1.4%
0.9%
1.6%
0.3%
3.5%
1.4%
5-38
-------
Table 4-19. Number of people with 1 or more 8-hour exposures above different levels (ppb-8hr), All people (moderate exertion)
City
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
Atlanta
Denver
Los Angeles
Philadelphia
myear
2006
2006
2006
2006
2007
2007
2007
2007
2008
2008
2008
2008
2009
2009
2009
2009
2010
2010
2010
2010
Mean
Mean
Mean
Mean
Above 0
base
3,080,000
2,040,000
12,100,000
4,000,000
3,080,000
2,060,000
12,000,000
3,990,000
3,080,000
2,070,000
12,100,000
3,970,000
3,080,000
2,070,000
12,100,000
3,970,000
3,080,000
2,070,000
12,100,000
3,980,000
3,080,000
2,060,000
12,100,000
3,980,000
Above 60
75/4
2006-8
294,000
221,000
474,000
240,000
309,000
119,000
310,000
333,000
82,000
95,300
409,000
224,000
228,000
145,000
398,000
266,000
Above 60
75/4
2008-10
304,000
264,000
496,000
492,000
135,000
128,000
360,000
89,200
200,000
159,000
183,000
541,000
213,000
184,000
347,000
374,000
Above 60
base
826,000
465,000
3,140,000
888,000
874,000
347,000
2,780,000
1,050,000
466,000
349,000
3,130,000
808,000
249,000
184,000
2,680,000
277,000
356,000
227,000
1,990,000
922,000
554,000
314,000
2,750,000
789,000
Above 70
75/4
2006-8
45,200
23,500
62,500
11,900
42,800
5,730
20,300
44,700
4,280
2,640
28,000
18,500
30,800
10,600
37,000
25,000
Above 70
75/4
2008-10
53,500
24,200
47,000
114,000
21,100
11,400
22,100
2,000
19,600
8,380
3,500
85,500
31,400
14,700
24,200
67,300
Above 70
base
396,000
161,000
1,720,000
339,000
418,000
71,200
1,300,000
480,000
119,000
51,000
1,550,000
316,000
51,500
20,000
1,240,000
24,900
58,400
20,300
722,000
315,000
209,000
64,700
1,310,000
295,000
Above 80
75/4
2006-8
3,040
527
223
785
1,990
52
0
3,330
76
39
0
78
1,700
206
74
1,400
Above 80
75/4
2008-10
5,750
1,210
447
10,900
1,200
584
0
0
1,200
0
0
3,820
2,720
598
149
4,910
Above 80
base
138,000
24,200
805,000
69,900
139,000
5,920
501,000
160,000
21,700
2,630
639,000
76,900
6,560
947
477,000
987
5,270
91
211,000
44,500
62,200
6,750
527,000
70,400
5-39
-------
5.6.3 Characterization Of Factors Influencing High Exposures
In this analysis, we investigated the particular factors that influence estimated exposures
with a focus on persons experiencing the highest daily maximum 8-hour exposures within each
study area. This analysis required the generation of detailed APEX output files having varying
time intervals, that is, the daily, hourly, and minute-by-minute (or events) files. Given that the
size of these time-series files is dependent on the number of persons simulated, we simulated
5,000 persons and restricted the analysis to a single year (2006) to make this evaluation
tractable.8 Both the base case (unadjusted or 'as is' recent air quality conditions) and ambient O3
adjusted to just meet the current standard (0.075 ppm) air quality scenarios were evaluated in
each of the four study areas. All APEX conditions (e.g., ME descriptions, AERs, MET data)
were consistent with the 200,000 person APEX simulations that generated all of summary output
discussed in the main body of this chapter.
We were interested in identifying the specific microenvironments and activities most
important to 63 exposure and evaluating their duration and particular times of the day persons
were engaged in them. Because ambient 63 concentrations peak mainly in the afternoon hours,
we focused our microenvironmental time expenditure analysis on the hours between 12PM and
8PM. For every day of the exposure simulation, we aggregated the time spent outdoors, indoors,
near-roadways, and inside vehicles during these afternoon hours (i.e., the time of interest
summed to 480 minutes per person day). Data from several APEX output files were then
combined to generate a single daily file for each person containing a variety of personal
attributes (e.g., age, sex), their daily maximum 8-hour ambient and exposure concentrations, and
the aforementioned time expenditure metrics.
We performed an analysis of variance (ANOVA) using SAS PROC GLM (SAS, 2012) to
determine the factors contributing most to variability in the dependent variable, i.e., each
person's daily maximum 8-hour Os exposure concentrations. This analysis was distinct for five
We recognize that there is year-to-year variability in ambient O3 concentrations and it is possible that fewer
persons simulated could result in differences in exposures compared to large-scale multi-year model simulations.
Based on a similar detailed evaluation performed for the Carbon Monoxide REA (US EPA, 2010), it is expected any
differences that exist between exposures estimated in a large simulation versus that using a smaller subset of persons
would be small and of limited importance to this particular evaluation.
5-40
-------
age-groups of interest (<5, 5-17, 18-35, 36-65, >65 years of age). The final models9 included a
total of seven explanatory variables: the main effects of (1) daily maximum 8-hour ambient Os,
(2-4) afternoon time spent outdoors, near-roads, and inside vehicles,10 and (5) PAI, while also
including interaction effects from (6) afternoon time outdoors by daily maximum 8-hour ambient
concentration and (7) PAI by afternoon time outdoors. Two conditions were considered: all
person days of the simulation, and only those days where a person's 8-hour maximum exposure
concentration was >0.05 ppm.11 Selected output from this ANOVA included parameter
estimates for each variable, model R-square statistic (R2), and Type III model sums of squares
(SS3).12
Model fits, as indicated by an R2 value, were reasonable across each of the study areas
(Table 5-5). The selected factors explain about 40-80% of the total variability in 8-hour daily
maximum exposures. Model fits were best when using all person days of the simulation and
results were similar for both air quality scenarios. When considering only those days where
persons had 8-hour daily maximum 63 exposures >0.05 ppm, consistently less variability was
explained by the factors included in each model, though overall model fits were acceptable.
Furthermore, the most robust models were those developed using either children aged 5-17 or
adults 18-35 years old (e.g., see Table 5-6 for Los Angeles model R2 by age groups).
Table 5-5. Range of ANOVA model R fit statistics by study area, air quality scenario, and
exposure level.
Study Area
Atlanta
Denver
Los Angeles
Philadelphia
Base Case Model R2
All Person
Days
0.64-0.75
0.62-0.69
0.72-0.79
0.65-0.71
Person Days with 8-hour
Exposure > 0.05 ppm
0.55-0.63
0.41-0.62
0.47-0.68
0.43-0.64
Current Standard Model R2
All Person
Days
0.62-0.74
0.61-0.68
0.69-0.76
0.63-0.69
Person Days with 8-hour
Exposure > 0.05 ppm
0.52-0.64
0.45-0.62
0.54-0.66
0.41-0.64
9 In this investigation, we also evaluated the influence of sex, work and home districts, meteorological zones, each
with varying statistical significance, though overall adding little to explaining variability beyond the final
explanatory variables included.
10 Including indoor afternoon time creates a strict linear dependence among these four variables and generates biased
estimates, thus it was neither included nor needed in this analysis.
11 This breakpoint was selected due to the limited sample size (5,000 total simulated persons), an issue of increasing
importance when selecting for persons with the highest exposures.
12 In each of the ANOVA models constructed, type II = type III = type IV sums of squares.
5-41
-------
Table 5-6. ANOVA model R fit statistics in Los Angeles by age group, air quality
scenario, and exposure level.
Study Area
Los Angeles
Age
Group
(years)
<5
5-17
18-36
36-64
>65
Base Case Model R2
All Person
Days
0.74
0.79
0.73
0.73
0.72
Person Days with
8-hour Exposure
> 0.05 ppm
0.47
0.61
0.65
0.68
0.58
Current Standard Model R2
All Person
Days
0.71
0.76
0.70
0.70
0.69
Person Days with
8-hour Exposure
> 0.05 ppm
0.59
0.54
0.65
0.66
0.62
We evaluated the relative contribution each variable had on the total explained variability
using the SS3 in each respective model.13 As with the R2 statistics generated above, there were
four separate model results generated per study area, with relative contribution results for Los
Angeles illustrated in Figure 5-16. When considering all person days of the simulation (left side
of figure), the daily maximum 8-hour ambient 63 concentration variable contributes the greatest
to the explained model variance, consistently estimated to be about 80% across all age groups
and for either air quality scenario. The interaction of this variable with afternoon outdoor time
contributes an additional 10% to the explained variance, indicating that both ambient
concentration and time spent outdoors collectively contribute to 90% or more of the explained
model variance when evaluating all (both high, mid and low) daily maximum 8-hour Os
exposure concentrations. The main effect of outdoor time contributed very little to the explained
variance under these conditions as did contributions from the other included variables, except for
time spent near-roads (about a 5% contribution). These results suggest that when considering the
Los Angeles study population broadly, the daily maximum 8-hour ambient O3 concentration is
the most important driver in estimating population exposures Os, nearly regardless of specific
microenvironmental locations where exposure might occur.
When considering only person days having daily maximum 8-hour Os exposures > 0.05
ppm and for either air quality scenario in Los Angeles, collectively the main effects of ambient
concentration and outdoor time combined with their interaction similarly contribute to
approximately 80% of the total explained variance (right side of Figure 5-16). However, the
13 Type III sums of squares (SS3) for a given effect are adjusted for all other effects evaluated in the model,
regardless of whether they contain the given effect or not. Thus the SS3 for each variable represents the individual
effect sums of squares that sum to the total effect sums of squares (or the total model explained variance).
5-42
-------
main effect of the 8-hour daily maximum ambient 63 concentration variable has a sharply lower
contribution (generally about 5-15%) along with greater contribution from the main effects
variable outdoor time (15-20% contribution) and its interaction with the ambient concentration
variable (50-60%). These results suggest that for highly exposed persons, the most important
drivers are time spent outdoors corresponding with high daily maximum 8-hour ambient Os
concentrations.
Results for Atlanta were generally similar to Los Angeles (Figure 5-17), with notable
differences discussed here.14 The contribution of the maximum 8-hour ambient Os concentration
variable to the total explained variance (about 40-50%) was less than that observed in Los
Angeles when considering all person days (left side of figures 5-16 and 5-17), while the
contribution from the outdoor time/ambient O3 interaction variable was greater in Atlanta (about
20-40% versus 10% in Los Angeles). This dissimilarity is likely driven by the differences in
A/C prevalence rates and AER distributions used for each study area. Los Angeles has lower
A/C prevalence and higher AERs, thus a greater contribution to exposure is expected from
ambient concentrations by infiltrating to indoor microenvironments and hence, reflected in the
strong main effects for the 8-hour daily maximum ambient Os concentration variable in Los
Angeles. Afternoon time spent near Atlanta roads was estimated to contribute to about 20-30%
of the total explained variance when considering all person days and exposures, a value greater
than that estimated for Los Angeles (generally about 5%) again possibly reflecting an increased
importance of outdoor microenvironments in Atlanta relative to that in Los Angeles and the other
study locations (not shown).
Because afternoon outdoor time expenditure and 8-hour daily maximum ambient Os
concentrations are an important determinant for maximum Os exposures regardless of air quality
scenario, we compared the distributions of the two variables considering person day exposures
below and at or above 0.05 ppm. Figure 5-18 presents an example of this comparison for Los
14 This discussion regarding the relative contribution of the variables to the total explained model variance also
applies to the other two study areas, whereas results for Denver and Philadelphia were generally similar to Los
Angeles. While A/C prevalence is greatest in Philadelphia compared to LA and Denver, the AER distributions are
identical to those used for Denver and similar to LA.
5-43
-------
I
(3 18 to 35
Los Angeles, All Person Days, Base Air Quality
• Max8-hourAmbO3
• outdoors 12-8PM
• MaxS-hour Amb O3 *
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
Relative Percent of Model Explained Variance
Los Angeles, All Person Days, Current Std.
• MaxS-hour Amb O3
• outdoors 12-8PM
• MaxS-hour Amb O3 *
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
Relative Percent of Model Explained Variance
Los Angeles, Person Days w/8-hr Exposure >0.05 ppm, Base Air Quality
• MaxS-hour AmbO3
• outdoors 12-8PM
• MaxS-hour AmbO3*
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
0% 20% 40% 60% 80% 100%
Relative Percent of Model Explained Variance
LOG Angeles, Person Days w/8-hr Exposures >0.05 ppm,
Current Std.
• MaxS-hour Amb O3
• outdoors 12-8PM
• MaxS-hour AmbO3*
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
20% 40% 60% 80% 100%
Relative Percent of Model Explained Variance
Figure 5-16. Contribution of individual variables to total model explained variance by age group, air quality scenario, exposure level in
Los Angeles.
5-44
-------
Atlanta, All Person Days, Base Air Quality
• Max 8-hour AmbO3
• outdoors 12-8PM
• Max 8-hour AmbO3*
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
0% 20% 40% 60% 80%
Relative Percent of Model Explained Variance
Atlanta, All Person Days, Current Std.
• Max 8-hour Amb O3
• outdoors 12-8PM
• Max 8-hour Amb O3 *
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
20% 40% 60% 80%
Relative Percent of Model Explained Variance
anta, Person Days w/8-nr Exposures >0.05 ppm, Bas
; Air Quality
• Max8-hourAmbO3
• outdoors 12-8PM
• Max8-hourAmbO3*
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
20% 40% 60% 80% 100%
Relative Percent of Model Explained Variance
tlanta, Person Days w/8-hr Exposures> 0.05 ppm, Current Std
• Max 8-hour Amb O3
• outdoors 12-8PM
• Max 8-hour Amb O3 *
outdoors 12-8PM
• PAI
• outdoors 12-8PM*PAI
near road 12-8PM
vehicle 12-8PM
20% 40% 60% 80% 100%
Relative Percent of Model Explained Variance
Figure 5-17. Contribution of individual variables to total model explained variance by age group, air quality scenario, exposure level in
Atlanta.
5-45
-------
1 Angeles children15 and considering the base air quality for year 2006 (top). Not surprising, the
2 distributions for both the outdoor time and ambient concentration variables are shifted to the
3 right of the figure for person days where 8-hour daily maximum exposures > 0.05 ppm, as more
4 than half of the days, simulated persons spend about 250 minutes outdoors during the afternoon
5 hours along with experiencing daily maximum 8-hour ambient Os concentrations > 0.075 ppm.
6 For days where daily maximum 8-hour Os exposure < 0.05 ppm, greater than half of the person
7 days had no time spent outdoors and 8-hour daily maximum ambient Os concentrations < 0.045
8 ppm. By design, when air quality is simulated to just meet the current standard (Figure 5-18,
9 bottom) upper percentile ambient concentrations are dramatically reduced compared to those
10 comprising the base air quality such that the majority of concentrations fall well below the
11 current standard level of 0.075 ppm. Given so few occurrences of very high 8-hour ambient Oj
12 concentrations for this air quality scenario, only those persons having a majority of their time
13 spent outdoors experienced the highest 8-hour Os exposure concentrations.
14 By definition, an 8-hour exposure is time-averaged across all microenvironmental
15 concentrations therefore several different microenvironments may contribute to each person's
16 daily maximum level. Understandably based on the above analysis, the outdoor
17 microenvironment is the most important for those having the highest Os exposures, but we are
18 also interested in the percentage of time expenditure spent among detailed indoor, outdoor, and
19 vehicular locations people may inhabit during the afternoon. As an example, Figure 5-19
20 presents this information for Los Angeles children (ages 5-17) having daily maximum 8-hour
21 average Os exposures > 0.05 ppm and considering base air quality conditions. On average,
22 approximately 50% of total afternoon time is spent outdoors, of which half of this portion is
23 spent outdoors at home, with parks and other non-residential outdoor locations comprising the
24 remaining portion. Approximately 40% of the children's time on high exposure days is spent
25 indoors, while only 10% of time is spent near-roads or inside motor vehicles. Afternoon
26 microenvironmental time expenditure for highly exposed adults in Los Angeles was generally
27 similar with these estimates (data not shown).
28
15 The overall features of these two outdoor time and ambient concentration distributions are similar in the other
study areas (data not shown).
5-46
-------
2
O
4
5
Los Angeles, Children 5-17 Years Old, Base Air Quality
Maximum 8-Hour Ambient Ozone (ppm)
0 0.03 0.06 0.09 0.12 0.15
100
90 -
100
- 90
Out time: Max 8-H Expos <0.05 ppm
Out time: Max 8-H Expos >=0.05 ppm
- Max 8-H Amb: Max 8-H Expos < 0.05 ppm
100 200 300 400 500
Time Spent Outdoors Between 12-8PM (minutes)
Los Angeles, Children 5-17 Years Old, Current Std.
Maximum 8-Hour Ambient Ozone (ppm)
0 0.03 0.06 0.09 0.12 0.15
0)
0)
u
i.
0)
Out time: Max 8-H Expos <0.05 ppm
Out time: Max 8-H Expos >=0.05 ppm
----- Max 8-H Amb: Max 8-H Expos < 0.05 ppm
• Max 8-H Amb: Max 8-H Expos >= 0.05 ppm
60 *>
0)
u
i.
0)
50
40
- 30
20
10
0
100 200 300 400
Time Spent Outdoors Between 12-8PM (minutes)
500
Figure 5-18. Distributions of afternoon outdoor time expenditure and 8-hour daily maximum
ambient 63 concentrations for Los Angeles children (0-17) person days with 8-hour daily
maximum exposures > 0.05 ppm.
5-47
-------
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
Afternoon Time Expenditure: Children 5-17, 8-hr Exposures >
0.05 ppm, Los Angeles, Base Air Quality
Outdoor (School.
grounds)
1%
Outdoor (Park or Golf
course)
11%
Vehicle (Cars and Light
Duty Trucks)
5%
Vehicle Other (5 MEs)
0%
Indoor (Residence)
31%
Outdoor Other (Non-_
residential)
14%
Outdoor (Residential)
25%
Near-road Other (3
MEs)
Indoor (School)
3%
Indoor Other (11 MEs)
6%
.Near-road (Within 10
yards of street)
3%
-Near-road (Service
station)
1%
Figure 5-19. Afternoon microenvironmental time expenditure for Los Angeles children (ages 5-
17) experiencing 8-hour daily maximum 63 exposures > 0.05 ppm, base air quality.
A person's activity level plays an important role in estimating the risk of adverse health
responses. As such, we evaluated the activities performed by highly exposed individuals while
they spent time outdoors during the afternoon hours. Note there are over 100 specific activity
codes used in CHAD/APEX, though not all of these will be used in an exposure modeling
simulation depending on the diaries that are selected to represent the simulated population. We
summed the time spent in each specific activity across all highly exposed persons that spent time
outdoors, ranked them, and identified the top ten activities performed. An aggregate of any
remaining less often performed activities was generated to complete this analysis of activity time
expenditure.
Figure 5-20 shows results for Los Angeles children, indicating that greater than half of
the time highly exposed children spent outdoors specifically involves performing a moderate or
greater exertion level activity, such as a sporting activity. The same type of analysis was done
5-48
-------
1 for highly exposed adults in Los Angeles (Figure 5-21), whereas about 25% of the outdoor time
2 expenditure was spent engaged in a paid work related activity (though not necessarily a high
3 exertion level activity), 20% of the time was spent playing sports or other moderate or greater
4 exertion level activity, with much of the remaining specific activities associated with low
5 exertion level (e.g., eating, sitting, visiting) or other less frequently performed activities of
6 variable exertion level.
7 These results support our earlier assessment results in identifying children as an
8 important exposure population group, largely a result of the combined outdoor time expenditure
9 along with concomitantly performing moderate or high exertion level activities. However, one
10 issue not explicitly addressed in the exposure modeling and remaining as a limitation to the
11 results is that outdoor workers are not addressed by our modeling.
12
Afternoon Outdoor Activities: Children 5-17, 8-hr Exposures >
0.05 ppm, Los Angeles, Base Air Quality
Other (67 activities) _^^ Participate in sports
23% """ ^^^ 25%
Outdoor chores
2%
.Play/Outdoor Leisure
Travel, All -Bunting J ^B ' 12%
3%
Playgames.
4%
Exercise/Walk^ ^^^ »^^ Leisure/Active
7/o Other sports and aTO Leisure/Sports
leisure
13
14
15 Figure 5-20. Activities performed during afternoon time outdoors for Los Angeles children
16 (ages 5-17) experiencing 8-hour daily maximum 63 exposures > 0.05 ppm, base air quality.
17
5-49
-------
2
3
4
5
Afternoon Outdoor Activities: Adults 18-35, 8-hr Exposures >
0.05 ppm, Los Angeles, Base Air Quality
Work, unspecified
20%
Other (81 activities)
32%
Other repairs
3%
Participate in sports
10%
^J^Attend sports events P^e sitting
Outdoor chores
7%
Other entertainment /
social events
6%
Figure 5-21. Activities performed during afternoon time outdoors for Los Angeles adults (ages
18-35) experiencing 8-hour daily maximum Os exposures > 0.05 ppm, base air quality.
5-50
-------
1 5.6.4 Discussion of Exposure Modeling Results
2 The patterns of estimated exposures are variable from city to city, primarily due to
3 differences in air quality (local emissions and meteorology affect these), the rollback procedure
4 as applied to each separate area, and people's time-location-activity patterns. Inspection of
5 Figures 4-1 to 4-15 shows marked differences between urban areas in the levels of exposures,
6 both for the base case and current standard scenarios. For example, under the current standard, it
7 is estimated that 14 percent of the Denver children but very few of the Los Angeles children
8 experience 8-hr Os exposures above 0.06 ppm-8hr while engaged in moderate exertion based on
9 2006. In 2007, the percents of exposures above 0.06 ppm-8hr ranged from 14 percent in Atlanta
10 to 3 percent in Los Angeles; in 2010 the percents ranged from 18 percent in Philadelphia to 2
11 percent in Los Angeles. Los Angeles in most cases has a smaller percent of children with
12 exposures above 0.06 and 0.07 ppm-8hr than the other cities. In Los Angeles, because of the
13 highly skewed nature of the distribution of ozone concentrations, much more of the upper range
14 of the air quality distribution needed to be rolled back to allow for the meeting of the current
15 standards, thus significantly reducing the frequency of occurrence of high ambient
16 concentrations (and therefore exposures).
17 After simulating just meeting the current standard, estimates of exposures above 0.07
18 ppm-8hr while engaged in moderate exertion are 2 percent or below, except for Philadelphia,
19 which has estimates of 4 percent in 2008 and 3 percent in 2010 for children. Estimates of
20 exposures above 0.08 ppm-8hr while engaged in moderate exertion are less than 0.5 percent for
21 all cities and years after simulating just meeting the current standard.
22 As discussed in Chapter 3, multiple exposures pose a greater health concern than single
23 exposures. However, multiple repeated exposures are greatly underestimated by APEX
24 (Langstaff, 2007, p. 49-50). This underestimation results primarily from the way that people's
25 activities are modeled using CHAD, which does not properly account for repeated behavior of
26 individuals. Repeated routine behavior from one weekday to the next is not simulated. For
27 example, there are no simulated individuals representing children in summer camps who spend a
28 large portion of their time outdoors, or adults with well-correlated weekday schedules. These
29 limitations apply to both children and adults, and therefore multiple exposures to children are
30 also expected to be underestimated by APEX. The second draft REA will provide quantitative
5-51
-------
1 estimates of the extent of repeated exposures for selected populations for which sequences of
2 daily activities can be reliably constructed.
3 The year-to-year variability in exposures in recent years, due in varying degrees to
4 changes in weather and emissions of precursors to Os, can be seen in Figures 5-22 to 5-25, which
5 show results for the 2006 to 2010 base case scenarios for each urban area and illustrate the range
6 of exposures generated by the use of multiple years of ambient air quality data. These figures
7 show the percent of school-age children who experience at least one 8-hour average exposure
8 above levels ranging from 0.04 to 0.08 ppm-8hr, with all five years presented in each graph.
9 Figure 5-22 illustrates the estimates of the percent of children in Atlanta who experience 8-hr 63
10 exposures above levels ranging from 0.04 to 0.08 ppm-8hr while engaged in moderate exertion.
11 Each line represents the estimates for one year, from 2006 to 2010. In Atlanta, 2007 had the
12 most exposures, while 2009 saw the least. Figures 5-23, 24, and 25 illustrate these results for
13 Denver, Los Angeles, and Philadelphia. These figures demonstrate that, while different years
14 have the highest and lowest numbers of exposed children for different cities, the trends across
15 exposure levels are similar, both across cities and across years.
16 The exposure modeling results are discussed further in Chapter 9.
17
Figure 5-22. Percent of Children (moderate exertion) in Atlanta with at least one 8-hour
exposure above different levels, across years
D
90% -
80% -
70% -
60% -
50%-
40% -
30%-
20%
0.04
D
2006
0.05 0.06 0.07
Exposure level (ppm-8hr)
2007 2008 2009
0.08
2010
5-52
-------
Figure 5-23. Percent of Children (moderate exertion) in Denver with at least one 8-hour
exposure above different levels, across years
D
90% -
80% :
70% :
60% :
50% -
40% -
30%-
20% -
0.04
D
2006
0.05 0.06 0.07
Exposure level (ppm-8hr)
2007 2008 2009
0.08
2010
Figure 5-24. Percent of Children (moderate exertion) in Los Angeles with at least one 8-
hour exposure above different levels, across years
D
90%:
80% -
70% -
60% -
50% -
40% -
30%-
20% -
10%-
0.04
D
2006
0.05 0.06 0.07
Exposure level (ppm-8hr)
2007 2008 2009
0.08
2010
5-53
-------
1
2
4
5
6
10
11
12
13
14
15
16
17
18
Figure 5-25. Percent of Children (moderate exertion) in Philadelphia with at least one 8-
hour exposure above different levels, across years
D
90%-"
80% -
70% -
60% -
50% -
40% -
30%-
20% -
10%-
0.04
D
2006
0.05 0.06 0.07
Exposure level (ppm-8hr)
2007 2008 2009
0.08
2010
REFERENCES
Ainsworth, B. E., Haskell, W. L., Leon, A. S., Jacobs, D. R., Jr., Montoye, H. J., Sallis, J. F.,
Paffenbarger, R. S., Jr. (1993). Compendium of physical activities: classification of energy
costs of human physical activities. Med Sci Sports Exer. 25:71-80.
Akland, G. G., Hartwell, T. D., Johnson, T. R., Whitmore, R. W. (1985). Measuring human
exposure to carbon monoxide in Washington, D. C. and Denver, Colorado during the
winter of 1982-83. Environ Sci Technol. 19:911-918.
Blanken PD, Dillon J, Wismann G. (2001). The impact of an air quality advisory program on
voluntary mobile source air pollution reduction. Atmos Environ. 35:2417-2421.
Bresnahan BW, Dickie M, Gerking S. (1997). Averting behavior and air pollution. LandEcon.
73:340-357.
Dey, A. N., and Bloom B. (2005). Summary Health Statistics for U.S. Children: National Health
Interview Survey, 2003. National Center for Health Statistics. Vital Health Stat 10(223).
Available at: http://www.cdc.gov/nchs/data/series/sr_10/sr 10_223.pdf
Freedson, P. S. (1989). Field monitoring of physical activity in children. Pediatr Exerc Sci.
1:8-18
5-54
-------
1 Graham SE (2012). Comprehensive review of published averting behavior studies and available
2 technical documents. Memo to Bryan Hubbell, Group Leader, Risk and Benefits Group,
3 Office of Air Quality Planning and Standards. EPA docket #
4 Graham, S. E. and McCurdy, T. (2004). Developing meaningful cohorts for human exposure
5 models. JExpo Anal Environ Epidemiol. 14: 23-43.
6 Hartwell, T. D., Clayton, C. A., Ritchie, R. M., Whitmore, R. W., Zelon, H. S., Jones, S. M.,
7 Whitehurst, D. A. (1984). Study of Carbon Monoxide Exposure of Residents of
8 Washington, DC and Denver, Colorado. Research Triangle Park, NC: U.S. Environmental
9 Protection Agency, Office of Research and Development, Environmental Monitoring
10 Systems Laboratory. EPA-600/4-84-031.
11 Houtven GV, Johnson FR, Mansfiled C, Yang J-C, Pyles A. (2003). Parental Averting Behavior
12 With Respect to Ozone Alerts. For presentation at AEA/ASSA Meeting, San Diego CA,
13 January 2004.
14 Isaacs , Kristin, Graham Glen, Thomas Mccurdy and Luther Smith. Modeling energy
15 expenditure and oxygen consumption in human exposure models: accounting for fatigue
16 and EPOC. Journal of Exposure Science and Environmental Epidemiology (2008) 18, 289-
17 298.
18 Johnson, T. (1984). A Study of Personal Exposure to Carbon Monoxide in Denver, Colorado.
19 Research Triangle Park, NC: U.S. Environmental Protection Agency, Environmental
20 Monitoring Systems Laboratory. EPA-600/4-84-014.
21 Johnson, T. (1989). Human Activity Patterns in Cincinnati, Ohio. Palo Alto, CA: Electric
22 Power Research Institute. EPRIEN-6204.
23 Johnson, T., Capel, J., McCoy, M. (1996a). Estimation of Ozone Exposures Experienced by
24 Urban Residents Using a Probabilistic Version of NEM and 1990 Population Data.
25 Prepared by International Technology Air Quality Services for Office of Air Quality
26 Planning and Standards, U.S. Environmental Protection Agency, Research Triangle Park,
27 NC. Contract no. 63-D-30094. April 1996. Available at:
28 http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_pr_td.html.
29 Johnson, T., Capel, J., McCoy, M., Warnasch, J. (1996b). Estimation of Ozone Exposures
30 Experienced by Outdoor Children in Nine Urban Areas Using a Probabilistic Version of
31 NEM. Prepared by International Technology Air Quality Services for Office of Air
32 Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle
33 Park, NC. Contract no. 63-D-30094. April 1996. Available at:
34 http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3_prtd.html
35 Johnson, T., Capel, J., McCoy, M., Warnasch, J. (1996c). Estimation of Ozone Exposures
36 Experienced by Outdoor Workers in Nine Urban Areas Using a Probabilistic Version of
37 NEM. 1996. Prepared by International Technology Air Quality Services for Office of Air
38 Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle
39 Park, NC. Contract no. 63-D-30094. April, 1996. Available at:
40 http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3 pr td.html
5-55
-------
1 Klepeis, N. E., Tsang, A. M., Behar, J. V. (1996). Analysis of the National Human Activity
2 Pattern Survey (NHAPS) Respondents from a Standpoint of Exposure Assessment.
3 Washington, DC: U.S. Environmental Protection Agency, Office of Research and
4 Development. EPA/600/R-96/074.
5 Knowledge Networks. (2009). Field Report: National Scale Activity Survey (NSAS).
6 Conducted for Research Triangle Institute. Submitted to Carol Mansfield November 13,
7 2009.
8 KSDOH. (2006). Environmental Factors, Outdoor Air Quality, and Activity Level. Results
9 from 2005 Kansas Behavioral Risk Factor Surveillance System. Office of Health
10 Promotion, Kansas Department of Health and Environment. Available at:
11 http ://www.kdheks. gov^rfss/PDF/cste_report_final .pdf
12 Langstaff, J. E. (2007). OAQPS Staff Memorandum to Ozone NAAQS Review Docket (OAR-
13 2005-0172). Subject: Analysis of Uncertainly in Ozone Population Exposure Modeling.
14 [January 31,2007]. Available at:
15 http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3crtd.html
16 Lioy, PJ. (1990). The analysis of total human exposure for exposure assessment: A multi-
17 discipline science for examining human contact with contaminants. Environ Sci Technol.
18 24: 938-945.
19 Liu L-JS, Box M, Kalman D, Kaufman J, Koenig J, Larson T, Lumley T, Sheppard L, Wallace
20 L. (2003). Exposure assessment of particulate matter for susceptible populations in
21 Seattle. Environ Health Persp. 111:909-918.
22 Mansfield C, Houtven GV, Johnson FR, Yang J-C. (2009). Environmental Risks and Behavior:
23 Do children spend less time outdoors when ozone pollution is high? ASSA annual
24 meeting, January 5, 2009. Update of Houtven et al. (2003), using the OAB CHAD data set,
25 and related to Mansfield et al. (2006).
26 Mansfield C, Johnson FR, Van Houtven G. (2006). The missing piece: averting behavior for
27 children's ozone exposures. Resource Energy Econ. 28:215-228.
28 McCurdy, T. (2000). Conceptual basis for multi-route intake dose modeling using an energy
29 expenditure approach. JExpo Anal Environ Epidemiol. 10: 1-12.
30 McCurdy, T., Glen, G., Smith, L., Lakkadi, Y. (2000). The National Exposure Research
31 Laboratory's Consolidated Human Activity Database. J Expo Anal Environ Epidemiol. 10:
32 566-578.
33 McDermott M, Srivastava R, Croskell S. (2006). Awareness of and compliance with air
34 pollution advisories: A comparison of parents of asthmatics with other parents. J Asthma.
35 43:235-239.
36 Montoye, H. J., Kemper, H. C. G., Saris, W.H.N., Washburn, R.A. (1996). Measuring Physical
37 Activity and Energy Expenditure. Champaign IL: Human Kinetics.
38 National Research Council (1991). Human Exposure Assessment for Airborne Pollutants:
39 Advances and Opportunities. Washington, DC: National Academy of Sciences.
5-56
-------
1 Neidell M. (2010). Air quality warnings and outdoor activities: evidence from Southern
2 California using a regression discontinuity approach design. JEpidemiol Community
3 Health. 64:921-926.
4 Richmond H., Palma, T, Langstaff, J., McCurdy, T., Glenn, G., Smith, L. (2002). Further
5 Refinements and Testing of APEX (3.0): EPA's population exposure model for criteria and
6 air toxic inhalation exposures. Joint meeting of the International Society of Exposure
7 Analysis and International Society of Environmental Epidemiology, Vancouver, CAN.
8 August 11-15, 2002.
9 Robinson, J. P., Wiley, J. A., Piazza, T., Garrett, K., and Cirksena, K. (1989). Activity Patterns
10 of California Residents and their Implications for Potential Exposure to Pollution.
11 California Air Resources Board, Sacramento, CA. CARB-A6-177-33.
12 SAS (2012). SAS/STAT 9.2 User's Guide, Second Edition. Available at:
13 http://support.sas.com/documentation/cdl/en/statug/63033/PDF/default/statug.pdf
14 Semenza JC, Wilson DJ, Parra J, Bontempo BD, Hart M, Sailor DJ, George LA. (2008). Public
15 perception and behavior change in relationship to hot weather and air pollution. Environ
16 Res. 107:401-411.
17 Sexton AL. (2011). Responses to Air Quality Alerts: Do Americans Spend Less Time
18 Outdoors? Available at: http://www.apec.umn.edU/prod/groups/cfans/@pub/@cfans/
19 @apec/documents/asset/cfans asset 365645.pdf
20 Spier, C. E., Little, D. E., Trim, S. C., Johnson, T. R., Linn, W. S., Hackney, J. D. (1992).
21 Activity patterns in elementary and high school students exposed to oxidant pollution. J
22 Expo Anal Environ Epidemiol. 2: 277-293.
23 Tsang A. M., and Klepeis, N. E. (1996). Descriptive Statistics Tables from a Detailed Analysis
24 of the National Human Activity Pattern Survey (NHAPS) Data. U.S. Environmental
25 Protection Agency. EPA/600/R-96/148.
26 U.S. Environmental Protection Agency (1986). Air Quality Criteria for Ozone and Other
27 Photochemical Oxidants. Office of Health and Environmental Assessment, Environmental
28 Criteria and Assessment Office, U.S. Environmental Protection Agency, Research Triangle
29 Park, NC. EPA-600/8-84-020aF-eF. Available from: NTIS, Springfield, VA., PB87-
30 142949.
31 U.S. Environmental Protection Agency (1996a). Review of National Ambient Air Quality
32 Standards for Ozone: Assessment of Scientific and Technical Information - OAQPS Staff
33 Paper. Office of Air Quality Planning and Standards, U.S. Environmental Protection
34 Agency, Research Triangle Park, NC. EPA/452/R-96-007. Available at:
35 http://www.epa.gov/ttn/naaqs/standards/ozone/s o3_pr sp.html
36 U.S. Environmental Protection Agency (1996b). Air Quality Criteria for Ozone and Related
37 Photochemical Oxidants. Office of Research and Development, National Center for
38 Environmental Assessment, U.S. Environmental Protection Agency, Research Triangle
39 Park,NC. EPA/600/P-93/004aF-cF. Available at:
40 http: //cfpub. epa. gov/ncea/cfm/recordi spl ay. cfm? dei d=2 831
5-57
-------
1 U.S. Environmental Protection Agency (2002). Consolidated Human Activities Database
2 (CHAD) Users Guide. Database and documentation available at:
3 http://www.epa.gov/chadnet 1 /
4 U.S. Environmental Protection Agency (2006). Air Quality Criteria for Ozone and Related
5 Photochemical Oxidants (Final). National Center for Environmental Assessment, U.S.
6 Environmental Protection Agency, Research Triangle Park, NC. EPA/600/R-05/004aF-cF.
7 Available at: http://cfpub.epa.gov/ncea/cfm/recordisplay.cfm?deid=149923
8 U.S. Environmental Protection Agency (2007a). Review of National Ambient Air Quality
9 Standards for Ozone: Policy Assessment of Scientific and Technical Information - OAQPS
10 Staff Paper. Office of Air Quality Planning and Standards, U.S. Environmental Protection
11 Agency, Research Triangle Park, NC. EPA-452/R-07-007. Available at:
12 U.S. Environmental Protection Agency (2007b). Ozone Population Exposure Analysis for
13 Selected Urban Areas. Office of Air Quality Planning and Standards, U.S. Environmental
14 Protection Agency, Research Triangle Park, NC. Available at:
15 http://www.epa.gOv/ttn/naaqs/standards/ozone/s _o3_cr_td. html
16 U.S. Environmental Protection Agency (2008). Risk and Exposure Assessment to Support the
17 Review of the NO2 Primary National Ambient Air Quality Standard. Report no. EPA-
18 452/R-08-008a. November 2008. Available at:
19 http://www.epa.gov/ttn/naaqs/standards/nox/data/20081121 NO2REA final.pdf.
20 U.S. Environmental Protection Agency (2009). Risk and Exposure Assessment to Support the
21 Review of the SO2 Primary National Ambient Air Quality Standard. Report no. EPA-
22 452/R-09-007. August 2009. Available
23 athttp ://www. epa. gov/ttn/naaq s/standards/so2/data/200908 SO2REAFinalReport.pdf.
24 U.S. Environmental Protection Agency (2010). Quantitative Risk and Exposure Assessment for
25 Carbon Monoxide - Amended. EPA Office of Air Quality Planning and Standards. EPA-
26 452/R-10-009. July 2010. Available at:
27 http://www.epa.gov/ttn/naaqs/standards/co/data/CO-REA-Amended-July2010.pdf
28 U.S. Environmental Protection Agency (2012a). Integrated Science Assessment of Ozone and
29 Related Photochemical Oxidants (Third External Review Draft). U.S. Environmental
30 Protection Agency, National Center for Environmental Assessment, U.S. Environmental
31 Protection Agency, Research Triangle Park, N. EPA/600/R-10/076C, 2012. Available at:
32 http://cfpub.epa.gov/ncea/isa/recordisplay.cfm?deid=242490
33 U.S. Environmental Protection Agency (2012b). Total Risk Integrated Methodology (TRIM) -
34 Air Pollutants Exposure Model Documentation (TRIM.Expo / APEX, Version 4.4) Volume
35 I: User's Guide. Office of Air Quality Planning and Standards, U.S. Environmental
36 Protection Agency, Research Triangle Park, NC. EPA-452/B-12-00la. Available at:
37 http ://www. epa. gov/ttn/fera/human_apex.html
38 U.S. Environmental Protection Agency (2012c). Total Risk Integrated Methodology (TRIM) -
39 Air Pollutants Exposure Model Documentation (TRIM.Expo / APEX, Version 4.4) Volume
40 II: Technical Support Document. Office of Air Quality Planning and Standards, U.S.
5-58
-------
1 Environmental Protection Agency, Research Triangle Park, NC. EPA-452/B-12-001b.
2 Available at: http://www.epa.gov/ttn/fera/human apex.html
3 University of Michigan. (2012). The Panel Study of Income Dynamics (PSID). Data and
4 documentation available at http://psidonline.isr.umich.edu .
5 Wells, B., Wesson, K., Jenkins, S. (2012). Analysis of Recent U.S. Ozone Air Quality Data to
6 Support the O3 NAAQS Review and Quadratic Rollback Simulations to Support the First
7 Draft of the Risk and Exposure Assessment. Available at:
8 http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3 2008 rea.html
9 Wen X-J, Balluz L, Mokdad A. (2009). Association between media alerts of Air Quality Index
10 and change of outdoor activity among adult asthma in six states, BRFSS, 2005. JComm
11 Health. 34:40-46.
12 Whitfield, R., Biller, W., Jusko, M., Keisler, J. (1996). A Probabilistic Assessment of Health
13 Risks Associated with Short- and Long-Term Exposure to Tropospheric Ozone. Argonne
14 National Laboratory, Argonne, IL.
15 Wiley, J. A., Robinson, J. P., Piazza, T., Garrett, K., Cirksena, K., Cheng, Y.-T., Martin, G.
16 (1991a). Activity Patterns of California Residents: Final Report. California Air Resources
17 Board, Sacramento, CA. ARB/R93/487. Available from: NTIS, Springfield, VA., PB94-
18 108719.
19 Wiley, J. A., Robinson, J. P., Cheng, Y.-T., Piazza, T., Stork, L., Pladsen, K. (1991b). Study of
20 Children's Activity Patterns: Final Report. California Air Resources Board, Sacramento,
21 CA. ARB-R-93/489.
22 Williams R, Suggs J, Rea A, Leovic K, et al. (2003a). The Research Triangle paniculate panel
23 study: PM mass concentrations relationships. AtmosEnviron. 37:5349-5363.
24 Williams R, Suggs J, Rea A, Sheldon L, et al. (2003b). The Research Triangle paniculate panel
25 study: modeling ambient source contributions to personal and residential PM mass
26 concentrations. Atmos Environ. 37:5365-5378.
27 Williams, R., Suggs, J., Creason, J., Rodes, C., Lawless, P., Kwok, R., Zweidinger, R., Sheldon,
28 L. (2000). The 1998 Baltimore paniculate matter epidemiology-exposure study: Part 2.
29 Personal exposure associated with an elderly population. J Expo Anal Environ Epidemiol.
30 10(6): 533-543.
31 Zivin JG, Neidell M. (2009). Days of haze: Environmental information disclosure and
32 intertemporal avoidance behavior. JEnvironmentEconManag. 58:119
33
5-59
-------
1 6 CHARACTERIZATION OF HEALTH RISK BASED ON
2 CONTROLLED HUMAN EXPOSURE STUDIES
[This chapter is still under development and will be submitted separately in August]
-------
1 7 CHARACTERIZATION OF HEALTH RISK BASED ON
2 EPIDEMIOLOGICAL STUDIES
3 This section provides an overview of the methods used in the urban study area risk
4 assessment. Section 7.1 discusses the basic structure of the risk assessment, identifying the
5 modeling elements and related sources of input data needed for the analysis and presenting an
6 overview of the approach used in calculating health effect incidence using concentration-
7 response functions based on epidemiological studies. Section 7.2 discusses air quality
8 considerations. Section 7.3 discusses the selection of model inputs including: (a) selection and
9 delineation of urban study areas, (b) selection of epidemiological studies and specification of
10 concentration-response functions (C-R functions), (c) defining Os concentration ranges for which
11 there is increased confidence in estimating risk (d) specification of baseline health effect
12 incidence and prevalence rates, and (e) estimation of population (demographic) counts. Section
13 7.4 describes how uncertainty and variability are addressed in the risk assessment. Section 7.5
14 summarizes the risk estimates that are generated. Section 7.6 provides and integrative discussion
15 of risk estimates with consideration for key sources of variability and uncertainty associated with
16 the overall analysis. Finally, Section 7.7 describes potential refinements to the first draft analysis
17 described here which will be considered for the second draft risk and exposure analysis (REA).
18 7.1 GENERAL APPROACH
19 7.1.1 Basic Structure of the Risk Assessment
20 This risk assessment involves the estimation of the incidence of specific health effect
21 endpoints associated with exposure to ambient Os for defined populations located within a set of
22 urban study areas. Because the risk assessment focuses on health effect incidence experienced by
23 defined populations, it represents a form of population-level risk assessment. This analysis does
24 not estimate risks to individuals within the population.
25 The general approach used in both the prior and current 63 risk assessments rely on C-R
26 functions based on effect estimates and model specifications obtained from epidemiological
27 studies. Since these studies derive effect estimates and model specifications using ambient air
28 quality data from fixed-site, population-oriented monitors, uncertainty in the application of these
29 functions in an Os risk assessment is minimized if, in modeling risk, we also use ambient air
30 quality data at fixed-site, population-oriented monitors to characterize exposure. Therefore, we
31 developed a composite monitor for each urban study area to represent population by averaging
32 across the monitors in that study area to produce a single composite hourly time series of
33 averaged values. The 63 metrics used in evaluating risk are derived form the composite monitor
7-1
-------
1 hourly time series distribution (see sections 7.2 and Chapter 4 for additional detail on the
2 characterization of ambient Os levels).
3 The general Os health risk model, illustrated in Figure 7-1, combines Os air quality data,
4 C-R functions, baseline health incidence and prevalence data, and population data (all specific to
5 a given urban study area) to derive estimates of the annual incidence of specified health effects
6 for that urban study area. This first draft exposure and risk assessment (first draft REA) models
7 risk for 12 urban study areas selected to provide coverage for the types of urban O3 scenarios
8 likely to exist across the U.S. (see section 7.4.1).
9 The analyses conducted for this review focus on estimating risks associated with recent
10 Os air quality and estimating changes in risk associated with air quality simulated to just meet the
11 current Os ambient air quality standard (simulation of risk associated with meeting alternative Os
12 standard levels will be completed for second Draft of the risk assessment). In simulating just
13 meeting the current Os standard level, we assume that reductions in Os precursor emissions
14 would only apply to U.S. anthropogenic emissions sources. This was implemented by using
15 modeled estimates of U.S. background O3, (i.e. O3 concentrations in the absence of continental
16 emissions of U.S. anthropogenic NOx and VOC), as a lower bound in conducting the rollback of
17 hourly 63 levels to simulate just meeting the current standard. In other words, we did not allow
18 any single hourly monitored value to be rolled down below U.S. background. We were able to
19 simulate just meeting the current standard in all twelve urban study areas through the reduction
20 of U.S.-anthropogenic 63 alone. The procedures for modeling U.S. background 63 and
21 simulating attainment with the current Os standards are discussed in Chapter 4 and in the Air
22 Quality Appendices accompanying this REA.
23 As discussed in Chapters 2 and 3, in modeling risk we employ continuous non-threshold
24 C-R functions relating ozone exposure to health effect incidence. The use of non-threshold
25 functions reflects the conclusion reached in the ISA based on a thorough review of available
26 evidence (see Oj ISA, section 2.5.4.4, U.S. EPA 2012). However, also consistent with the
27 conclusions of the ISA, we recognize that there is less confidence in specifying the shape of the
28 C-R function at 63 levels towards the lower end of the distribution of data used in fitting the
29 curve. In particular, we would expect our overall confidence in specifying the magnitude of risk
30 associated with each unit of Os exposure to be significantly reduced at levels below the lowest
31 measured level (LML) used in the epidemiological study. Similarly, we would expect our
32 confidence in specifying the magnitude of risk to be increasing with the level of ozone above the
33 LML, and become appreciably greater at ozone concentrations closer to the central mass of
34 measurements used in the underlying epidemiological study. In order to reflect considerations of
35 the differences in relative confidence above and below the LML, we generate two types of risk
7-2
-------
1 estimates for a particular scenario which when considered together inform consideration of
2 uncertainty related to application of the C-R function at low Os levels:
3 • Risk modeled down to the LML: This is a higher confidence estimate of risk since it
4 only considers exposure levels within the range of the O3 data used in the derivation
5 of the C-R function (i.e., exposures down to the LML). However, given that there is
6 no evidence of a threshold for these health effects, and that the statistical models used
7 in the epidemiology studies did not specific a cutoff at the LML, exclusion of
8 exposures below the LML is likely to result in a low-biased risk estimate.
9 • Risk modeled down to zero O3: With this estimate, consistent with the underlying
10 statistical models used in the epidemiology studies, we apply the C-R function across
11 the full range of ambient O3 levels in the study area. While this estimate will reflect
12 the full range of potential exposure and risk (all the way down to zero O3), there is a
13 higher degree of uncertainty about the estimates because they include risks based on
14 extrapolating the C-R function beyond the range of observed O3.
15 Due to data limitations, we were not able to specify LMLs for the full set of
16 epidemiological studies supporting C-R functions used in the risk assessment. Therefore, we
17 used a surrogate metric as a stand-in for the actual study-based LMLs. Specifically, we used the
18 lowest Os values from the composite monitor Os distribution used in modeling risk for a
19 particular combination of urban study area, health endpoint and simulation year to represent the
20 LML for that combination. We recognize that these estimates are not the best surrogates for the
21 true study-specific LMLs, and are evaluating alternative approaches for the second draft REA.
22 While the surrogate LMLs in most cases match the Os metric and ozone season used in the
23 underlying epidemiological study, the surrogate LMLs are based on composite monitor
24 distributions specified for the two years included in the risk assessment (2007 and 2009), while
25 63 levels used in the epidemiological studies typically reflect several years from an earlier time
26 period (varies across studies). This mismatch in timeframes between the surrogate LMLs and
27 actual study-specific LMLs introduce uncertainty into the analysis. For the second draft REA, we
28 are working to obtain actual LML values used in the source epidemiological studies underlying
29 C-R functions used in the risk assessment (see section 7.7). The specific technical approach used
30 to integrate the LMLs into the generation of risk estimates is discussed in section 7.1.2.1.
31 In modeling risk for all health endpoints included in the analysis, for recent O3
32 conditions and just meeting the current standard, we estimated total risk, both above zero and
33 above the LML. For meeting the current standard, we estimated both total risk as well as the
34 difference in risk, or the risk delta, representing the degree of risk reduction (benefit) associated
35 with just meeting the current standard.
36 In previous NAAQS-related risk assessments, we have generated two categories of risk
37 estimates, including a set of core (or primary) estimates and an additional set of sensitivity
7-3
-------
1 analyses. The core risk estimates utilize C-R functions based on epidemiological studies for
2 which we have relatively greater overall confidence. While it is generally not possible to assign
3 quantitative levels of confidence to these core risk estimates, they are generally based on inputs
4 having higher overall levels of confidence relative to risk estimates that are generated using other
5 C-R functions. Therefore, emphasis is placed on the core risk estimates in making observations
6 regarding total risk and risk reductions associated with recent conditions and the simulated just
7 meeting the current and alternative standard levels. By contrast, the sensitivity analysis results
8 typically reflect application of C-R functions covering a wider array of design elements which
9 can impact risk (e.g., copollutants models, lag structures, statistical modeling methods etc). The
10 sensitivity analysis results provide insights into the potential impact of these design elements on
11 the core risk estimates, thereby informing our characterization of overall confidence in the core
12 risk estimates.
7-4
-------
1
2
3
Figure 7-1. Major components of Os health risk assessment.
Air Quality
Ambient Population-
oriented Monitoring
and Estimated US
Background Levelsfor
Selected Cities
Air Quality Adjustment
Procedures
| CurrentStandard Level
Concentration-Response
Human Epidemiological
Studies (various health
endpoints)
Estimates of City-specific
Baseline Health Effects
Incidence and
Prevalence Rates
(various health
endpoints) and
Population Data
RecentAir
Quality Analysis
Changes in
Distribution
of Ozone Air
Quality
Concentration-
Response
Relationships
BenMAP
Health
Risk
Model
Risk Estimates:
• RecentAir
Quality
• Current
Standard
7-5
-------
1 For first draft of this analysis, we have focused primarily on generating a robust set of
2 core risk estimates and have not developed a comprehensive set of sensitivity analyses due to
3 limitations in the available data from published epidemiology studies. Specifically, for mortality,
4 we obtained Bayes-adjusted city-specific effect estimates which reflected single pollutant models
5 based on 8-hour Os metrics for a common lag structure directly from the authors and
6 incorporated those into city-specific risk simulations to generate risk estimates for each of the 12
7 urban study areas. However, we were not able to obtain similar estimates for other model
8 specifications (e.g. co-pollutant models, alternative lags, etc) typically considered in sensitivity
9 analyses. For the second draft REA, we are investigating methods for obtaining alternative
10 model specifications for use in sensitivity analyses. However, we would note that the set of core
11 risk estimates for short-term exposure morbidity generated for this first draft include coverage
12 for a variety of design elements (including multi-/single-pollutant models and lag structures) and
13 therefore, the array of core risk estimates informs consideration of the impact that these design
14 elements have on risk estimates (see section 7.5).
15 The risk assessment reflects consideration for five years of recent air quality data from
16 2006 through 2010, with these five years reflecting two three-year attainment simulation periods
17 that share a common overlapping year (i.e., 2006-2008 and 2008-2010 - see section 7.2). These
18 two attainment periods were selected to provide coverage for a more recent time period with
19 relatively elevated Os levels (2006-2008) and recent time period with relatively lower Os levels
20 (2008-2010). For the first draft analysis, we modeled risk for the middle year of each three-year
21 attainment simulation period in order to provide estimates of risk for a year with generally higher
22 Os levels (2007) and a year with generally lower Os levels (2009). In modeling risk, we matched
23 the population data used in the risk assessment to the year of the air quality data. For example,
24 when we used 2007 air quality data, we used 2007 population estimates. For baseline incidence
25 and prevalence, rather than interpolating rates for the two specific years modeled in the risk
26 assessment, we selected the closest year for which we had existing incidence/prevalence data
27 (i.e., for simulation year 2007, we used available data for 2005 and for simulation year 2009, we
28 used data from 2010). The calculation of baseline incidence and prevalence rates is described in
29 detail in section 7.3.4.
30 The risk assessment procedures described in more detail below are diagramed in Figure
31 7-2. To estimate the change in incidence of a given health effect resulting from a given change
32 in ambient Os concentrations in an assessment location, the following analysis inputs are
33 necessary:
34 • Air quality information including: (1) 63 air quality data from each of the
35 simulation years included in the analysis (2007 and 2009) from population-oriented
36 monitors in the assessment location, (2) estimates of U.S.-background Os
7-6
-------
1 concentrations appropriate to this location, and (3) a method for adjusting the air
2 quality data to simulate just meeting the current or alternative suite of Os standards.
3 (These air quality inputs are discussed in more detail in section 7.2).
4 • C-R function(s) which provide an estimate of the relationship between the health
5 endpoint of interest and Os concentrations (for this analysis, the majority of C-R
6 functions used were applied to urban study areas matching the assessment locations
7 from the epidemiological studies used in deriving the functions, in order to increase
8 overall confidence in the risk estimates generated - see section 7.3.2). For 63,
9 epidemiological studies providing information necessary to specify C-R functions are
10 readily available for Os-related health effects associated with short-term exposures
11 (Section 7.1.2 describes the role of C-R functions in estimating health risks associated
12 with 63). For the first draft analysis, we have not modeled any endpoints associated
13 with long-term Os exposure (the potential for modeling these health endpoints is
14 discussed in sections 7.7).
15 • Baseline health affects incidence and prevalence rates and population. The
16 baseline incidence provides an estimate of the incidence rate (number of cases of the
17 health effect per year or day, depending on endpoint, usually per 10,000 or 100,000
18 general population) in the assessment location corresponding to recent ambient O3
19 levels in that location. The baseline prevalence rate describes the prevalence of a
20 given disease state or conditions (e.g., asthma) within the population (number of
21 individuals with the disease state/condition, usually per 10,000 or 100,000 general
22 population). To derive the total baseline incidence or prevalence per year, this rate
23 must be multiplied by the corresponding population number (e.g., if the baseline
24 incidence rate is number of cases per year per 100,000 population, it must be
25 multiplied by the number of 100,000s in the population). (Section 7.3.4 summarizes
26 considerations related to the baseline incidence and prevalence rates and population
27 data inputs to the risk assessment).
28
7-7
-------
Figure 7-2. Flow diagram of risk assessment for short-term exposure studies.
Air Quality Data
Compute daily
distribution of
composite
monitorvalues
(for each study area)
Compute total
ozone (above
surrogate LML)
and delta (change)
in ozone between
current conditions
and current/
alternative standard
Concentration-Response Functions
Ident
locati
spec
stud
ify
on-
fic
es
Identify
Relative Risk
(RR)orslope
coefficents (li)
i T
1 Convert RR |
t
| (if necessary)'
Identify
functional form
Compute % change
in baseline rate
of d isease f o r each d ay
in the modeling period
(different ozone periods
used fordifferent
health effect end point/
epidemiology study
combinations)
Baseline Health Incidence
Compute total
annual
number of
disease cases
associated
with
an ozone level
and the change
in numberof
cases associated
with attaining
a current or
alternative standard
Identify
US background
levels
Specify surrogate
LML values
I Specify roll back
method
I (for certain
I analyses)
Estimate of
percent change in
total incidence
Estimate of
ozone-associated
incidence
7-S
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
This risk assessment was implemented using the EPA's Benefits Mapping and Analysis
Program (BenMAP) (Abt, 2010). This GIS-based computer program draws upon a database of
population, baseline incidence/prevalence rates and effect coefficients to automate the
calculation of health impacts. For this analysis, the standard set of effect coefficients and health
effect incidence data available in BenMAP has been augmented to reflect the latest studies and
data available for modeling 63 risk. EPA has traditionally relied upon the BenMAP program to
estimate the health impacts avoided and economic benefits associated with adopting new air
quality rules. For this analysis, EPA used the model to estimate Os-related risk for the suite of
health effects endpoints described in section 7.3.2. The following figure summarizes the data
inputs (in black text) and outputs (in blue text) for a typical BenMAP analysis.
Census
Population
Data
Air Quality
Monitoring
Health
Functions
Population
Estimates
Population
Exposure
Adverse
Health Effects
Population
Projections
Air Quality
Modeling
Incidence and
Prevalence
Rates
There are three primary advantages to using BenMAP for this analysis, as compared to
the procedure for estimating population risk followed in the last review. First, once we have
configured the BenMAP software for this particular 63 analysis, the program can produce risk
estimates for an array of modeling scenarios across a large number of urban areas. Second, the
program can more easily accommodate a variety of sensitivity analyses (which we are evaluating
for inclusion in second Draft). Third, BenMAP allowed us to complete the national assessment
of Os mortality described in Chapter 8, which plays in important role in assessing the
representativeness of the urban study area analysis.
7-9
-------
1 7.1.2 Calculating O3-Related Health Effects Incidence
2 The C-R functions used in the risk assessment are empirically estimated associations
3 between average ambient concentrations of Os and the health endpoints of interest (e.g.,
4 mortality, hospital admissions, emergency department visits). This section describes the basic
5 method used to estimate changes in the incidence of a health endpoint associated with changes in
6 Os, using a "generic" C-R function of the most common functional form.
7 Although some epidemiological studies have estimated linear C-R functions and some
8 have estimated logistic functions, most of the studies used a method referred to as "Poisson
9 regression" to estimate exponential (or log-linear) C-R functions in which the natural logarithm
10 of the health endpoint is a linear function of Os:
11
12 y = Befk (1)
13
14 where x is the ambient Os level, y is the incidence of the health endpoint of interest at Os
15 level x, p is the coefficient relating ambient Os concentration to the health endpoint, and B is the
16 incidence at x=0, i.e., when there is no ambient Os. The relationship between a specified ambient
17 Os level, XQ, for example, and the incidence of a given health endpoint associated with that level
18 (denoted as yo) is then
19
20 y0=Beflc" (2)
21
22 Because the log-linear form of a C-R function (equation (1)) is by far the most common
23 form, we use this form to illustrate the "health impact function" used in the Os risk assessment.
24 If we let XQ denote the baseline (upper) Os level, and xi denote the lower Os level, and yo
25 and yi denote the corresponding incidences of the health effect, we can derive the following
26 relationship between the change in x, Ax= (x0- xi), and the corresponding change in y, Ay, from
27 equation (I).1
28 *y = (y0-yl) = y0V-e-p&x]. (3)
29
30 Alternatively, the difference in health effects incidence can be calculated indirectly using
31 relative risk. Relative risk (RR) is a measure commonly used by epidemiologists to characterize
32 the comparative health effects associated with a particular air quality comparison. The risk of
1 If Ax < 0 - i.e., if Ax = (xr x0) - then the relationship between Ar and Ay can be shown to be
Ay = (y\ - y0) = y0[effa -1]. If Ax < 0, Ay will similarly be negative. However, the magnitude of Ay will be the
same whether Ar>OorAx<0- i.e., the absolute value of Ay does not depend on which equation is used.
7-10
-------
1 mortality at ambient 63 level x0 relative to the risk of mortality at ambient 63 level xi, for
2 example, may be characterized by the ratio of the two mortality rates: the mortality rate among
3 individuals when the ambient O?, level is x0 and the mortality rate among (otherwise identical)
4 individuals when the ambient Os level is xi. This is the RR for mortality associated with the
5 difference between the two ambient Os levels, XQ and xi. Given a C-R function of the form
6 shown in equation (1) and a particular difference in ambient 63 levels, Ax, the RR associated
7 with that difference in ambient O3, denoted as RRAx, is equal to epAx. The difference in health
8 effects incidence, Ay, corresponding to a given difference in ambient Os levels, Ax, can then be
9 calculated based on this RRAx as:
10
11 Ay = (y0-yi) = y0[l-(l/RR^]. (4)
12
13 Equations (3) and (4) are simply alternative ways of expressing the relationship between
14 a given difference in ambient Os levels, Ax > 0, and the corresponding difference in health
15 effects incidence, Ay. These health impact equations are the key equations that combine air
16 quality information, C-R function information, and baseline health effects incidence information
17 to estimate ambient Os health risk.
18 7.1.2.1 Incorporating LMLs into the estimation of risk
19 This risk analysis provides two types of risk estimates for each scenario evaluated
20 including: (a) risk modeled down to zero 63 concentration and (b) risk modeled down to the
21 LML from the epidemiological study providing the C-R function. When considered together
22 these two types of risk estimates inform consideration of uncertainty related to application of the
23 C-R functions at low O3 levels. As noted in section 7.1.1, due to data limitations, we are using
24 surrogate LML values for the first draft REA in place of actual LMLs from the studies
25 underlying the C-R functions. Specifically, we used the composite monitor dataset used in
26 modeling risk for a particular health endpoint (e.g., the 8hr max set of hourly values used in
27 modeling short-term exposure-related mortality for L.A.) as a surrogate for the set of measured
28 Os levels used in deriving the C-R function for that endpoint/city combination. The LML of the
29 composite monitor dataset was used to define an Os exposure range of increased confidence in
30 estimating risk for a particular endpoint/location combination.
31 The LMLs were incorporated in calculation risk as follows. In modeling absolute risk for
32 the recent conditions scenario, we modeled risk for the Os increment from the recent conditions
33 down to the LML. Similarly, when estimating the delta (risk reduction) in going from recent
34 conditions to just meeting the current standard, we model risk only for that increment of the
35 change in Os that occurred above the LML. As would be expected, application of the LML did
7-11
-------
1 affect estimates of total Os-attributable risk for both the recent conditions and meeting the
2 current standard scenarios, with the LML-based estimates being lower. However, estimates of
3 the change in risk between these two air quality scenarios (i.e., in going from recent conditions to
4 meeting the current standard) was not significantly affected by application of the LML since on a
5 daily basis, the recent conditions and current standard values typically occurred above the LML,
6 which meant that the differences between the two levels (on a particular day) nearly always
7 occurred at levels of absolute O3 well above the LML. The surrogate LMLs used in the first draft
8 REA are presented in section 7.3.3.
9 7.2 AIR QUALITY CONSIDERATIONS
10 Air quality data are discussed in detail in Chapter 4 of this report. Here we describe those
11 air quality considerations that are directly relevant to the estimation of health risks in the
12 epidemiology based portion of the risk assessment. As described in section 7.1.1, the risk
13 assessment uses composite monitor values derived for each urban study area as the basis for
14 characterizing population exposure in modeling risk. The use of composite monitors reflects
15 consideration for the way ambient 63 data are used in the epidemiological studies providing the
16 C-R functions (see section 7.1.1). Because the Os risk assessment focuses on short-term exposure
17 related health endpoints, the composite monitor values derived for this analysis include hourly
18 time series for each study area (where the O3 value for each hour is the average of measurements
19 across the monitors in that study area reporting values for that hour).
20 For this analysis, reflecting consideration for available evidence in the published
21 literature (see section 7.3.2), we have focused the analysis on short-term peak O3 metrics
22 including Ihr max, 8hr mean and 8hr max. The more generalized 24 hour average has been
23 deemphasized for this analysis, although it is still used in risk modeling when use of C-R
24 functions based on this metric allow us to cover a specific health effect endpoint/location of
25 particular interest - see section 7.3.2).
26 For the first draft REA, we estimate risk associated with recent conditions as well as risk
27 associated with simulating just meeting the current standard. While the derivation of composite
28 monitor hourly 63 distributions (and associated peak exposure metrics) for recent conditions is
29 relatively straightforward, the generation of these estimates for the scenario of just meeting the
30 current standard is more complex. Simulating meeting the current Os standard involves
31 application of modeled U.S. background 63 levels as a floor for hourly 63 concentrations in the
32 quadratic rollback procedure. The procedure for generating composite monitor values for the
33 recent conditions scenario, along with a summary of the resulting composite monitor values is
34 presented in section 7.2.1. We then describe the procedure used to estimate U.S. background
35 levels for each urban study area, in section 7.2.2. Finally, in section 7.2.3, we briefly describe the
7-12
-------
1 quadratic rollback approach used to simulate just meeting the current standard level and we
2 provide a summary of the resulting composite monitor Os metrics. A more complete discussion
3 of these procedures is provided in the air quality chapter (see Chapter 4).
4 7.2.1 Characterizing Recent Conditions
5 Recent conditions were characterized using composite monitor-based peak 63 metrics
6 generated for each of the five years considered in the simulation (additional detail on the
7 generation of composite monitor values is presented in Chapter 4). As noted in section 7.1.1,
8 risk estimates where only generated for 2007 and 2009, which represent the middle years for
9 each of the 3-year attainment periods considered in the analysis. The composite monitors were
10 specified as hourly time series with each hour reflecting the average of available measurements
11 across monitors in a particular study area. The 12 urban study areas included in the analysis are
12 based on the set of counties used in one of the two epidemiology studies providing C-R functions
13 for modeling short-term exposure-related mortality (Zanobetti and Schwartz., 2008b). This
14 county-level specification of the urban study areas resulted in each study area having between
15 one and five counties, with a composite monitor being developed for each study area. The
16 composite monitors for each area were derived using the ambient Os monitors falling within each
17 urban area, with the number ranging from three to seventeen monitors per study area. Table 7-1
18 identifies (a) the counties used in specifying each urban study area, (b) the number of Os
19 monitors associated with each and (c) the 63 season for each study area.
7-13
-------
Table 7-1 Information on the 12 Urban Case Study Areas in the Risk Assessment
Study Area
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
Counties
Cobb County, GA
DeKalb County, GA
Fulton County, GA
Gwinnett County, GA
Baltimore City, MD
Baltimore County, MD
Middlesex County, MA
Norfolk County, MA
Suffolk County, MA
Cuyahoga County, OH
Denver County, CO
Wayne County, MI
Harris County, TX
Los Angeles County, CA
Bronx County, NY
Kings County, NY
New York County, NY
Queens County, NY
Richmond County, NY
Philadelphia County, PA
Sacramento County, CA
St. Louis City, MO
St. Louis County, MO
#ofO3
Monitors
5
3
5
4
o
6
4
17
17
8
4
8
8
Required Oj
Monitoring Season
March - October
April - October
April - September
April - October
March - September
April - September
January - December
January - December
April - October
April - October
January - December
April - October
2
3 The Os season is an important factor in the risk assessment. In modeling risk for a
4 particular health endpoint, we attempted to match the 63 season used in deriving the composite
5 monitor value to the Os period utilized in the epidemiology study supplying the underlying C-R
6 function. Consequently, there were several versions of the daily peak Os metrics generated for
7 the risk assessment (to match the various 63 periods used in the underlying epidemiology
8 studies). To keep the task of deriving the daily peak Os metrics tractable, rather than explicitly
9 matching the 63 periods used in each of the mortality and morbidity studies providing C-R
10 functions used in the analysis, we elected to match the sets of O3 periods used in the two
11 epidemiology studies providing C-R functions used in the core analysis for modeling short-term
12 exposure-related mortality (i.e., the Zanobetti and Schwartz 2008b and Bell et al., 2004 studies).
13 The Zanobetti and Schwartz 2008b study used a fixed Os period of June-August (combined with
14 an 8hr mean daily Os measurement), while the Bell et al., 2004 study reflected the O3 monitoring
15 period (essentially the 63 season) specific to each study area - this is the period reflected in Table
16 7-1 (combined with an 8hr max daily Oj measurement).2 For all other health effects endpoints
2 The ozone monitoring periods used in these two studies are reflected in modeling risk based on C-R
functions derived from these studies. Therefore, because the Zanobetti and Schwartz (2008b) study uses a notably
shorter monitoring period relative to the Bell et al., (2005) study, risk estimates generated based on C-R functions
7-14
-------
1 modeled for the first draft REA, we then matched up each study to whichever of these two 63
2 periods provided the closest match, although we also included a Ihr max daily Os metric and a
3 24hr average metric to comply with the metrics used in several of the studies (see section 7.3.2
4 for a description of the studies used including their air metrics).
5 In deriving the composite monitor values, we did not interpolate any missing data and
6 instead took the average of available measurements for each hour. We are evaluating this
7 approach and for the second draft, and may consider application of interpolation methods as a
8 sensitivity analysis to evaluate the potential bias introduced into the analysis by not interpolating
9 missing measurements - see section 7.7. Peak Os daily metrics including Ihr max, 8hr mean and
10 8hr max values were derived from the composite monitor values and used in generating risk
11 estimates. In addition, 24hr average values were also derived as note earlier.
12 Table 7-2 presents a summary of the composite monitor-based daily metrics for the two
13 short-term exposure-related mortality studies used in the analysis: Zanobetti and Schwartz 2008b
14 (8hr mean metric for June-August) and Bell et al., 2004 (8hr max metric for the city-specific Os
15 seasons). These two metrics were selected for illustrating composite monitor values used in the
16 analysis since they provide Os air metrics for the majority of health endpoints used in the
17 analysis. These composite monitor summary statistics, which represent recent 63 conditions for
18 the 12 urban study areas, are presented for 2007 and 2009, reflecting the two simulation years
19 included in the first draft.
20
21
22
23
24
25
26
27
28
29
30
31
32
obtained from the former study will be notably smaller (other factors equal) than risk estimates generated using C-R
functions based on the latter study. This is an important factor which is considered when we review the mortality
risk estimates that are generated (see section 7.1.5).
7-15
-------
1 Table 7-2 Composite monitor values (recent conditions) for 2007 and 2009 for air metrics
2 used in modeling short-term exposure-related mortality
Urban
study area
8hr (mean) (June-August) (ppb)
Min
10th
Mean
90th
Max
8hr max (city-specific Oj, season) (ppb)
Min
10th
Mean
90th
Max
2007 Simulation year
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
24
13
19
6
21
19
10
31
10
12
30
22
36
31
25
25
36
29
17
42
22
27
37
38
60
48
43
43
50
48
33
54
43
49
51
56
81
64
64
65
60
69
56
67
66
68
65
77
104
81
89
79
72
86
72
80
82
96
99
93
17
13
12
12
4
13
6
9
10
13
13
8
32
25
26
27
27
30
18
21
19
26
23
32
53
43
43
44
44
47
35
40
38
45
41
50
73
62
65
65
57
70
56
60
62
66
59
71
106
81
89
88
72
89
79
87
85
96
99
93
2009 Simulation year
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
21
24
17
16
22
11
15
22
12
14
30
22
29
32
24
25
36
20
22
33
23
23
35
32
49
48
37
40
48
40
37
52
40
41
52
44
65
62
50
58
58
56
57
68
57
57
71
56
81
70
70
66
68
84
76
91
73
77
82
68
5
9
12
15
16
14
7
8
8
9
5
7
24
25
26
24
31
26
18
22
19
21
20
24
42
42
39
40
45
42
35
42
36
38
41
41
60
58
53
56
56
57
55
63
55
55
66
57
83
72
76
73
68
86
90
91
73
78
90
68
3
4
5
6
7
7.2.2 Estimating U.S. Background
Model based estimates of U.S. Background 63 levels specific to each urban study area
are used as a lower bound for hourly Os concentrations in the quadratic rollback procedure used
to simulate just meeting the current standard level. This approach reflects the assumption that
reductions in 63 precursor emissions would only apply to U.S. anthropogenic emissions sources.
The derivation of the model-based U.S. Background estimates is described in detail in Chapter 4
7-16
-------
1 and consequently, we only provide a brief discussion here, focusing on aspects particularly
2 relevant to the risk assessment.
3 U.S. background Os was modeled at the 70km grid cell level of spatial resolution using a
4 combination of GEOS-Chem (for international transport) with a nested CMAQ model (for more
5 refined transport and atmospheric chemistry within the U.S.). The simulation provides hourly-
6 level estimates of U.S. background 63 for 2006 (no other years were simulated). Each of the 63
7 monitors within a given urban study area is then assigned the U.S. Background hourly profile
8 associated with the 70km grid within which that monitor falls. Because the characterization of
9 U.S. background is model-based and only simulated for 2006, we could not directly match up
10 absolute U.S. background values to absolute measured Os levels at a particular monitor on an
11 hour-by-hour basis. Therefore, we developed a more generalized representation of U.S.
12 background levels in the form of U.S. background ratios for each hour/month combination at
13 each monitor. For example we would have a ratio of U.S. background to total Os for the 2pm
14 hour in October at a particular monitor. These more generalized U.S. Background ratios can then
15 be multiplied by the actual measured O3 level at a given monitor for a particular hour (at any
16 time during the 5 year simulation period) to generate the U.S. background estimate for that
17 specific hour/monitor combination. This procedure is repeated for all 63 measurements
18 associated with a particular monitor within a study area. This distribution of estimated U.S.
19 background levels then serves as the lower bound floor when applying quadratic rollback to that
20 monitor. Additional detail on the derivation of U.S. background values to support quadratic
21 rollback is provided in Chapter 4.
22 7.2.3 Simulating Air Quality to Just Meet Current and Alternative Standards
23 Simulating just meeting the current standard uses the same quadratic rollback method as
24 was used in the risk assessment completed for the last Os NAAQS review (U.S.EPA, 2007).
25 However for this analysis, we use model-derived estimates of U.S. Background as a lower bound
26 for application of the quadratic rollback.
27 Quadratic rollback uses a quadratic equation to reduce high concentrations at a greater
28 rate than low concentrations. The intent is to simulate reductions in Os resulting from
29 unspecified reductions in precursor emissions, without greatly affecting concentrations near
30 ambient background levels (Duff et al., 1998) (see Chapter 4 for additional detail on application
31 of the quadratic rollback). We are considering the use of a more sophisticated and representative
32 method for the second Draft analysis (the DDM method). Specifically, we are evaluating the
33 Decoupled Direct Method (DDM) approach implemented using the Community Multi-scale Air
34 Quality (CMAQ) model. This approach simulates just meeting the current (as well as alternative)
35 standard levels based on modeling the response of ozone concentrations to reduction in
7-17
-------
1 anthropogenic NOx and VOC emissions (see Chapter 4 for additional detail). In the risk
2 assessment, quadratic rollback is applied to adjust the distribution of Os levels at each monitor
3 within a study area such that the Os standard is attained at the design monitor within that study
4 area. The rollback procedure is applied to each of the three years of monitoring data associated
5 with each attainment period considered in the analysis (i.e., 2006-2008 and 2008-2010). Once
6 the rollback has been fully implemented and the current 63 standard is just met for that study
7 area, we then recompute the composite monitor with its daily peak O3 metrics. This procedure is
8 described in section 7.2.1.
9 Table 7-3 presents summary statistics for the composite monitor values at each of the
10 urban study areas (for 2006 and 2009) following simulation of just meeting the current standard
11 level.
12
7-18
-------
1 Table 7-3 Composite monitor values (simulation of meeting current standard) for 2007
2 and 2009 for air metrics used in modeling short-term exposure-related mortality
Urban
study area
8hr (mean) (June-August) (ppb)
Min
10th
Mean
90th
Max
8hr max (city-specific Oj, season) (ppb)
Min
10th
Mean
90th
Max
2007 Simulation year
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
23
13
18
6
21
19
10
27
11
13
27
22
33
29
23
24
34
28
16
35
20
24
33
35
51
43
40
40
45
45
30
43
39
43
43
51
67
55
60
59
54
64
50
52
58
58
53
69
79
68
81
71
64
78
62
57
70
82
74
81
16
13
12
11
4
12
6
8
11
14
13
8
29
23
25
25
26
29
17
19
20
25
21
30
46
39
41
41
41
44
32
33
35
40
36
46
61
54
60
59
52
64
50
47
55
57
49
64
81
68
81
78
64
81
67
61
71
82
74
81
2009 Simulation year
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
20
22
16
15
22
11
14
20
11
13
27
21
28
30
23
24
35
20
21
29
22
22
32
31
46
43
36
39
47
40
35
43
37
38
44
43
62
55
49
56
56
56
52
53
52
53
57
55
76
61
69
64
65
84
68
64
66
70
65
66
5
9
12
15
16
14
6
8
7
8
5
6
23
23
25
24
30
26
17
20
18
20
19
23
40
38
38
38
44
42
33
36
34
35
36
40
57
52
52
55
55
57
50
50
51
52
55
55
78
63
75
70
65
86
79
64
66
71
69
67
4 7.3 SELECTION OF MODEL INPUTS
5 7.3.1 Selection and Delineation of Urban Study Areas
6 This analysis focuses on modeling risk for a set of urban study areas, reflecting the goal
7 of providing risk estimates that have higher overall confidence due to the use of location-specific
7-19
-------
1 data when available for these urban locations. In addition, given the greater availability of
2 location-specific data, a more rigorous evaluation of the impact of uncertainty and variability can
3 be conducted for a set of selected urban study areas than would be possible for a broader regional
4 or national-scale analysis. The following factors were considered in selecting the 12 urban study
5 areas included in this analysis:
6 • Air quality data: An urban area has reasonably comprehensive monitoring data for the
7 period of interest (2006-2010) to support the risk assessment. This criterion was
8 evaluated qualitatively by considering the number of monitors within the attainment area
9 associated with prospective urban areas. Locations with one or two monitors would be
10 excluded since they had relatively limited spatial coverage in characterizing Os levels.
11 Ideally, at least three monitors and upwards of five would be present to provide
12 reasonable spatial coverage, but the determination of "reasonable coverage" is
13 complicated since it reflects consideration for population density together with potential
14 gradients in Os (and commuting patterns). A rigorous analysis of the degree of effective
15 coverage of monitoring networks for urban populations (and prospective exposure and
16 risk) would not only support a more rigorous selection of urban study areas, but also a
17 better understanding of potential measurement error associated with the epidemiological
18 studies used in risk modeling.
19 • Elevated ambient O3 levels: Because we are interested in evaluating the potential
20 magnitude of risk reductions associated with just meeting the current and alternative 63
21 standard levels, we need to include study areas with elevated ambient Os levels such that
22 they are not currently meeting the current Os standard, or at least have ambient levels
23 close to the current standard, such that alternative 63 standard levels to be simulated in
24 the second Draft risk assessment would result in some degree of risk reduction.
25 Consequently, in selecting urban study areas, we considered their status regarding just
26 meeting the current standard, favoring locations that are either not in attainment, or are
27 just barely attaining the standard
28 • Location-specific C-R functions: Given the health endpoints selected for inclusion in
29 the analysis (see section 7.3.2), there are epidemiological studies of sufficient quality
30 available for these urban study areas to provide the C-R functions necessary for modeling
31 risk. This criterion primarily applies to short-term epidemiological studies since the
32 associated health effect endpoints are the primary focus of the first draft REA. Note, that
33 short-term exposure-related epidemiological studies often include city-specific effect
34 estimates, and in some cases are multi-city studies that provide estimates for multiple
35 cities. This is case for mortality where, for this analysis, we have obtained city-specific
36 Bayesian adjusted effect estimates for all selected cities from multi-city studies, (see
37 section 7.3.2).
38 • Baseline incidence rates and demographic data: The required urban area-specific
39 baseline incidence rates and population data are available for a recent year for at least one
40 of the health endpoints.
7-20
-------
1 • Geographic heterogeneity: Because 63 distributions and population characteristics vary
2 geographically across the U.S., we selected urban study areas to provide coverage for
3 regional variability in factors related to O3 risk including inter-urban gradients in O3, co-
4 pollutant concentrations, population exposure (differences in residential housing density,
5 air conditioning use and commuting patterns), population vulnerability (baseline
6 incidence rates, SES demographics) and variability in effect estimates. The degree to
7 which the set of urban study areas provided coverage for regional differences across the
8 U.S. in many of these 63 risk-related factors was evaluated as part of the
9 representativeness analysis presented in Chapter 8.
10 Application of the above criteria resulted in the selection of 12 urban study areas for
11 inclusion in the risk assessment including:
12 • Atlanta, GA
13 • Baltimore, MD
14 • Boston. MA
15 • Cleveland, OH
16 • Denver, CO
17 • Detroit, MI
18 • Houston, TX
19 • Los Angeles, CA
20 • New York, NY
21 • Philadelphia, PA
22 • Sacramento, CA
23 • St. Louis, MO
24
25 The footprint of each urban study area was based on the set of counties included in one of
26 the two epidemiological studies providing city-specific C-R functions for modeling short-term
27 exposure related mortality (Zanobetti and Schwartz., 2008b). This decision reflects the fact that
28 this health endpoint is considered the most important endpoint modeled in this first draft REA
29 and consequently, matching the shape of the study areas to the specific set of counties modeled
30 in one of the two studies supporting modeling of this critical health endpoint, would increase
31 overall confidence in modeling that endpoint. Note, we had considered developing a second set
32 of study area delineations to match the other epidemiology study used in modeling short-term
33 exposure related mortality (Bell et al., 2004), however, this was not feasible given resources and
34 time, and would add an additional difference between the risk estimates for the two studies and
35 reduce the ability to compare risk estimates across the studies. We would point out however, that
36 the two studies have relatively similar county-level delineations of these urban study areas and
37 therefore, the degree of uncertainty introduced into modeling mortality using the Bell et al., 2004
38 C-R functions (matched to study areas delineations reflecting the Zanobetti and Schwartz, 2008b
7-21
-------
1 study) is expected to be low. The specific set of counties used in defining each of the 12 urban
2 study areas is presented in Table 7-1.
3 7.3.2 Selection of Epidemiological Studies and Specification of Concentration-Response
4 Functions
5 Once the set of health effect endpoints to be included in the risk assessment has been
6 specified, the next step was to select the set of epidemiological studies that will provide the
7 effect estimates and model specifications used in the C-R functions. This section describes the
8 approach used in completing these tasks and presents a summary of the epidemiological studies
9 and associated C-R functions specified for use in the risk assessment.
10 In Chapter 2, section 2.5 we identified the set of health effect categories and associated
11 endpoints to be included in the first draft REA, based on review of the evidence provided in the
12 Os ISA (U.S. EPA, 2012). The selection of specific health effect endpoints to model within a
13 given health effect endpoint category is an iterative process involving review of both the strength
14 of evidence (for a given endpoint) as summarized in the Os ISA together with consideration for
15 the available epidemiological studies supporting a given endpoint and the ability to specific key
16 inputs needed for risk modeling, including effect estimates and model forms. Ultimately,
17 endpoints are only selected if (a) they are associated with an overarching effect endpoint
18 category selected for inclusion in the risk assessment and (b) they have sufficient
19 epidemiological study support to allow their modeling in the risk assessment. Health effect
20 endpoints selected for inclusion in the first draft REA include, specifically for short-term related
21 Os exposure:
22 • Mortality (likely casual relationship)
23 o Non-accidental
24 o All-cause
25 o Cardiovascular
26 o Respiratory
27 • Respiratory effects (causal relationship)
28 o ED (asthma, wheeze, all respiratory symptoms)
29 o HA (unscheduled pulmonary illness, asthma)
30 o Respiratory symptoms
31
32 In addition, as noted in section 2.5, long-term O3 exposure, represented primarily by
33 studies of peak exposures averaged over longer time periods, was associated with respiratory
34 effects (likely causal relationship), including both respiratory mortality and morbidity. While we
35 have not modeled any long-term exposure related health endpoints for the first draft risk
36 assessment, we are considering the estimation of long-term exposure related respiratory mortality
37 for the second Draft risk assessment (see section 7.7). The remainder of this section deals
38 exclusively with the selection of epidemiological studies and specification of C-R functions for
7-22
-------
1 health effect endpoints associated with short-term 63 exposure. We provide an evaluation of
2 potential endpoints associated with long term exposures in Section 7.7.
3 The selection of epidemiological studies to support modeling of the health effect
4 endpoints listed above reflected application of a number of criteria including3:
5 • The study was peer-reviewed, evaluated in the Os ISA, and judged adequate by EPA
6 staff for purposes of inclusion in the risk assessment. Criteria considered by staff
7 include: whether the study provides C-R relationships for locations in the U.S.,
8 whether the study has sufficient sample size to provide effect estimates with a
9 sufficient degree of precision and power, and whether adequate information is
10 provided to characterize statistical uncertainty.
11 • The study is multicity and ideally, includes Bayes-adjusted city-specific effect
12 estimates (or provides data that supports their derivation) since these effect estimates
13 combine local signals with broader regional or national signals. However, in the case
14 of respiratory morbidity endpoints, in most cases we did not have multicity studies
15 and instead, relied upon city-specific studies to provide coverage for these important
16 endpoints.
17 • The study design is considered robust and scientifically defensible, particularly in
18 relation to methods for covariate adjustment (including confounders and effects
19 modifiers). For example, if a given study used ecological-defined variables (e.g.,
20 smoking rates) as the basis for controlling for confounding, concerns may be raised as
21 to the effectiveness of that control.
22 • The study is not superseded by another study (e.g., if a later study is an extension or
23 replication of a former study, the later study would effectively replace the former
24 study), unless the earlier study has characteristics that are clearly preferable.
25 While the first draft REA applies results from epidemiological studies using composite
26 monitors, we are also evaluating studies which utilized more sophisticated and potentially
27 representative exposure surrogates in characterizing population exposure (e.g., linking exposures
28 in individual counties or U.S. Census tracts to the nearest monitor, rather than using a composite
29 monitor value to represent the entire study area). Depending on the results of our evaluation, we
30 may include these types of epidemiology studies as sensitivity analyses in the second Draft risk
31 assessment (see section 7.7). If we are to use effect estimates from these studies that reflect more
32 sophisticated exposure surrogates, it is important that we also utilize those same exposure
33 surrogates in our risk assessment and not link effect estimates (based on more refined exposure
34 surrogates) with the more generalized composite monitors used in modeling most endpoints in
3 In addition to the criteria listed here, we also attempted to include studies that provide coverage for
populations considered particularly at-risk for a particular health (e.g., children, individuals with preexisting
disease). However, a study would have to meet the criteria listed here (in addition to providing coverage for an at-
risk population) in order for that study to be used to derive C-R functions.
7-23
-------
1 the risk assessment. As part of the evaluation of these types of studies, we are determining the
2 feasibility of generating these more customized exposure surrogates to match specific
3 epidemiological studies.
4 Application of the above criteria resulted in the set of epidemiological studies presented
5 in Table 7-4 being identified for use in specifying C-R functions for the first draft analysis (Note,
6 that Table 7-4 also describes elements of the C-R functions specified using each epidemiological
7 study, as discussed below).
8 Once the set of epidemiology studies was selected, the next step was to specify C-R
9 functions for use in the risk assessment using those studies. Several factors were considered in
10 identifying the effect estimates and model forms used in specifying C-R functions for each
11 endpoint. These factors are described below:
12 • Os exposure metric: In the risk assessment supporting the previous 63 NAAQS
13 review, for short-term exposure, we had included C-R functions based on both 24hr
14 averages as well as a number of peak Os measurements. However, based on review of
15 information provided in the Os ISA (U.S. EPA, 2012), we now believe there is
16 increased confidence associated with modeling short-term exposure-related health
17 endpoints using peak Os metrics (i.e., Ihr max, 8hr max and 8hr means) relative to
18 modeling risk using 24hr averages. Consequently, for the first draft REA, we have
19 focused on the peak 63 metrics and excluded C-R functions based on 24hr averages
20 (with one exception).4 The rational for focusing on peak metrics reflects
21 consideration for a number of factors. A study of respiratory ED visits in Atlanta
22 (Darrow et al., 2011) found stronger associations with peak metrics (including Ihr
23 and 8hr max measurements) compared with 24hr averages (see Os ISA section 6.2.7.3
24 and Figure 6-16, U.S. EPA, 2012). Controlled human exposure studies have also
25 demonstrated effects on FEV1, respiratory symptoms, and inflammatory responses
26 associated with exposures up to 8hr (see ISA section 2.5.3). With regard to mortality,
27 the picture is not as clear, primarily due to limitations in the number of
28 epidemiological studies comparing the association of peak 63 metrics and the 24hr
29 average metric with mortality. However, when we consider the other information
30 described here, we conclude that it is generally appropriate to place greater emphasis
31 on C-R functions (for both mortality and morbidity) that utilize peak exposure
32 metrics.5
4 As noted earlier, in order to provide estimates of respiratory-related HA for LA, we did include a C-R
function based on Linn et al., 2000, which utilizes a 24hr average exposure metric.
5 In addition, peak ozone metrics, by focusing on daily ozone levels, avoid the issue where simulation of
meeting the current standard results in nighttime ozone levels actually increasing in some situations (this is a
concern for the 24hr ozone metrics, where these increases in nighttime ozone can dampen predicted reductions in
daytime ozone).
7-24
-------
Table 7-4 Overview of Epidemiological Studies Used in Specifying C-R Functions
Epidemiological
study
(stratified by
short-term
exposure-related
health
endpoints)
Health
endpoints
Location (urban
study area(s)
covered)
Exposure metric
(and modeling
period)
Additional study
design details
Notes regarding application in first Draft analysis
Mortality
Bell et al., 2004
Zanobetti and
Schwartz
(2008b)
Non-
accidental,
respiratory,
cardiovascular
Non-
accidental,
respiratory,
cardiovascular
95 large urban
communities
(provides
coverage for all
12 urban study
areas)
48 U.S. cities
(provides
coverage for the
12 urban study
areas)
24hr avg, 8hr max,
Ihrmax. April
through October and
all year
8hr max. June-
August
Adjusting for time-
varying confounders
(PM, weather,
seasonality). Lag
structure included 0,1,
2 and day 3 lag as well
as 0-6 day distributed
lag. Age range: all
ages.
Effect controlled for
season, day of week,
and temperature. Lag
structure included 0-
3d, 0-20 and 4-20
day). Age range: all
ages
Obtained Bayes-adjusted city-specific effect estimates for non-
accidental mortality from Dr. Bell (personal communication, Dr.
Michelle Bell, December 22, 201 1). Effect estimates based on
constrained distributed lag (0-6 days) for the 8hr max peak metric
evaluated for the fullest of monitored data associated with each urban
area (for most urban areas, this represents measurements taken
during city-specific ozone season). For this reason, we constrained
risk modeling using these effect estimates to the ozone season
specific to each urban study area (see Table 7-1).
Obtained Bayes-adjusted city-specific effect estimates for non-
accidental, respiratory and cardiovascular from Dr. Zanobetti
(personal communication, Dr. Antonella Zanobetti, January 5, 2012).
These effect estimates reflect a 0-3 day distributed lag and are based
on 8hr mean ozone levels measured between June and August.
Consequently, we constrained modeling of risk with these effect
estimates to June-August for each urban study area.
Morbidity - HA for respiratory effect)
Medina-Ramon
etal.,2006.
Linn etal., 2000
Lin et al., 2008
Katsouyanni et
al 2009
HA: COPD,
pneumonia
HA:
unscheduled
for pulmonary
illness
HA:
respiratory
disease
HA:
cardiovascular
disease,
chronic
36 cities
(provides
coverage for all
12 urban study
areas)
LA only
NY State (used to
cover NYC)
14 cities
(provides
coverage for
Detroit only)
8hr mean, warm
(May -August), cool
(October-April), all
year
24hr mean, LA ozone
season (all year)
Ihr max (for 10am-
6pm interval), warm
season (April-
October)
Ihrmax. Summer
only and all year
Distributed lag (0-1
day). Age range: >
65yrs.
Lag 0. Age range: all
ages
LagO, 1,2, 3. Age
range: <18yrs
Lag 0-1 day. Age
range: > 65yrs.
Generated risk estimates based on warm season (used existing June-
August composite monitor 8hr mean values).
Included effect estimate based on 24hr avg metric since this provided
additional coverage for HA in L.A.
Used Ihr max metric applied to the city-specific ozone season for
NYC (April-October).
C-R function applied only for all respiratory endpoint. Used June-
August-based composite monitor.
7-25
-------
Epidemiological
study
(stratified by
short-term
exposure-related
health
endpoints)
Silvermanet al.,
2010
Health
endpoints
obstructive
pulmonary
disease,
pneumonia,
all respiratory
HA: asthma
(ICU and non-
ICU)
Location (urban
study area(s)
covered)
NYC
Exposure metric
(and modeling
period)
8hr max. Warm
season (April-
August)
Additional study
design details
Includes control for
PM2.5. Lag 0-1 day.
Age range: children 6-
18yrs
Notes regarding application in first Draft analysis
Applied C-R function (for ozone and ozone with control for PM2 5)
to the city-specific ozone season for NYC (slightly longer than the
modeling period used in the study).
Morbidity - ED andER visits (respiratory)
Ito etal., 2007
Tolbertetal.,
2007
Strickland etal.,
2010
Darrow etl al.,
2011
ED: asthma
ED: all
respiratory
ER:
respiratory
ED: all
respiratory
NYC
Atlanta
Atlanta
Atlanta
8hr max. Warm
season (April-
September)
8hr max. Summer
(March-October)
8hr max (based on
population weighted
average across
monitors). Warm
season (May to
October) and cool
(November to April)
Shrmax, Ihrmax,
24hr avg for summer
(March-October).
Includes models
controlling for SO2,
NO2, CO and PM2.5.
Lag: 0, 1, and
distributed lag (0-1
day). Age range: all
ages
Includes models
controlling forNO2,
CO, PMio.and NO2/
NO2. Age range: all
ages
Lag: average of 0-2
day, distributed lag 0-7
day. Age range: 5-
17yrs
Lag: Iday. Age range:
all ages
Applied C-R functions (for ozone alone and ozone with control for
listed pollutants) to the city-specific ozone season for NYC (slightly
longer than the modeling period used in the study).
Applied C-R functions (for ozone alone and ozone with control for
listed pollutants) to the city-specific ozone season for Atlanta.
Included effect estimates based on both lag structures and used
composite monitor values for city-specific ozone season.
Used city-specific ozone season-based composite monitor values.
Morbidity - respiratory symptoms
Gent etal., 2003
Respiratory
symptoms:
wheeze,
persistent
Springfield MA
(study used to
cover Boston)
Ihr max, 8hr max
Lag: 0 and 1 day. Age
range: asthmatic
children < 12 yrs.
Included effect estimates for different symptoms based on both 8hr
max and Ihr max metrics (for city-specific ozone season composite
monitor values for Boston). The study area (which focuses on
Springfield and the northern portion of Connecticut) does not
7-26
-------
Epidemiological
study
(stratified by
short-term
exposure-related
health
endpoints)
Health
endpoints
cough, chest
tightness,
shortness of
breath
Location (urban
study area(s)
covered)
Exposure metric
(and modeling
period)
Additional study
design details
Notes regarding application in first Draft analysis
encompass Boston. However, we are willing to accept uncertainty
associated with using effect estimates from this study to provide
coverage for Boston given the goal of providing coverage for this
morbidity endpoint. However, there is increased uncertainty
associated with modeling for this endpoint. .
7-27
-------
1
2 • Single- and multi-pollutant models (pertains to both short-term and long-term
3 exposure studies): Epidemiological studies often consider health effects associated with
4 ambient 63 independently as well as together with co-pollutants (e.g., 63, nitrogen
5 dioxide, sulfur dioxide, carbon monoxide). To the extent that any of the co-pollutants
6 present in the ambient air may have contributed to health effects attributed to Os in single
7 pollutant models, risks attributed to O3 may be overestimated or underestimated if C-R
8 functions are based on single pollutant models. This would argue for inclusion of models
9 reflecting consideration of co-pollutants. Conversely, in those instances where co-
10 pollutants are highly correlated with Os, inclusion of those pollutants in the health impact
11 model can produce unstable and statistically insignificant effect estimates for both O3 and
12 the co-pollutants. This situation would argue for inclusion of a model based exclusively
13 on 03. Given that single and multi-pollutant models each have potential advantages and
14 disadvantages, to the extent possible, given available information we have included both
15 types of C-R functions in the risk assessment.
16 • Single-city versus multi-city studies: All else being equal, we judge C-R functions
17 estimated in the assessment location as preferable to a function estimated in some other
18 location, to avoid uncertainties that may exist due to differences associated with
19 geographic location. There are several advantages, however, to using estimates from
20 multi-city studies versus studies carried out in single cities. Multi-city studies are
21 applicable to a variety of settings, since they estimate a central tendency across multiple
22 locations. Multi-city studies also tend to have more statistical power and provide effect
23 estimates with relatively greater precision than single-city studies due to larger sample
24 sizes, reducing the uncertainty around the estimated health coefficient. By contrast,
25 single-city studies, while often having lower statistical power and varying study designs
26 which can make comparison across cities challenging, reflect location-specific factors
27 such as differences in underlying health status, and differences in exposure-related factors
28 such as air conditioner use and urban density with larger populations exposed near high-
29 traffic roads. There is a third type of study design that generates Bayes-adjusted city-
30 specific effect estimates, thereby combining the advantages of both city-specific and
31 multi-city studies. Bayes-adjusted city-specific estimates begin with a city-specific effect
32 estimate and shrink that towards a multi-city mean effect estimate based on consideration
33 for the degree of variance in both estimates. For the first draft REA, we have elected to
34 place greater confidence on these types of Bayesian-adjusted effect estimates when they
35 are available. Otherwise, given the advantages for both city-specific and multi-city effect
36 estimates, we have used both types when available. In those instances where a multi-city
37 study only provides aggregated effect estimates, but does differentiate those estimates
38 regionally, we would use those regional-specific estimates rather than a single national -
39 level estimate by matching selected urban study areas to these regions. For the
40 epidemiological studies we identified for this first draft analysis, none included these
41 types of regional effect estimates - see Table 7-4.
42 • Multiple lag models: Based on our review of evidenced provided in the ISA, we believe
43 there is increased confidence in modeling both mortality and respiratory morbidity risk
44 based on exposures occurring up to a few days prior to the health effect, with less support
7-28
-------
1 for associations over longer exposure periods or effects lagged more than a few days
2 from the exposure (see O3 ISA section 2.5.4.3, U.S. EPA, 2012). Consequently, we have
3 favored C-R functions reflecting shorter lag periods (e.g., 0, 1 or 1-2 days). With regard
4 to the specific lag structure (e.g, single day versus distributed lags), the O3 ISA notes that
5 epidemiological studies involving respiratory morbidity have suggested that both single
6 day and multi-day average exposures are associated with adverse health effects (see O3
7 ISA section 2.5.4.3). Therefore, when available both types of lag structures where
8 considered in specifying C-R functions.
9 • Seasonally-differentiated effects estimates: The previous O3 AQCD (published in
10 2006) concluded that aggregate population time-series studies demonstrates a positive
11 and robust association between ambient O3 concentrations and respiratory-related
12 hospitalizations and asthma ED visits during the warm season (see O3 ISA section 2.5.3m
13 U.S. EPA, 2012). The current O3 ISA notes that recent studies of short-term exposure-
14 related respiratory mortality in the U.S. suggest that the effect is strengthened in the
15 summer season (O3 ISA section 6.6.2.5, U.S. EPA, 2012). In addition, we note that many
16 of the key epidemiological studies exploring both short-term exposure related mortality
17 and morbidity discussed in the current O3 ISA have larger (and more statistically
18 significant) effect estimates when evaluated for the summer (O3) season, relative to the
19 full year (see O3 ISA Figures 6-18 and 6-26, U.S. EPA, 2012). Given that we anticipate
20 O3 levels to be elevated during the O3 season resulting in increased exposure and risk, we
21 favored C-R functions based on O3 measurements taken during the O3 (or warm/summer)
22 season and placed less emphasis on C-R functions reflecting O3 measured over the entire
23 year (unless, as with L. A. the O3 period is the entire year).
24 • Shape of the functional form of the risk model (including threshold): The current O3
25 ISA concludes that there is little support in the literature for a population threshold for
26 short-term exposure-related effects, although in the case of mortality, the O3 ISA notes
27 that the nature of the mortality effect as well as study design may mean that these studies
28 are not well suited to identify a threshold should it exist (see O3 ISA, section 2.5.4.4, U.S.
29 EPA, 2012). Given the above observation from the ISA regarding the potential for
30 thresholds, we did not include C-R functions for any of the short-term exposure-related
31 health endpoints modeled that incorporated a threshold.
32 Application of the above criteria resulted in an array of C-R functions specified for the
33 risk assessment (see Table 7-4). In presenting the C-R functions in Table 7-4, we have focused
34 on describing key attributes of each C-R function (and associated source epidemiological study)
35 relevant to a review of their use in the risk assessment. More detailed technical information
36 including effect estimates and model specification is provided in Appendix 7-A (Table 7A-1).
37 Specific summary information provided in Table 7-4 includes:
38 • Health endpoints: identifies the specific endpoints evaluated in the study. Generally
39 we included all of these in our risk modeling, however, when a subset was modeled,
40 we reference that in the "Notes" column (last column in the table).
7-29
-------
1 • Location: identifies the specific urban areas included in the study and maps those to
2 the set of 12 urban study areas included in the risk assessment.
3 • Exposure metric: describes the exposure metric used in the study, including the
4 specific modeling period (e.g., 63 season, warm season, full year). As noted earlier,
5 for the first draft REA, we developed two categories of composite monitor values to
6 match the modeling periods used in the two short-term exposure-related mortality
7 studies providing C-R functions for the analysis. For the remaining morbidity
8 endpoints, we mapped specific C-R functions to whichever of these two composite
9 monitor categories most closely matched the modeling period used in the underlying
10 epidemiological study. This mapping (for morbidity endpoint C-R functions) is
11 described in the "Notes" column (the seasons reflecting in modeling for each C-R
12 function are also presented in Appendix 7-A, Table 7A-1).
13 • Additional study design details: this column provides additional information primarily
14 covering the lag structure and age ranges used in the study.
15 • Notes regarding application in first draft analysis: as the name implies, this column
16 provides notes particular to the application of a particular epidemiological study and
17 associated C-R functions in the risk assessment.
18 7.3.3 Defining Os concentration ranges (down to the LML) for which there is increased
19 confidence in estimating risk
20 As discussed in section 7.1.1 and 7.3.2, for this first draft REA, we did not incorporate
21 thresholds in modeling risk, reflecting consideration of the evidence as summarized in the Oj
22 ISA (see section 2.5.4.4, U.S. EPA, 2012). However, we did identify Os concentration ranges for
23 which there is increased confidence in estimating risk. Specifically, we note that modeling risk
24 within the range of Os levels used in the derivation the C-R function has increased confidence
25 relative to modeling risk for Os levels below that range. Therefore, we can use the LML
26 associated with the derivation of a particular C-R function to help define an 63 concentration
27 range with increased confidence in estimating risk. Overall confidence is further increased as we
28 model risk closer to the central mass of Os levels used in the derivation of the C-R function.
29 Ideally we would have access to the 63 monitor-based datasets used in each of the
30 epidemiological studies providing C-R functions used in this analysis so that we could define
31 these ranges of increased confidence accordingly. In the case of city-specific effect estimates
32 ideally we would obtain the underlying O3 measurement data stratified by urban study area.
33 Note, also that, when we reference "measurement data" we are actually referring to the specific
34 exposure surrogate used in deriving the C-R function and not simply the array of hourly values
35 for each monitor. However, data limitations prevented us from identifying the LML each study
36 and therefore, as noted in section 7.1.1, we used the distributions of composite monitor values
7-30
-------
1 calculated for each of the two simulation years included in the analysis (2007 and 2009) to
2 estimate surrogates for the LML.
3 Given the different dimensions associated with risk estimates generated for this analysis
4 (e.g., 12 urban study areas, two simulation years, several different daily peak 63 level metrics
5 associated with different C-R functions) an array of LMLs had to be extracted from the
6 composite monitor values used in the risk assessment. The set of LML values used to define 63
7 concentration ranges for which there is increased confidence in estimating risk is presented
8 below in Table 7-5. The set of LMLs is also provided as part of the full set of model inputs
9 presented in Appendix 7A, Table 7A-1.
10 LML values presented in Table 7-5 were linked to a given C-R function based on the air
11 quality metric used by the C-R function. For example, with short-term exposure related mortality
12 estimated for Baltimore in 2007 using the Bell et al., (2004) study and associated C-R function,
13 we used the LML value for the 8hr max metric (city-specific Os season), reflecting the metric
14 used for that C-R function (see Table 7-4 and Appendix 7A, Table 7A-1). Consequently, we
15 would identify 13 ppb from Table 7-5 (and Table 7A-1) as the LML for modeling that endpoint.
16 As noted earlier in section 7.1.2, we then use the LML as a lower bound on the C-R function
17 (i.e., risk would not be modeled below 13ppb), in generating higher confidence risk estimates.6
18
19
20
21
22
23
24
25
26
6 The values presented in Table 7-5 allowed us to define exposure ranges with increased confidence for
most of the endpoints included in this analysis (see Table 7-4 and Appendix 7A, Table 7A-1 for details on which air
metrics were used in modeling specific health endpoints and consequently, which of the values from Table 7-5
would be used in specifying regions of increased confidence). However, several short-term exposure-related
morbidity studies used ozone metrics different form the 8hr mean (June-August) and 8hr max (city-specific ozone
season) reflected in the statistics presented in Table 7-5 and therefore, we had to identify LML values from different
composite monitors in order to specify regions of increased confidence for these endpoints (the full set of LMLs for
all C-R functions is presented in Appendix 7A, Table 7A-1).
7-31
-------
1 Table 7-5 Composite Monitor
2 Modeling Risk
LML Used in Defining Ranges of Increased Confidence in
Urban Study
Area
8r max (city-specific Os
season) ppb
8hr mean (reflects June-
August levels) ppb
Metrics Based on 2007 Composite Monitors
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
17
13
12
12
4
13
6
9
10
13
13
8
24
13
19
6
21
19
10
31
10
12
30
22
Metrics Based on 2009 Composite Monitors
Atlanta
Baltimore
Boston
Cleveland
Denver
Detroit
Houston
Los Angeles
New York
Philadelphia
Sacramento
St. Louis
5
9
12
15
16
14
7
8
8
9
5
7
21
24
17
16
22
11
15
22
12
14
30
22
4 7.3.4 Baseline health effect incidence and prevalence data
5 As noted earlier (section 7.1.2), the most common epidemiological-based health risk
6 model expresses the reduction in health risk (Ay) associated with a given reduction in 63
7 concentrations (Ax) as a percentage of the baseline incidence (y). To accurately assess the
8 impact of 63 air quality on health risk in the selected urban areas, information on the baseline
7-32
-------
1 incidence of health effects (i.e., the incidence under recent air quality conditions) in each
2 location is needed. In some instances, health endpoints are modeled for a population with an
3 existing health condition, necessitating the use of a prevalence rate. Where at all possible, we use
4 county-specific incidences or incidence rates (in combination with county-specific populations).
5 In some instances, when county-level incidence rates were not available, BenMAP can calculate
6 and employ more generalized regional rates (see BenMAP Guidance Manual for additional
7 detail, Abt Associates, Inc. 2010). For prevalence rates (which were only necessary for modeling
8 respiratory symptoms among asthmatic children using Gent et al., (2008) - see Table 7-4), we
9 utilized a national-level prevalence rate appropriate for the age group being modeled. A
10 summary of available baseline incidence data for specific categories of effects (and prevalence
11 rates for asthma) is presented below:
12 • Baseline incidence data on mortality: County-specific (and, if desired, age-and race-
13 specific) baseline incidence data are available for all-cause and cause-specific
14 mortality from CDC Wonder.7 The most recent year for which data are available
15 online is 2005 and this was the source of incidence data for the risk assessment.8
16 • Baseline incidence data for hospital admissions and emergency room (ER) visits:
17 Cause-specific hospital admissions baseline incidence data are available for each of
18 40 states from the State Inpatient Databases (SID). Cause-specific ER visit baseline
19 incidence data are available for 26 states from the State Emergency Department
20 Databases (SEDD). SID and SEDD are both developed through the Healthcare Cost
21 and Utilization Project (HCUP), sponsored by the Agency for Healthcare Research
22 and Quality (AHRQ). In addition to being able to estimate State-level rates, SID and
23 SEDD can also be used to obtain county-level hospital admission and ER visit counts
24 by aggregating the discharge records by county.
25 • Asthma prevalence rates: state-level prevalence rates that are age group stratified are
26 available from the Centers for Disease Control and Prevention (CDC) Behavioral
27 Risk Factor Surveillance System (BRFSS) (U.S. CDC, 2010).
28 Incidence and prevalence rates used in the first draft REA are presented as part of the full
29 set of model inputs documented in Appendix 7 A, Table 7A-1. The incidence rates and
30 prevalence rates provided in Table 7A-1 are weighted average values for the age group
31 associated with each of the C-R functions. These weighted averages are calculated within
32 BenMAP using more refined age-differentiated incidence and prevalence rates originally
33 obtained from the data sources listed in the bullets above.
7 http ://wonder. cdc.gov/mortsql. html
8
Note: For years 1999 - 2005, CDC Wonder uses ICD-10 codes; for years prior to 1999, it uses ICD-9
codes. Since most of the studies use ICD-9 codes, this means that EPA will have to create or find a mapping from
ICD-9 codes to ICD-10 codes if the most recent data available are to be used.
7-33
-------
1 7.3.5 Population (demographic) data
2 To calculate baseline incidence rate, in addition to the health baseline incidence data we
3 also need the corresponding population. We obtained population data from the U.S. Census
4 bureau (http://www.census.gov/popest/counties/asrh/). These data, released on May 14, 2009, are
5 the population estimates of the resident populations by selected age groups and sex for counties
6 in each U.S. state from 2000 to 2008. Total population counts used in modeling each of the
7 health endpoints evaluated in the analysis (differentiated by urban study area and simulation
8 year) are provided as part model inputs presented in Appendix 7A, Table 7A-1.
9 7.4 ADDRESSING VARIABILITY AND UNCERTAINTY
10 An important component of a population risk assessment is the characterization of both
11 uncertainty and variability. Variability refers to the heterogeneity of a variable of interest within
12 a population or across different populations. For example, populations in different regions of the
13 country may have different behavior and activity patterns (e.g., air conditioning use, time spent
14 indoors) that affect their exposure to ambient Os and thus the population health response. The
15 composition of populations in different regions of the country may vary in ways that can affect
16 the population response to exposure to O3 - e.g., two populations exposed to the same levels of
17 Os might respond differently if one population is older than the other. Variability is inherent and
18 cannot be reduced through further research. Refinements in the design of a population risk
19 assessment are often focused on more completely characterizing variability in key factors
20 affecting population risk - e.g., factors affecting population exposure or response - in order to
21 produce risk estimates whose distribution adequately characterizes the distribution in the
22 underlying population(s).
23 Uncertainty refers to the lack of knowledge regarding the actual values of inputs to an
24 analysis. Models are typically used in analyses, and there is uncertainty about the true values of
25 the parameters of the model (parameter uncertainty) - e.g., the value of the coefficient for Os in a
26 C-R function. There is also uncertainty about the extent to which the model is an accurate
27 representation of the underlying physical systems or relationships being modeled (model
28 uncertainty) - e.g., the shapes of C-R functions. In addition, there may be some uncertainty
29 surrounding other inputs to an analysis due to possible measurement error—e.g., the values of
30 daily Os concentrations in a risk assessment location, or the value of the baseline incidence rate
31 for a health effect in a population.9 In any risk assessment, uncertainty is, ideally, reduced to the
32 maximum extent possible through improved measurement of key variables and ongoing model
9 It is also important to point out that failure to characterize variability in an input used in modeling can
also introduce uncertainty into the analysis. This reflects the important link between uncertainty and variability with
the effort to accurately characterize variability in key model inputs actually reflecting an effort to reduce uncertainty.
7-34
-------
1 refinement. However, significant uncertainty often remains, and emphasis is then placed on
2 characterizing the nature of that uncertainty and its impact on risk estimates. The
3 characterization of uncertainty can be both qualitative and, if a sufficient knowledgebase is
4 available, quantitative.
5 The selection of urban study areas for the Os risk assessment was designed to cover the
6 range of (Vrelated risk experienced by the U.S. population and, in general, to adequately reflect
7 the inherent variability in those factors affecting the public health impact of O3 exposure.
8 Sources of variability reflected in the risk assessment design are discussed in section 7.4.1, along
9 with a discussion of those sources of variability which are not fully reflected in the risk
10 assessment and consequently introduce uncertainty into the analysis.
11 The characterization of uncertainty associated with risk assessment is often addressed in
12 the regulatory context using a tiered approach in which progressively more sophisticated
13 methods are used to evaluate and characterize sources of uncertainty depending on the overall
14 complexity of the risk assessment (WHO, 2008). 3Guidance documents developed by EPA for
15 assessing air toxics-related risk and Superfund Site risks (U.S.EPA, 2004 and 2001, respectively)
16 as well as recent guidance from the World Health Organization (WHO, 2008) specify multi-
17 tiered approaches for addressing uncertainty.
18 The WHO guidance, in particular, presents a four-tiered approach for characterizing
19 uncertainty (see Chapter 3, section 3.2.6 for additional detail on the four tiers included in the
20 WHO's guidance document). With this four-tiered approach, the WHO framework provides a
21 means for systematically linking the characterization of uncertainty to the sophistication of the
22 underlying risk assessment. Ultimately, the decision as to which tier of uncertainty
23 characterization to include in a risk assessment will depend both on the overall sophistication of
24 the risk assessment and the availability of information for characterizing the various sources of
25 uncertainty. EPA staff has used the WHO guidance as a framework for developing the approach
26 used for characterizing uncertainty in this risk assessment.
27 The overall analysis in the Os NAAQS risk assessment is relatively complex, thereby
28 warranting consideration of a full probabilistic (WHO Tier 3) uncertainty analysis. However,
29 limitations in available information prevent this level of analysis from being completed at this
30 time. In particular, the incorporation of uncertainty related to key elements of C-R functions
31 (e.g., competing lag structures, alternative functional forms, etc.) into a full probabilistic WHO
32 Tier 3 analysis would require that probabilities be assigned to each competing specification of a
33 given model element (with each probability reflecting a subjective assessment of the probability
34 that the given specification is the "correct" description of reality). However, for many model
35 elements there is insufficient information on which to base these probabilities. One approach that
36 has been taken in such cases is expert elicitation; however, this approach is resource- and time-
7-35
-------
1 intensive and consequently, it was not feasible to use this technique in the current 63 NAAQS
2 review to support a WHO Tier 3 analysis.10
3 For most elements of this risk assessment, rather than conducting a full probabilistic
4 uncertainty analysis, we have included qualitative discussions of the potential impact of
5 uncertainty on risk results (WHO Tierl). As discussed in section 7.1.1, we had originally
6 planned to complete a comprehensive sensitivity analysis exploring the potential impact of
7 various design elements on the core risk estimates being generated (WHO Tier 2). However, the
8 effort required to complete a comprehensive set of core risk estimates for the mortality and
9 morbidity endpoints included in the analysis prevented us from completing a comprehensive
10 sensitivity analysis for the first draft REA. We do note however, that the set of core risk
11 estimates generated for the analysis does provide, for some of the health endpoints (i.e.,
12 respiratory morbidity) an array of estimates that covers a number of modeling elements (e.g.,
13 copollutants models, lag structure, air quality metric). Insights into the potential impact of these
14 design elements on the core risk estimates are discussed as those risk estimates are summarized
15 in sections 7.1.4.2. Sensitivity analyses being considered for the second draft REA are described
16 in section 7.7.1.
17 In addition to the qualitative and quantitative treatment of uncertainty and variability
18 which are described here, we have also completed an analysis to evaluate the representativeness
19 of the selected urban study areas against national distributions for key O3 risk-related attributes
20 to determine whether they are nationally representative or more focused on a particular portion
21 of the distribution for a given attribute (see Chapter 8, section 8.2.1). In addition, we have
22 completed a second analysis addressing the representativeness issue, which identified where the
23 12 urban study areas included in this risk assessment fall along a distribution of national-level
24 long-term exposure-related mortality risk (see Chapter 8, section 8.2.2). This analysis allowed us
25 to assess the degree of which the 12 urban study areas capture locations within the U.S. likely to
26 experience elevated levels of risk related to O3 exposure.
27 The remainder of this section is organized as follows. Key sources of variability which
28 are reflected in the design of the risk assessment, along with sources excluded from the design,
29 are discussed in section 7.1.4.1. A qualitative discussion of key sources of uncertainty associated
30 with the risk assessment (including the potential direction, magnitude and degree of confidence
31 associated with our understanding of the source of uncertainty - the knowledge base) is
32 presented in section 7.1.4.2.
10 Note, that while a full probabilistic uncertainty analysis was not completed for this risk assessment, we
were able to use confidence intervals associated with effects estimates (obtained from epidemiological studies) to
incorporate statistical uncertainty associated with sample size considerations in the presentation of risk estimates.
7-36
-------
1 7.4.1 Treatment of Key Sources of Variability
2 The risk assessment was designed to cover the key sources of variability related to
3 population exposure and exposure response, to the extent supported by available data. Here, the
4 term key sources of variability refers to those sources that the EPA staff believes have the
5 potential to play an important role in impacting population incidence estimates generated for this
6 risk assessment. Specifically, EPA staff has concluded that these sources of variability, if fully
7 addressed and integrated into the analysis, could result in adjustments to the core risk estimates
8 which might be relevant from the standpoint of interpreting the risk estimates in the context of
9 the Os NAAQS review. The identification of sources of variability as "key" reflects
10 consideration for sensitivity analyses conducted for previous O3 NAAQS risk assessments,
11 which have provided insights into which sources of variability (reflected in different elements of
12 those earlier sensitivity analyses) can influence risk estimates, as well as information presented
13 in the O3 IS A.
14 As with all risk assessments, there are sources of variability which have not been fully
15 reflected in the design of the risk assessment and consequently introduce a degree of uncertainty
16 into the risk estimates. While different sources of variability were captured in the risk
17 assessment, it was generally not possible to separate out the impact of each factor on population
18 risk estimates, since many of the sources of variability are reflected collectively in a specific
19 aspect of the risk model. For example, inclusion of urban study areas from different regions of
20 the country likely provides some degree of coverage for a variety of factors associated with 63
21 risk (e.g., air conditioner use, differences in population commuting and exercise patterns,
22 weather). However, the model is not sufficiently precise or disaggregated to allow the individual
23 impacts of any one of these sources of variability on the risk estimates to be characterized.
24 Key sources of potential variability that are likely to affect population risks are discussed
25 below, including the degree to which they are captured in the design of the risk assessment:
26 • Heterogeneity in the effect of 63 on health across different urban areas: A
27 number of studies cited in the ISA have found evidence for regional heterogeneity in
28 the short-term exposure-related mortality effect (Smith et al., 2009 and Bell and
29 Dominici, 2008, Bell et al., 2004, Zanobetti an Schwartz 2008b - see O3 ISA section
30 6.6.2.2, U.S. EPA, 2012). These studies have demonstrated that the cross-city
31 differences in effect estimates can be quite substantial (see ISA Figures 6-31 and 6-
32 32). For the short-term exposure-related mortality endpoint, we have used Bayes-
33 adjusted city-specific effect estimates which are intended to capture cross-city
34 differences in effect estimates for the mortality endpoint (while still utilizing
35 information provided by a more stable national-level estimate). However, Smith et
36 al., 2009 had recommended that Bayes-adjusted city-specific effect estimates such as
37 those cited in Bell et al., 2004, utilize regionally-differentiate effect estimates for
38 updating the city specific effect estimates, rather than a national-level effect estimate,
39 in order to more fully capture spatial heterogeneity in the Os effect. This
7-37
-------
1 recommended refinement by Smith et al., 2009 to the derivation of effect estimates
2 using the Bayes-adjustment technique has not been implemented, but may be
3 considered for the second draft analysis (see section 7.7.1). For short-term morbidity
4 endpoints, typically we have used city-specific effect estimates, however, for most
5 endpoints, we only have estimates for a subset of the urban study areas (typically
6 NYC, Atlanta and/or LA). Therefore, while our risk estimates do reflect the
7 application of city-specific effect estimates, because we do not have estimates for all
8 12 urban study areas, we do not provide comprehensive coverage for heterogeneity in
9 modeling the respiratory morbidity endpoint category.
10 • Intra-urban variability in ambient Os levels: The picture with regard to within city
11 variability in ambient O3 levels and the potential impact on epidemiologic-based
12 effect estimates is somewhat more complicated. The ISA notes that spatial variability
13 in Os levels is dependent on spatial scale with O3 levels being more homogeneous
14 over a few kilometers due to the secondary formation nature of O3, while levels can
15 vary substantially over tens of kilometers. Community exposure may not be well
16 represented when monitors cover large areas with several subcommunities having
17 different sources and topographies as exemplified by Los Angeles which displays
18 significantly greater variation in inter-monitor correlations than does for example,
19 Atlanta or Boston (see O3 ISA section 4.6.2.1 U.S. EPA 2012). Despite the potential
20 for substantial variability across monitors (particularly in larger urban areas with
21 greater variation in sources and topography), the ISA notes that studies have tended to
22 demonstrate that monitor selection has only a limited effect on the association of
23 short-term O3 exposure with health effects. The likely explanation for this is that,
24 while absolute values for a fixed point in time can vary across monitors in an urban
25 area, the temporal patterns of O3 variability across those same monitors tends to be
26 well correlated. Given that most of the O3 epidemiological studies are time series in
27 nature, the O3 ISA notes that the stability of temporal profiles across monitors within
28 most urban areas means that monitor selection will have little effect on the outcomes
29 of an epidemiological study examining short-term exposure-related mortality or
30 morbidity. For this reason, we conclude that generally intra-city heterogeneity in O3
31 levels is not a significant factor likely to impact the risk assessment. One exception is
32 LA which, due to its size and variation in O3 sources and other factors impacting O3
33 patterns such as topography, may display significant variation in ambient O3 levels
34 with a subsequent impact on risk. However, in the case of LA (as with the other
35 urban study areas), we model risk using composite monitors which do not provide
36 spatially-differentiated representations of exposure and consequently, we do not
37 address this source of variability in the first draft analysis.
38 • Variability in the patterns of ambient Os reduction across urban areas: In
39 simulating just meeting the current or alternative suites of standards, there can be
40 considerable variability in the patterns of ambient O3 reductions that result from
41 different simulation approaches (i.e., they can be more localized, more regional, or
42 some combination thereof). Given the secondary formation of O3, variation in the
43 spatial pattern of O3 reductions is likely to be dampened somewhat. For the first draft
44 REA, we have only included one strategy for simulating the just meeting the current
45 O3 standard (quadratic rollback). As noted in section 7.2.3, we may employ a more
7-38
-------
1 sophisticated method for predicting ambient 63 under current and alternate standard
2 levels for the second Draft analysis. Therefore, while we have not rigorously
3 evaluated potential variability in the reduction of O3 levels in response to simulating
4 the current standard level for the first draft analysis, we may have a more
5 comprehensive treatment of the issue for the second Draft analysis.
6 • Copollutant concentrations: Recent studies examining the potential for
7 confounding by PM (and it constituents) of the short-term exposure-related mortality
8 effect yielded mixed results with some studies showing little attenuation, while other
9 studies suggest modest attenuation (O3 ISA section 6.6.3, U.S. EPA, 2012). However,
10 the ISA concludes that"... across studies, the potential impact of PM indices on Ch-
11 mortality risk estimates tended to be much smaller than the variation in Ch-mortality risk
12 estimates across cities suggesting that Ch effects are independent of the relationship between
13 PM and mortality. Although some studies suggest that Ch-mortality risk estimates may be
14 confounded by PM or its chemical components the interpretation of these results requires
15 caution due to the limited PM datasets used as a result of the every-3rd- and 6th-day PM
16 sampling schedule." (O3ISA, section 6.6.3). While these observations suggest that
17 copollutants confounding may not be a significant issue, stated concerns regarding the every
18 3rd and 6th day sampling schedule leave the possibility that the sampling strategy is masking a
19 copollutants effect. Due to limits in available data from the multi-city O3 mortality studies,
20 we did not include multipollutant model specifications for mortality. Multipollutant effect
21 estimates were available for a number of the respiratory morbidity endpoints, and we include
22 risk results based on those estimates in the array of core results. Therefore, we are in a
23 position to evaluate to some extent the potential impact of copollutants confounding on the
24 respiratory effects category.
25 • Demographics and socioeconomic-status (SES)-related factors: Variability in
26 population density, particularly in relation to elevated levels of O3 has the potential to
27 influence population risk, although the significance of this factor also depends on the
28 degree of intra-urban variation in O3 levels (as discussed above). In addition,
29 community characteristics such as pre-existing health status, ethnic composition, SES
30 and the age of housing stock (which can influence rates of air conditioner use thereby
31 impacting rates of infiltration of O3 indoors) can contribute to observed differences in
32 (Vrelated risk (discussed in O3 ISA - section 2.5.4.5, U.S. EPA, 2012). Some of the
33 heterogeneity observed in effect estimates between cities in the multicity studies may
34 be due to these demographic and SES factors, and while we cannot determine how
35 much of that heterogeneity is attributable to these factors, the degree of variability in
36 effect estimates between cities in our analysis should help to capture some of the
37 latent variability in SES and demographics.
38 • Baseline incidence of disease: We collected baseline health effects incidence data
39 (for mortality and morbidity endpoints) from a number of different sources (see
40 section 7.3.4). Often the data were available at the county-level, providing a
41 relatively high degree of spatial refinement in characterizing baseline incidence given
42 the overall level of spatial refinement reflected in the risk assessment as a whole.
43 Otherwise, for urban study areas without county-level data, either (a) a surrogate
44 urban study area (with its baseline incidence rates) was used, or (b) less refined state-
45 level incidence rate data were used.
7-39
-------
1 7.4.2 Qualitative Assessment of Uncertainty
2 As noted in section 7.4, we have based the design of the uncertainty analysis carried out
3 for this risk assessment on the framework outlined in the WHO guidance document (WHO,
4 2008). That guidance calls for the completion of a Tier 1 qualitative uncertainty analysis,
5 provided the initial Tier 0 screening analysis suggests there is concern that uncertainty associated
6 with the analysis is sufficient to significantly impact risk results (i.e., to potentially affect
7 decision making based on those risk results). Given previous sensitivity analyses completed for
8 prior Os NAAQS reviews, which have shown various sources of uncertainty to have a potentially
9 significant impact on risk results, we believe that there is justification for conducting a Tier 1
10 analysis.
11 For the qualitative uncertainty analysis, we have described each key source of uncertainty
12 and qualitatively assessed its potential impact (including both the magnitude and direction of the
13 impact) on risk results, as specified in the WHO guidance. Similar to our discussion of
14 variability in the last section, the term key sources of uncertainty refers to those sources that the
15 EPA staff believes have the potential to play an important role in impacting population incidence
16 estimates generated for this risk assessment (i.e., these sources of uncertainty, if fully addressed
17 could result in adjustments to the core risk estimates which might impact the interpretation of
18 those risk estimates in the context of the Os NAAQS review). These key sources of uncertainty
19 have been identified through consideration for sensitivity analyses conducted for previous Os
20 NAAQS risk assessments, together with information provided in the final Os ISA and comments
21 provided by CAS AC on the analytical plan for the risk assessment.
22 As shown in Table 7-6, for each source of uncertainty, we have (a) provided a
23 description, (b) estimated the direction of influence (over, under, both, or unknown) and
24 magnitude (low, medium, high) of the potential impact of each source of uncertainty on the risk
25 estimates, (c) assessed the degree of uncertainty (low, medium, or high) associated with the
26 knowledge-base (i.e., assessed how well we understand each source of uncertainty), and (d)
27 provided comments further clarifying the qualitative assessment presented. Table 7-6 includes
28 all key sources of uncertainty identified for the Os REA.
29 The categories used in describing the potential magnitude of impact for specific sources
30 of uncertainty on risk estimates (i.e., low, medium, or high) reflect EPA staff consensus on the
31 degree to which a particular source could produce a sufficient impact on risk estimates to
32 influence the interpretation of those estimates in the context of the Os NAAQS review.11 Sources
11 For example, if a particular source of uncertainty were more fully characterized (or if that source was
resolved, potentially reducing bias in a core risk estimate), could the estimate of incremental risk reduction in going
from the current to an alternative standard level change sufficiently to produce a different conclusion regarding the
magnitude of that risk reduction in the context of the O3 NAAQS review?
7-40
-------
1 classified as having a "low" impact would not be expected to impact the interpretation of risk
2 estimates in the context of the Os NAAQS review; sources classified as having a "medium"
3 impact have the potential to change the interpretation; and sources classified as "high" are likely
4 to influence the interpretation of risk in the context of the Os NAAQS review. Because this
5 classification of the potential magnitude of impact of sources of uncertainty is qualitative and not
6 informed directly by any type of analytical results, it is not possible to place a quantitative level
7 of impact on each of the categories. Therefore, the results of the qualitative analysis of
8 uncertainty have limited utility in informing consideration of overall confidence in the core risk
9 estimates and, instead, serve primarily as a means for guiding future research to reduce
10 uncertainty related to Os risk assessment.
11 As with the qualitative discussion of sources of variability included in the last section, the
12 characterization and relative ranking of sources of uncertainty addressed here is based on
13 consideration by EPA staff of information provided in previous Os NAAQS risk assessments
14 (particularly past sensitivity analyses), the results of risk modeling completed for the current 63
15 NAAQS risk assessment and information provided in the third draft O3 ISA as well as earlier O3
16 Criteria Documents. Where appropriate, in Table 7-6, we have included references to specific
7-41
-------
1
2 Table 7-6 Summary of Qualitative Uncertainty Analysis of Key Modeling Elements in the O3 NAAQS Risk Assessment.
Source
Description
Potential influence of
uncertainty on risk
estimates
Direction
Magnitude
Knowledge-
Base
uncertainty*
Comments
(KB: knowledge base, INF: influence of uncertainty on risk
estimates)
A. Characterizing
ambient Os
levels for study
populations using
the existing
ambient
monitoring
network
If the set of monitors used in a
particular urban study area to
characterize population
exposure as part of an ongoing
risk assessment do not match
the ambient monitoring data
used in the original
epidemiological study, then
uncertainty can be introduced
into the risk estimates.
Both
Low-
medium
Low-medium
KB and INF: In modeling risk, we used a study area definition for
each urban area based on the set of counties used in the Zanobetti and
Schwartz (2008b) study of short-term exposure-related mortality. In
those instances where other epidemiological studies used different
county definitions in specifying the set of Os monitors used in
characterizing uncertainty, then uncertainty may be introduced into
the risk assessment and it is challenging to evaluate the nature and
magnitude of the impact that that uncertainty would have on risk
estimates, given the complex interplay of factors associated with
mismatched monitoring networks (i.e., differences in the set of
monitors used in modeling risk and those used in the underlying
epidemiological study).
B. Characterizing
U.S. Background
C>3 levels
For this analysis, we have used
modeling to estimate U.S.
background levels for each
urban study area. Depending
on the nature of errors reflected
in that modeling, uncertainty (in
both directions) may be
introduced into the analysis.
Both
Low
Low
INF: Given that the risk assessment focuses primarily on the
reduction in risk associated with moving from recent conditions to
simulated just meeting the current standard, the impact of uncertainty
in U.S. background levels on the risk estimates is expected to be low,
since generally, both recent conditions and current standard Oj, levels
occur well above U.S. Background (for a particular day) and
consequently, consideration of U.S. background does not factor into
estimating the magnitude of deltas (risk reductions).
C. Characterizing
intra-urban
population
exposure in the
context of
epidemiology
studies linking
C>3 to specific
health effects
Exposure misclassification
within communities that is
associated with the use of
generalized population monitors
(which may miss important
patterns of exposure within
urban study areas) introduces
uncertainty into the effect
estimates obtained from
epidemiology studies.
Under
(generally)
Low-
medium
Medium
KB and INF: Despite the potential for substantial variability in Oj,
levels across monitors (particularly in larger urban areas with greater
variation in sources and topography such as L.A.), the ISA notes that
studies have tended to demonstrate that monitor selection has only a
limited effect on the association of short-term Oj, exposure with
health effects (see ISA section???). However, s noted here, this issue
could be more of a concern in larger urban areas which may exhibit
greater variation in Oj, levels due to diverse sources, topography and
patterns of commuting.
D. Statistical fit
Exposure measurement error
Both
Medium
Medium
INF: For short-term mortality and morbidity health endpoints, there is
7-42
-------
Source
Description
Potential influence of
uncertainty on risk
estimates
Direction
Magnitude
Knowledge-
Base
uncertainty*
Comments
(KB: knowledge base, INF: influence of uncertainty on risk
estimates)
ofthe C-R
functions
combined with other factors
(e.g., size of the effect itself,
sample size, control for
confounders) can effect the
overall level of confidence
associated with the fitting of
statistical effect-response
models in epidemiological
studies.
(short-term
health
endpoints)
greater uncertainty associated with the fit of models given the smaller
sample sizes often involved, difficulty in identifying the etiologically
relevant time period for short-term O3 exposure, and the fact that
models tend to be fitted to individual counties or urban areas (which
introduces the potential for varying degrees of confounding and
effects modification across the locations). These studies can also have
effects estimates that are not statistically significant. Note, however
that for this risk assessment, in modeling short-term mortality, we are
not relying on location-specific models. Instead, we are using city-
specific effects estimates derived using Bayesian techniques (these
combine national-scale models with local-scale models).
E. Shape of the
C-R functions
Uncertainty in predicting the
shape of the C-R function,
particularly in the lower
exposure regions which are
often the focus in O3 NAAQS
regulatory reviews.
Both
Medium
Low-medium
KB and INF: Studies reviewed in the O3 ISA that attempt to
characterize the shape of the O3 C-R curve along with possible
"thresholds" (i.e., O3 concentrations which must be exceeded in order
to elicit an observable health response) have indicated a generally
linear C-R function with no indication of a threshold (for analyses
that have examined 8-h max and 24-h avg O3 concentrations).
However, the ISA notes that there is less certainty in the shape of the
C-R curve at the lower end of the distribution of O3 concentrations
due to the low density of data in this range. Therefore, while there is
increased uncertainty in specifying the nature of the C-R function at
lower exposure levels, we do not believe that the risk drops to zero
outside of the range of O3 data used in the underlying
epidemiological study providing the C-R function. As discussed in
section 7.1.1, we are including risk estimates where we model
exposure down to a surrogate for the LML of the underlying
epidemiological study in order to evaluate the impact of modeling
risk over a range of exposures where we have greater confidence
(relative to modeling all the way down to zero O3).
F. Surrogate
LMLs used in
defining ranges
of increased
confidence in
estimating risk
Ideally, we would use LMLs
from epidemiological studies
supporting the C-R functions
used in modeling risk to
identify a range of O3
concentrations with greater
Both
Medium
Low-medium
INF: Because the surrogate LMLs are based on individual years not
matched to the analysis periods used in the epidemiological studies
underlying the C-F functions, there is uncertainty associated with use
of the surrogate LMLs. In addition, there is the potential that that way
the composite monitor distributions were designed (surrogate LMLs
are obtained from these distributions) may differ from the way air
7-43
-------
Source
Description
Potential influence of
uncertainty on risk
estimates
Direction Magnitude
Knowledge-
Base
uncertainty*
Comments
(KB: knowledge base, INF: influence of uncertainty on risk
estimates)
confidence in modeling risk
(i.e., only modeling risk
matching the range of data used
in fitting the C-R function).
However, data limitations
meant that we used surrogate
LMLs in place of the study-
specific LMLs (the surrogate
LMLs were obtained from the
composite monitor distributions
used in risk modeling - see
section 7.1.1).
quality data were used in the epidemiological studies - this would
add additional uncertainty into the use of the surrogate LMLs.
KB: we do not have comprehensive LML data form any of the
epidemiological studies at this time and therefore, are not able to
rigorously evaluate the degree to which the surrogate LMLs match
actual study-based LMLs.12
G. Addressing
co-pollutants
The inclusion or exclusion of
co-pollutants which may
confound, or in other ways,
affect the Oj, effect, introduces
uncertainty into the analysis.
Both
Low-
medium
Medium
KB and INF: The Os ISA notes that across studies, the potential
impact of PM indices on O3-mortality risk estimates tended to be
much smaller than the variation in O3-mortality risk estimates across
cities. This suggests that O3 effects are independent of the
relationship between Os and mortality. However, interpretation of the
potential confounding effects of PM on O3-mortality risk estimates
requires caution. This is because the PM-O3 correlation varies across
regions, due to the difference in PM components, complicating the
interpretation of the combined effect of PM on the relationship
between O3 and mortality. Additionally, the limited PM or PM
component datasets used as a result of the every-3rd- and 6th-day PM
sampling schedule instituted in most cities limits the overall sample
size employed to examine whether PM or one of its components
confounds the 03-mortality relationship (ISA section 2.5.4.5).
H. Specifying la
structure (short-
term exposure
studies)
There is uncertainty associated
with specifying the exact lag
structure to use in modeling
short-term exposure-related
mortality and respiratory -
Both
Low-
Medium
Low
KB and INF: The majority of studies examining different lag models
suggest that Oj, effects on mortality occur within a few days of
exposure. Similar, studies examining the impact of Os exposure on
respiratory-related morbidity endpoints suggests a rather immediate
response, within the first few days of C>3 exposure (see ISA section
12 We are in the process of evaluating descriptive statistics (including LMLs) reflecting data used in Zanobetti and Schwartz (2008b). However at the
time of the first draft REA, we were not yet in a position to use these data to complete a rigorous performance evaluation of the surrogate LMLs developed for
this (or other) health endpoints modeled in the analysis.
7-44
-------
Source
Description
Potential influence of
uncertainty on risk
estimates
Direction Magnitude
Knowledge-
Base
uncertainty*
Comments
(KB: knowledge base, INF: influence of uncertainty on risk
estimates)
related morbidity.
2.5.4.3). Consequently, while the exact nature of the ideal lag models
remains uncertain, generally, we are fairly confident that they would
be on the order of a day to a few days following exposure.
I. Using studies
from one
geographic area
to cover urban
areas outside of
the study area
In the case of Gent et al., 2003
(used in modeling asthma
exacerbations in Boston), we
are using C-R functions based
on an epidemiological study of
a region (northern Connecticut
and Springfield) that does not
encompass the actual urban
study area assessed for risk
(Boston).
Both
Medium
Low
INF: Factors related to O3 exposure including commuting patterns,
exercise levels etc may differ between the region reflected in the
epidemiological study and Boston. If these differences are great, then
applying the effect estimate from the epidemiological study to Boston
could be subject to considerable uncertainty and potential bias. We
have not conducted a more rigorous comparison of the two locations
with regard to attributes impacting O3 (including monitor levels) but
that may be undertaken as part of the second draft ERA in order to
increase our understanding of potential uncertainty associated with
this category of risk estimate.
J. Characterizing
baseline
incidence rates
Uncertainty can be introduced
into the characterization of
baseline incidence in a number
of different ways (e.g., error in
reporting incidence for specific
endpoints, mismatch between
the spatial scale in which the
baseline data were captured and
the level of the risk
assessment).
Both
Low-
medium
Low
INF: The degree of influence of this source of uncertainty on the risk
estimates likely varies with the health endpoint category under
consideration. There is no reason to believe that there are any
systematic biases in estimates of the baseline incidence data. The
influence on risk estimates that are expressed as incremental risk
reductions between alternative standards should be relatively
unaffected by this source of uncertainty.
KB: The county level baseline incidence and population estimates at
the county level were obtained from data bases where the relative
degree of uncertainty is low.
1
2
3
4
5
* Refers to the degree of uncertainty associated with our understanding of the phenomenon, in the context of assessing and characterizing its uncertainty
(specifically in the context of modeling PM risk)
7-45
-------
1 sources of information considered in arriving at a ranking and classification for a particular
2 source of uncertainty.
3 7.5 URBAN STUDY AREA RESULTS
4 This section presents and discusses risk estimates generated for the set of 12 urban study
5 areas, including estimates generated to characterize recent Os conditions as well as estimates
6 generated after simulated just meeting the current Os standard level in each urban study area.
7 Risk estimates for alternative standard levels will be generated as part of the second draft
8 analysis.
9 A number of details regarding these risk estimates should be kept in mind when
10 reviewing the estimates presented in this section:
11 • All risk estimates presented represent core (higher confidence) estimates -
12 sensitivity analyses will be completed for the second draft analysis: As discussed
13 in section 7.1.1, the risk estimates generated for the first draft analysis focus on an
14 array of core (higher confidence) analyses. A supporting set of comprehensive
15 sensitivity analyses to help interpret overall confidence in the core estimates will be
16 included in the second draft analysis. However, specifically in the case of short-term
17 exposure-related morbidity, the array of core analyses includes coverage for a variety
18 of design elements (including multi-/single-pollutant models and lag structures) and
19 therefore, the array of core risk estimates does inform our consideration of the impact
20 that these design elements has on risk estimates for this category of morbidity
21 endpoints.
22 • Estimates are presented for two simulation years (2007 and 2009): Each
23 simulation year represents the middle year of a 3 year attainment period (2006-2008
24 and 2008-2010, respectively). The two attainment periods were selected to provide
25 coverage for generally lower and higher Os periods (i.e., 2006-2008 being relatively
26 higher in general terms compared with the 2008-2010 period although this does not
27 hold across all 12 urban study areas).
28 • All estimates reflect short-term exposure-related endpoints: Analysis of evidence
29 presented in the Os ISA combined with consideration for the availability of data
30 required to model specific health endpoints resulted in our designing the first draft
31 REA to cover the health endpoints listed at the beginning of this section which are all
32 related to short-term Os exposure. We also completed a review of evidence
33 supporting modeling of long-term exposure-related mortality and morbidity.
34 Treatment of those endpoints categories as planned for the second Draft analysis is
35 discussed below in section 7.7.3.
36 • Short-term exposure-related mortality estimates are generated for all 12 urban
37 study areas, while most morbidity estimates (depending on the specific health
38 endpoint) are generated for only a subset of urban study areas: All mortality
39 estimates are generated using Bayes-adjusted city-specific effect estimates obtained
40 form Bell et al., (2004) (for all-cause mortality only) and Zanobetti and Schwartz
7-46
-------
1 (2008b) (for all-cause, respiratory and cardiovascular-related mortality). For
2 morbidity endpoints, coverage for the urban study areas differed depending on the
3 specific endpoint with (a) ER visits evaluated for Atlanta and New York City, (b) HA
4 evaluated for all 12 urban study areas with additional coverage for New York, Detroit
5 and LA and (c) asthma exacerbations evaluated for Boston.
6 • For short-term exposure-related mortality, we include two types of risk
7 estimates for each scenario which, when considered together, inform
8 consideration of uncertainty related to application of the C-R functions at low O3
9 levels: For short-term exposure-related mortality, we include (a) estimates of risk
10 reflecting modeling of exposure down to zero Os and (b) higher confidence estimates
11 of risk reflecting exposures modeled down to a surrogate for the LML used in fitting
12 the C-R function (see 7.1.1). While risk modeled down to the LML has greater
13 overall confidence since we are modeling exposure reflected in the fitting of the C-R
14 function, estimates bounded by the LML are also likely biased low since they do not
15 include exposures between the LML and zero O3. By contrast, estimates of risk all
16 the way to zero 63 benefit from considering the full range of exposure, but also
17 incorporate a range of exposure associated with reduced confidence in modeling risk
18 (i.e., Os levels below those used in fitting the C-R function used in modeling risk).
19 When considered together these two types of risk estimates inform consideration of
20 uncertainty related to application of the C-R function at low 63 levels. It is important
21 to point out that only the LML-based risk estimates were generated for the short-term
22 exposure-related morbidity endpoints (these did not include estimates based on
23 modeling exposure down to zero O3).
24 There are several categories of risk metrics generated for the mortality and morbidity
25 endpoints modeled in this analysis. These metrics are described below (these descriptions are
26 separated into mortality-related tables and morbidity-related tables):
27
28 I. Tables presenting mortality estimates
29
30 • Heat map tables for mortality illustrating distribution of mortality across daily
31 Os levels (Tables 7-7 through 7-10): The heat map tables illustrate the distribution
32 of estimated Os-related deaths across daily Os levels for each city. The color gradient
33 reflects the distribution of mortality across the range of daily 8-hour ozone levels with
34 colors ranging from green (low) to red (high). The color gradients are a visual tool to
35 explore trends in mortality counts across daily Os levels and between cities. As an
36 example, with Table 7-7 (which presents recent conditions mortality risk estimates for
37 2007 based on Zanobetti and Schwartz, 2008b C-R functions), the value of 72 in the
38 "New York" row and "60-65" column represents the fact that 72 of the total of 708
39 deaths estimated for New York city occurred on days with Os levels between 60 and
40 65 ppb. Similarly, in that same table, we see that only 13 of the estimated deaths in
41 New York City occurred on days with 8hr mean Oj levels between 20 and 25 ppb.
42 The heat map tables allow us to evaluate which days (in terms of 63 levels) are
43 associated with the majority of estimated Os-related deaths. When we compare heat
7-47
-------
1 map tables between recent conditions and simulating just meeting the current
2 standard, we can look at how that distribution of estimated O3-related deaths across
3 daily O3 levels shifts (i.e., the entire distribution shifts to the left, reflecting the fact
4 that the distribution of daily O3 levels is reduced when we simulate just meeting the
5 current standard). Separate sets of heat map tables were generated using C-R
6 functions based on Bell et al., (2004) and Zanobetti and Schwartz (2008b). The heat-
7 map tables were only generated for the 2007 simulation year, given that the general
8 pattern displayed in these tables would also hold for 2009. In addition, heat-map
9 tables were only generated for all-cause mortality - the patterns displayed in the table
10 would hold for other mortality categories modeled in the analysis. Estimates
11 presented in the heat-map tables reflect application of the LMLs (i.e., risks were
12 modeled down to LML, and not down to zero).
13 • Tables presenting estimates of O3-related mortality with consideration for
14 ranges of increased confidence defined based on the composite monitor LMLs
15 (Tables 7-11 Through 7-14): As discussed in sections 7.1.1 and 7.1.2, rather than
16 incorporating a biological threshold into modeling risk, we have defined ranges of
17 increased confidence corresponding to levels of O3 similar to those used in the
18 epidemiological studies providing the C-R functions used in the analysis. However,
19 as noted in those earlier sections, due to data limitations we used statistics obtained
20 from the set of composite monitor values used in modeling risk as surrogates for
21 statistics that would have come from the actual epidemiological studies. Specifically,
22 we estimated risks down to LMLs from the composite monitor data sets. Estimates of
23 risk presented in these tables include estimates modeled all the way down to zero to
24 establish a baseline of the highest potential estimated risk. Estimates presented in
25 Tables 7-11 through 7-14, reflect all-cause mortality and include 95th percentile
26 confidence intervals representing uncertainty associated with the statistical fit of the
27 effect estimates used. Estimates are presented based both on Bell et al., (2004) and
28 Zanobetti and Schwartz (2008b) C-R functions. Note, that 95th% confidence intervals
29 are not presented for the delta (risk reduction) estimates since these were calculated
30 off of point estimates (for the recent conditions and current standard level) and were
31 not based on separate model runs for the delta O3 levels. Estimates presented in these
32 tables allow for consideration for the pattern of risk reduction (in incidence) in going
33 from recent conditions to just meeting the current standard level and how that pattern
34 varies across urban study areas. Estimates in these tables also illustrate how risk
35 changes when consideration is given to different levels of confidence about risks
36 attributable to O3 concentrations at the lower end of the observed O3 data used in the
37 underlying epidemiology studies.
38 • Tables comparing cause-specific mortality for the recent conditions (2007)
39 scenario: Table 7-15 presents estimates of cause-specific mortality (all-cause,
40 respiratory and cardiovascular) for the 2007 simulation year based on C-R functions
41 obtained from Zanobetti and Schwartz (2008b). These tables include consideration
42 for the range of increased confidence defined using the LMLs as cutoffs for modeling
43 risk. The estimates presented in these tables allow consideration for differences in the
44 magnitude of mortality risk associated with different mortality categories.
7-48
-------
1 • Tables presenting estimates of the percent of total mortality attributable to Os:
2 Tables 7-16 through 7-19 present estimates of the percent of total (all-cause)
3 mortality attributable to O3 for the recent conditions and simulation of the current
4 standard scenarios and for the delta (risk reduction) between these two scenarios.
5 Estimates presented in these tables include those generated with consideration for
6 ranges of increased confidence based on the composite monitor LMLs, as well as
7 estimates of risk based on modeling all the way to zero O3. Results are presented
8 based on estimates of mortality derived using C-R functions obtained both from Bell
9 et al., (2004) and Zanobetti and Schwartz (2008b). Estimates presented in these tables
10 allow for consideration for the pattern of risk reduction (in terms of the percent of
11 total mortality) in going from recent conditions to just meeting the current standard
12 level and how that pattern varies across urban study areas. Estimates in these tables
13 also illustrate how risk changes when consideration is given to different levels of
14 confidence about risks attributable to O3 concentrations at the lower end of the
15 observed O3 data used in the underlying epidemiology studies.
16 • Tables presenting estimates of the percent reduction in ozone-related mortality
17 incidence: Table 7-20 presents estimates of the reduction in ozone-related mortality
18 incidence in going from recent conditions to the simulation of the current ozone
19 standard level. This table includes consideration for the range of increased confidence
20 defined based on composite monitor LMLs, as well as estimates of risk based on
21 modeling all the way to zero O3. Results are presented based on estimates of
22 mortality derived using C-R functions obtained both from Bell et al., (2004) and
23 Zanobetti and Schwartz (2008b). Estimates presented in these tables allow
24 consideration for how the pattern of reductions in ozone-related mortality (in going
25 from recent conditions to meeting the current standard) varies across urban study
26 areas. Estimates in these tables also illustrate how risk changes when consideration is
27 given to different levels of confidence about risks attributable to O3 concentrations at
28 the lower end of the observed O3 data used in the underlying epidemiology studies.
29 II. Tables presenting morbidity estimates
30
31 • Table summarizing risk estimates for short-term exposure-related ER visits (for
32 respiratory symptoms including asthma): Table 7-21 presents estimates of the
33 incidence of ER visits for respiratory symptoms and asthma) specifically for New
34 York City and Atlanta based on C-R functions obtained from several epidemiological
35 studies. The C-R functions available for modeling this category of health effect
36 endpoints included consideration for a number of design elements (copollutants and
37 lag structure). Therefore, while the set of risk estimates presented in these tables does
38 collectively represent the core simulation for this endpoint, consideration for different
39 design elements also allows us to evaluate their potential impact on core risk
40 estimates. Risk estimates presented in these tables include: (a) point estimates and
41 95th percentile estimates for O3-attributable incidence, (b) percent of baseline
42 incidence (the increment of total ER attributable to O3 exposure), (c) risk reductions
43 (deltas) in both O3-related incidence and the fraction of total incidence attributable to
44 O3 and (d) reduction in O3-related mortality.
7-49
-------
1 • Tables summarizing risk estimates for short-term exposure-related HA visits
2 (for respiratory symptoms including asthma): Tables 7-22 and 7-23 present
3 estimates of the incidence of HA (for respiratory symptoms, chronic lung disease and
4 asthma). Risk estimates are generated for a subset of the urban study areas for some
5 of the health endpoints (e.g., New York City for HA [chronic lung disease and
6 asthma]), while HA (respiratory-related) estimates cover all 12 urban study areas.
7 These estimates include the same mix of risk metrics and other parameters described
8 for the ER-visit estimates (see above).
9 • Table summarizing risk estimates for short-term exposure-related asthma
10 exacerbation: Table 7-24 presents estimates of the incidence of asthma exacerbations
11 (including estimates for a range of symptoms) for Boston (the only urban study area
12 with C-R functions supporting modeling for this endpoint). Risk estimates presented
13 in Table 7-24 include consideration for a number of modeling elements (Os metrics,
14 lag structure and copollutants). The array of risk estimates presented in these tables
15 collectively represents the core simulation for this endpoint. Consideration for
16 different design elements allows us to evaluate their potential impact on core risk
17 estimates. As with the other short-term exposure-related morbidity risk estimates,
18 estimates presented in this tables include: (a) point estimates and 95th percentile
19 estimates for (Vattributable incidence, (b) percent of baseline incidence (the
20 increment of total ER attributable to Os exposure), (c) risk reductions (deltas) in both
21 Os-related incidence and the fraction of total incidence attributable to Os and (d)
22 reduction in O3-related mortality.
23 In reviewing the risk estimates generated for the first draft analysis we have focused on
24 developing a set of key observations reflecting consideration for goals originally set out for the
25 risk assessments in the Scope and Methods Plan (U.S. EPA, 2011). These goals included:
26 • Provide estimates of the potential magnitude of premature mortality and/or selected
27 morbidity health effects associated with recent conditions and with the simulated just
28 meeting just meeting the current suite of Os standards and any alternative standards
29 that might be considered in selected urban study areas (note, alternative standards will
30 be evaluated in the second Draft analysis).
31 • Develop a better understanding of the influence of various inputs and assumptions on
32 the risk estimates to more clearly differentiate alternative standards that might be
33 considered including potential impacts on various sensitive populations.
34 • Gain insights into the distribution of risks and patterns of risk reduction and
35 uncertainties in those risk estimates.
36 Typically, the last two bullets are addressed primarily through sensitivity analysis runs
37 that provide additional perspective on the impact of varying modeling elements (including
38 aspects of C-R function specification) on risk estimates. These sensitivity analyses will be
39 included in the second draft REA- see section 7.7.1. Therefore, the discussion presented below
40 focuses primarily on characterizing the magnitude of risk and risk reduction associated with the
7-50
-------
Os scenarios modeled and also provides some ally insights on the distribution of risks and
patterns of risk reduction.
7-51
-------
1
2
3
4
Table 7-7 Heat Map Table: Short-Term O3 Exposure-Related All-Cause Mortality - Recent Conditions (2007) (Zanobetti and
Schwartz, 2008b C-R functions) (illustrates distribution of O3-related all-cause mortality across distribution of daily 8hr mean O3 levels for each
urban study area - colors in cells reflect size of mortality estimate)
Study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Daily Shr Mean Ozone Level (ppb)
0-5
0
0
0
0
0
0
0
0
0
0
0
0
5-10
0
0
0
0
0
0
0
0
0
0
0
0
10-15
0
0
0
0
0
0
0
0
2
0
0
0
15-20
0
0
0
1
0
0
1
0
1
0
0
0
20-25
0
0
1
1
0
0
2
0
13
1
0
0
25-30
0
1
7
2
0
4
1
0
41
2
0
0
30-35
1
1
5
7
0
5
2
0
26
4
0
1
35-40
1
6
13
S
0
10
3
1
95
4
1
3
40-45
1
11
9
11
1
17
2
5
102
8
3
6
45-50
2
12
17
13
1
20
1
10
61
13
5
S
50-55
5
12
15
9
3
17
2
16
68
11
5
7
55-60
7
11
15
7
2
12
2
27
33
8
3
10
60-65
7
15
12
5
1
9
0
12
72
14
6
10
65-70
8
4
11
10
0
15
1
11
117
7
3
10
70-75
5
4
3
2
0
S
1
7
29
5
1
7
>75
19
3
14
3
0
19
0
6
47
9
3
24
Total
56
84
123
78
10
135
20
96
708
87
30
86
5
6
1
8
Table 7-8 Heat Map Table: Short-Term Os Exposure-Related All-Cause Mortality - Simulation of Meeting the Current
Standard (2007) (Zanobetti and Schwartz, 2008b C-R functions) (illustrates distribution of O3-related all-cause mortality across
distribution of daily Shr mean O3 levels for each urban study area - colors in cells reflect size of mortality estimate)
Study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Daily Shr Mean Ozone Level (ppb)
0-5
0
0
0
0
0
0
0
0
0
0
0
0
5-10
0
0
0
0
0
0
0
0
0
0
0
0
10-15
0
0
0
0
0
0
0
0
3
0
0
0
15-20
0
0
0
1
0
0
1
0
4
0
0
0'
20-25
0
1
3
2
0
0
2
0
17
2
0
0
25-30
0
1
5
2
0
5
1
0
35
2
0
0
30-35
1
3
7
9
0
5
2
0
64
4
0
2
35-40
1
12
15
10
1
11
3
6
113
7
2
6
40-45
3
12
11
12
1
24
1
14
44
16
4
6
45-50
5
12
17
10
3
11
2
16
102
10
5
6
50-55
7
16
15
8
2
17
2
9
39
7
5
12
55-60
a
7
11
7
1
12
1
5
130
13
2
9
60-65
9
4
12
6
0
13
1
0
48
4
1
10
65-70
4
3
5
3
0
7
0
0
13
3
0
10
70-75
2
0
3
1
0
11
0
0
14
2
1
10
>75
1
0
7
0
0
6
0
0
0
3
0
3
Total
42
71
110
72
9
122
17
50
626
72
20
73
Delta
14
13
13
7
2
14
3
46
81
14
11
13
9
10
7-52
-------
1
2
3
Table 7-9 Heat Map Table: Short-Term O3 Exposure-Related All-Cause Mortality - Recent Conditions (2007) (Bell et al, 2004
C-R functions) (illustrates distribution of O3-related all-cause mortality across distribution of daily 8hr max O3 levels for each urban study area -
colors in cells reflect size of mortality estimate)
Study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Daily Shr Max Ozone Level (ppb)
0-5
0
0
0
0
0
0
0
0
0
0
0
0
5-10
0
0
0
0
0
0
0
0
0
0
0
0
10-15
0
0
0
0
0
0
2
1
0
0
0
0
15-20
0
0
0
0
0
0
7
10
15
0
1
0
20-25
1
2
6
1
1
1
IS
26
22
1
3
1
25-30
3
4
20
5
1
2
23
41
60
4
5
3
30 35
10
6
26
7
2
6
34
69
69
6
9
8
35-40
13
14
32
13
3
9
30
66
70
10
14
14
40-45
17
14
43
13
5
13
26
99
99
11
14
18
45-50
22
14
39
15
6
17
28
87
60
11
19
16
50-55
42
14
23
14
6
5
21
103
49
12
17
24
55-60
51
10
33
9
6
10
15
88
40
11
8
21
60-65
39
16
13
6
2
6
9
46
50
11
8
24
65-70
40
4
26
10
1
6
19
40
73
8
3
10
70-75
34
5
19
7
0
5
12
24
27
7
4
9
>75
52
3
26
7
0
13
2
27
23
7
4
25
Total
323
106
307
109
32
94
244
729
658
98
110
174
4
5
6
7
Table 7-10 Heat Map Table: Short-Term Os Exposure-Related All-Cause Mortality - Simulation of Meeting the Current
Standard (2007) (Bell et al., 2004 C-R functions) (illustrates distribution of O3-related all-cause mortality across distribution of daily Shr
max O3 levels for each urban study area - colors in cells reflect size of mortality estimate)
Study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Daily Shr Max Ozone Level (ppb)
0-5
0
0
0
0
0
0
0
0
0
0
0
0
5-10
0
0
0
0
0
0
0
0
0
0
0
0
10-15
0
0
0
0
0
0
2
2
1
0
0
0
15-20
0
0
1
1
0
0
8
17
13
0
1
0
20-25
2
3
8
2
1
1
22
35
31
2
3
2
25 -JO
5
4
23
5
1
3
24
64
69
5
7
4
30-35
12
10
24
11
3
7
37
70
76
8
15
10
35-40
IS
15
48
13
3
10
31
119
103
12
14
20
40-45
21
15
26
16
7
17
28
113
55
12
21
16
45-50
57
15
39
14
7
§
21
81
63
12
13
23
50-55
43
14
32
10
5
9
12
44
50
8
8
24
55-60
45
8
14
9
2
7
IS
15
66
11
5
21
60-65
29
5
26
9
0
7
11
7
44
5
3
12
65-70
16
2
24
5
0
6
1
0
0
2
0
12
70-75
6
0
7
2
0
6
0
0
13
1
1
10
>75
4
0
11
2
0
4
0
0
0
2
0
4
Total
260
90
282
98
30
86
217
567
585
82
90
157
Delta
63
16
26
11
J
8
27
162
73
16
20
17
7-53
-------
1
2
3
Table 7-11 Short-Term O3 Exposure-Related All Cause Mortality Incidence (2007)
(Zanobetti and Schwartz, 2008b C-R Functions) (no cutoff"and LML cutoff columns present
Os-attributable risks modeled down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone-Exposure Related All-Cause Mortality (2007) (Zanobetti and Schwartz,
2008b)
Recent conditions
no cutoff
94
(-S3 -269)
117
(-23 - 252}
223
(13-426)
92
(-15-196)
IS
(-23 - 58)
226
(82-365}
29
(-63-11S)
227
(-121-566)
931
(544-1310)
116
(2-227)
74
(-26 - 170}
143
(-29 -SOS)
LML cutoff
56
(-51 - 160)
84
(-17 - 182)
123
(7-236)
78
(-13-167)
10
(-13-33)
135
(49 - 220}
20
(-43 - 81)
96
(-51 - 241)
70S
(412 - 997)
87
(1-170)
30
(-10-71)
36
(-17-186)
Current standard
no cutoff
SO
(-74-230)
104
(-21-225)
209
(12-401)
85
(-14-182)
16
(-21-53)
212
(77-344)
26
(-57-107)
ISO
(-96-451)
849
(495 - 1197)
102
(1-200)
63
(-22-146)
130
(-27-281)
LML cutoff
42
(-39 - 121)
71
(-14-155)
110
(6-211)
72
(-12-153)
9
(-11-28)
122
(44-198)
17
(-37-70)
50
(-27-126)
626
(365-884)
72
(1-142)
20
(-7-45)
73
(-15 - 159)
Delta (risk reduction)
no cutoff
14
LML cutoff
14
NA
13
13
NA
14
13
NA
7
6
NA
2
1
NA
14
13
NA
3
3
NA
47
46
NA
82
82
NA
14
15
NA
11
10
NA
13
13
NA
4
5
7-54
-------
1 Table 7-12 Short-Term O3 Exposure-Related All Cause Mortality Incidence (2009)
2 (Zanobetti and Schwartz, 2008b C-R Functions) (no cutoff"and LML cutoff columns present
3 Os-attributable risks modeled down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone-Exposure Related All-Cause Mortality (2007) (Zanobetti and Schwartz,
Recent conditions
no cutoff
77
(-71 - 221)
113
(-22-245)
135
(10-354)
31
(-14-174)
17
(-22-53)
173
(64-2SS)
32
(-70-132)
215
(-115-537)
835
(487 - 1176)
92
(1-1SO)
75
(-26 - 173)
108
(-22 - 234)
LML cutoff
43
(-40-125)
56
(-11-121)
93
(5 -189)
49
(-8 - 104)
9
(-12 - 29)
128
(46 - 207)
19
(-41 - 78)
123
(-66-309)
579
(337-817)
60
(1-117)
32
(-11-73)
53
(-11-116)
Current standard
no cutoff
73
(-68 - 211)
103
(-20 - 222)
ISO
(10-345)
79
(-13-168)
16
(-21 - 52}
173
(64-288)
30
(-65 - 122)
175
(-93-438)
777
(453 - 1095)
86
(1-169)
64
(-22-147)
105
(-21 - 228)
LML cutoff
40
(-37-115)
45
(-9-98)
93
(5 - 179)
46
(-8-99)
8
(-11-27)
127
(46 - 207)
17
(-36-69)
83
(-44-210)
521
(303 - 736)
54
(1-106)
21
(-7-48)
50
(-10 - 110)
Delta (risk reduction)
no cutoff
4
LML cutoff
3
MA
10
11
MA
5
5
\A
2
3
\A
1
1
MA
0
1
MA
2
2
NIA
40
40
\A
58
58
\A
6
6
\A
11
11
NA
3
3
MA
7-55
-------
1
2
3
Table 7-13 Short-Term O3 Exposure-Related All Cause Mortality Incidence (2007) (Bell et
al., 2004 C-R Functions) (no cutoff and LML cutoff columns present O3-attributable risks modeled
down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone-Exposure Related All-Cause Mortality (2007) (Zanobetti and Schwartz,
Recent conditions
no cutoff
479
(181-769)
153
(-70 - 370J
430
(105-748)
151
(-.59 - 355)
36
(-21-92)
132
(-71-330)
297
(-102-6S7)
950
(-379 - 2243)
901
(-168 - 1940)
139
(-97-368)
163
(-63 - 382)
210
(-106-516)
LML cutoff
323
(122-520)
106
(-48-257)
307
(75-535)
109
(-43-256)
32
(-19 - S3)
94
(-51-237)
244
(-83-564)
729
(-290-1725)
658
(-122 - 1421)
98
(-68-260)
110
(-42 - 259)
174
(-SS-429)
Current standard
no cutoff
415
(157-668)
137
(-63 - 332)
404
(98-704)
140
(-55-330)
33
(-19-85)
124
(-67-309)
270
(-92-625)
786
(-313 - 1862)
827
(-154-1784)
123
(-86-325)
142
(-55 - 335)
193
(-97-474)
LML cutoff
260
(98-419)
90
(-41-219)
282
(68-491)
98
(-38-231)
30
(-17-76)
86
(-46-216)
217
(-74-503)
567
(-225-1346)
585
(-109 - 1265)
82
(-57-217)
90
(-34-212)
157
(-79 - 387)
Delta (risk reduction)
no cutoff
64
LML cutoff
63
MA
16
16
MA
26
25
\A
11
11
\A
3
2
MA
8
8
NA
27
27
NIA
164
162
NA
74
73
NA
16
16
\A
21
20
NA
17
17
MA
4
5
7-56
-------
1 Table 7-14 Short-Term O3 Exposure-Related All Cause Mortality Incidence (2009) (Bell et
2 al., 2004 C-R Functions) (no cutoff and LML cutoff columns present O3-attributable risks modeled
3 down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone-Exposure Related All-Cause Mortality (2007) (Zanobetti and Schwartz,
Recent conditions
no cutoff
3S1
(144-614)
145
(-66-351J
378
(92-658)
129
(-.50 - 303)
35
(-20-88)
110
(-60 - 276)
292
(-100-674)
976
(-3S9 - 2303)
820
(-153 - 1767}
108
(-76 - 2S7)
162
(-62-380)
168
(-85 - 414)
LML cutoff
332
(125-534)
112
(-51-272)
259
(63-453)
79
(-31-187)
22
(-13-57)
72
(-39-182)
231
(-79-534)
781
(-311-1847)
630
(-117-1362)
SI
(-57-216)
140
(-54-330)
13 S
(-69 - 340}
Current standard
no cutoff
364
(138-586)
132
(-60-320)
369
(90-642)
125
(-49-295)
34
(-20-86)
110
(-60-276)
272
(-93 - 628}
821
(-326-1942)
764
(-142-1649)
102
(-71-270)
141
(-54-332)
164
(-83 - 404)
LML cutoff
315
(119 - 507)
99
(-45-242)
250
(61-437)
75
(-29 - 179)
21
(-12 - 54)
72
(-39-182)
211
(-72-489)
628
(-249-1488)
576
(-107-1245)
75
(-52-199)
120
(-46 - 282)
134
(-67-330)
Delta (risk reduction)
no cutoff
17
LML cutoff
17
\A
13
13
MA
9
9
\A
4
4
MA
1
1
MA
0
0
MA
20
20
NIA
155
153
\A
56
54
\A
6
6
\A
21
20
NA
4
4
MA
7-57
-------
1 Table 7-15 Pathway-Specific Mortality Incidence (2007 recent conditions) (Zanobetti and
2 Schwartz, 2008b, C-R functions) (no cutoff"and LML cutoff columns present Os-attributable risks modeled
3 down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone Exposure-Related Mortality (recent conditions - 2007) (Zanobetti and
Schwartz , 200Sb)
Total
no cutoff
94
117
223
92
IS
226
29
227
931
116
74
143
LML cutoff
56
84
123
78
10
135
20
96
70S
87
30
36
Respiratory
no cutoff
27
19
36
11
5
17
6
33
64
11
11
20
LML cutoff
16
13
20
10
3
10
4
14
49
9
5
12
Cardiovascular
no cutoff
52
SI
77
45
11
119
33
94
451
57
32
80
LML cutoff
31
58
42
38
6
72
22
40
343
42
13
48
4
5
7
8
7-58
-------
J
4
Table 7-16 Percent of Total All-Cause Mortality Attributable to O3 (2007) (Zanobetti and
Schwartz, 2008b C-R functions) (no cutoff'and LML cutoff columns present O3-attributable risks
modeled down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone Exposure-Related All Cause Mortality - PERCENT of total baseline (2007) (Zanobetti and
Schwartz, 2008b C-R functions)
Recent conditions
no cutoff
1.7
2.4
2.9
2 5
17
4.9
0.5
1.5
4.6
3.0
2 8
3.0
LML cutoff
1.0
1.8
1.6
2 2
1C
3.0
0.4
0.6
3.5
2.2
1.2
U
Current standard
no cutoff
1.5
2.2
2.7
2.4
1.6
4.6
0.5
1.2
4.2
2.6
2.4
2,7
LML cutoff
0.8
1.5
1.4
2 C
C.S
2.7
0.3
0.3
3.1
1.9
O.S
1.6
Delta (reduction)
no cutoff
0.3
0.3
0.2
C 2
c :
0.3
0.05
0.3
0.4
0.4
0.4
C B
LML cutoff
0.3
0.3
0.2
0.2
0.1
0.3
0.05
0.3
0.4
0.4
M
C.3
5
6
7
Table 7-17 Percent of Total All-Cause Mortality Attributable to O3 (2009) (Zanobetti and
Schwartz, 2008b C-R functions) (no cutoff'and LML cutoff columns present O3-attributable risks
modeled down to zero O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone Exposure-Related All Cause Mortality - PERCENT of total baseline (2007) (Zanobetti and
Schwartz, 2008b C-R functions)
Recent conditions
no cutoff
1.4
2.4
2.5
2.4
1.7
4 :
C 6
1.4
4.3
2.5
2.9
2 3
LML cutoff
0.8
1.2
1.3
1.4
0.9
3.0
C.,4
O.S
3.0
1.6
1.2
1.2
Current standard
no cutoff
1.4
2.2
2.4
2.3
1.6
4.1
C.6
1.2
4.0
2.3
2.5
2.3
LML cutoff
0.7
1.0
1.3
1.4
0.9
3.0
B.3
0.6
2.7
1.5
0.8
1.1
Delta (reduction)
no cutoff
0.1
0.2
0.1
0.1
0.05
0.003
C C-
0.3
0.3
0.2
0.4
0.1
LML cutoff
0.1
0.2
0.1
0.1
G.05
C CCi
c ex
0.3
0.3
0.2
0.4
0.1
10
7-59
-------
1
2
3
Table 7-18 Percent of Total All-Cause Mortality Attributable to O3 (2007) (Bell et al., 2004
C-R functions) (no cutoff and LML cutoff columns present O3-attributable risks modeled down to zero O3 and
the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone Exposure-Related All Cause Mortality- PERCENT of total baseline
(2007) (Bell et al., 2004 C-R functions)
Recent conditions
no cutoff
3.7
1.5
2.9
1.9
1.7
1.5
1.5
1.7
2.0
1.7
1.7
2.0
LML cutoff
2.5
1.0
2.1
1.4
1,5
1.1
1.3
1.3
1.5
1.2
1.2
1.7
Current standard
no cutoff
3.2
1.3
2.7
l.S
1.5
1,4
1.4
1.4
1.8
1.5
1.5
1.9
LML cutoff
2.0
0.9
1.9
1.2
1.4
1.0
1.1
1.0
1.3
1.0
1.0
1.5
Delta (reduction)
no cutoff
0.5
0.2
0.2
0.1
0.1
0.1
0.1
0.3
0.2
0.2
0.2
0.2
LML cutoff
0.5
0.2
0.2
0.1
0.1
0.1
0.1
0.3
0.2
0.2
0.2
0.2
4
5
6
7
Table 7-19 Percent of Total All-Cause Mortality Attributable to O3 (2009) (Bell et al., 2004
C-R functions) (no cutoff and LML cutoff columns present O3-attributable risks modeled down to zero
O3 and the surrogate LML, respectively)
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Ozone Exposure-Related All Cause Mortality- PERCENT of total baseline
(2007) (Zanobetti and Schwartz, 2008b C-R functions)
Recent conditions
no cutoff
2.9
1.4
2.7
1.7
1.7
1.4
1.5
1.7
1.9
1.4
1.7
1.7
LML cutoff
2.6
1.1
l.S
1.1
1.1
0.9
1.2
1.4
1.5
1.1
1.5
1.4
Current standard
no cutoff
2,8
1.3
2.6
1.7
1.7
1.4
1.4
1,5
1.8
1.3
1.5
1.6
LML cutoff
2.4
1.0
l.S
1.0
1.1
0.9
1.1
1.1
1.3
1.0
1.3
1.4
Delta (reduction)
no cutoff
0.1
0.1
0.1
0.05
0.05
0.0006
0.1
0.3
0.1
0.1
0.2
0.04
LML cutoff
0.1
0.1
0.1
0.05
0.05
0.0005
0.1
0.3
0.1
0.1
0.2
0.04
7-60
-------
1 Table 7-20 Percent Reduction in Ozone-Attributable Short-Term Exposure-Related
2 Mortality (no cutoff and LML cutoff columns present O3-attributable risks modeled down to zero O3 and
3 the surrogate LML, respectively)
Urban study area
Percent Reduction in Ozone Exposure-Related All-Cause Mortality
Zanobetti and Scnwartz 2008b-
based C-R functions
no cutoff
LML cutoff
Bell et al., 2004-based C-R
functions
no cutoff
LML cutoff
2007 Simulation Year
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
15%
11%
6%
7%
8%
6%
9%
20%
9%
12%
15%
9%
25%
15%
11%
8%
15%
10%
13%
48%
11%
16%
36%
15%
13%
10%
6%
7%
8%
6%
9%
17%
8%
12%
12%
8%
20%
15%
S%
10%
9%
9%
11%
22%
11%
17%
18%
10%
2009 Simulation Year
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
5%
9%
3%
3%
3%
0.08%
7%
19%
7%
6%
15%
3%
8%
19%
5%
5%
5%
0.11%
12%
32%
10%
9%
35%
5%
4%
9%
2%
3%
3%
0.04%
7%
16%
7%
6%
13%
3%
5%
11%
4%
4%
4%
0.06%
8%
20%
9%
S%
15%
3%
7-61
-------
1 Table 7-21 Short-Term Ozone Exposure-Related Morbidity (ER visits)
Urban study area (endpoint)
Study author
Effect
estimator
differentiators
Recent conditions
point
estimate
95th Confidence
Interval
2.5
97.5
%of
baseline
Simulation of meeting current standard
point
estimate
95th Confidence
Interval
2.5
97.5
%of
baseline
Delta (risk reduction)
point
estimate
Wof
baseline
% reduction in
ozone-related
morbidity
2007 Simulation
Atlanta GA
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
New York
ER visits (asthma)
ER visits (asthma)
ER visits (asthma)
ER visits (asthma)
ER visits (astrr'iu
Tolbert
Tolbert
Tolbert
Tolbert
Tolbert
Darrow
Strickland
Strickland
Ito
ltd
Ito
Ito
Ito
CO
N02
PM10
PM10, NIO2
Darrow
distlagO-7
avg day lag 0-2
PM2.5
NO2
CO
SO2
5,054
4,498
4,066
3,190
3,080
2,728
5,978
3,522
10,232
7,974
6,572
10,818
8,233
3,496
2,760
2,147
1,125
1,010
1,657
4,248
1,922
6,951
4,270
2,939
7,630
4,766
-
-
-
-
-
-
-
-
-
-
-
6,586
6,202
5,944
5,209
5,103
3,787
7,603
5,037
13,312
11,433
9,972
13,814
11,483
4.5
4.0
3.6
2.9
2.8
2.4
15.6
9.2
13.5
10.5
8.6
14.2
10.8
4,076
3,625
3,275
2,567
2,478
2,194
4,894
2,858
9,199
7,149
5,881
9,733
7,383
2,814
2,220
1,726
903
810
1,331
3,455
1,551
6,223
3,810
2,619
6,837
4,256
-
-
-
-
-
-
-
-
-
-
-
-
-
5,320
5,008
4,798
4,201
4,115
3,049
6,262
4,109
12,015
10,294
8,962
12,476
10,340
3.7
3.3
2.9
2.3
2.2
2.0
12.8
7.5
12.1
9.4
7.7
12.8
9.7
979
873
791
623
602
534
1,084
664
1,034
826
691
1,085
850
0.9
0.8
0.7
0.6
0.5
0.5
2.8
1.7
1.4
1.1
0.9
1.4
1.1
19
19
19
2C
20
2C
18
19
10
10
11
10
10
2009 Simulation
Atlanta GA
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
ER visits (Resp)
New York
ER visits (asthma)
ER visits (asthma)
ER visits (asthma)
ER visits (asthma)
ER visits (asthma:
Tolbert
Tolbert
Tolbert
Tolbert
Tolbert
Darrow
Strickland
Strickland
Ito
Ito
Ito
Ito
Ito
CO
NO2
PM10
PM10, N02
Darrow
distlagO-7
avg day lag 0-2
PM2.5
NO2
CO
SO2
6,063
5,397
4,879
3,330
3,697
3,276
7,056
4,171
12,945
10,115
8,350
13,677
10,441
4,197
3,314
2,579
1,352
1,213
1,991
5,026
2,281
S,S28
5,439
3,750
9,682
6,068
-
-
-
-
-
-
-
-
-
-
7,895
7,436
7,127
6,248
6,121
4,545
8,951
5,953
16,777
14,443
12,620
17,399
14,505
5.3
4.7
4.2
3.3
3.2
2.8
18.0
10.7
16.9
13.2
10.9
17.9
13.7
5,795
5,157
4,662
3,658
3,532
3,129
6,768
3,992
12,152
9,479
7,316
12,844
9,786
4,009
3,165
2,463
1,290
1,153
1,901
4,813
2,180
8,266
5,032
3,500
9,071
5,672
-
-
-
-
-
-
-
-
-
-
-
-
7,548
7,109
6,813
5,972
5,850
4,342
8,599
5,706
15,737
13,571
11,344
16,379
13,630
5.0
4.5
4.0
3.2
3.1
2.7
17.3
10.2
15.9
12.4
10.2
16.8
12.8
269
240
218
172
166
147
287
179
793
636
534
832
655
0.2
0.2
0.2
0.1
0.1
0.1
0.7
0,5
1.0
0.8
0.7
1.1
0.9
4
4
4
4
4
4
4
4
6
6
6
6
6
2
3
7-62
-------
1 Table 7-22 Short-Term Ozone Exposure-Related Morbidity (Hospital Admissions - 2007 simulation year)
Urban study area
(endpoint)
New York
HA (chronic lung dis
HA (asthma)
HA (asthma;
Detroit
HA (respiratory)
HA (respiratory)
HA (respiratory)
Atlanta, GA
Baltimore, MO
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
LA
HA (respiratory)
Study author
Lin
Silverman
Silverman
Katsouyanni
Katsouyanni
Medina-Ramon
_ -i-i
Effect estimator
differentiators
PM2.5
Ihr max, penalized
spines
Ihr max, natural
spines
Shr mean
Ihr max, penalized
spines
Current conditions simualtion (2007)
point
estimate
133
694
508
55
53
35
21
30
15
3
21
13
55
61
13
7
23
106
95th Confidence
2.5
78
49
-175
-13
-16
10
6
S
4
1
6
4
15
17
4
2
6
-137
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
_
97.5
188
1,192
1,036
121
120
60
36
51
25
5
36
23
93
105
21
11
39
344
%of
baseline
2.2
19.0
13. S
l.S
1.8
3.2
2.5
2.3
2.3
2.6
2.5
l.S
2.9
2.3
2.6
2.7
3.0
0.7
Current standard simulation (2007)
point
estimate
115
628
457
49
47
26
16
24
11
2
17
10
37
47
9
5
18
62
95th Confidence
2.5
67
43
-155
-11
-14
7
5
7
3
1
5
3
10
13
3
1
5
-SO
-
-
-
-
-
-
-
-
-
-
-
-
-
-
_
97.5
162
1,094
946
108
107
45
28
41
20
4
29
17
64
SI
16
8
31
202
%of
baseline
1.9
17.3
12.4
1.6
1.6
2.4
2.0
l.S
l.S
2.1
2.1
1.3
2.0
l.S
1.9
2.0
2.4
0.4
Delta (risk reduction)
point
estimate
IS
66
51
6
6
9
5
6
3
1
4
4
17
14
3
2
4
44
%of
baseline
0.3
1.8
1.3
0.2
0.2
0.8
0.6
0.5
0.5
0.5
0.5
0.5
0.9
0.5
0.6
0.7
0.6
0.3
% reduction in
ozone-related
morbidity
14
10
1C:
11
11
25%
23%
20%
21°:
20%
19%
27%
31%
23%
25%
26%
19%
42
2
3
7-63
-------
1 Table 7-23 Short-Term Ozone Exposure-Related Morbidity (Hospital Admissions - 2009 simulation year)
Urban study area (endpoint)
New York
HA (chronic lung disease)
HA (asthma)
HA (asthma)
Detroit
HA (respiratory)
HA (respiratory)
HA (respiratory)
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
LA
HA (respiratory)
Study author
Lin
Siluerman
Silverman
Katsouyanni
Katsouyanni
Medina-Ramon
Linn
Effect estimator
differentiators
3M2.5
Ihr max, penalized
spines
Ihr max, natural
spines
Shrmax
Ihr max, penalized
spines
Current conditions simualtion (2007)
point
estimate
192
876
644
75
72
31
22
26
13
3
17
16
55
57
10
7
IS
272
95th Confidence
2.5
112
62
-226
-IS
-22
9
6
7
4
1
5
4
15
16
3
2
5
-35S
-
-
97.5
271
1,482
1,294
165
163
52
37
44
23
5
29
27
93
98
17
12
31
876
%of
baseline
3.2
23.7
17.2
2.6
2.5
2.6
2.5
1.9
2.2
2.6
2.1
2.0
2.S
2.1
2.2
2.8
2.3
1.7
Current standard simulation (2007)
point
estimate
179
825
605
75
72
29
20
25
13
3
17
15
44
53
10
6
IS
255
95th Confidence
2.5
104
58
-210
-IS
-22
S
5
7
4
1
5
4
12
15
3
2
5
-334
-
-
-
-
-
-
-
-
-
-
-
-
-
-
-
_
97.5
252
1..409
1,227
165
163
50
33
43
22
5
29
25
76
91
16
10
30
822
%of
baseline
3.0
22.4
16.2
2.6
2.5
2.5
2.3
1.9
2.1
2.5
2.1
l.S
2.3
2.0
2.0
2.4
2.3
1.6
Delta (risk reduction)
point
estimate
13
51
39
C
C
1
2
1
0.4
0.1
0.0
1
10
4
1
1
0
17
%of
baseline
0
1
1
0
0
0.1
0.2
0.1
0.1
0.1
0.0
0.1
0.5
0.2
0.1
0.4
0.1
0
% reduction in
ozone-related
morbidity
7
6
6
C
C
5%
9%
3%
3%
3%
0%
7%
19%
7%
6%
15%
3%
6
2
3
7-64
-------
1 Table 7-24 Short-Term Ozone Exposure-Related Morbidity (Asthma Exacerbations)
Urban study area (endpoint)
Study author
Effect estimator
differentiators
Current conditions simualtion (2007)
point
estimate
95th Confidence
2.5
97.5
%of
baseline
Current standard simulation (2007)
point
estimate
95th Confidence
2.5
97.5
%of
baseline
Delta (risk reduction)
point
estimate
%of
baseline
% reduction in
ozone-related
morbidity
2007 Simulation
Boston MA
asthma exacer (Chest tightness)
asthma exacer (shortness of breath)
asthma exacer (Chest tightness)
asthma exacer (shortness of breath)
Asthma exacer (chest tightness;
asthma exacer (Chest tightness;
asthma exacer (wheeze)
Gent
Gent
Gent
Gent
Gent
Gent
Gent
Ihrmax, lagl
Ihrmax, lagl
Shrmax, lagl
Shrmax, lagl
Ihrmax PM2.5lagO
Ihr max PM2.5 lagl
Ihrmax, PM2.5, lag
0
28,639
20,035
20,493
23,700
28,949
26,701
53,682
14,989
2,485
6,722
4,749
13,374
10,632
19,682
-
-
-
-
40,322
35,259
32,412
39,922
42,008
40,156
82,795
22.0
12.2
15.7
14.4
22.1
20.4
17.6
26,401
18,348
18,932
21,878
26,691
24,539
49,333
13,720
2,260
6,172
4,354
12,231
9,711
17,956
-
-
-
-
-
-
37,412
32,495
30,114
37,082
39,014
37,255
76,598
20.2
11.2
14.5
13.4
20.4
18.8
16.2
2,238
1,687
1,562
1,822
2,258
2,112
4,350
1.7
1.0
1.2
1.1
1.7
1.6
1.4
8
8
8
S
8
8
8
2009 Simulation
Boston MA
asthma exacer (Chest tightness)
asthma exacer (shortness of breath)
asthma exacer (Chest tightness)
asthma exacer (shortness of breath)
Asthma exacer (chest tightness)
asthma exacer (Chest tightness)
asthma exacer (wheeze)
Gent
Gent
Gent
Gent
Gent
Gent
Gent
ihr max
Ihrmax
Shrmax
Shrmax
Ihr max PM2.5 lag 0
Ihr max PM2.5 lag 1
Ihrmax, PM2.5
24,387
16,799
IS, 340
21,180
24,661
22,682
45,379
12,553
2,050
5,943
4,188
11,179
S,859
16,356
34,361
30,007
29,329
36,107
36,402
34,710
71,096
1S.5
10.2
13.9
12.8
1S.7
17.2
14.7
23,538
16,226
17,726
20,467
23,853
21,934
43,862
12,124
1,978
5,736
4,040
10,795
8,552
15,786
-
-
-
-
-
-
-
33,767
29,021
23,389
34,947
35,267
33,620
68,819
17.9
9.8
13.4
12.4
18.1
16.6
14.2
799
573
614
713
807
748
1,517
0.6
0.3
0.5
0.4
0.6
0.6
0.5
3
3
3
3
3
3
3
7-65
-------
1 The presentation of key observations drawn from review of the risk estimates is divided
2 into two sections including: the assessment of health risks associated with recent conditions
3 (section 7.5.1) and with just meeting the current and alternative standards (sections 7.5.2). As
4 noted earlier, for the first draft REA we are only presenting results for the simulation of just
5 meeting the current standard. Risks under simulated just meeting alternative standards will be
6 presented in the second draft analysis. The presentation of key observations (for both recent
7 conditions and the simulated just meeting the suite of current O3 standards) is further separated
8 into those associated with mortality estimates and morbidity estimates.
9 7.5.1 Assessment of Health Risk Associated with Recent conditions
10 The assessment of risk for the recent conditions scenario for the 12 urban study areas (for
11 short-term exposure-related mortality) focuses on characterizing absolute risk using two types of
12 risk estimates (a) risk modeled down to zero Os, which reflects consideration for the full range of
13 exposure and (b) risk modeled down to the LML, which represents a higher confidence estimate
14 with the caveat that it excludes exposures below the LML (and is therefore likely biased low).
15 For short-term exposure-related morbidity endpoints, we only included estimates of risk down to
16 the LML. Estimates of the reduction in risk (deltas) are not relevant in evaluating the recent
17 conditions scenario, but are an important part of the analysis completed for the simulation of just
18 meeting the current standard level (presented in the next section).
19
20 Short-term Ch exposure-related mortality
21
22 • Higher confidence estimates of CVrelated all-cause mortality (modeled down to
23 LML) range 0.4 to 3.5% of total mortality across the 12 urban study areas (for 2007)
24 using Zanobetti and Schwartz (2008b) C-R functions. Estimates of CVrelated all-
25 cause mortality (modeled down to zero Os) range from 0.5 to 4.9% of total mortality
26 (for 2007) using Zanobetti and Schwartz (2008b) C-R functions (see Table 7-16).
27 This translates into from 10 to 710 (Vrelated deaths across the 12 urban study areas
28 when exposure is modeled down to the LML and from 20 to 930 deaths when
29 exposure is modeled down to zero (V Of particular note regarding the mortality
30 estimates based on the Zanobetti and Schwartz (2008b) C-R functions are the higher
31 risk estimates generated for Detroit and New York (see Table 7-16 and 7-17). In both
32 cases, these higher estimates reflect the use of effect estimates which are substantially
33 larger than estimates used for other urban study areas. As part of the second draft
34 REA, we will explore this observation (regarding higher risk related to notably higher
35 effect estimates) in greater detail (see section 7.7).
36 • Higher confidence estimates of CVrelated all-cause mortality (modeled down to
37 LML) range from 1.0 to 2.5% of total mortality across the 12 urban study areas (for
38 2007) using Bell et al., (2004) C-R functions. Estimates of (Vrelated all-cause
39 mortality (modeled down to zero Os) range from 1.5 to 3.7% of total mortality (for
7-66
-------
1 2007) using Bell et al., (2004) C-R functions (see Table 7-16). This translates into
2 from 30 to 730 Os-related deaths across the 12 urban study areas when exposure is
3 modeled down to the LML and from 40 to 950 deaths when exposure is modeled
4 down to zero O3.
5 • While we have a high degree of overall confidence in estimates generated using C-R
6 functions based on Zanobetti and Schwartz (2008b) and Bell et al (2004), resulting in
7 both sets of risk estimates being considered core estimates, we would note that
8 Zanobetti and Schwartz (2008b)-based estimates, only included exposures associated
9 with June-August and therefore may bias estimates of Os-related deaths low by not
10 considering Os exposure occurring during the rest of the Os season defined for each
11 urban study area. By contrast, Bell et al (2004)-based C-R functions provide coverage
12 for 63 exposure occurring across the full 63 season defined for each urban study area.
13 This potential low-bias in the Zanobetti and Schwartz (2008b)-based risk estimates
14 effects incidence count metrics.
15 • For a number of the urban study areas, confidence intervals (but not point estimates)
16 for short-term all-cause mortality (using C-R functions derived both from Zanobetti
17 and Schwartz 2008b and Bell et al., 2004) include values that fall below zero (see
18 Tables 7-11 through 7-14). Population incidence estimates with negative lower-
19 confidence bounds do not imply that additional exposure to 63 has a beneficial effect,
20 but only that the estimated Os effect estimate in the C-R function was not statistically
21 significantly different from zero.
22 • Cause-specific mortality could only be evaluated using C-R functions based on
23 Zanobetti and Schwartz (2008b) (Bayes-shrunken city-specific estimates for cause
24 specific mortality were not available for Bell et al, 2004). For 2007, estimates of
25 cardiovascular-related mortality incidence (associated with Os exposure) were
26 substantially larger than estimates of respiratory-related mortality incidence (see
27 Table 7-15). The sum of cardiovascular and respiratory does not equal total mortality
28 for most of the urban study areas and in some cases can be substantially lower than
29 total mortality (see Table 7-15). We may explore potential explanations for this as
30 part of the second draft REA.
31 • All-cause mortality estimates derived using C-R functions from both Zanobetti and
32 Schwartz (2008b) and Bell et al (2004) are driven largely by days with total O3 levels
33 falling in the range of 35 to 70 ppb, with a substantial portion of the mortality
34 estimate associated with days having 63 levels above 60 ppb (for 2007 - see Tables 7-
35 7 and 7-9, respectively).13
36 • Generally, all-cause mortality risks decrease somewhat for simulation year 2009
37 compared with estimates generated for 2007, reflecting the lower measured 63 levels
13 Characterization of ozone level ranges in Tables 7-7 through 7-10 is based on the air metric used by each
C-R function. The Zanobetti and Schwartz (2008b) based C-R functions uses daily 8hr mean daily values for the
composite monitor in a given urban area for the simulation period June through August. The Bell et al (2004) based
C-R functions uses 8hr max daily values for the composite monitor in a given urban area for the ozone season
specific to that urban area.
7-67
-------
1 in the later simulation year (with the exception of Atlanta, Baltimore, Los Angeles,
2 Sacramento and Houston, depending on the C-R function used, which did not have
3 lower O3 levels in 2009) - compare LML-based estimates presented in Table 7-16
4 with estimates in 7-17 and/or compare estimates in Table 7-18 with those in Table 7-
5 19.
6 Short-term O3 exposure-related morbidity
7 • Estimates of O3- attributable ER visits (respiratory symptoms) for 2007 in Atlanta
8 (based on modeling exposure down to the LML) range from roughly 2.4 to 15.6% of
9 total baseline incidence which translates into from 3,100 to 6,000 visits depending on
10 the model formulation (i.e., epidemiological study providing the C-R function and the
11 treatment of lag and copollutants) (see Table 7-21). Estimates of O3- attributable ER
12 visits (for asthma) for 2007 in New York range form roughly 8.6 to 14.2% of total
13 baseline which translates into 6,600 to 10,800 visits again depending on the treatment
14 of copollutants in the model (see Table 7-21).
15 • Estimates of ER visits in both urban study areas are modestly larger for 2009,
16 reflecting higher O3 levels (for the O3 metrics involved in modeling these endpoints)
17 (see Table 7-21).
18 • Estimates of O3- attributable HA (for asthma) in New York in 2007 (based on
19 modeling risk down to LML) range form roughly 13.8 to 19% of baseline incidence
20 which translates into roughly 500 to 700 admissions depending whether PM2.5 is
21 included in the model (see Table 7-22). Estimates of HA (for chronic lung disease) in
22 New York in 2007 are approximately 2.2% of baseline which translates into 130
23 admissions (see Table 7-22). Estimates of O3- attributable HA (respiratory symptoms)
24 across the 12 urban study areas range from 0.7 to 3.2% of baseline, which translates
25 into 3 to 110 admissions (see Table 7-22).
26 • Estimates of HA visits for simulation year 2009 are generally marginally lower across
27 most cities, reflecting lower measured O3 levels (with the notable exception of New
28 York, which had notably higher estimates of HA for asthma in 2009, reflecting higher
29 O3 levels in 2009 for the metric used in modeling risk) (compare estimates in Table 7-
30 23 to those in Table 7-22).
31 • Estimates of O3- attributable asthma exacerbations for Boston in 2007 (based on
32 modeling risk down to LML) range from roughly 12.2 to 22.1% of baseline incidence
33 which translates into 20,000 to 29,000 events (for chest tightness or shortness of
34 breath). This range reflects differences in model specification (e.g., lag structure and
35 peak O3 metric used). Estimates of O3- attributable asthma exacerbation (wheeze) was
36 17.6% of baseline which translates into 55,000 events (see Table 7-24). These
37 estimates were somewhat lower in 2009 (see Table 7-24).
38 • While estimates for both ER visits and asthma exacerbations included 95th percentile
39 confidence intervals that did not include negative values, several of the analyses
40 involving HA did include negative lower estimates for the 2.5th percentile values (i.e.,
7-68
-------
1 the lower bound of the 95th percentile intervals for the incidence estimates) (see Table
2 7-22 and 7-23). The negative lower bound values for the subset of HA estimates
3 likely reflects, at least in part, the considerably smaller sample size associated with
4 modeling for this endpoint compared with other HA-related endpoints and both ER
5 and asthma exacerbation endpoints included in this analysis. And, as was discussed
6 above in relation to short-term exposure-related mortality, negative values for lower
7 bound statistics does not imply that Os is beneficial, but rather speaks to the lower
8 sample size, as discussed here.
9 7.5.2 Assessment of Health Risk Associated with Simulating Meeting the Current Suite
10 of O3 Standards
11 The analysis of risk after simulating just meeting the current standard includes both (a)
12 assessment of absolute risk remaining and (b) the risk reduction (delta) associated with a
13 comparison of Os levels for recent conditions with Os levels after simulating just meeting the
14 current primary 63 standard. In both cases, we generated two types of risk estimates including an
15 assessment of risk based on modeling exposure down to zero Os and a higher confidence
16 estimate based on modeling risk down to the surrogate LML. As noted earlier in section 7.1.2.1,
17 constraining the analysis to only consider exposures above the LML did not have a substantial
18 impact on delta (risk reduction) estimates, since most of the daily reductions in Os occurred at
19 levels well above the applicable LML. Our discussion of risk estimates presented below focuses
20 primarily on the level of Oj- attributable risk remaining after simulation of meeting the current
21 standard level.
22
23 Short-term Ch exposure-related mortality
24
25 • Higher confidence estimates of (Vrelated all-cause mortality (modeled down to
26 LML) range 0.3 to 3.1% of total mortality across the 12 urban study areas (for 2007)
27 using Zanobetti and Schwartz (2008b) C-R functions. Estimates of total CVrelated
28 all-cause mortality (modeled down to zero O3) range from 0.5 to 4.6% of total
29 mortality (for 2007) using Zanobetti and Schwartz (2008b) C-R functions (see Table
30 7-16). This translates into from 10 to 630 (Vrelated deaths across the 12 urban study
31 areas when exposure is modeled down to the LML and from 20 to 850 deaths when
32 exposure is modeled down to zero O3. As with risk estimated for recent conditions,
33 the mortality estimates generated for Detroit and New York are notably higher than
34 those for the remaining 10 study areas (see Table 7-16 and 7-17). As stated earlier,
35 these higher estimates reflect the use of effect estimates which are substantially larger
36 than estimates used for other urban study areas. As part of the second draft REA, we
37 will explore this issue in greater detail (see section 7.7).
38 • Higher confidence estimates of CVrelated all-cause mortality (modeled down to
39 LML) range 0.9 to 2.0% of total mortality across the 12 urban study areas (for 2007)
40 using Bell et al., (2004) C-R functions. Estimates of total (Vrelated all-cause
41 mortality (modeled down to zero 63) range from 1.3 to 3.2% of total mortality (for
7-69
-------
1 2007) using Bell et al., (2004) C-R functions (see Table 7-16). This translates into
2 from 30 to 590 O3-related deaths across the 12 urban study areas when exposure is
3 modeled down to the LML and from 3 to 830 deaths when exposure is modeled down
4 to zero O3.
5 • Delta risk reductions for all-cause mortality associated with the simulation of the
6 current standard level (for 2007 using Zanobetti and Schwartz (2008b)-based C-R
7 functions) range roughly from 1 to 80 deaths averted across the 12 urban study areas
8 whether we model risk down to the LML, or down to zero. As noted above, this risk
9 metric is fairly invariant to consideration of the LML, since most reductions in O3
10 occur at levels well above the LML. If we use C-R functions based on Bell et al.,
11 (2004), then delta risk ranges from 2 to 160 deaths averted across the 12 urban study
12 areas.
13 • As noted earlier, estimates generated using C-R functions based on Zanobetti and
14 Schwartz (2008b) may be biased low since they only considered exposures between
15 June and August. By contrast, Bell et al (2004)-based C-R functions model risk for
16 the entire ozone season specific to each urban study area. This potential low-bias in
17 the Zanobetti and Schwartz (2008b)-based risk estimates effects incidence count
18 metrics.
19 • As noted earlier, population incidence estimates with negative lower-confidence
20 bounds do not imply that additional exposure to O3 has a beneficial effect, but only
21 that the estimated O3 effect estimate in the C-R function was not statistically
22 significantly different from zero.
23 • As with risk estimates generated for the recent conditions scenario, estimates of O3-
24 attributable cardiovascular-related mortality incidence were substantially larger than
25 estimates of respiratory-related mortality incidence (see Table 7-15).
26 • Even after simulation of urban study areas meeting the current ozone standard, all-
27 cause mortality estimates derived using C-R functions from both Zanobetti and
28 Schwartz (2008b) and Bell et al (2004) continue to be driven largely by days with
29 total O3 levels falling in the range of 35 to 70 ppb, with a substantial portion of the
30 mortality estimate associated with days having O3 levels above 60 ppb (for 2007 - see
31 Tables 7-7 and 7-9, respectively).
32 • Generally, O3-attributable all-cause mortality risks continue to be lower for the 2009
33 simulation year as compared with the 2007 simulation year (with the exception of
34 Atlanta, Baltimore, Los Angeles, Sacramento and Houston, depending on the C-R
35 function used, which did not have lower O3 levels in 2009) - compare LML-based
36 estimates presented in Table 7-16 with estimates in 7-17 and/or compare estimates in
37 Table 7-18 with those in Table 7-19.
38
39
7-70
-------
1 Short-term O^ exposure-related morbidity
2 • Estimates of Os-attributable ER visits (respiratory symptoms) for 2007 in Atlanta
3 (based on modeling exposure down to the LML) range from roughly 2.0 to 12.8% of
4 total baseline incidence which translates into from 2,200 to 4,900 visits depending on
5 the model formulation (i.e., epidemiological study providing the C-R function and the
6 treatment of lag and copollutants) (see Table 7-21). Estimates of Os- attributable ER
7 visits (for asthma) for 2007 in New York range form roughly 7.7 to 12.8% of total
8 baseline which translates into 5,900 to 9,700 visits again depending on the treatment
9 of copollutants in the model (see Table 7-21).
10 • Estimates of ER visits in both urban study areas are larger for 2009, reflecting higher
11 Os levels (for the Os metrics involved in modeling these endpoints) (see Table 7-21).
12 • Estimates of Os-attributable HA (for asthma) in New York in 2007 (when modeling
13 risk down to LML) range form roughly 12.4 to 17.3% of baseline incidence which
14 translates into roughly 500 to 600 admissions depending whether PM2.5 is included in
15 the model (see Table 7-22). Estimates of HA (for chronic lung disease) in New York
16 in 2007 are approximately 1.9% of baseline which translates into 120 admissions (see
17 Table 7-22). Estimates of Os- attributable HA (respiratory symptoms) across the 12
18 urban study areas range from 0.4 to 2.4% of baseline, which translates into 2 to 60
19 admissions (see Table 7-22).
20 • Estimates of Os- attributable asthma exacerbations for Boston in 2007 (based on
21 modeling risk down to LML) range from roughly 11.2 to 20.4% of baseline incidence
22 which translates into 18,000 to 27,000 events (for chest tightness or shortness of
23 breath). This range reflects differences in model specification (e.g., lag structure and
24 peak Os metric used). Estimates of Os- attributable asthma exacerbation (wheeze) was
25 16.2% of baseline which translates into 49,000 events (see Table 7-24). These
26 estimates were somewhat lower in 2009 (see Table 7-24).
27 • Risk reductions (comparing recent conditions to meeting the current standard) for ER
28 visits (respiratory) in Atlanta (2007) range from 500 to 1,100 visits averted (see Table
29 7-21). Delta risk for ER visits (asthma) in New York (2007) range from 700 to 1,100
30 visits averted (see Table 7-21). Risk reductions for HA (asthma) in New York (2007)
31 range from 50 to 70 admissions averted (see Table 7-22). Risk reduction for HA
32 (chronic lung disease) in New York is estimated at 18 admissions averted. Risk
33 reductions for HA (respiratory) across the 12 urban study areas range from 1 to 40
34 admissions averted (see Table 7-22). Risk reduction for asthma exacerbations
35 (shortness of breath or chest tightness) in Boston (2007) ranges from 1,600 to 2,300
36 cases averted (see Table 7-24). We estimate that in 2007 in Boston, we would see
37 4,400 fewer asthma exacerbations (wheeze) the city was in attainment. All risk
38 reduction estimates summarized in this bullet reflect modeling of risk down to the
39 LML.
40 • Estimates of HA visits for simulation year 2009 are generally marginally lower across
41 most cities, reflecting lower measured Os levels (with the notable exception of New
7-71
-------
1 York, which had notably higher estimates of HA for asthma in 2009, reflecting higher
2 Os levels in 2009 for the metric used in modeling risk) (compare estimates in Table 7-
3 23 to those in Table 7-22).
4 • As noted earlier, negative lower bound values for the subset of HA estimates likely
5 reflects, at least in part, the considerably smaller sample size associated with
6 modeling for this endpoint compared with other HA-related endpoints as well as both
7 ER and asthma exacerbation endpoints included in this analysis. And, as was
8 discussed above in relation to short-term exposure-related mortality, negative values
9 for lower bound statistics does not imply that Os is beneficial, but rather reflect the
10 lower sample size.
11 7.6 KEY OBSERVATIONS DRAWN FROM THE URBAN CASE STUDY ANALYSIS
12 OF OS-RELATED RISK
13 This chapter provides key observations regarding: (a) overall confidence in the analysis
14 reflecting both the design of the risk assessment and the degree to which variability and
15 uncertainty have been addressed (section 7.6.1) and (b) risk estimates generated for both the
16 recent conditions and just meeting the current standard level (including the distribution of risks
17 and pattern of risk reduction across the 12 urban study areas and two simulation years evaluated)
18 (section 7.6.2).
19 7.6.1 Overall Confidence in the Risk Assessment and Risk Estimates
20 Based on consideration for observations listed as bullets below, EPA staff preliminarily
21 concludes that there is a reasonable degree of confidence in the core risk estimates generated for
22 mortality associated with short-term 63 exposure. However, we differentiate between the
23 estimates of risk based on modeling exposure down to zero O3 and those based on modeling risk
24 down to the LML. Generally, we have higher confidence in the estimates of risk based on
25 modeling risk down to the LML, since these reflect the Os levels used in fitting the C-R
26 functions underlying the risk estimates. However, the LML estimates are likely low-biased given
27 that they exclude exposures below the LML. In this context, the estimates of risk down to zero
28 63 may be particularly useful in gaining perspective on the potential magnitude of this excluded
29 risk (i.e., the risk associated with exposures below the LML).
30 Overall confidence in estimating mortality risk will likely be increased further with the
31 inclusion of a sensitivity analysis in the second draft REA, exploring the potential impact of
32 design elements on these risk estimates. Confidence in risk estimates generated for all of the
33 health endpoints will be further increased if we can obtain the actual LMLs associated with the
34 studies underlying the C-R functions, since that will allow us to estimate risk with consideration
35 for the actual range of data used in fitting the C-R functions (and not a surrogate).
7-72
-------
1 Confidence in our characterization of short-term exposure-related morbidity risk is
2 somewhat lower (but still reasonable) given that morbidity effects are only evaluated (for most
3 endpoints) for a subset of urban study areas and because we do not have multiple C-R functions
4 from multiple studies for the same endpoint. In addition, most of the epidemiological studies
5 covering respiratory morbidity endpoints are city-specific and it would be preferable to also have
6 Bayes-shrunken estimates which combine both a local and broader-scale regional or national
7 signal in modeling risk for each urban area.
8 Key observations addressing overall confidence in the analysis include:
9 • A deliberative process was used in specifying each of the analytical elements
10 comprising the risk model. This process included first identifying specific goals for
11 the analysis, and then designing the analysis to meeting those goals, given available
12 information and methods. Specific analytical elements reflected in the design
13 include: selection of urban study areas, characterization of ambient air Os levels,
14 selection of health endpoints to model and selection of epidemiological studies (and
15 specification of C-R functions) (see sections 7.1.1 and 7.3).
16 • Modeling of short-term exposure-related mortality (the key endpoint in the analysis)
17 utilized Bayes-adjusted city-specific effect estimates (see section 7.1.1 and section
18 7.3.2). These effect estimates are considered to have increased overall confidence
19 since they combine elements of the local city-specific signal with a broader scale
20 (national) signal.
21 • Review of available literature (as specified in the 63 ISA, U.S. EPA. 2012), resulted
22 in a decision not to incorporate a true (no effect) threshold into our risk modeling.
23 Conversely, the literature supports a log-linear, no-threshold relationship down to
24 concentrations at the lower end of the range of ambient Os concentrations. To explore
25 the impact of focusing risk modeling on ranges of increased confidence, we generated
26 risk estimates reflecting the range of exposures used in deriving the C-R functions
27 underlying the risk estimates (see section 7.1.1). However, we also included estimates
28 of risk reflecting the full range of exposures down to zero Os. Together, these two
29 types of risk estimates inform consideration of uncertainty related to application of
30 the C-R function at low ozone levels.
31 • Evaluation of the degree to which key sources of variability impacting Os risk were
32 incorporated into the design of the analysis (see section 7.4.1). Some of the key
33 sources considered in the design include: heterogeneity in effect 63 across cities,
34 intra-urban variability in 63 levels, variability in the pattern of 63 reductions within
35 urban areas when simulating just meeting the current standard, inter-urban and intra-
36 urban variability in copollutants levels and their role as potential confounders,
37 variability in demographic and SES-related factors, and variability in baseline
38 incidence rates.
39 • Application of a strategy based on the WHO's 4-tiered approach for characterizing
40 uncertainty to evaluate the potential impact of uncertainty on risk estimates (see
7-73
-------
1 section 7.4.2). This approach involves both a quantitative sensitivity analysis to
2 evaluate the potential impact of specific design elements on risk estimates and
3 completion of a qualitative analysis to provide additional coverage for potential
4 sources of uncertainty. For the first draft analysis, we completed the qualitative
5 analysis, however, we did not complete the sensitivity analysis (that is planned for the
6 second draft analysis). The qualitative analysis of uncertainty suggested that the
7 statistical fit and shape of C-R functions together with the use of surrogate LMLs to
8 define ranges of increased confidence in estimating risk could have a medium impact
9 on risk estimates. Other factors (e.g., characterization of ambient air O3 levels,
10 addressing copollutants in the context of deriving C-R functions) could have a low-
11 medium impact (see section 7.4.2).
12 7.6.2 Risk Estimates Generated for Both the Recent Conditions and Simulation of
13 Meeting the Current Standard
14 Key observations regarding risk estimates generated for both the recent conditions and
15 simulating just meeting the current standard level are presented below:
16 • Estimates of short-term exposure-related all-cause mortality attributable to O3 under
17 recent conditions vary widely across urban study areas, reflecting differences both in
18 ambient O3 levels and population counts, as well as differences in effect estimates.
19 Risk based on modeling exposure down to the LML (for simulation year 2007) is
20 estimated to range from 0.4 to 3.5% of total baseline mortality across the 12 urban
21 study areas which translate into from roughly 10 to 710 deaths across the 12 urban
22 study areas. When risk is modeled for ozone exposures down to zero O3 (i.e.,
23 considering the full range of potential exposures), then O3-related risk (again for
24 2007) ranges from 0.5 to 4.9% of total mortality, which translates into from roughly
25 20 and 930 deaths.
26 • Estimates of O3-attributable all-cause mortality under recent conditions in 2007 are
27 driven largely by days with O3 levels falling in the range of 35 to 70ppb (for the
28 metrics involved in risk modeling - 8hr max and 8hr averages). A substantial portion
29 of the mortality risk is associated with days having O3 levels even higher, above 60
30 ppb. This observation accounts for the notable magnitude of risk reduction seen with
31 simulation of just meeting the current standard (see below).
32 • For most of the study areas, estimates of short-term exposure-related all-cause
33 mortality attributable to O3 are somewhat (but not substantially) smaller for
34 simulation year 2009 as compared with simulation year 2007. This reflects primarily
35 the lower O3 levels seen in 2009.
36 • Estimates of short-term exposure-related morbidity attributable to O3 under recent
37 conditions for 2007 include: (a) ER visits (for respiratory symptoms in Atlanta)
38 range from roughly 2.4 to 15.6% of total baseline incidence which translates into
39 from 3,100 to 6,000, (b) ER visits (for asthma in New York City) range form roughly
40 8.6 to 14.2% of total baseline which translates into 6,600 to 10,800, (c) HA (for
41 asthma in New York City) range form roughly 13.8 to 19% of baseline incidence
42 which translates into roughly 500 to 700 admissions, (d) HA (for chronic lung disease
7-74
-------
1 New York City) are roughly 2.2% of baseline which translates into 130 admissions,
2 (e) HA (respiratory symptoms across the 12 urban study areas) range from 0.7 to
3 3.2% of baseline, which translates into 3 to 110 admissions, (f) asthma exacerbations
4 (chest tightness or shortness of breath for Boston) range from roughly 12.2 to 22.1%
5 of baseline incidence which translates into 20,000 to 29,000 events (for chest
6 tightness or shortness of breath) and (g) asthma exacerbation (wheeze in Boston) was
7 17.6% of baseline which translates into 55,000 events. All these estimates reflect
8 modeling exposure down to the applicable LML value (and not down to zero O3).
9 • Estimates of short-term exposure-related all-cause mortality attributable to O3 after
10 simulating meeting the current standard vary widely across urban study areas,
11 reflecting differences both in ambient O3 levels and population counts, as well as
12 differences in effect estimates. Risk based on modeling exposure down to the LML
13 (for simulation year 2007) is estimated to range from 0.3 to 3.1% of total baseline
14 mortality across the 12 urban study areas which translate into from roughly 10 to 630
15 deaths across the 12 urban study areas. If we model risk all the way down to zero O3
16 (i.e., considering the full range of potential exposures), then O3-related risk (again for
17 2007) ranges from 0.5 to 4.6% of total mortality, which translates into from roughly
18 20 and 850 deaths.
19 • Estimates of O3-attributable all-cause mortality after simulating meeting the current
20 standard in 2007 are driven largely by days with O3 levels falling in the range of 35 to
21 70ppb (for the metrics involved in risk modeling - 8hr max and 8hr averages). A
22 substantial portion of the mortality risk continues to be associated with days having
23 O3 levels even higher, above 60 ppb. This observation accounts for the notable
24 magnitude of risk reduction seen with simulation of just meeting the current standard
25 (see below).
26 • For most of the study areas, estimates of short-term exposure-related all-cause
27 mortality attributable to O3 are somewhat (but not substantially) smaller for
28 simulation year 2009 as compared with simulation year 2007. This reflects primarily
29 the lower O3 levels seen in 2009.
30 • Estimates of short-term exposure-related morbidity attributable to O3 after simulating
31 meeting the current standard for 2007 include: (a) ER visits (for respiratory symptoms
32 in Atlanta) range from roughly 2.0 to 12.8% of total baseline incidence which
33 translates into from 2,200 to 4,900, (b) ER visits (for asthma in New York City) range
34 form roughly 7.7 to 12.8% of total baseline which translates into 5,900 to 9,700, (c)
35 HA (for asthma in New York City) range form roughly 12.4 to 17.3% of baseline
36 incidence which translates into roughly 500 to 600 admissions, (d) HA (for chronic
37 lung disease New York City) are roughly 1.9% of baseline which translates into 120
38 admissions, (e) HA (respiratory symptoms across the 12 urban study areas) range
39 from 0.4 to 2.4% of baseline, which translates into 2 to 60 admissions, (f) asthma
40 exacerbations (chest tightness or shortness of breath for Boston) range from roughly
41 11.2 to 20.4% of baseline incidence which translates into 18,000 to 27,000 events (for
42 chest tightness or shortness of breath) and (g) asthma exacerbation (wheeze in
43 Boston) was 16.2% of baseline which translates into 49,000 events. All these
7-75
-------
1 estimates reflect modeling exposure down to the applicable LML value (and not
2 down to zero O^).
3 • Under simulation of just meeting the current standard, we see a shift in the daily
4 metric profile for 63, as would be expected given application of the quadratic rollback
5 method in predicting reductions in 03. However, we still see that all-cause mortality
6 attributable to Os is driven by days in the higher Os ranges (i.e., 30 to 70ppb, with a
7 significant portion associated with days above 60 ppb).
8 • Generally, for most of the urban study areas, reductions in all-cause mortality risk
9 associated with simulated just meeting the current standard is significantly lower for
10 simulation year 2009 compared with estimates generated for 2007, reflecting the
11 lower measured Os levels in the later simulation year.
12 • Risk reductions for all-cause mortality associated with the simulation of the current
13 standard level (for 2007) range roughly from 1 to 160 deaths averted across the 12
14 urban study areas whether we model risk down to the LML, or down to zero. Risk
15 reductions for morbidity endpoints are: (a) ER visits (respiratory) in Atlanta (2007)
16 range from 500 to 1,100 visits averted, (b) ER visits (asthma) in New York (2007)
17 range from 700 to 1,100 visits averted, (c) HA (asthma) in New York (2007) ranges
18 from 50 to 70 admissions averted, (d) HA (chronic lung disease) in New York is
19 estimated at 18 admissions averted, (e) HA (respiratory) across the 12 urban study
20 areas ranges from 1 to 40 admissions averted, (f) asthma exacerbations (shortness of
21 breath or chest tightness) in Boston (2007) ranges from 1,600 to 2,300 cases averted,
22 and (g) in Boston, we estimate 4,400 fewer asthma exacerbations (wheeze). All risk
23 reduction estimates summarized in this bullet reflect modeling of risk down to the
24 LML.
25 7.7 POTENTIAL REFINEMENTTS FOR SECOND DRAFT RISK ASSESSMENT
26 This section describes potential refinements for the second draft REA which include: (a)
27 sensitivity analyses intended to enhance our understanding of the impact of design elements on
28 core risk assessments, (c) additional refinements to the core sets of risk estimates presented in the
29 first draft REA, and (c) treatment of both long-term exposure-related mortality and morbidity
30 endpoints. Each of theses topics is discussed separately.
31 7.7.1 Potential sensitivity analyses
32 As noted earlier in section 7.1.1, we did not complete a comprehensive set of sensitivity
33 analyses for the first draft REA due to emphasis being placed on generating a set of core risk
34 estimates. The following set of sensitivity analyses will be considered for the second Draft risk
35 assessment, in order to gain further insights into the potential impact of modeling design choices
36 on risk estimates.
37 • Interpolation of missing air quality data: For the first draft risk assessment, we did
38 not fill in any missing monitoring data in generating the composite monitor
7-76
-------
1 distributions (see section 7.2.1). For the second draft REA, we may explore this issue
2 of interpolating missing measurement data as part of the sensitivity analysis. The goal
3 would be to determine whether incorporating interpolation of missing data has a
4 significant impact on risk estimates. The sensitivity analysis could consider (a)
5 interpolation methods used in key epidemiological studies supporting the C-R
6 functions used in the risk assessment (to the extent that those studies used
7 interpolation) and/or (b) interpolation methods used in the first draft exposure
8 analysis (see section 5.5.6)..
9 • Short-term exposure-related mortality: Because we believe that greater confidence is
10 associated with the use of Bayes-adjusted city-specific effect estimates, we also
11 believe that ideally, sensitivity analyses (examining different model design options)
12 should also be based on Bayes-adjusted city-specific effect estimates. This would
13 necessitate that, if we are to conduct sensitivity analyses for this endpoint group, we
14 obtain Bayes-adjusted city-specific effect estimates reflecting different design options
15 (e.g., lag structures, copollutants models). While, we would consider using regional-
16 and national-level effect estimates differentiated for different design element options,
17 insights gained through these use of these non-city specific effect estimates would be
18 more limited. Possible design element choices considered for sensitivity analyses
19 included: (a) lag structure, (b) copollutants models, (c) regional versus national
20 adjustment (in the context of generating Bayes-adjusted city-specific effect estimates)
21 and (d) modeling period and air quality metric combinations (summer versus ozone
22 season for 8hr mean and 8hr max metrics).
23 • Short-term exposure-related morbidity (hospital admissions, emergency visits and
24 asthma exacerbations): Additional coverage for lag structure, copollutants models
25 and combinations of modeling periods and air quality metrics would be considered,
26 depending on coverage in the available literature.
27 While sensitivity analyses described above would both provide additional insights into
28 overall confidence in both short-term exposure-related morbidity and mortality, given the
29 emphasis placed on mortality in this risk assessment (as the most significant health endpoint), we
30 would focus on completing sensitivity analyses for the mortality endpoint group.
31 7.7.2 Additional refinements to the core risk estimates completed for the first draft
32 REA
33 A number of refinements to the set of core risk estimates would be considered for the
34 second draft risk assessment, including:
35 • Generate confidence intervals for the delta (risk reduction) estimates: The method
36 used for generating delta (risk reduction) estimates in the first draft REA, while
37 providing sound point estimates, did not allow for the generating of confidence
38 intervals reflecting the impact of statistical uncertainty associated with the fit of the
39 effect estimates used (CIs were only generated for absolute risk for both the recent
40 conditions and simulated attainment of the current standard scenarios). For the
41 second draft, we will consider also generating CIs for the delta risk estimates.
7-77
-------
1 • Rigorous comparison of OB air quality data used in source epidemiological studies
2 and the design of the composite monitors used in risk assessment: For the second
3 draft analysis, we will complete a more rigorous comparison of the composite
4 monitor design used in the first draft REA with the methods used in the
5 epidemiological studies underlying the C-R functions used in the risk assessment. It is
6 likely that there will be varying degrees of agreement across the C-R functions (in
7 relation to the way air quality data are integrated), leading to different degrees of
8 uncertainty being introduced into the analysis. As part of the second draft analysis,
9 we will characterize this uncertainty and will consider using alternate composite
10 monitor designs if (a) they would more closely match the approach used in a given
11 epidemiological study and (b) EPA staff believes this refinement is likely to make a
12 substantial difference in risk characterization. Aspects of this task may fall into the
13 category of sensitivity analysis, depending on how they are implemented, in which
14 case they will be presented as part of the sensitivity analysis.
15 • Further exploration of patterns of potential interest in the risk estimates: We will
16 complete a more thorough review of the risk estimates generated with emphasis on
17 explaining any patterns of particular interest. An excellent example of this involves
18 short-term exposure-related mortality modeled using C-R functions based on
19 Zanobetti and Schwartz (2008b). As noted in section 7.5.1, these risk estimates in the
20 form of percent of baseline mortality (which is normalized on population count) are
21 up to 50% for New York City and Detroit compared with the other urban study areas.
22 In this case, these larger risk estimates directly reflect larger effect estimates specified
23 for these two cities in the underlying epidemiological study. As part of the second
24 draft REA, we would provide a more thorough assessment of regionality in effect
25 estimates reflected in this example and its impact on risk.
26 • Characterizing "ranges ofOs concentrations with increased confidence " using data
27 from the underlying epidemiological studies rather than the use of composite
28 monitor-basedLMLs: depending on available data, we may use LMLs values from
29 the actual epidemiological studies underlying C-R functions to define ranges of
30 increased confidence used in the risk assessment (in place of the surrogate values
31 obtained form the composite monitor distributions used in the first draft REA). In the
32 event that we are not able to obtain LMLs for all of the epidemiological studies used
33 in the risk assessment, we may also consider generating surrogate LMLs based on
34 obtaining 63 monitoring data that matches the measurement period (range of years)
35 used in a particular epidemiological study, rather than using the composite monitor-
36 based LML values from the modeled (simulation) years as was done here for the first
37 draft REA. Based on consideration for CASAC and public comments we will also
38 consider using additional metrics, besides the LML, in specifying ranges of increased
39 confidence in estimating risk. For example, we could include estimates of risk down
40 to Os levels higher than the LML, to explore modeling of risk closer to the central
41 mass of measurement data used in the epidemiological studies supporting the C-R
42 functions.
7-78
-------
1 7.7.3 Treatment of both long-term exposure-related mortality and morbidity endpoints
2 For the second draft REA, based on review of the evidence as summarized in the Oj ISA
3 (U.S. EPA, 2012), we are planning to model risk for long-term exposure-related mortality. Our
4 rationale for this decision is laid out in greater detail in section 8.1.1.5 (Chapter 8 discusses the
5 national-scale risk assessment, but the rationale for including long-term exposure-related
6 mortality as presented there, also applies for the urban study area risk assessment). In summary,
7 the decision to model long-term exposure-related mortality reflects consideration for evidence
8 supporting the endpoint category which is suggestive of a casual association (for long-term
9 mortality), but likely to be causal for the broader category of long-term exposure-related
10 respiratory health effects (which includes mortality). Given that our analysis would focus on
11 respiratory mortality (see below), we conclude that modeling long-term exposure-related
12 (respiratory) mortality would be reasonably well-supported by the evidence. In modeling the
13 endpoint for the urban study area risk assessment, as with the national-scale analysis, we would
14 use the national-level respiratory effect estimate reflecting control for PM2.5 (from Jarrett et al.,
15 2009), with that single effect estimate being applied to each of the urban study areas. In
16 addition, as a sensitivity analysis, we would consider modeling risk using the regional-level
17 respiratory effect estimates presented in Table 4 of the study, although it is important to note that
18 (a) these regional effect estimates do not include control for PM2.5 and (b) regional differences in
19 the ozone effect may reflect to a great extent, differing degrees of exposure measurement error
20 (e.g., related to temperature, differing residential/commuting patterns).
21 With regard to long-term exposure-related morbidity, after careful review of the available
22 evidence as summarized in the Os ISA (U.S. EPA, 2012), we have concluded that, while the
23 overall body of evidence supports a likely causal association between long-term exposure and
24 respiratory health effects, limitations in the study-level data required to support risk assessment
25 prevents us at this point from completing a quantitative risk assessment for this category of
26 health endpoints with a reasonable degree of confidence. It is important to emphasize that these
27 limitations do not prevent the use of this evidence from informing consideration of the levels of
28 exposure at which specific types of health effects may occur (i.e., the evidence analysis, which is
29 an important aspect of the ozone NAAQS review). Rather, these limitations only prevent the
30 quantitative estimation of risk with a reasonable degree of confidence.
31 In considering the potential for modeling risk for long-term exposure-related morbidity,
32 we first identified a subset of epidemiological studies as candidates for supporting the
33 specification of C-R functions including: (a) Meng et al., 2010 (HA and ED visits by asthmatics
34 in San Joaquin Valley, CA), (b) Akinbami et al., 2010 (current asthma and asthma attack
35 prevalence in children in U.S metropolitan areas), (c) Lin et al., 2008 (first asthma HA in
7-79
-------
1 children in NYC and NY state), and (d) Moore et al., 2008 (hospital discharges for asthma in
2 Southern CA). The discussion of limitations in the evidence focuses on these studies:
3 • When considering these studies and their potential use in quantitative analyses it is
4 important to recognize that Meng et al. (2010) and Akinbami et al. (2010) are both
5 cross-sectional studies. CASAC has advised us on numerous occasions to place less
6 emphasis on the results from this type of study design due to implicit limitations and
7 difficulty in interpreting the results.
8 • It is also important to consider the age range included in some of these studies that are
9 relying on an asthma diagnosis. Diagnosing asthma in very young children (<4) is
10 difficult. Both Lin et al. (2008) and Akinbami et al. (2010) recognize this, and to
11 account for it exclude children under the age of 1 and 3, respectively. Still, the
12 majority of the children included in the analysis by Lin et al (2008) are between 1 and
13 2 years of age, which introduces uncertainly into the diagnosis.
14 • Moore et al. (2008) includes a series of cross-sectional studies, where the exposure is
15 limited to a quarterly average and linked to hospital admissions during that quarter;
16 the analysis includes two quarters each year (spring and summer) over an 18 year
17 period. This type of longitudinal cross-sectional study design is unusual. Although
18 the research group behind this study has published multiple papers in high quality
19 journals it remains unclear if using DSA in the model building step is appropriate -
20 mostly because it is unclear how this approach selects the appropriate model.
21 • Lin et al. (2008a) represents the strongest of the long-term Os exposure and
22 respiratory morbidity studies and its strengths are discussed in the 63 ISA (U.S. EPA,
23 2012). Lin et al. is a retrospective cohort study that focuses on first time asthma
24 hospital admission in NY state. Never the less, there are concerns related to this study
25 when considering as the basis for C-R functions used in risk assessment:
26 o Enrollment and follow-up of the cohort was done using administrative
27 records; follow-up questionnaires were not sent out to each child that entered
28 the cohort so that children that may have moved out of state are considered to
29 be part of the cohort, even though they may have had a hospital admission in
30 another state. It is unclear how this influences the overall results of the study.
31 o The majority of admissions are for children between the age of 1-2, as stated
32 previously it is sometimes difficult to diagnose asthma in children of this age.
33 Therefore, the study may more accurately represent hospital admissions for a
34 respiratory condition and not necessarily asthma alone. It is not known what
35 level of uncertainty this might introduce and if the discharge diagnosis might
36 impact this.
37 o Finally, this study could be compared with Lin et al. (2008b) (Environmental
38 Research, 108 (2008): 42-47), which examined short-term O3 exposure and
39 respiratory hospital admissions in NY state to compare the risk estimates
40 obtained in both studies. Further CASAC comments note the issue of
7-80
-------
1 controlling for effects due to short-term exposures in such long term studies.
2 The revised ISA notes this but does not further inform the level of uncertainty
3 related to this issue (i.e., a long-term exposure-related capturing a short-term
4 exposure-related signal).
5 Taken together, the limitations presented above resulted in EPA staff concluding that,
6 at this time, we could not generate risk estimates for the long-term exposure-related
7 respiratory morbidity effect category (specifically the set of health effect reflected in the four
8 studies identified above) with a reasonable degree of confidence.
9
7-81
-------
1 REFERENCES
2
3 Abt Associates Inc. (1996). A Participate Matter Risk Assessment for Philadelphia and Los Angeles.
4 Prepared for Office of Air Quality Planning and Standards, U.S. Environmental Protection
5 Agency, Research Triangle Park, NC. July 1996. Available electronically on the internet at:
6 http: //www .epa.gov/ttnnaaqs/standards/pm/data/j ly 3 amd .pdf
7 Abt Associates, Inc. 2010. Environmental Benefits and Mapping Program (Version 4.0). Bethesda, MD.
8 Prepared for U.S. Environmental Protection Agency Office of Air Quality Planning and
9 Standards. Research Triangle Park, NC. Available on the Internet at
10 .
11 Akinbami, LJ; Lynch, CD; Parker, JD; Woodruff, TJ. (2010). The association between childhood asthma
12 prevalence and monitored air pollutants in metropolitan areas, United States, 2001-2004. Environ
13 Res 110:294-301.
14 Bell, ML; Dominici, F. (2008). Effect modification by community characteristics on the short-term
15 effects of ozone exposure and mortality in 98 U.S. communities. Am J Epidemiol 167: 986-997.
16 Bell, ML; McDermott, A; Zeger, SL; Samet, JM; Dominici, F. (2004). Ozone and short-term mortality in
17 95 U.S. urban communities, 1987-2000. JAMA 292: 2372-2378.
18 Darrow, L. A., Klein, M., Sarnat, J. A., Mulholland, J. A., Strickland, M. J., Sarnat, S. E., et al. (2011).
19 The use of alternative pollutant metrics in time-series studies of ambient air pollution and
20 respiratory emergency department visits. Journal of Exposure Science and Environmental
21 Epidemiology, 21,10-19.
22 Gent, J. F., Triche, E. W., Holford, T. R., Belanger, K., Bracken, M. B., Beckett, W. S., et al. (2003).
23 Association of low-level ozone and fine particles with respiratory symptoms in children with
24 asthma. Jama, 290(14), 1859-1867.
25 Ito, K., Thurston, G. D., & Silverman, R. A. (2007). Characterization of PM2.5, gaseous pollutants, and
26 meteorological interactions in the context of time-series health effects models. J Expo Sci
27 Environ Epidemiol, 17 Suppl 2, S45-60.
28 Katsouyanni, K., Samet, J. M., Anderson, H. R., Atkinson, R., Tertre, A. L., Medina, S., et al. (2009). Air
29 Pollution and Health: A European and North American Approach (APHENA): Health Effects
30 Institute
31 Lin, S., Bell, E. M., Liu, W., Walker, R. J., Kim, N. K., & Hwang, S. A. (2008a). Ambient ozone
32 concentration and hospital admissions due to childhood respiratory diseases in New York State,
33 1991- 2001. Environmental Research, 108, 42-47.
34 Lin, S; Liu, X; Le, LH; Hwang, SA. (2008b). Chronic exposure to ambient ozone and asthma hospital
35 admissions among children. Environ Health Perspect 116: 1725-1730.
36 Linn, W. S., Szlachcic, Y., Gong, H., Jr., Kinney, P. L., & Berhane, K. T. (2000). Air pollution and daily
37 hospital admissions in metropolitan Los Angeles. Environ Health Perspect, 108(5), 427-434.
7-82
-------
1 Medina-Ramon, M., Zanobetti, A., & Schwartz, J. (2006). The effect of ozone and PM10 on hospital
2 admissions for pneumonia and chronic obstructive pulmonary disease: a national multicity study.
3 Am J Epidemiol, 163(6), 579-588.
4 Meng, YY; Rull, RP; Wilhelm, M; Lombardi, C; Balmes, J; Ritz, B. (2010). Outdoor air pollution and
5 uncontrolled asthma in the San Joaquin Valley, California. J Epidemiol Community Health 64:
6 142-147.
7 Moore, K; Neugebauer, R; Lurmann, F; Hall, J; Brajer, V; Alcorn, S; Tager, I. (2008). Ambient ozone
8 concentrations cause increased hospitalizations for asthma in children: An 18-year study in
9 Southern California. Environ Health Perspect 116: 1063-1070.
10 Silverman, R. A., & Ito, K. (2010). Age-related association of fine particles and ozone with severe acute
11 asthma in New York City. J Allergy Clin Immunol, 125(2), 367-373 e365.
12 Smith, RL; Xu, B; Switzer, P. (2009b). Reassessing the relationship between ozone and short- term
13 mortality in U.S. urban communities. Inhal Toxicol 21: 37-61.
14 Strickland, M. J., Darrow, L. A., Klein, M., Flanders, W. D., Sarnat, J. A., Waller, L. A., et al. (2010).
15 Short-term Associations between Ambient Air Pollutants and Pediatric Asthma Emergency
16 Department Visits. Am J Respir Crit Care Med, 182, 307-316.
17 Tolbert, P. E., Klein, M., Peel, J. L., Sarnat, S. E., & Sarnat, J. A. (2007). Multipollutant modeling issues
18 in a study of ambient air quality and emergency department visits in Atlanta. J Expo Sci Environ
19 Epidemiol, 17 Suppl 2, S29-3 5.
20 U.S. CDC, 2010. Centers for Disease Control and Prevention, Behavioral Risk Factor Surveillance
21 System (BRFSS), 2010, Table "Table Cl Adult Self-Reported Current Asthma Prevalence Rate
22 (Percent) and Prevalence (Number) by State or Territory", available at:
23 http://www.cdc.gov/asthma/brfss/2010/current/tableCl.htm
24 U.S. Environmental Protection Agency (2001) Risk assessment guidance for Superfund. Vol. Ill, Part A.
25 Process for conducting probabilistic risk assessment (RAGS 3A). Washington, DC, United States
26 Environmental Protection Agency (EPA 540-R-02-002; OSWER 9285.7-45; PB2002 963302;
27 http://www.epa.gov/oswer/riskassessment/rags3adt/index.htm).
28 U.S. Environmental Protection Agency (2004) EPA's risk assessment process for air toxics: History and
29 overview. In: Air toxics risk assessment reference library. Vol. 1. Technical resource manual.
30 Washington, DC, United States Environmental Protection Agency, pp. 3-1 - 3-30 (EPA-453-K-
31 04-001A; http://www.epa.gov/ttn/fera/data/risk/vol_l/chapter_03.pdf).
32 U.S. Environmental Protection Agency (2007) Ozone Health Risk Assessment for Selected Urban Areas.
33 Office of Air Quality Planning and Standards, Research Triangle Park, EPA 452/R-07-009.
34 Available at: http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3cr.html
35 U.S. Environmental Protection Agency. (2011). Ozone National Ambient Air Quality Standards: Scope
36 and Methods Plan for Health Risk and Exposure Assessment, U.S. Environmental Protection
37 Agency, Research Triangle Park, NC. EPA-452/P-11-001.
38
7-83
-------
1 U.S. Environmental Protection Agency. (2012). Integrated Science Assessment for Ozone and Related
2 Photochemical Oxidants: Third External Review Draft, U.S. Environmental Protection Agency,
3 Research Triangle Park, NC. EPA/600/R-10/076C.
4 World Health Organization. (2008). Part 1: Guidance Document on Characterizing and Communicating
5 Uncertainty in Exposure Assessment, Harmonization Project Document No. 6. Published under
6 joint sponsorship of the World Health Organization, the International Labour Organization and
7 the United Nations Environment Programme. WHO Press, World Health Organization, 20
8 Avenue Appia, 1211 Geneva 27, Switzerland (tel.: +41 22 791 2476).
9 Zanobetti, A; Schwartz, J. (2008b). Mortality displacement in the association of ozone with mortality: An
10 analysis of 48 cities in the United States. Am J Respir Crit Care Med 177: 184-189.
7-84
-------
1 8 NATIONAL-SCALE RISK ASSESSMENT AND
2 REPRESENTATIVENESS ANALYSIS
3 8.1 INTRODUCTION
4 In this section we estimate nationwide premature mortality resulting from recent
5 exposures to ambient O3. There are two main goals for this assessment: (1) estimate the
6 incidence of premature mortality within the U.S. attributable to recent O3 concentrations (Section
7 7.3); (2) identify where the subset of counties assessed in the urban case study areas analysis fall
8 along the distribution of national county-level risk (Section 7.4). Compared with the urban scale
9 analysis in Section 7.2, this analysis includes full spatial coverage across the U.S. but has less
10 specificity in the risk-related attributes that are inputs to the health impact calculation. The
11 national scale analysis is therefore intended as a complement to the urban scale analysis,
12 providing both a broader assessment of O3-related health risks across the U.S. as well as an
13 evaluation of how well the urban study areas examined in Section 7.2 represent the full
14 distribution of O3-related health risks in the U. S. To perform this assessment we use a national-
15 scale "fused" spatial surface of seasonal average O3 concentrations from a 2007 simulation from
16 the Community Multiscale Air Quality (CMAQ) model (Byun and Schere, 2006) and 2006-2008
17 O3 air quality data. These gridded seasonal average O3 concentrations are input into the
18 environmental Benefits Mapping and Analysis Program (BenMAP; Abt Associates, 2010) to
19 estimate short-term O3-related premature mortality nationwide using city-specific mortality risk
20 estimates from the Bell et al. (2004) study of 95 urban communities and from the Zanobetti and
21 Schwartz (2008) study of 48 U.S. cities.
22 Using these methods, we estimate the total all-cause deaths associated with average
23 2006-2008 O3 levels across the continental U.S. We provide three analyses to give perspective
24 on the confidence in the estimates of O3-related mortality: (1) risk bounded by applying the
25 concentration-response functions down to zero (no O3 concentration cutoff) and down to the
26 lowest measured levels in Zanobetti and Schwartz (2008), (2) risk estimated only within the
27 urban areas included by Bell et al. (2004) and Zanobetti and Schwartz (2008); and (3) the
28 distribution of O3-related deaths across the range of 2006-2008 average O3 concentrations.
29 For the application of Bell et al. (2004) effect estimates for May-September, we estimate
30 18,000 (95% CI, 5,700-30,000) premature O3-related deaths with no concentration cutoff and
31 15,000 (95% CI, 4,800-25,000) with the LML cutoff of 7.5 ppb. The estimated percentage of
32 total county-level mortality attributable to O3 ranges from 0.4% to 4.2% (median 1.9%) with no
33 concentration cutoff and 0.3% to 3.5% (median 1.6%) with the LML cutoff of 7.5 ppb. For the
34 application of Zanobetti and Schwartz (2008) effect estimates for lune-August, we estimate
35 15,000 (95% CI, 5,800-24,000) premature O3-related deaths with no concentration cutoff and
8-1
-------
1 13,000 (95% CI, 4,900-21,000) with the LML cutoff of 7.5 ppb. The estimated percentage of
2 total county-level mortality attributable to Os ranges from 0.5% to 5.2% (median 2.5%) with no
3 concentration cutoff and 0.4% to 4.4% (median 2.1%) with the LML cutoff of 7.5 ppb. For both
4 epidemiology studies, we find that 85-90% of (Vrelated deaths occur in locations where the
5 May to September average 8-hr daily maximum or the June-August average 8-hr daily mean
6 (10am-6pm) 63 concentration is greater than 40 ppb, corresponding to 4th high 8-hr daily
7 maximum O3 concentrations ranging from approximately 50 ppb to 100 ppb.
8
9 8.1.1 Methods
10 This assessment combines information regarding estimated O3 concentrations, population
11 projections, baseline mortality rates, and mortality risk coefficients to estimate (Vrelated
12 premature mortality. Figure 1.1 below provides a conceptual diagram detailing each of the key
13 steps involved in performing this health impact assessment.
14 8.1.1.1 Estimates of Population Exposures to Ambient 63 Concentrations
15 BenMAP uses projections of the size and geographic distribution of the potentially
16 exposed population along with estimates of the ambient O3 concentrations to estimate population
17 exposure1. In contrast to the urban study areas analysis, the national scale analysis employed a
18 data fusion approach to take advantage of the accuracy of monitor observations and the
19 comprehensive spatial information of the CMAQ modeling system to create a national-scale
20 "fused" spatial surface of seasonal average 03. The spatial surface is created by fusing 2006-
21 2008 measured Os concentrations with the 2007 CMAQ model simulation, which was run for a
22 12 km gridded domain, using the EPA's Model Attainment Test Software (MATS; Abt
23 Associates, 2010), which employs the enhanced Voronoi Neighbor Averaging (eVNA) technique
24 (Timin et al., 2010). More details on the ambient measurements and the 2007 CMAQ model
25 simulation, as well as the spatial fusion technique, can be found in Wells et al. (2012). It should
26 also be noted that this same spatial fusion technique was employed for a national-scale risk
27 assessment by Fann et al. (2012) to produce "fused" spatial fields for Os and PM2.5 and in the PM
28 NAAQS REA to produce a national-scale spatial field for PM2.5 (U.S. EPA, 2010). Two "fused"
29 spatial surfaces were created for: (1) the May-September mean of the 8-hr daily maximum
30 (consistent with the metric used by Bell et al. 2004); and (2) the June-August mean of the 8-hr
31 daily mean from 10am to 6pm (consistent with the metric used by Zanobetti and Schwartz 2008)
32 O3 concentrations across the continental U.S. Figure 1.2 and Figure 1.3 show the geographic
1 Population exposure refers to the ambient concentrations estimated for populations living in specific
locations, rather than individual personal exposure to ozone (see Chapter 5 for a discussion of personal exposure
modeling).
8-2
-------
1 distribution of these spatial surfaces. Figure 1.4 shows the frequency and cumulative percent of
2 the seasonal average Os concentrations by gridcell, using both metrics. May-September average
3 8-hr daily maximum concentrations are most frequently in the 40-50 ppb range, while June-
4 August average 8-hr daily mean concentrations are more evenly distributed across a range of 20-
5 70 ppb. Maximum concentrations for the June-August mean of the 8-hr daily mean
6 concentrations from 10am to 6pm are generally higher than for the May-September mean of the
7 8-hr daily maximum concentrations since the seasonal definition is limited to the summer
8 months when Oj tends to be highest. The maximum, minimum, mean, median, and 95th
9 percentile concentrations for both 8-hr daily maximum and 8-hr daily mean are shown in Table
10 1.1. These seasonal average metrics are not equivalent to the averaging time for the current
11 NAAQS, which is based on the 4th highest value rather than seasonal mean, so the values should
12 not be directly compared against the NAAQS.
13
Census population
data
Population projections
14
15
16
17
18
Population estimates
Air quality monitoring
Population exposure
Mortality functions
Airquality modeling
Incidenceand
prevalence rates
Figure 1.1
Ozone-related
premature mortality
Conceptual diagram of data inputs and outputs for national short-term
mortality risk assessment
8-3
-------
2 Figure 1.2 Seasonal (May-September) average 8-hr, daily maximum baseline Os
3 concentrations (ppb) at the surface, based on a 2007 CMAQ model
4 simulation fused with average 2006-2008 observations from the Os monitor
5 network.
7
8
9
10
11
& <§>
S *T
ppb
Figure 1.3
Seasonal (June-August) average 8-hr, daily mean (10am-6pm) baseline Os
concentrations (ppb) at the surface, based on a 2007 CMAQ model
simulation fused with average 2006-2008 observations from the Os monitor
network.
8-4
-------
1
2
0)
3
tr
m
0.4
0.3
0.2
0.1
0
20
100%
0%
20
•June-August average 8hr daily mean IQam-Gpm
•May-September average 8hr daily maximum
30
40
50
60
70
30
40 50
Concentration (ppb)
60
70
4
5
6
7
8
9
10
Figure 1.4
Frequency and cumulative percent of May-September average 8-hr daily
maximum and the June-August 8-hr daily mean (10am-6pm) Os
concentration (ppb) by gridcell, based on 2006-2008 monitor observations
fused with 2007 CMAQ-modeled O3 levels.
§-5
-------
4
5
Table 1.1 Statistical characterization of the May-September average 8-hr daily
maximum and the June-August 8-hr daily mean (10am-6pm) Os
concentration (ppb), based on 2006-2008 monitor observations fused with
2007 CMAQ-modeled O3 levels.
Maximum
Minimum
Mean
Median
95th Percentile
May-September average 8-hr daily
maximum concentration (ppb)
65.0
19.7
41.8
42.6
51.6
June-August average daily 10am -
6pm daily mean concentration
(ppb)
85.5
18.0
40.4
41.3
55.1
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
8.1.1.2 Baseline incidence estimates
Epidemiological studies of the association between pollution levels and adverse health
effects generally provide a direct estimate of the relationship between air quality changes and the
relative risk of a health effect, rather than estimating the absolute number of avoided cases. For
example, a typical result might be that a 10 ppb decrease in daily 63 levels might, in turn,
decrease hospital admissions by 3%. The baseline incidence of the health effect is necessary to
convert this relative change into a number of cases. A baseline incidence rate is the estimated
number of cases of the health effect per year in the assessment location, as it corresponds to
baseline pollutant levels in that location. To derive the total baseline incidence per year, this rate
must be multiplied by the corresponding population number. For example, if the baseline
incidence rate is the number of cases per year per million people, that number must be multiplied
by the millions of people in the total population. We derive baseline incidence rates for mortality
from the CDC Wonder database (CDC, 2004-2006). The CDC Wonder database provides
baseline mortality estimates that are age-, cause-, and county-specific. As this database only
provides baseline incidence rates in 5-year increments, we use data for the year 2005, the closest
year to the analysis year 2007 used for the population and air quality modeling.
8.1.1.3 Population estimates
The starting point for estimating the size and demographics of the potentially exposed
population is the 2000 census-block level population, which BenMAP aggregates up to the same
grid resolution as the air quality model. BenMAP projects this 2000 population to the analysis
year of 2007 using county-level growth factors based on economic projections (Woods and
-------
1 Poole Inc., 2008). We use 2007 population because it matches both the year of the emissions
2 inventory and meteorology used for the air quality modeling.
O
4 8.1.1.4 Premature mortality estimates
5 To quantify the impact of 63 concentrations on mortality, we applied risk estimates
6 drawn from two major short-term epidemiological studies. These studies are consistent with
7 those used in the analysis of Os-related risk in selected urban areas (Section 7.2). We use city-
8 specific and national average risk estimates drawn from the Bell et al. (2004) study of 63 and
9 mortality in 95 U.S. urban communities between 1987 and 2000, and the Zanobetti and Schwartz
10 (2008) study of O3 and mortality in 48 U.S. cities between 1989 and 2000. City-specific effect
11 estimates for both studies are provided in Appendix 4-A.
12 Bell et al. (2004) found that the average non-accidental mortality increase across all 95
13 urban areas was 0.64% (95% posterior interval [PI], 0.41%-0.86%) for a 15 ppb increase in the
14 previous week's 8-hr daily maximum 63 concentration (equivalent to 0.43% for a 10 ppb
15 increase), based on yearly Os observations (often just the Os season, April to October). As the
16 national-scale analysis requires a single modeling period definition, the corresponding city-
17 specific effect estimates are applied to each day from May to September in BenMAP using the
18 2006-2008 average May to September mean 8-hr daily maximum Os concentration. The length
19 of the 63 season can affect the magnitude of mortality effect estimates. Bell et al. (2004)
20 reported that a 10 ppb increase in 24-hr average Os concentration was associated with a 0.52%
21 (95% PI, 0.27%-0.77%) increase in mortality using all O3 data and a 0.39% (95% PI, 0.13%-
22 0.65%) increase in mortality using only days from April to October. Since 63 values are
23 typically higher during the summer season, the higher effect estimate derived from year-round
24 63 data may yield an equivalent 63 mortality impact as the lower effect estimate derived from
25 the warm season O3 data only. For the second draft Risk and Exposure Assessment, EPA staff
26 proposes to use city-specific 8-hr daily maximum effect estimates for the warm season only, if
27 available, to model risk for the corresponding months.
28 Zanobetti and Schwartz (2008) found that the average total mortality increase across all
29 48 cities was 0.53% (95% confidence interval, 0.28%-0.77%) for a 10 ppb increase in June-
30 August 8-hr daily mean 63 concentration from 10 am to 6 pm, using a 0-3 day lag. We apply the
31 city-specific effect estimates that correspond to this national average effect estimate each day
32 from June to August in BenMAP using the 2006-2008 June to August mean 8-hr daily mean Os
33 concentration.
34 As this national assessment applies to the entire geographical scale of the continental
35 U.S. in a gridded format, it includes locations not covered by the Bell et al. (2004) and Zanobetti
3-7
-------
1 and Schwartz (2008) studies. For gridcells outside of the urban areas included by the
2 epidemiological studies, we assign the average effect estimate derived from all the urban areas
3 included in each of the studies ("national average"). Applying the national average estimate
4 takes advantage of a broader population and the variability among population response to 63
5 introduced by effect modifying characteristics, compared with an alternative approach of
6 assigning these gridcells the effect estimate from the nearest urban area. Since both national
7 average estimates from these studies are based on urban areas only, we have higher confidence in
8 their application to other U.S. urban areas than to rural areas. To demonstrate the magnitude of
9 the results for which we have the highest confidence, we present the percentage of estimated
10 deaths occurring within the urban areas included in the epidemiological studies. It should be
11 noted, however, that we also have high confidence in the magnitude of results in U.S. urban
12 areas that were excluded from the epidemiological studies, since results from the 48 city study by
13 Zanobetti and Schwartz (2008) were generally comparable to results from the larger 90 city
14 study by Bell et al. (2004). In addition, lower confidence in the results for rural areas does not
15 indicate that the mortality risk among populations living in such areas is unaffected by O3
16 pollution. Rather, the level of understanding for the Os-mortality relationship in these areas is
17 simply lower due to a lack of available epidemiological data at these levels.
18 The current literature does not support the existence of concentration thresholds below
19 which Os is not associated with health effects (U.S. EPA 2012a). However, the concentration-
20 response relationship is less certain at lower 63 concentrations since fewer observations at those
21 levels exist to inform the epidemiology studies. Consistent with the approach used in the urban
22 case studies (see Chapter 7), in addition to estimating risk for the full distribution of
23 concentrations (i.e. down to zero), we estimate risk occurring above the lowest measured level
24 (LML) in the underlying epidemiological studies. In order to apply the LML in all locations in
25 the U.S., we use the average LML across all cities in the Zanobetti and Schwartz (2008) study,
26 7.5 ppb, as a surrogate for the location specific LML. In the second draft REA we will explore
27 the implications of variability in the LML on the national mortality risk estimates. We apply the
28 LML of 7.5 ppb in estimating mortality risks using the C-R functions from both Zanobetti and
29 Schwartz (2008) and Bell et al. (2004) because the data on LMLs were not available for the Bell
30 et al. (2004) study. We also show the distribution of Os-related deaths by baseline Os
31 concentration to provide context for interpreting confidence in the magnitude of the mortality
32 estimates.
33
-------
1 8.1.1.5 Consideration of long-term O3-related mortality
2 The Integrated Science Assessment for O3 and Related Photochemical Oxidants (O3 ISA)
3 concluded that the evidence supports a likely to be causal relationship between long-term O3
4 exposure and respiratory effects, including respiratory morbidity and respiratory-related
5 mortality (U.S. EPA, 2012a). One major national-scale cohort study has found a significant
6 positive relationship between long-term O3 exposure and mortality (Jerrett et al. 2009). Another
7 study with a cohort limited to individuals with chronic conditions that might predispose to O3
8 effects (chronic obstructive pulmonary disease, diabetes, congestive heart failure, and
9 myocardial infarction) also found that long-term O3 exposure is associated with increased risk of
10 death in these groups (Zanobetti and Schwartz 2011). The O3 ISA concluded that these findings
11 are consistent and coherent with the evidence from the epidemiologic, controlled human
12 exposure, and animal toxicological studies for the effects of long-term exposure to O3 on
13 respiratory effects (U.S. EPA 2012a, Section 7.7.1).
14 After considering its strengths and weaknesses, EPA staff considers the Jerrett et al.
15 (2009) study to be an appropriate basis for estimating long-term O3-related respiratory mortality
16 risk in the 2nd draft REA. Key strengths of this study are that it included 1.2 million participants
17 in the American Cancer Society cohort from all 50 states, DC, and Puerto Rico; included O3 data
18 from 1977 (5 years before enrollment in the cohort began) to 2000; considered co-pollutant
19 models that controlled for PM^.s; and evaluated for threshold concentrations. Key limitations are
20 possible exposure misclassification and uncontrolled confounding by PM2.5 and temperature,
21 which are endemic to most long-term epidemiological studies. We note that while Jerrett et al.
22 (2009) found negative associations between O3 exposure and cardiovascular mortality when
23 controlling for PM2.5, null or negative associations are consistent with the evidence that PM2.5 is
24 strongly associated with cardiovascular disease (EPA 2009 PM ISA). Based largely on the
25 findings of this study and considering its strengths and weaknesses, the O3 ISA concluded that
26 the evidence was strong enough to be suggestive of a causal relationship for long-term O3
27 exposure and mortality.
28 Recent studies have used long-term O3-mortality relationships found by Jerrett et al.
29 (2009) to quantify the burden of mortality due to anthropogenic O3 globally (Anenberg et al.
30 2010, 2011) and for the U.S. specifically (Fann et al. 2012). These studies have found that using
31 Jerrett et al. (2009) long-term effect estimates yields O3-related mortality burden estimates that
32 are approximately two to four times larger than estimates based on Bell et al. (2004) short-term
33 effect estimates. Since long-term mortality relationships include both acute and chronic
34 exposure effects, the significantly larger mortality estimates calculated using long-term
35 concentration-mortality relationships suggest that considering only short-term mortality may
36 exclude a substantial portion of O3-related risk.
8-9
-------
1 EPA staff plans to quantify long-term O3-attributable respiratory-mortality in the 2nd draft
2 Risk and Exposure Assessment to be completed in November 2012 for two main reasons: (1) the
3 Os ISA has concluded that evidence indicates a likely to be causal relationship for long-term
4 ozone exposure and respiratory effects, including respiratory morbidity and respiratory-related
5 mortality, and (2) long-term respiratory-related mortality estimates may provide a more
6 comprehensive estimate of O3-related health risks, as they include both acute and chronic
7 exposure effects. To quantify long-term (Vattributable respiratory-related mortality risks, EPA
8 staff plans to use the respiratory mortality effect estimates from the Jerrett et al. (2009) two-
9 pollutant model that controlled for PM2.5 concentrations, applied to each gridcell across the entire
10 United States. This model found that a 10 ppb increase in the May-September average of the 1-
11 hr daily maximum O3 concentration was associated with a 4% (95% confidence interval, 1.0%-
12 6.7%) increase in respiratory mortality.
13
14 8.1.2 Results
15 Table 1.2 summarizes the estimated O3-related premature mortality associated with 2006-
16 2008 average O3 concentrations under various assumptions for the health impact function. For
17 the application of Bell et al. (2004) effect estimates for May-September, we estimate 18,000
18 (95% CI, 5,700-30,000) premature O3-related deaths with no concentration cutoff and 15,000
19 (95% CI, 4,800-25,000) with the LML cutoff of 7.5 ppb. For the application of Zanobetti and
20 Schwartz (2008) effect estimates for June-August, we estimate 15,000 (95% CI, 5,800-24,000)
21 premature O3-related deaths with no concentration cutoff, and 13,000 (95% CI, 4,900-21,000)
22 with the LML cutoff of 7.5 ppb. These results are calculated by applying the city-specific risk
23 estimates from each epidemiological study to the gridcells corresponding to each urban area, and
24 applying the national average risk estimate (based on all urban areas included in the study) from
25 the same study to all other gridcells. Figure 1.5 and Figure 1.6 show that estimated O3-related
26 mortality is most concentrated in highly populated counties or those counties with urban areas
27 found to have high effect estimates by Bell et al. (2004) or Zanobetti and Schwartz (2008).
28 Because the epidemiological studies included only selected urban areas, we are more
29 confident in the magnitude of the estimated O3-related deaths occurring within those urban areas.
30 Approximately 35% and 30% of the estimated O3-related deaths occur in the urban locations
31 included by Bell et al. (2004; 95 urban areas) and Zanobetti and Schwartz (2008; 48 urban
32 areas), respectively. We also have high confidence in extrapolating the national average effect
33 estimates to other urban areas, as the national average estimates are based on all urban areas
34 included by the study. While our confidence is lower when the national average effect estimates
8-10
-------
1 are extrapolated to rural areas, it is important to note that less certainty in the magnitude of (V
2 related deaths in rural areas does not imply a null effect of Os on health in these areas.
3 Table 1.2 also shows Os-related deaths estimated by applying the national average risk
4 estimate from the epidemiological studies to all gridcells in the United States. Compared with
5 applying city-specific effect estimates to the gridcells corresponding to each urban area, using
6 the national average effect estimate for all gridcells yields equivalent central estimates.
7 However, applying the national average also results in tighter confidence intervals since the
8 national average effect estimates had higher statistical power and thus tighter confidence bounds
9 compared with the effect estimates for individual cities.
10 Table 1.3 shows the mean, median, minimum, and maximum of the estimated percentage
11 of mortality attributable to ambient Os across all counties in the U.S. Using Bell et al. (2004)
12 effect estimates, the estimated percentage of total county-level mortality attributable to 63 ranges
13 from 0.4% to 4.2% (median 1.9%) with no concentration cutoff and from 0.3% to 3.5% (median
14 1.6%) with the LML cutoff of 7.5 ppb. For results using Zanobetti and Schwartz (2008) effect
15 estimates, the estimated percentage of total county-level mortality attributable to O3 ranges from
16 0.5% to 5.2% (median 2.5%) with no concentration cutoff and from 0.4% to 4.4% (median
17 2.1%) with the LML cutoff of 7.5 ppb. Figure 1.7 and Figure 1.8 show that the counties with the
18 highest percentage of mortality attributable to Os are typically those with the highest Os levels
19 (see Figure 1.2 and Figure 1.3).
20 Figure 1.9 displays the cumulative distribution of the percent of county-level total
21 mortality attributable to ambient Os using effect estimates from both epidemiological studies
22 with no concentration cutoff and using the LML cutoff. For the results based on Bell et al.
23 (2004) effect estimates with no concentration cutoff, 1.5% to 2.2% of total mortality is
24 attributable to Os for approximately 95% of U.S. counties. For the results based on Zanobetti
25 and Schwartz (2008) effect estimates with no concentration cutoff, between 2% and 3% of total
26 mortality is attributable to O3 for approximately 90% of U.S. counties.
27
28
8-11
-------
1 Table 1.2
2
Estimated O3-related premature mortality associated with 2006-2008 average
Os concentrations (95th percentile confidence interval)
Risk estimate and concentration cutoff
Bell et al. (2004), May-September
None
7.5 ppb (LML)
Zanobetti and Schwartz (2008), June-August
None
7.5 ppb (LML)
City-specific
effect
estimates1
18,000
(5,700-30,000)
15,000
(4,800-25,000)
15,000
(5,800-24,000)
13,000
(4,900-21,000)
National
average effect
estimate2
18,000
(12,000-24,000)
15,000
(8,200-22,000)
% reduced
from no
concentration
cutoff
-
17%
-
28%
4
5
6
7
1 City-specific effect estimates are applied to the gridcells lying within the cities defined in the epidemiological
studies. Average effect estimates across all cities included in the epidemiological studies (national average) are
applied to all other gridcells.
2National average effect estimates are based on the average of all cities included in the epidemiological studies.
8-12
-------
1
2
3
4
5
(b)
Figure 1.5 Estimated non-accidental deaths associated with average 2006-2008 May-
September average 8-hr daily maximum Os levels by county using Bell et al.
(2004) effect estimates and (a) no concentration cutoff, (b) LML cutoff of 7.5
ppb.
8-13
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
(b)
Figure 1.6 Estimated all-cause deaths associated with average 2006-2008 June-August
average 8-hr daily mean (10am-6pm) Os levels by county using Zanobetti
and Schwartz (2008) effect estimates and (a) no concentration cutoff, (b)
LML cutoff of 7.5 ppb.
8-14
-------
1
2
Table 1.3 Mean, median, minimum, and maximum of the estimated percentage of
mortality attributable to ambient Os for all U.S. counties.
Risk estimate and concentration cutoff
Bell et al. (2004), May-September
None
7.5 ppb (LML)
Zanobetti and Schwartz (2008), June-August
None
7.5 ppb (LML)
Mean
(%)
1.9
1.6
2.5
2.1
Median
(%)
1.9
1.6
2.5
2.1
Minimum
(%)
0.4
0.3
0.5
0.4
Maximum
(%)
4.2
3.5
5.2
4.4
4
5
Figure 1.7 Estimated percentage of May-September total mortality attributable to 2006-
2008 average Os levels by county using Bell et al. (2004) effect estimates and
(a) no concentration cutoff, (b) LML cutoff of 7.5 ppb.
8-15
-------
(b)
2 Figure 1.8 Estimated percentage of June-August total mortality attributable to 2006-
3 2008 average Os levels by county using Zanobetti and Schwartz (2008) effect
4 estimates and (a) no concentration cutoff, (b) LML cutoff of 7.5 ppb.
8-16
-------
100%
90%
IS)
0)
80%
60%
50%
« 70%
"S
§5,
2
c
0)
u
§. 40%
•| 30%
I 20%
3
10%
0%
0.5
1.5 2 2.5
Percentage of total mortality attributable to ozone
•Belletal. (2004), no cutoff
•Bell et al. (2004), cutoff=7.5 ppb
— — — Zanobetti and Schwartz (2008), no cutoff
--- Zanobetti and Schwartz (2008), cutoff=7.5 ppb
Figure 1.9 Cumulative distribution of county-level percentage of total mortality attributable to 2006-2008 average Os for
the U.S., using city-specific effect estimates. Results based on Bell et al. (2004) effect estimates are for non-
accidental mortality, while those based on Zanobetti and Schwartz (2008) effect estimates are for all-cause
mortality.
8-17
-------
Figure 1.10 shows the cumulative distribution of the county-level percent of total O3-
related deaths by O3 concentration. The mortality results based on Bell et al. (2004)
concentration-response functions are compared with the May-September average of the 8-hr
daily maximum O3 concentration, while those based on Zanobetti and Schwartz (2008)
concentration-response functions are compared with the June-August average of the 8-hr mean
O3 concentration from 10am to 6pm, consistent with the O3 concentration metrics used in each
study. The mortality results based on Zanobetti and Schwartz (2008) effect estimates are shifted
to the right of the mortality results based on the Bell et al. (2004) concentration response
functions because the seasonal averaging time for the results based on Zanobetti and Schwartz
(2008) is limited to the summer months when O3 tends to be highest. The 4th highest 8-hr daily
maximum O3 concentrations are typically 50% higher than the corresponding May-September
average of the 8-hr daily maximum concentration, with a range across all gridcells of 14% to
270% (Figure 1.11). For the June-August average of the 8-hr daily mean from 10am-6pm, the
corresponding 4th high 8-hr daily maximum concentrations are typically 60% higher, with a
range from 13% to 360% (Figure 1.12). For both epidemiology studies, we find that 85-90% of
O3-related deaths occur in locations where the May to September average 8-hr daily maximum or
June to August 8-hr daily mean (10am-6pm) O3 concentrations are greater than 40 ppb. When
the May to September average of the 8-hr daily maximum is 40 ppb, the 4th high 8-hr daily
maximum ranges from approximately 50 ppb to 90 ppb (Figure 1.11). When the June to August
average of the 8-hr daily mean from 10am-6pm is 40 ppb, the 4th high 8-h daily maximum ranges
from approximately 50 ppb to 100 ppb (Figure 1.12).
8-18
-------
100
IS)
4-1
5 80
•a
20
30 40 50 60
Seasonal average ozone concentration (ppb)
70
Figure 1.10
—Results based on Bell et al. (2004) effect estimates
^Results based on Zanobetti and Schwartz (2008) effect estimates
Cumulative percentage of total O3 deaths by baseline O3 concentration, using
city-specific effect estimates. O3 concentrations are reported as May-
September average 8-hr daily maximum for results based on Bell et al. (2004)
effect estimates and June-August average 8-hr mean (10am to 6pm) for
results based on Zanobetti and Schwartz (2008) effect estimates.
8-19
-------
160
2:1 line/
10 20 30 40 50 60 70 80 90
May-September average of 8-hr daily maximum ozone (ppb)
th
Figure 1.11 Gridcell values of 4 high 8-hr daily maximum Os concentrations versus
May-September average of 8-hr daily maximum Os concentrations for the
average of 2006-2008.
8-20
-------
.0
Q.
a.
0)
s
o
E
3
E
'x
re
E
_>
'ro
•o
oo
00
40
20
0 10 20 30 40 50 60 70 80 90
June-August average of daily 10am-6pm mean ozone (ppb)
Figure 1.12 Gridcell values of 4th high 8-hr daily maximum Os concentrations versus
June-August average of 8-hr daily 10am-6pm mean O3 concentrations for the
average of 2006-2008.
8.1.3 Discussion
We estimated the total all-cause deaths associated with short-term exposure to recent Os
levels across the continental U.S., using average 2006-2008 observations from the Os monitoring
network fused with a 2007 CMAQ simulation and city-specific (Vmortality effect estimates
from two short-term epidemiology studies. For the application of Bell et al. (2004) effect
estimates for May-September, we estimate 18,000 (95% CI, 5,700-30,000) premature O3-related
deaths with no concentration cutoff and 15,000 (95% CI, 4,800-25,000) with the LML cutoff of
7.5 ppb. The estimated percentage of total county-level mortality attributable to Os ranges from
0.4% to 4.2% (median 1.9%) with no concentration cutoff and from 0.3% to 3.5% (median
1.6%) with the LML cutoff of 7.5 ppb. For the application of Zanobetti and Schwartz (2008)
effect estimates for June-August, we estimate 15,000 (95% CI, 5,800-24,000) premature O3-
related deaths with no concentration cutoff and 13,000 (95% CI, 4,900-21,000) with the LML
8-21
-------
cutoff of 7.5 ppb. The estimated percentage of total county-level mortality attributable to 63
ranges from 0.5% to 5.2% (median 2.5%) with no concentration cutoff, and from 0.4% to 4.4%
(median 2.1%) with the LML cutoff of 7.5 ppb. For both epidemiology studies, we find that 85-
90% of (Vrelated deaths occur in locations where the seasonal average 8-hr daily maximum or
8-hr daily mean (10am-6pm) Oj concentration is greater than 40 ppb, corresponding to 4th high
8-hr daily maximum 63 concentrations ranging from approximately 50 ppb to 100 ppb.
A previous analysis estimated that short-term O3 exposure was associated with 4,700
(95% CI, 1,800-7,500) premature deaths nationwide, based on 2005 Os concentrations and Bell
et al. (2004) national average effect estimates (Fann et al. 2012). The results estimated here are
generally higher, depending on the concentration cutoff. These methods differ from those of
Fann et al. (2012) in two important ways. First, Fann et al. (2012) estimated risk only above
North American background, simulated 63 concentrations in the absence of North American
anthropogenic emissions, which was set to 22 ppb in the east and 30 ppb in the west. The
mortality results shown in Table 1.2 that are based on the most comparable concentration cutoff
of 29 ppb (10th percentile of O3 concentrations observed by Zanobetti and Schwartz 2008) are
approximately 40% larger than the estimate by Fann et al. (2012). Another important difference
is that Fann et al. (2012) used a national average mortality effect estimate for 8-hr daily
maximum Os during the warm season only, calculated using ratios of 24-hr mean concentrations
to 8-hr daily maximum concentrations (see Abt Associates 2010). The Bell et al. (2004) national
average beta used here, 0.000425, is based on yearly 63 data and is approximately 60% larger
than that used by Fann et al. (2012), 0.000261. Since the risk modeling period (and the seasonal
definition for the seasonal average 8-hr daily maximum concentration) was May to September
for both studies, the higher beta used here yields a larger 63 mortality estimate. These two
differences in methods explain the larger Os mortality estimates of this analysis compared with
the previous estimate by Fann et al. (2012). As previously mentioned, for the second draft Risk
and Exposure Assessment, EPA staff proposes to use city-specific 8-hr daily maximum effect
estimates for the warm season only, if available, to model risk for the corresponding months.
8.2 EVALUATING THE REPRESENTATIVENESS OF THE URBAN STUDY AREAS
IN THE NATIONAL CONTEXT
The goal in selecting the 12 urban study areas included in this risk assessment was
twofold: (1) to choose urban locations with relatively elevated ambient Os levels (in order to
evaluate risk for locations likely to experience some degree of risk reduction under alternative
standards) and (2) to include a range of urban areas reflecting heterogeneity in other Os risk
related attributes across the country. When selecting the cities, we took into account the
8-22
-------
following criteria:(l) availability of data; (2) 63 concentrations measured between 2006-2010;
(3) inclusion of sensitive populations; and (4) geographical heterogeneity. The "data availability"
criteria reflected the need for the urban area to have short-term mortality and morbidity study
data that could be used in the risk and exposure assessment, detailed air conditioning prevalence
data (that could be used in the exposure assessment analyses described in Chapter 5), and
baseline health information. The other selection criteria reflect the desire to include urban areas
that had relatively elevated ambient O3 levels and that geographically represented the different
regions of the U.Ss, as well as the desire to include sensitive population in the risk and exposure
assessment.
To further support interpretation of risk estimates generated in Section 7.2, we included
two analyses that assess the representativeness of the 12 urban study areas in the national
context. First, we assessed the degree to which the urban study areas represent the range of key
63 risk-related attributes that spatially vary across the nation. We have partially addressed this
issue by selecting urban study areas that provide coverage for different O3 regions of the country
(see Section 7.2). In addition, we have evaluated how well the selected urban areas represent the
overall U.S. for a set of spatially-distributed 63 risk related variables (e.g. weather,
demographics including socioeconomic status, baseline health incidence rates). This analysis,
which is discussed in Section 7.4.1, helps inform how well the urban study areas reflect national-
level variability in these key 63 risk-related variables. The second representativeness analysis,
which is discussed in Section 7.4.2, identified where the 23 counties comprising our 12 urban
study areas fall along the distribution of national county-level Os-attributable mortality risk. This
analysis allowed us to assess the degree of which the 12 urban study areas capture locations
within the U.S. likely to experience elevated levels of risk related to ambient Os.
We observe that the 23 counties for the 12 urban study areas considered in Section 7.2
capture urban areas that are among the most populated in the U.S., have relatively high O3 levels,
and represent the range of city-specific effect estimates found by Bell et al. (2004) and Zanobetti
and Schwartz (2008). These three factors suggest that the urban study areas capture overall risk
for the nation well, with a potential for better characterization of the high end of the risk
distribution. We find that the urban study areas are not capturing areas with the highest baseline
mortality rates, those with the oldest populations, and those with the lowest air conditioning
prevalence. These areas tend to have relatively low Os concentrations and low total population,
suggesting that the urban study areas are not missing high risk populations that have high 63
concentrations in addition to greater susceptibility per unit O3. The second representativeness
analysis demonstrated that the 12 urban study areas represent the full range of county-level Os-
related risk across the entire U.S.
-------
8.2.1 Analysis Based on Consideration of National Distributions of Risk-Related
Attributes
As noted above, the first representativeness analysis evaluated how well the urban study
areas reflect national -level variability in a series of O3 risk-related variables. For this analysis,
we first generated distributions for risk-related variables across U.S. counties and for the specific
counties considered in Section 7.2 from generally available data (e.g. from the 2000 Census,
Centers for Disease Control (CDC), or other sources). We then plotted the specific values of
these variables for the selected urban study areas on these distributions, and evaluated how
representative the selected study areas are of the national distributions for these individual
variables.
Estimates of risk (either relative or absolute, e.g. number of cases) within our risk
assessment framework are based on four elements: population, baseline incidence rates, air
quality, and the coefficient relating air quality and the health outcome (i.e. the O3 effect
estimates). Each of these elements can contribute to heterogeneity in risk across urban locations,
and each is variable across locations. In addition, there may be additional identifiable factors that
contribute to the variability of the four elements across locations. In this assessment, we examine
the representativeness of the selected urban area locations for the four main elements, as well as
factors that have been identified as influential in determining the magnitude of the C-R function
across locations.
While personal exposure is not incorporated directly into O3 epidemiology studies,
differences in the O3 effect estimates between cities is impacted by differing levels of exposure
which in turn are related to a number of exposure determinants. The correlation between
monitored O3 and personal O3 exposure also varies between cities. The O3 ISA has
comprehensively reviewed epidemiological and toxicological studies to identify variables which
may affect the O3 effect estimates used in the city-specific risk analysis in Section 7.2 and the
national-scale risk analysis in Section 7.3 (U.S. EPA 2012a Section 6.6). Broadly speaking,
determinants of the O3 effect estimates used in risk assessment can be grouped into three areas:
• Demographics: education, income, age, unemployment rates, race, body mass index and
physical conditioning, public transportation use, and time spent outdoors.
• Baseline health conditions: asthma, chronic obstructive pulmonary disease,
cardiovascular disease (atherosclerosis, congestive heart disease, atrial fibrillation,
stroke), diabetes, inflammatory diseases, and smoking prevalence.
• Climate and air quality: O3 levels, co-pollutant levels (annual mean PM^.s), temperatures
(days above 90 degrees, mean summer temp, 98th percentile temp), and air conditioning
prevalence.
8-24
-------
Based on these identified potential risk determinants, we identified datasets that could be
used to generate nationally representative distributions for each parameter. We were not able to
identify readily available national datasets for all variables. In these cases, if we were able to
identify a broad enough dataset covering a large enough portion of the U.S., we used that dataset
to generate the parameter distribution. In addition, we were not able to find exact matches for all
of the variables identified through our review of the literature. In cases where an exact match
was not available, we identified proxy variables to serve as surrogates. For each parameter, we
report the source of the dataset, its degree of coverage, and whether it is a direct measure of the
parameter or a proxy measure. The target variables and sources for the data are provided in Table
1.4. Summary statistics for the most relevant variables are provided in Table 1.5.
Figure 1.13 through Figure 1.19 show the cumulative distribution functions (CDF)
plotted for the nation for the four critical risk function elements (population, air quality, baseline
incidence, and the Os effect estimate), as well as where the urban study areas fall on the
distribution. These figures focus on critical variables representing each type of risk determinant,
e.g. we focus on all-cause and non-accidental mortality rates, but we also have conducted
analyses for cardiovascular and respiratory mortality separately. The vertical black lines in each
graph show the values of the variables for the individual urban study areas. The city-specific
values that comprise the national CDF for mortality risks found by Zanobetti and Schwartz
(2008) are also displayed on the graphs of those attributes, as the number of cities included in
that study is smaller (48 cities). The complete set of analyses is provided in Appendix 4-A.
These figures show that the selected urban study areas represent the upper percentiles of
the distributions of population and do not represent the locations with lower populations (urban
study areas are all above the 90th percentile of U.S. county populations). This is consistent with
the objectives of our case study selection process, e.g. we are characterizing risk in areas that are
likely to be experiencing excess risk due to 63 levels above alternative standards. The urban
study areas span the full range of seasonal average 8-hr daily maximum O3 concentrations in
monitored U.S. counties and the full distribution of Os risk coefficients across the cities included
by Bell et al. (2004) and Zanobetti and Schwartz (2008). We have included the two cities with
the highest risk coefficients found by Zanobetti and Schwartz (2008), New York City and
Detroit. We have not included the two highest found by Bell et al. (2004), Albuquerque and
Honolulu, but have included the 3rd and 4th highest, Atlanta and Boston. The urban study areas
do not capture the upper end of the distribution of baseline all-cause and non-accidental
mortality. The interpretation of this is that the case study risk estimates may not capture the
additional risk that may exist in locations that have the highest baseline mortality rates.
8-25
-------
Table 1.4 Data sources for O3 risk-related attributes
Potential risk
determinant
Metric
Year
Source
Degree of
national
coverage
Demographics
Age
Age
Age
Education
Income
Race
Percent age 85 years and
older
Percent age 65 years and
older
Percent age 14 years and
younger
Population with less than
high school diploma
Unemployment Percent unemployed
Per capita personal income
Percent nonwhite
2005 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
2005 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
2005 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
2000 USDA/ERS, All counties
http://www.ers.usda.gov/Data/Education/
2005 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
2005 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
2006 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
8-26
-------
Population
Total population
2008
Population density Population/square mile
2008
Urbanicity
ERS Classification Code
2003
Cumulative Estimates of Resident All counties
Population Change for the United States,
States, Counties, Puerto Rico, and Puerto
Rico Municipios: April 1, 2000 to July 1,
2008, Source: Population Division, U.S.
Census Bureau
Cumulative Estimates of Resident All counties
Population Change for the United States,
States, Counties, Puerto Rico, and Puerto
Rico Municipios: April 1, 2000 to July 1,
2008, Source: Population Division, U.S.
Census Bureau
County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
Climate and Air Quality
63 levels
63 levels
Os levels
O3 levels
PM2.5 levels
Monitored 4th high 8-hr
daily maximum
Seasonal mean 8-hr daily
maximum
Seasonal mean 1-hr daily
maximum
Seasonal mean
Monitored annual mean
2007
EPA Air Quality System (AQS)
Avg. 2006-2008 AQS
Avg. 2006-2008 AQS
Avg. 2006-2008 AQS
2007
AQS
725 Monitored
counties
671 Monitored
counties
671 Monitored
counties
671 Monitored
counties
617 Monitored
counties
8-27
-------
Temperature
Mean July temp
Relative Humidity Mean July RH
Ventilation
Percent residences with no
air conditioning
1941-1970 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
1941-1970 County Characteristics, 2000-2007 Inter- All counties
university Consortium for Political and
Social Research
2004 American Housing Survey 76 cities
Baseline Health Conditions
Baseline mortality
Baseline mortality
Baseline mortality
Baseline mortality
Baseline morbidity
Baseline morbidity
Baseline morbidity
Baseline morbidity
Obesity
Level of exercise
Level of exercise
All Cause
Non Accidental
Cardiovascular
Respiratory
Acute myocardial
infarction prevalence
Diabetes prevalence
Stroke prevalence
Congestive heart disease
prevalence
Body Mass Index
Vigorous activity 20
minutes
Moderate activity 30
minutes or vigorous
activity 20 minutes
CDC Wonder 1999-2005
CDC Wonder 1999-2006
CDC Wonder 1999-2007
CDC Wonder 1999-2008
2007 Behavioral Risk Factor Surveillance System
(BRFSS)
2007 BRFSS
2007 BRFSS
2007 BRFSS
2007 BRFSS
2007 BRFSS
2007 BRFSS
All counties
All counties
All counties
All counties
184 metropolitan
statistical areas
(MSA)
184 MS A
184 MS A
184 MS A
184 MS A
184 MS A
184 MS A
8-28
-------
Respiratory risk
factors
Smoking
Current asthma
Ever smoked
2007
2007
BRFSS
BRFSS
184 MS A
184 MS A
C-R Estimates
Mortality risk
Mortality risk
Mortality risk
Mortality risk
Non Accidental
All Cause
Cardiovascular
Respiratory
2004
2008
2008
2008
Bell et al. (2004)
Zanobetti and Schwartz (2008)
Zanobetti and Schwartz (2008)
Zanobetti and Schwartz (2008)
95 cities
48 cities
48 cities
48 cities
8-29
-------
Table 1.5 Summary statistics for selected O3 risk-related attributes
Risk Attribute
Demographics
Population
Population density (Pop/sq mile)
Median age (Years)
% Age 0 to 14 years
% Age 65+ years
% Age 85+ years
Unemployment rate (%)
% with less than high school diploma
Income ($)
% Non-white
% Commute by public transportation*
Health Conditions
Prevalence of CHD (%) *
Prevalence of asthma (%) *
Prevalence of diabetes (%) *
Prevalence of AMI (%) *
Prevalence of obesity (%) *
Prevalence of stroke (%) *
Prevalence of ever smoked (%)*
Prevalence of exercise (20 minutes,
%)*
Prevalence of exercise (30
minutes,%)*
Non-accidental mortality (deaths per
100,000 people)
Average
Urban
Study U.S.
Areas Dataset
1,642,198 97,020
10,378 258
35.7 38.6
20.7 19.0
11.3 14.9
1.7 2.1
5.7 5.4
20.9 22.6
40305 27367
36.4 13.0
7.1 1.6
3.6 4.3
8.5 8.1
8.1 8.5
3.6 4.1
24.7 26.0
2.6 2.7
18.3 19.6
29.5 28.0
50.2 49.7
756.2 950.6
Standard Deviation
Urban
Study U.S.
Areas Dataset
1,972,403 312,348
16,550 1,757
2.3 4.4
2.4 2.9
2.5 4.1
0.6 0.9
1.2 1.8
7.9 8.8
14238 6604
15.3 16.2
8.1 2.5
0.8 1.3
1.3 1.9
1.2 2.1
0.6 1.3
4.0 4.1
0.7 1.0
3.1 4.0
2.7 4.8
2.3 5.4
204.1 249.6
Maximum
Urban
Study U.S.
Areas Dataset
9,862,049 9,862,049
71,758 71,758
40.0 55.3
24.6 36.8
15.2 34.7
2.5 7.7
8.6 20.9
37.7 65.3
93377 93377
86.7 95.3
30.7 30.7
4.6 8.7
11.2 13.2
10.6 16.5
4.8 10.2
32.7 35.7
3.7 6.5
23.1 34.4
33.8 44.1
55.3 67.1
1139.5 1958.4
Minimum
Urban
Study U.S.
Areas Dataset
354,361 42
1,313 0
32.1 20.1
14.7 0.0
5.8 2.3
0.5 0.1
4.1 1.9
8.7 3.0
23513 5148
31.7 0.0
1.5 0.0
2.6 1.8
6.0 3.6
5.4 2.2
2.8 1.7
18.7 14.0
1.5 0.7
14.2 6.5
23.7 15.4
47.4 37.3
361.6 117.7
Sample Size
(# of counties or
cities)
Urban
Study U.S.
Areas Dataset
23 3143
23 3143
23 3141
23 3141
23 3141
23 3141
23 3133
23 3141
23 3086
23 3141
12 366
11 184
11 184
11 184
11 184
11 182
11 184
11 184
11 183
11 182
23 3142
8-30
-------
All cause mortality (deaths per
100,000 people)
Cardiovascular mortality (deaths per
100,000 people)
Respiratory mortality (deaths per
100,000 people)
Air Quality and Climate
C>3 4th high maximum 8-hr average
(ppb)
C>3 seasonal mean (ppb)
Os seasonal mean of maximum 8-hr
average (ppb)
O?, seasonal mean of 1-hr daily
maximum (ppb)
PM2 5 annual mean (ug/m3)
PM2 .5 98th %ile daily average (ug/m3)
Average temperature (°F)
July temperature long term average
(°F)
July Relative Humidity long term
average (%)
% No air conditioning*
C-R Estimates
Non-accidental mortality OT, risk*
All Cause mortality Os risk*
Respiratory mortality Oj risk*
Cardiovascular mortality Os risk*
810.1 1022.3
310.5 392.1
66.2 97.3
0.087 0.077
33.9 34.5
50.7 48.6
58.8 54.7
14.1 11.7
35.8 30.7
57.2 57.2
76.0 75.9
61.5 56.2
15.5 16.6
0.000515 0.000423
0.000627 0.000527
0.000877 0.000800
0.000898 0.000825
217.4 258.6
93.9 121.0
17.0 32.3
0.009 0.010
5.4 6.6
7.5 7.2
7.5 8.0
2.6 3.1
8.1 9.3
5.0 7.9
3.4 5.4
10.2 14.6
85.7 79.1
0.000138 0.000133
0.000314 0.000205
0.000282 0.000186
0.000173 0.000124
1257.8 2064.2
459.6 970.4
90.1 351.0
0.105 0.126
51.0 64.8
70.2 79.7
85.1 92.4
16.9 22.5
59.0 81.1
70.3 76.2
83.3 93.7
70.0 80.0
42.9 86.7
0.000705 0.000940
0.001092 0.001092
0.001424 0.001424
0.001064 0.001064
402.5 176.8
122.4 37.5
34.8 13.3
0.072 0.033
25.8 8.6
40.8 13.3
46.5 17.6
8.4 3.4
21.2 9.1
50.1 39.0
68.5 55.5
28.0 14.0
0.4 0.0
0.000331 0.000088
0.000163 0.000096
0.000307 0.000307
0.000418 0.000418
23 3142
23 3142
23 3136
23 725
22 671
22 671
22 671
23 617
23 617
23 202
23 3104
23 3104
12 76
12 95
12 48
12 48
12 48
* Attribute for which only city-specific data were available
8-31
-------
Comparison of Urban Case Study Area with U.S. Distribution (3143 U.S.
Counties) - Population
Urban case study areas
are all above the 90th
percentile of county
populations
1000
10000 100000
Population, 2008
1000000
10000000
•All Counties CDF
Case Study Counties
Figure 1.13 Comparison of distributions for key elements of the risk equation: Total
population.
8-32
-------
Comparison of Urban Case Study Area with U.S. Distribution
(671 U.S. Counties with Ozone Monitors) -
Seasonal Mean 8-hr Daily Max Ozone
1/1
.0)
C
3
O
u
0)
O
'E
O
^
'o
SB
J.UU/O
90%
80%
70%
60%
50%
40%
30%
20%
10%
no/
30
40
50
60
70
Seasonal Mean 8-hr Daily Max Ozone Concentration, Average 2006-2008
(ppb)
•All Counties CDF
Case Study Counties
Figure 1.14 Comparison of distributions for key elements of the risk equation: Seasonal
mean 8-hr daily maximum Os concentration.
8-33
-------
Comparison of Urban Case Study Area with U.S. Distribution (725 U.S.
Counties with Ozone Monitors) -
4th High 8-hr Daily Maximum Ozone
40
50 60 70 80 90 100
4th High 8-hr Daily Maximum Ozone, 2007 (ppb)
110
•All Counties CDF
Case Study Counties
Figure 1.15 Comparison of distributions for key elements of the risk equation: 4th
highest 8-hr daily maximum Os concentration.
8-34
-------
Comparison of Urban Case Study Area with U.S. Distribution (3137 U.S.
Counties) - All Cause Mortality
Urban case
study areas are
all below the
85th percentile
of county all
cause mortality
500
600 700 800 900 1000 1100 1200 1300 1400
All Cause Mortality per 100,000 Population, 1999-2005
1500
•All Counties CDF
Case Study Counties
Figure 1.16 Comparison of distributions for key elements of the risk equation: Baseline
all-cause mortality rate.
8-35
-------
Comparison of Urban Case Study Area with U.S. Distribution (3135 U.S.
Counties) - Non Accidental Mortality
100%
Urban case study
counties are all
below the 80th
percentile of non
accidental
mortality
500
700
900
^~ ,~~ ^~~ 1100 1300 1500
Non Accidental Mortality per 100,000 Population, 1999-2005
•All Counties CDF • Case Study Counties
Figure 1.17 Comparison of distributions for key elements of the risk equation: Baseline
non-accidental mortality rate.
8-36
-------
Comparison of Urban Case Study Area with U.S. Distribution (48 Z&S
Cities) - All Cause Mortality Risk (p)
0%
0.0002
0.0004 0.0006 0.0008 0.001
All Cause Mortality Risk Coefficient ((J)
0.0012
All cities
•All Cities CDF
Case Study Cities
Figure 1.18 Comparison of distributions for key elements of the risk equation: All-cause
mortality risk coefficient from Zanobetti and Schwartz (2008).
8-37
-------
Comparison of Urban Case Study Area with U.S. Distribution (95
NMMAPS Cities) -
Non Accidental Mortality Risk ((J)
0%
0.0002
0.0004 0.0006 0.0008 0.001
Non Accidental Mortality Risk Coefficient ((J)
0.0012
•All Cities CDF
Case Study Cities
Figure 1.19 Comparison of distributions for key elements of the risk equation: Non-
accidental mortality risk coefficient from Bell et al. (2004).
Figure 1.20 through Figure 1.25 show national CDFs and the urban study area values for
several selected potential risk attributes. These potential risk attributes do not directly enter the
risk equations, but have been identified in the literature as potentially affecting the magnitude of
the Os C-R functions reported in the epidemiological literature. Comparison graphs for other risk
attributes are provided in Appendix 4-A. The selected urban study areas do not capture the
higher end percentiles of several risk characteristics, including populations 65 years and older,
baseline cardiovascular disease prevalence, baseline respiratory disease prevalence, and smoking
prevalence. Summarizing the analyses of the other risk attributes, we conclude that the urban
study areas provide adequate coverage across population, population density, Os levels (seasonal
mean, seasonal mean 8-hr daily maximum, and seasonal mean 1-hr daily maximum), PM2 5 co-
pollutant levels, temperature and relative humidity, unemployment rates, percent non-white
population, asthma prevalence obesity prevalence, income, and less than high school education.
We also conclude that while the urban study areas cover a wide portion of the distributions, they
do not provide coverage for the upper end of the distributions of percent of population 65 and
8-38
-------
older (below 60th percentile), percent of population 85 years and older (below 75th percentile),
prevalence of angina/coronary heart disease (below 70th percentile), prevalence of diabetes
(below 85th percentile), stroke prevalence (below 90th percentile), prevalence of heart attack
(below 80th percentile), prevalence of smoking (below 85th percentile), all-cause mortality rates
(below 85th percentile), non-accidental mortality rates (below 80th percentile), cardiovascular
mortality rates (below 75th percentile) and respiratory mortality rates (below 50th percentile), and
percent of residences without air conditioning (below 90th percentile). In addition, the urban
study areas do not capture the highest or lowest ends of the distribution of exercise prevalence
and do not capture the low end of the distribution of public transportation use (above the 65th
percentile).
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Percent Younger than 15 Years Old
14
16
18 20 22 24 26
% Younger than 15 Years Old, 2005
28
30
•All Counties CDF
Case Study Counties
Figure 1.20 Comparison of distributions for selected variables expected to influence the
relative risk from Os: Percent of population younger than 15 years old.
8-39
-------
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Percent 65 Years and Older
Urban case study
counties are all
below the 60th
percentile of county
% of population 65
years and older
11 13 15 17 19 21
% 65 Years and Older, 2005
23
•All Counties CDF
Case Study Counties
25 27
Figure 1.21 Comparison of distributions for selected variables expected to influence the
relative risk from Os: Percent of population age 65 years and older.
8-40
-------
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Income per capita
100%
90%
80%
.2 70%
4-1
§ 60%
i/5 50%
% 40%
° 30% -
20%
10%
0%
10000 20000 30000 40000 50000 60000 70000 80000 90000 100000
Income per capita, 2005 ($)
•All Counties CDF
Case Study Counties
Figure 1.22 Comparison of distributions for selected variables expected to influence the
relative risk from Os: Income per capita.
8-41
-------
Comparison of Urban Case Study Area with U.S. Distribution (All U.S.
Counties) - July Temperature
66
68
70 72 74 76 78 80 82
Mean Temperature for July, 1941-1970 (F)
84
86
•All Counties CDF
Case Study Counties
Figure 1.23 Comparison of distributions for selected variables expected to influence the
relative risk from Os: July temperature.
8-42
-------
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) - Asthma Prevalence
8 10
Asthma Prevalence, 2007(%)
12
•All Counties CDF
Case Study Counties
14
Figure 1.24 Comparison of distributions for selected variables expected to influence the
relative risk from Os: Asthma prevalence.
8-43
-------
Comparison of Urban Case Study Area with U.S. Distribution (76 Cities) -
Air Conditioning Prevalence
Urban study areas
are all below the
90th percentile of
percent of
residences with no
air conditioning
10
20
30 40 50 60 70
No air conditioning, 2004 (%)
80
90
100
•All Cities CDF
Case Study Cities
Figure 1.25 Comparison of distributions for selected variables expected to influence the
relative risk from Os: Air conditioning prevalence.
Based on the above analyses, we can draw several inferences regarding the
representativeness of the urban case studies. First, the case studies represent urban areas that are
among the most populated in the U.S. Second, they represent areas with relatively high levels of
Os (4th high 8-hr daily maximum, seasonal mean 8-hr daily maximum, seasonal mean 1-hr daily
maximum, and seasonal mean). Third, they capture well the range of city-specific effect
estimates found by Bell et al. (2004) and Zanobetti and Schwartz (2008) studies. These three
factors would suggest that the urban study areas should capture well overall risk for the nation,
with a potential for better characterization of the high end of the risk distribution. However, there
are several other factors that suggest that the urban study areas may not be representing areas that
may have a high risk per ppb of Os. The analysis suggests that the urban study areas are not
capturing areas with the highest baseline mortality rates nor those with the oldest populations.
These areas may have higher risks per ppb of 63, and thus the high end of the risk distribution
may not be captured. However, the impact on characterization of overall Os risk may not be as
large, since overall Os risk depends on a combination of factors, including O?, levels and total
population, in addition to age distribution and baseline mortality rates.
8-44
-------
It should be noted that several of the factors with underrepresented tails, including age
and baseline mortality are spatially correlated (R=0.81), so that certain counties which have high
proportions of older adults also have high baseline mortality and high prevalence of underlying
chronic health conditions. Because of this, omission of certain urban areas with higher
percentages of older populations, for example, cities in Florida, may lead to underrepresentation
of high risk populations. However, with the exception of areas in Florida, most locations with
high percentages of older populations have low overall populations, less than 50,000 people in a
county. And even in Florida, the counties with the highest Os levels do not have a high percent of
older populations. This suggests that while the risk per exposed person per ppb of 63 may be
higher in these locations, the overall risk to the population is likely to be within the range of risks
represented by the urban case study locations.
The urban study areas also do not capture the highest end of percent of residences
without air conditioning. If the cities with the lowest air conditioning prevalence also have high
63 levels, we could be missing a high risk portion of the population that is exposed to 63 indoors
as air infiltrates indoors from outdoors. However, 4th highest 8-hr daily maximum Oj, levels in
the cities in the top 10th percentile of percentage of residences without air conditioning (mainly
in northern California and Washington) are approximately average (0.08 ppm) or lower than
average. The relatively low Os concentrations in these areas with low air conditioning
prevalence suggests that we are not excluding a high risk population that has both low air
conditioning prevalence and high 63 concentrations, and the overall risk to the population is
likely to be within the range of risks represented by the urban case study locations.
There is no nationally representative data base that will allow us to compare the time
spent outdoors among persons residing in each of the urban case study areas. As time spent
outdoors is an important personal attribute that influences exposure to Os (US EPA, 2007), EPA
staff is considering evaluating data from the American Time Use Survey (ATUS) for the 2nd
draft REA. ATUS is a recent (2003-2011) nationally representative survey that contains
information on people's time expenditure, many of whom reside in the urban case study areas
modeled in this assessment. ATUS does however have a few noteworthy limitations: (1) there
are no survey participants under 15 years of age, (2) time spent at home locations is neither
distinguished as indoors or outdoors, (3) missing or unknown location data can comprise a
significant portion of a persons' day (on average, about 40% (George and McCurdy, 2009)), (4)
only a single day is available for each participant, and (5) influential meteorological conditions
affecting time expenditure were not recorded (e.g., daily temperature and precipitation (Graham
and McCurdy, 2004)). To overcome a few of the ATUS limitations, EPA staff is planning to (1)
use particular activity codes (e.g., participation in a sport) to better approximate outdoor time
expenditure, (2) link National Climatic Data Center (NCDC) meteorological data to each ATUS
8-45
-------
diary, and (3) control for diaries having significant missing or unknown location information to
allow for a relative comparison of outdoor time across the urban case study areas.
8.2.2 Analysis Based on Consideration of National Distribution of O3-Related Mortality
Risk
In this section we discuss the second representativeness analysis which identified where
the counties comprising the 12 urban study areas fall along a distribution of estimated national-
scale mortality risk. This assessment reveals whether the baseline O3 mortality risks in the 12
urban case study areas represent more typical or higher end risk relative to the national risk
distribution (see Section 7.3). For ease of comparison, we use only the estimates of mortality
associated with total O3 (i.e. no concentration cutoff). Applying a concentration cutoff is
unlikely to change the conclusions of this assessment.
The results of this representativeness analysis are presented graphically in Figure 1.26
and Figure 1.27, which display the cumulative distribution of total mortality attributable to
ambient O3 at the county level developed as part of the national-scale analysis (see Figure 1.9).
Values for the 23 counties included in the urban case study analysis are then superimposed on
top of the cumulative distribution to assess the representativeness of the urban case study areas.
For the results based on Bell et al. (2004) effect estimates, Atlanta and Boston have the highest
percentage of total mortality attributable to ambient O3 of the 12 urban study areas and are
located at the highest end of the distribution of U.S. O3-related mortality risk. Of the 12 urban
study areas, these two cities had the highest effect estimates found by Bell et al. (2004; See
Appendix 4-A). Overall, O3 mortality risk in the 12 urban study areas are representative of the
full distribution of U.S. O3-related mortality risk, with the percentage of total mortality
attributable to O3 ranging from 1.4% to 3.6%, assuming no concentration cutoff.
For the results based on Zanobetti and Schwartz (2008) effect estimates, Detroit and New
York City are at the very highest end of the U.S. distribution of county-level risk of mortality
due to ambient O3. These two cities had the highest effect estimates of the 48 cities included in
the study (see Appendix 4-A). For this study, Houston and Los Angeles had the lowest risk and
were located at the very lowest end of the U.S. distribution of county-level risk of mortality due
to ambient O3. These two cities had the lowest effect estimates found by Zanobetti and Schwartz
(2008). The low effect estimates in Houston and Los Angeles could be due to several factors.
Both cities cover a large spatial extent and have high rates of time spent driving, possibly leading
to exposure misclassification in the underlying epidemiologic study. Houston also has a very
high rate of air conditioning use (nearly 100% of residences) and Los Angeles has been shown to
have high rates of adaptive behavior on high ambient O3 days (i.e. more time spent indoors as a
8-46
-------
result of high ambient Os concentrations; Neidell 2009, 2010), both of which would lead to
lower personal Os exposure relative to other cities. Overall, Os mortality risk in the 12 urban
study areas are representative of the full distribution of U.S. Os-related mortality risk, with the
percentage of total mortality attributable to 63 ranging from 0.6% to 4.8%, assuming no
concentration cutoff.
1/1
Q)
C
3
o
u
0)
00
ro
0)
u
QJ
Q.
0)
S
"5
E
3
U
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
0.5 1 1.5 2 2.5 3 3.5 4
Percentage of baseline mortality attributable to ozone
^—Results based on Bell et al. (2004) effect estimates, no LML
• Selected urban study area, no LML
Figure 1.26. Cumulative distribution of county-level percentage of total non-accidental
mortality attributable to 2006-2008 average O3 for the U.S. and the locations
of the selected urban study areas along the distribution, using Bell et al.
(2004) effect estimates.
8-47
-------
1/1
.0)
C
|
'o
0)
00
ro
0)
i_
0)
Q.
0)
•^
ro
|
u
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
0.5 1.5 2.5 3.5 4.5 5.5
Percentage of baseline mortality attributable to ozone
^^Results based on Zanobetti and Schwartz (2008) effect estimates, no LML
• Selected urban study areas, no LML
Figure 1.27. Cumulative distribution of county-level percentage of total all-cause
mortality attributable to 2006-2008 average Os for the U.S. and the locations
of the selected urban study areas along the distribution, using Zanobetti and
Schwartz (2008) effect estimates.
8-48
-------
8.2.3 Discussion
We conducted two analyses to assess the representativeness of the 12 urban study areas
examined in Section 7.2 in the national context. First, we assessed the degree to which the urban
study areas represent the range of key Os risk-related attributes that spatially vary across the
nation. We examined both the specific elements of our risk assessment framework (population,
baseline incidence rates, air quality, and the coefficient relating air quality and the health
outcome) in addition to factors that have been identified as influential in determining the
magnitude of the C-R function across locations (demographics, baseline heath conditions, and
climate and air quality attributes). The second representativeness analysis, which is discussed in
Section 7.4.2, identified where the 12 urban study areas fall along the distribution of national
county-level (Vattributable mortality risk. This analysis allowed us to assess the degree of
which the 12 urban study areas capture locations within the U.S. likely to experience elevated
levels of risk related to Os exposure.
We observe that the 23 counties for the 12 urban study areas considered in Section 7.2
capture urban areas that are among the most populated in the U.S., have relatively high Os levels,
and represent the range of city-specific effect estimates found by Bell et al. (2004) and Zanobetti
and Schwartz (2008). These three factors suggest that the urban study areas capture overall risk
for the nation well, with a potential for better characterization of the high end of the risk
distribution. We find that the urban study areas are not capturing areas with the highest baseline
mortality rates, those with the oldest populations, and those with the lowest air conditioning
prevalence. These areas tend to have relatively low Os concentrations and low total population,
suggesting that the urban study areas are not missing high risk populations that have high 63
concentrations in addition to greater susceptibility per unit Os. The second representativeness
analysis demonstrated that the 12 urban study areas represent the full range of county-level O^-
related risk across the entire U.S. We conclude from these analyses that the 12 urban study areas
adequately represent Os-related risk across the U.S.
8-49
-------
8.3 REFERENCES
Abt Associates, Inc. (2010). Model Attainment Test Software (Version 2). Bethesda, MD.
Prepared for the U.S. Environmental Protection Agency Office of Air Quality Planning
and Standards. Research Triangle Park, NC. Available on the Internet at:
http://www.epa.gov/scram001/modelingapps.mats.htm.
Abt Associates, Inc. (2010). Environmental Benefits and Mapping Program (Version 4.0).
Bethesda, MD. Prepared for U.S. Environmental Protection Agency Office of Air Quality
Planning and Standards. Research Triangle Park, NC. Available on the Internet at
.
Anenberg, S.C., JJ. West, L.W. Horowitz, D.Q. Tong. (2010). An estimate of the global burden
of anthropogenic 03 and fine particulate matter on premature human mortality using
atmospheric modeling. Environ Health Perspect, 118:1189-1195.
Anenberg, S.C., JJ. West, L.W. Horowitz, D.Q. Tong. (2011). The global burden of air pollution
mortality: Anenberg et al. respond. Environ Health Perspect, 119:A158-A425.
Bell, M.L., A. McDermott, S.L. Zeger, J.M. Samet, F. Dominici. (2004). 03 and short-term
mortality in 95 US urban communities, 1987-2000. JAMA, 292:2372-2378.
Byun, D., and K.L. Schere. (2006). Review of Governing Equations, Computational Algorithms,
and Other Components of the Models-3 Community Multiscale Air Quality (CMAQ)
Modeling System. Applied Mechanics Reviews, 59:51-77.
Centers for Disease Control: Wide-ranging OnLine Data for Epidemiological Research (CDC-
Wonder) (data from years 2004-2006), Centers for Disease Control and Prevention
(CDC), U.S. Department of Health and Human Services, Available on the Internet at
http://wonder.cdc.gov.
Fann N, Lamson AD, Anenberg SC, Wesson K, Risley D, Hubbell BJ. (2012). Estimating the
national public health burden associated with exposure to ambient PM2.5 and 03. Risk
Analysis, 32:81-95.
George BJ and McCurdy T. (2009). Investigating the American Time Use Survey from an
exposure modeling perspective. JESEE. 21:92-105.
Graham S and McCurdy T. (2004). Developing meaningful cohorts for human exposure
models. JEAEE. 14:23-43.
8-50
-------
Hollman, F.W., TJ. Mulder, and I.E. Kalian. (2000). Methodology and Assumptions for the
Population Projections of the United States: 1999 to 2100. Population Division Working
Paper No. 38, Population Projections Branch, Population Division, U.S. Census Bureau,
Department of Commerce.
Jerrett, M., R.T. Burnett, C.A. Pope III, K. Ito, G. Thurston, D. Krewski, Y. Shi, E. Calle, M.
Thun. (2009). Long-term 03 exposure and mortality. N. Eng. J. Med., 360:1085-1095.
NeidellM. (2009). Information, avoidance behavior and health. J Human Res. 44:450-478.
Neidell M. (2010). Air quality warnings and outdoor activities: evidence from Southern
California using a regression discontinuity approach design. J Epidemiol Community
Health. 64:921-926.
Timin B, Wesson K, Thurman J. Application of Model and Ambient Data Fusion Techniques to
Predict Current and Future Year PM2.5 Concentrations in Unmonitored Areas. (2010). Pp.
175-179 in Steyn DG, Rao St (eds). Air Pollution Modeling and Its Application XX.
Netherlands: Springer.
U.S. Environmental Protection Agency. (2012a). Integrated Science Assessment for 03 and
Related Photochemical Oxidants: Third External Review Draft, U.S. Environmental
Protection Agency, Research Triangle Park, NC.
U.S. Environmental Protection Agency. (2010). Quantitative Health Risk Assessment for
Particulate Matter, U.S. Environmental Protection Agency, Research Triangle Park, NC,
EPA-452/R-10-005.
U.S. Environmental Protection Agency. (2009). Integrated Science Assessment for Parti culate
Matter, U.S. Environmental Protection Agency, Research Triangle Park, NC,
EPA/600/R-08/139F.
US EPA. (2007). 03 Population Exposure Analysis for Selected Urban Areas. U.S.
Environmental Protection Agency, Research Triangle Park, NC, USA, (EPA-452/R-07-
010). Available at:
http://www.epa.gov/ttn/naaqs/standards/03/data/2007_07_03_exposure_tsd.pdf
8-51
-------
Wells, B., Wesson, K., Jenkins, S. (2012). Analysis of Recent U.S. Ozone Air Quality Data to
Support the O3 NAAQS Review and Quadratic Rollback Simulations to Support the First
Draft of the Risk and Exposure Assessment. Available on the Internet at:
http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3 2008 rea.html
Woods and Poole Inc. (2008). Population by Single Year of Age CD. CD-ROM. Woods and
Poole Economics, Inc.
Zanobetti, A., and J. Schwartz. (2008). Mortality displacement in the association of 03 with
mortality: An analysis of 48 cities in the United States. Am J Resp Crit Care Med,
177:184-189.
Zanobetti, A., and J. Schwartz. (2011). 03 and survival in four cohorts with potentially
predisposing diseases. Am J Resp Crit Care Med, 194:836-841.
Zhang, L., DJ. Jacob, N.V. Smith-Downey, D.A. Wood, D. Blewitt, C.C. Carouge, A. van
Donkelaar, D.B. A. Jones, L.T. Murray, Y. Wang. (2011). Improved estimate of the
policy-relevant background 03 in the United States using the GEOS-Chem global model
with l/2°x2/3° horizontal resolution over North America. Atmos Environ, 45:6769-6776.
8-52
-------
1 9 SYNTHESIS
2 This assessment has estimated exposures to 63 and resulting health risks for both current
3 Os levels and Os levels after simulating just meeting the current primary Os standard of 0.075
4 ppm for the 4th highest 8-hour daily maximum, averaged over 3 years. The results from these
5 assessments will help inform consideration of the adequacy of the current Os standards in the
6 first draft Policy Assessment.
7 The remaining sections of this chapter provide key observations regarding the exposure
8 assessment (Section 9.1), lung function risk assessment (Section 9.2), epidemiology based risk
9 assessment (Section 9.3), and a set of integrated findings providing insights drawn from
10 evaluation of the full assessment (Section 9.4).
11 9.1 SUMMARY OF KEY RESULTS OF POPULATION EXPOSURE ASSESSMENT
12 The first draft population exposure assessment evaluated exposures to 03 using the
13 APEX exposure model for the general population, all school-aged children (ages 5-18), and
14 asthmatic children, with a focus on populations engaged in moderate or greater exertion, for
15 example, children engaged in outdoor recreational activities. The strong emphasis on children
16 reflected the finding of the last O3 NAAQS review (EPA, 2007) and the ISA (EPA, 2012,
17 Chapter 8) that children are an important at-risk group. Children breathe more air per pound of
18 body weight, are more likely than adults to have asthma, and their lungs continue to develop
19 until they are fully grown.
20 In this first draft, exposure is assessed for 4 cities, Atlanta, Denver, Los Angeles, and
21 Philadelphia, for recent air quality (2006-2010) and for air quality simulated to just meet the
22 current standard. The analysis provided estimates of the percent of children exposed to
23 concentrations above three health-relevant 8-hour average Os exposure benchmarks: 0.060,
24 0.070, and 0.080 ppm. The ISA includes studies showing significant effects at each of these
25 benchmark levels (U.S. EPA, 2012). These benchmarks were selected so as to provide some
26 perspective on the public health impacts of Os-related health effects that have been demonstrated
27 in human clinical and toxicological studies, but cannot currently be evaluated in quantitative risk
28 assessments, such as lung inflammation and increased airway responsiveness. In addition, the
29 first draft exposure assessment also identified the specific microenvironments and activities most
30 important for exposure and evaluated their duration and time of the day persons were engaged in
31 them, with a focus on persons experiencing the highest daily maximum 8-hour exposure within
32 each study area.
33 It should also be noted that with regard to the exposure estimates, the APEX model is not
34 proficient at modeling activity patterns that lead to repeated exposures to elevated ozone
9-1
-------
1 concentrations. As a result, while we are able to report the percent of children with at least one
2 exposure greater than the alternative exposure benchmarks, we are not able to report with
3 confidence the percent of children with more than one exposure. Children with repeated
4 exposures may be at greater risk of significant health effects. In addition, we were only able to
5 model exposure in four cities for this first draft assessment. It is likely that variation in exposure
6 will be larger when we have modeled the full set of 16 cities in the second draft REA.
7 The key results of the first draft exposure assessment include:
8 • Exposure Assessment for Recent Conditions
9 o The average (i.e., average across years 2006 to 2010) percentages of
10 school age children estimated to experience one or more exposures per
11 year to 8-hour 63 concentrations at and above 0.060 ppm, while at
12 moderate or greater exertion, were approximately 20% for Denver
13 (corresponding to 109,000 children), 22% for Atlanta (corresponding to
14 189,000 children), 26% for Philadelphia (corresponding to 297,000
15 children), and 32% for Los Angeles (corresponding to 1,150,000
16 children). There was considerable variability in these percentages across
17 the years evaluated, ranging from approximately 12 to 30% in Denver, 10
18 to 36% in Atlanta, 9 to 34% in Philadelphia, and 23 to 37% in Los
19 Angeles. When considering exposures at and above 0.060 ppm in
20 asthmatic children at moderate or greater exertion, the results were similar
21 in term of percentages, corresponding to average numbers of exposed
22 asthmatic children of approximately 10,000 per year in Denver, 19,000 per
23 year in Atlanta, 35,000 per year in Philadelphia, and 110,000 per year in
24 Los Angeles.
25 o The average (i.e., average across years 2006 to 2010) percentages of
26 school age children estimated to experience one or more exposures per
27 year to 8-hour 63 concentrations at and above 0.070 ppm, while at
28 moderate or greater exertion, were approximately 4% for Denver
29 (corresponding to 22,000 children), 9% for Atlanta (corresponding to
30 75,000 children), 10% for Philadelphia (corresponding to 117,000
31 children), and 15% for Los Angeles (corresponding to 559,000 children).
32 There was considerable variability in these percentages across the years
33 evaluated, ranging from approximately 1 to 10% in Denver, 2 to 19% in
34 Atlanta, 1 to 16% in Philadelphia, and 8 to 21% in Los Angeles. When
35 considering exposures at and above 0.070 ppm in asthmatic children at
36 moderate or greater exertion, the results were similar in term of
9-2
-------
1 percentages, corresponding to average numbers of exposed asthmatic
2 children of approximately 2,000 per year in Denver, 8,000 per year in
3 Atlanta, 14,000 per year in Philadelphia, and 54,000 per year in Los
4 Angeles.
5 o The average (i.e., average across years 2006 to 2010) percentages of
6 school age children estimated to experience one or more exposures per
7 year to 8-hour Oj, concentrations at and above 0.080 ppm, while at
8 moderate or greater exertion, were approximately 0.4% for Denver
9 (corresponding to 2,000 children), 2% for Philadelphia (corresponding to
10 28,000 children), 3% for Atlanta (corresponding to 24,000 children), and
11 6% for Los Angeles (corresponding to 218,000 children). There was
12 considerable variability in these percentages across the years evaluated,
13 ranging from approximately 0 to 1% in Denver, 0 to 7% in Atlanta, 0 to
14 6% in Philadelphia, and 2 to 10% in Los Angeles. When considering
15 exposures at and above 0.080 ppm in asthmatic children at moderate or
16 greater exertion, the results were similar in term of percentages,
17 corresponding to average numbers of exposed asthmatic children of
18 approximately 200 per year in Denver, 2,000 per year in Atlanta, 3,000 per
19 year in Philadelphia, and 22,000 per year in Los Angeles.
20 o Between years, the pattern of exposures across cities differed. Generally,
21 from 2006 to 2009, 63 exposures fell, but in 2010, exposures increased
22 somewhat with the exception of Los Angeles. In the worst Os year
23 (2006), the percent of children exposed, while at moderate or greater
24 exertion, to concentrations at and above the lowest health benchmark,
25 0.060 ppm, ranged from 30 to 37% across the 4 study areas. The percent
26 at and above 0.070 ppm ranged from 10 to 21%, and the percent at and
27 above 0.080 ppm ranged from 1 to 10%. In the best O3 year (2009), the
28 percent of children ranged from 9 to 32% for exposures at and above
29 0.060 ppm, from 1 to 15% for exposures at and above 0.070 ppm, and
30 from 0 to 5% for exposures at and above 0.080 ppm, while at moderate or
31 greater exertion.
32 • Exposure Assessment for Simulating Meeting the Current 63 Standard
33 o Simulating just meeting the current Os standard reduces exposures such
34 that across the 5 years the estimated percent of children exposed to
35 concentrations at and above the lowest health benchmark, 0.060 ppm,
36 while at moderate or greater exertion, ranged from 3 to 14% for Atlanta
-------
1 (corresponding to approximately 24,000 to 123,000 children), 6 to 16%
2 for Denver (corresponding to approximately 31,000 to 89,000 children), 2
3 to 5% for Los Angeles (corresponding to approximately 66,000 to 186,000
4 children), and 3 to 18% for Philadelphia (corresponding to approximately
5 34,000 to 213,000 children).
6 o Just meeting the current standard in Los Angeles has the largest impact
7 across the four cities on the percent of children exposed above 0.060 ppm.
8 After simulating just meeting the current standard, the estimated percent
9 of children exposed above 0.060 ppm for the five years falls to a
10 maximum of 5% (with a range between 2 to 5%), compared with a
11 minimum of 23% (with a range between 23 to 37%) under recent
12 conditions.
13 o After just meeting the current 63 standard, the estimated percent of
14 children exposed to concentrations above 0.070 ppm, while at moderate or
15 greater exertion, ranged across the 5 years from 0.2 to 2% for Atlanta
16 (corresponding to approximately 1,000 to 18,000 children), 0.2 to 1.4%
17 for Denver (corresponding to approximately 1,000 to 7,000 children), 0 to
18 0.5% for Los Angeles (corresponding to approximately 1,000 to 17,000
19 children), and 0 to 4.0% for Philadelphia (corresponding to approximately
20 300 to 44,000 children).
21 o After just meeting the current 63 standard, the estimated percent of
22 children exposed to concentrations above 0.080 ppm, while at moderate or
23 greater exertion, ranged across the 5 years from 0 to 0.2% for Atlanta
24 (corresponding to approximately 0 to 2,000 children), 0 to 0.1% for
25 Denver (corresponding to approximately 0 to 400 children), 0% for Los
26 Angeles (corresponding to approximately 0 to 200 children), and 0 to
27 0.3% for Philadelphia (corresponding to approximately 0 to 3,000
28 children).
29 • Characterization of Factors Influencing High Exposures
30 o Children are an important exposure population subgroup, largely a result
31 of the combined outdoor time expenditure along with concomitantly
32 performing moderate or high exertion level activities.
33 o Persons having a majority of their time spent outdoors experienced the
34 highest 8-hour 03 exposure concentrations given that O^ concentrations in
35 other microenvironments were simulated to be lower than ambient
36 concentrations.
9-4
-------
1 o Simulations of highly exposed children in Los Angeles estimate that they
2 spend half of their outdoor time engaged in moderate or greater exertion
3 levels, such as in sporting activities. Highly exposed adults are estimated
4 to have lower activity levels during time spent outdoors.
5 o For populations experiencing one or more exposures per year to 8-hour 63
6 concentrations above 0.050 ppm, the highest modeled exposures are
7 determined primarily by amount of time spent outdoors in locations with
8 high ambient Os concentrations. There are differences in the influence of
9 outdoor time relative to ambient concentrations between locations, likely
10 due to air conditioning prevalence.
11
12 9.2 SUMMARY OF KEY RESULTS FOR HEALTH RISKS BASED ON
13 CONTROLLED HUMAN EXPOSURE STUDIES
14 The first draft lung function risk assessment evaluated risks of lung function decrements
15 due to Os exposure for all children and children with asthma. The analysis applies probabilistic
16 exposure-response relationships for lung function decrements (measured as percent reductions in
17 FEV1) associated with 8-hour moderate exertion exposures. The analysis provides estimates of
18 the percent of children experiencing a reduction in lung function for three different levels of
19 impact, 10, 15, and 20 percent decrements in FEV1. These levels of impact were selected based
20 on the literature discussing the adversity associated with these types of lung function decrements
21 (US EPA, 2012, Section 6.2.1.1; Henderson, 2006). For the first draft assessment, lung function
22 risks were estimated for 4 cities, Atlanta, Denver, Los Angeles, and Philadelphia. Key results
23 include: [To be provided in an updated draft anticipated to be available in August, 2012]
24
25 •
26 •
27
28 9.3 SUMMARY OF KEY RESULTS FOR HEALTH RISKS BASED ON
29 EPIDEMIOLOGICAL STUDIES
30 The first draft risk assessment also evaluated risks of mortality and morbidity from short-
31 term exposures to Os based on application of concentration-response functions derived from
32 epidemiology studies. The analysis included both a set of urban area case studies and a national
33 scale assessment. The urban case study analyses evaluated mortality and morbidity risks,
34 including emergency department (ED) visits, hospitalizations, and respiratory symptoms
35 associated with recent Os concentrations (2006-2010) and with Os concentrations simulating just
9-5
-------
1 meeting the current Os standard. Mortality and hospital admissions (HA) were evaluated in 12
2 urban areas, while ED visits and respiratory symptoms were evaluated in a subset of areas.
3 These 12 urban areas were: Atlanta, GA; Baltimore, MD; Boston, MA; Cleveland, OH; Denver,
4 CO; Detroit, MI; Houston, TX; Los Angeles, CA; New York, NY; Philadelphia, PA;
5 Sacramento, CA; and St. Louis, MO. The urban case study analyses focus on risk estimates for
6 the middle year of each three-year attainment simulation period (2006-2008 and 2008-2010) in
7 order to provide estimates of risk for a year with generally higher Oj, levels (2007) and a year
8 with generally lower Oi levels (2009).
9 The national scale assessment evaluated only mortality associated with recent O?,
10 concentrations across the entire U.S for 2006-2008. The national scale assessment is a
11 complement to the urban scale analysis, providing both a broader assessment of Os-related health
12 risks across the U.S., as well as an evaluation of how well the 12 urban study areas represented
13 the full distribution of ozone-related health risks in the U.S.
14 Both the urban area and national scale assessments provide the absolute incidence and
15 percent of incidence attributable to Os. Risk estimates are presented for ozone concentrations
16 down to zero, as well as down to the lowest measured levels (LML) of O3 in the year of the
17 analysis, as a weak surrogate for the LML in the epidemiology studies. The approach most
18 consistent with the statistical models reported in the epidemiological studies is to apply the
19 concentration-response functions to all ozone concentrations down to zero. However, consistent
20 with the conclusions of the ISA, we also recognize that confidence in the nature of the
21 concentration-response function and the magnitude of the risks associated with very low
22 concentrations of ozone is reduced because there are few ozone measurements at the lowest
23 levels in many of the urban areas included in the studies. As a result, the LML provides a cutoff
24 value above which we have higher confidence in the estimated risks. In our judgment, the two
25 sets of estimates based on estimating risk down to zero and estimating risk down to the LML
26 provide a reasonable bound on estimated total risks, reflecting uncertainties about the C-R
27 function below the lowest ozone levels evaluated in the studies.
28 Key results of the urban area case studies include:
29 • Short-term Mortality Risks Associated with Recent Air Quality
30 o There are significant differences in the spatial pattern of mortality risks
31 based on application of results from the two large multi-city epidemiology
32 studies. The estimates based on Zanobetti and Schwartz (2008) show the
33 largest impacts in Boston, Detroit, Los Angeles, and New York, while the
34 estimates based on Bell et al (2004) show the largest impacts in Atlanta,
35 Boston, Houston, Los Angeles, and New York.
9-6
-------
1 o Estimates of mortality attributable to short term Os exposure under recent
2 conditions vary widely across urban study areas, reflecting differences in
3 ambient Os levels and populations, as well as differences in city-specific
4 effect estimates. The patterns of variability across cities differs between
5 the Zanobetti and Schwartz (2008) and Bell et al (2004) based results
6 because of differences in the effect estimates and differences in the O3
7 metrics (daily 8-hour maximum vs fixed 8-hour mean).
8 o The Os attributable mortality risk estimates for 2007 based on the two
9 epidemiology studies range across the 12 urban areas from 20 to
10 approximately 930 deaths and approximately 0.5 to 4.9% of total baseline
11 all-cause mortality, with no concentration cutoff, and 10 to approximately
12 730deaths and approximately 0.4 to 3.5% of total baseline all-cause
13 mortality, with a concentration cutoff of the estimated LML. For 2009,
14 the Os attributable mortality risk estimates range across the 12 urban study
15 areas from 20 to approximately 980 deaths and approximately 0.6 to 4.3%
16 of total baseline all-cause mortality, with no concentration cutoff, and 10
17 to approximately 780 deaths and approximately 0.4 to 3.0% of total
18 baseline all-cause mortality, with a concentration cutoff of the estimated
19 LML. For most (but not all, e.g. Los Angeles) of the urban areas, Os-
20 attributable mortality risks are somewhat smaller in 2009 as compared
21 with 2007. This reflects primarily the lower Os levels seen in 2009.
22 o Twenty-five to 80% of the mortality risk is associated with days having Os
23 levels above 55 to 60 ppb.
24 • Short-term Mortality Risks Associated with Simulating Meeting the Current Os
25 Standard
26 o After simulating just meeting the current standard in 2007 across the 12
27 urban study areas, we estimate Os attributable mortality to vary from 20 to
28 850 deaths and approximately 0.5 to 4.6% of total baseline all-cause
29 mortality, with no concentration cutoff, and 10 to approximately 630
30 deaths and approximately 0.3 to 3.1% of total baseline all-cause mortality,
31 with a concentration cutoff of LML. After simulating just meeting the
32 current standard in 2009, we estimate Os attributable mortality across the
33 12 urban study areas to vary from 20 to 820 deaths and approximately 0.6-
34 4.1% of total baseline all-cause mortality, with no concentration cutoff,
35 and 10 to approximately 630 deaths and approximately 0.3 to 3.0% of total
36 baseline all-cause mortality, with a concentration cutoff of LML.
9-7
-------
1 o Five to 60% of mortality reductions occur due to reductions in Os on days
2 when 8-hour Os is greater than 55 to 60 ppb. As is expected, after
3 simulating just meeting the current standard, the percent of risk occurring
4 on days with 8-hour O3 greater than 55 to 60 ppb falls to 4 to 59%.
5 • Short-term Morbidity Risks Associated with Recent Conditions
6 o Estimates of morbidity attributable to short-term Os exposure in 2007
7 include: (a) 3,000 to 6,000 respiratory ED visits for Atlanta and 7,000 to
8 11,000 for asthma ED visits in New York, (b) 20,000 to 30,000 asthma
9 exacerbations in Boston, (c) 500 to 700 asthma HA in New York and (d)
10 up to 60 COPD and pneumonia HA in each of the 12 urban study areas.
11 • Short-term Morbidity Risks Associated with Simulating Meeting the Current Os
12 Standard
13 o Morbidity risks decrease after simulating just meeting the current
14 standards in 2007, although greater than 80% of ED visits remain in
15 Atlanta, 90% of ED visits remain in New York, and greater than 70% of
16 HA remain in most of the other urban areas.
17 o Ozone-related hospital admissions for respiratory causes remaining upon
18 just meeting the current standard, ranging across the 12 case study
19 locations, are estimated to be between 1.3 to 2.4% of all respiratory-
20 related hospital admissions. Further, in New York City, additional
21 information is available on ozone-related hospital admissions for asthma,
22 which upon just meeting the current standard are estimated to be
23 approximately 12 to 17% of total asthma-related hospital admissions.
24
25 Key results of the national scale assessment of mortality risk for recent (2006-2008) Os
26 concentrations:
27 • National-scale Short-term Mortality Risk
28 o The central estimates of the national burden of total Os attributable
29 mortality based on Zanobetti and Schwartz (2008) and Bell et al (2004)
30 and recent Os levels are estimated to be 13,000 and 18,000, respectively,
31 in 2006-2008.
32 o There is considerable variation between estimates based on the Zanobetti
33 and Schwartz (2008) results and those based on the Bell et al (2004)
34 results. The estimated percentage of total county-level mortality
35 attributable to Os across all counties for the Zanobetti and Schwartz based
36 estimates ranges from 0.5 to 5.2%, with a median of 2.5%, with no
9-8
-------
1 concentration cutoff and from 0.4 to 4.4%, with a median of 2.1%, with a
2 concentration cutoff at 7.5 ppb, which is the average LML across cities as
3 reported by Zanobetti and Schwartz. The estimated percentage of total
4 county-level mortality attributable to O^ for the Bell et al (2004) based
5 estimates ranges from 0.4 to 4.2%, with a median of 1.9%, with no
6 concentration cutoff and from 0.3 to 3.5%, with a median of 1.6%, with a
7 concentrati on cutoff at 7.5 ppb.
8 o For estimates based on both epidemiology studies, we find that 85-90% of
9 Os-related deaths occur in locations where the seasonal average 8-hr daily
10 maximum or 8-hr daily mean (10am-6pm) Os concentration is greater than
11 40 ppb, corresponding to 4th high 8-hr daily maximum Os concentrations
12 ranging from approximately 50 ppb to 100 ppb.
13 • Representativeness of the Urban Study Areas in the National Context
14 o We observe that the 23 counties for the 12 urban study areas considered
15 capture urban areas that are among the most populated in the U.S., have
16 relatively high ozone levels, and represent the range of city-specific effect
17 estimates found by Bell et al. (2004) and Zanobetti and Schwartz (2008).
18 These three factors suggest that the urban study areas represent the overall
19 distribution of risk across the nation well, with a potential for better
20 characterization of the high end of the risk distribution.
21 o We find that the urban study areas are not capturing areas with the highest
22 baseline mortality rates, those with the oldest populations, and those with
23 the lowest air conditioning prevalence. These areas tend to have relatively
24 low ozone concentrations and low total population, suggesting that the
25 urban study areas are not missing high risk populations that have high
26 ozone concentrations in addition to greater susceptibility per unit ozone.
27 o The second representativeness analysis demonstrated that the 12 urban
28 study areas represent the the overall distribution of ozone-related risk
29 across the entire U.S.
30
31 9.4 OBSERVATIONS
32 [These observations have been prepared based on the exposure and epidemiological risk
33 estimates available for the July public release of the first draft REA. We anticipate providing
34 Chapter 9, with additional observations based on the lung function risk analysis, when we
3 5 provide supplemental REA materials along with the submissions of the first draft Policy
36 Assessment for public review in August]
9-9
-------
1 Recent Os concentrations have in general been declining over the period of analysis, 2006
2 to 2010. As a result, the risks and exposures associated with Oj, have also been declining.
3 However, while the overall trend in Os has been downward, for some locations, Os has displayed
4 a more variable pattern, for example, while most study locations saw a decrease in O^ between
5 2007 and 2008, Sacramento saw an increase to its highest level in 2008. In addition, the
6 downward trend generally did not hold in 2010, which saw slightly higher Os concentrations in
7 almost all of the study areas. Thus, while 2007 and 2009 generally represent worst case and best
8 case years within this five-year period, it should be recognized that additional variability in
9 results exists. In general, year to year variability in results is as significant as variability between
10 urban areas for both exposure and risk.
11 The results of the risk and exposure assessment suggest that while Os concentrations have
12 generally been declining over the analytical period from 2006 to 2010, there are still remaining
13 exposures to elevated levels of 63, and health risks associated with those exposures. These
14 exposures and health risks vary across the urban case study areas, but are generally consistent in
15 showing exposures above health benchmarks and risks associated with recent 63 concentrations.
16 On a national scale, recent 63 concentrations (2006-2008) are associated with a significant public
17 health burden, and risks are widespread across the U.S., with 50% of counties experiencing at
18 least 0.7 to 1.0% mortality attributable to recent 63 concentrations.
19 There are several important factors to consider when evaluating exposures and risks
20 associated with recent exposures to 03. First, with regard to the epidemiology based risk
21 estimates, while we have included a number of different model specifications to begin
22 understanding how variability in the underlying epidemiological studies can affect results, there
23 are still a number of variables that might affect risk results that we have not been able to include
24 in this first draft assessment, particularly in the case of modeling short-term exposure-related
25 mortality risk. Some of these include alternative lag structures and treatment of co-pollutants.
26 Second, with regard to the exposure estimates, the APEX model is not proficient at
27 modeling repeated exposures. As a result, while we are able to report the percent of children
28 with at least one exposure greater than the alternative exposure benchmarks, we are not able to
29 report with confidence the percent of children with more than one exposure. Children with
30 repeated exposures may be at greater risk of significant health effects. In addition, we were only
31 able to model exposure in four cities for this first draft assessment. It is likely that variation in
32 exposure will be larger when we have modeled the full set of 16 cities in the second draft REA.
33 Third, for this first draft of the REA, while we used a relatively simple roll-back
34 approach tor simulating just meeting the current standard, we also discussed the use of other
35 approaches that are based on modeling the response of Os concentrations to reductions in
36 anthropogenic NOx and VOC emissions, using the Higher-Order Decoupled Direct Method
9-10
-------
1 (HDDM) capabilities in the Community Multi-scale Air Quality (CMAQ) model. This modeling
2 incorporates all known emissions, including emissions from non-anthropogenic sources and
3 anthropogenic emissions from sources in and outside of the U.S. As a result, the need to specify
4 values for U.S. background concentrations is not necessary, as it is incorporated in the modeling
5 directly. We plan to further explore the use of this methodology in the second draft of the REA.
6 Application of this approach also addresses the recommendation by the National Research
7 Council of the National Academies (NRC, 2008) to explore how emissions reductions might
8 effect temporal and spatial variations in 63 concentrations, and to include information on how
9 NOX versus VOC control strategies might affect risk and exposure to 03.
10 This first draft REA provides preliminary estimates of exposures and risks which provide
11 information that can be used to begin discussions in the Policy Assessment regarding the
12 adequacy of the current standard. The second draft REA will further refine the estimates of
13 exposure and risk by incorporating additional urban areas into the exposure and lung function
14 risk analyses, and by expanding the sensitivity analyses supporting the epidemiology based risk
15 estimates. In addition, based on advice and comments received on this first draft REA, the
16 second draft REA may include additional health endpoints associated with longer-term
17 exposures to 03. The second draft REA will also evaluate any alternative 03 standards identified
18 in the first draft Policy Assessment following evaluation of any advice and comments on those
19 potential alternative standards provided during the review by the CAS AC OT, Panel. Finally, we
20 anticipate that the second draft REA will incorporate an improved approach to adjusting 03
21 concentrations based on simulations of just meeting the current and alternative 63 standards.
9-11
-------
1 Appendix 5-A
2
3 Description of the Air Pollutants Exposure Model (APEX)
4
5 1. Overview
6
7 APEX estimates human exposure to criteria and toxic air pollutants at local, urban, or regional
8 scales using a stochastic, microenvironmental approach. That is, the model randomly selects
9 data on a sample of hypothetical individuals in an actual population database and simulates each
10 individual's movements through time and space (e.g., at home, in vehicles) to estimate their
11 exposure to the pollutant. APEX can assume people live and work in the same general area (i.e.,
12 that the ambient air quality is the same at home and at work) or optionally can model commuting
13 and thus exposure at the work location for individuals who work.
14
15 The APEX model is a microenvironmental, longitudinal human exposure model for airborne
16 pollutants. It is applied to a specified study area, which is typically a metropolitan area. The
17 time period of the simulation is typically one year, but can easily be made either longer or
18 shorter. APEX uses census data, such as gender and age, to generate the demographic
19 characteristics of simulated individuals. It then assembles a composite activity diary to represent
20 the sequence of activities and microenvironments that the individual experiences. Each
21 microenvironment has a user-specified method for determining air quality. The inhalation
22 exposure in each microenvironment is simply equal to the air concentration in that
23 microenvironment. When coupled with breathing rate information and a physiological model,
24 various measures of dose can also be calculated.
25
26 The term microenvironment is intended to represent the immediate surroundings of an
27 individual, in which the pollutant of interest is assumed to be well-mixed. Time is modeled as a
28 sequence of discrete time steps called events. In APEX, the concentration in a microenvironment
29 may change between events. For each microenvironment, the user specifies the method of
30 concentration calculation (either mass balance or regression factors, described later in this
31 paper), the relationship of the microenvironment to the ambient air, and the strength of any
32 pollutant sources specific to that microenvironment. Because the microenvironments that are
33 relevant to exposure depend on the nature of the target chemical and APEX is designed to be
5A-1
-------
1 applied to a wide range of chemicals, both the total number of microenvironments and the
2 properties of each are free to be specified by the user.
3
4 The ambient air data are provided as input to the model in the form of time series at a list of
5 specified locations. Typically, hourly air concentrations are used, although temporal resolutions
6 as small as one minute may be used. The spatial range of applicability of a given ambient
7 location is called an air district. Any number of air districts can be accommodated in a model
8 run, subject only to computer hardware limitations. In principle, any microenvironment could be
9 found within a given air district. Therefore, to estimate exposures as an individual engages in
10 activities throughout the period it is necessary to determine both the microenvironment and the
11 air district that apply for each event.
12
13 An exposure event is determined by the time reported in the activity diary; during any event the
14 district, microenvironment, ambient air quality, and breathing rate are assumed to remain fixed.
15 Since the ambient air data change every hour, the maximum duration of an event is limited to
16 one hour. The event duration may be less than this (as short as one minute) if the activity diary
17 indicates that the individual changes microenvironments or activities performed within the hour.
18
19 The APEX simulation includes the following steps:
20 1. Characterize the study area - APEX selects sectors (e.g., census tracts) within a study area
21 based on user-defined criteria and thus identifies the potentially exposed population and
22 defines the air quality and weather input data required for the area.
23 2. Generate simulated individuals - APEX stochastically generates a sample of simulated
24 individuals based on the census data for the study area and human profile distribution data
25 (such as age-specific employment probabilities). The user must specify the size of the
26 sample. The larger the sample, the more representative it is of the population in the study
27 area and the more stable the model results are (but also the longer the computing time).
28 3. Construct a long-term sequence of activity events and determine breathing rates - APEX
29 constructs an event sequence (activity pattern) spanning the period of simulation for each
30 simulated person. The model then stochastically assigns breathing rates to each event, based
31 on the type of activity and the physical characteristics of the simulated person.
32 4. Calculate pollutant concentrations in microenvironments - APEX enables the user to define
33 any microenvironment that individuals in a study area would visit. The model then
34 calculates concentrations of each pollutant in each of the microenvironments.
5A-2
-------
1 5. Calculate pollutant exposures for each simulated individual - Microenvironmental
2 concentrations are time weighted based on individuals' events (i.e., time spent in the
3 microenvironment) to produce a sequence of time-averaged exposures (or minute by minute
4 time series) spanning the simulation period.
5 6. Estimate dose - APEX can also calculate the dose time series for each of the simulated
6 individuals based on the exposures and breathing rates for each event. For CO there is a
7 physiologically-based dosimetry module that estimates blood carboxyhemoglobin (COHb)
8 levels resulting from CO exposure. When modeling particulate matter, the rate of mass
9 deposition in the respiratory system is calculated using an empirical model (ICRP 1994). For
10 all other pollutants, an intake dose can be estimated using the exposure concentration
11 multiplied by breathing rate.
12
13 The model simulation continues until exposures are determined for the user-specified number of
14 simulated individuals. APEX then calculates population exposure statistics (such as the number
15 of exposures exceeding user-specified levels) for the entire simulation and writes out tables of
16 distributions of these statistics.
17
18 2. Model Inputs
19 APEX requires certain inputs from the user. The user specifies the geographic area and the
20 range of ages and age groups to be used for the simulation. Hourly (or shorter) ambient air
21 quality and hourly temperature data must be furnished for the entire simulation period. Other
22 hourly meteorological data (humidity, wind speed, wind direction, precipitation) can be used by
23 the model to estimate microenvironmental concentrations, but are optional.
24
25 In addition, most variables used in the model algorithms are represented by user-specified
26 probability distributions which capture population variability. APEX provides great flexibility in
27 defining model inputs and parameters, including options for the frequency of selecting new
28 values from the probability distributions. The model also allows different distributions to be
29 used at different times of day or on different days, and the distribution can depend conditionally
30 on values of other parameters. The probability distributions available in APEX include beta,
31 binary, Cauchy, discrete, exponential, extreme value, gamma, logistic, lognormal, loguniform,
32 normal, off/on, Pareto, point (constant), triangle, uniform, Weibull, and nonparametric
33 distributions. Minimum and maximum bounds can be specified for each distribution if a
34 truncated distribution is appropriate. There are two options for handling truncation. The
35 generated samples outside the truncation points can be set to the truncation limit; in this case,
5A-3
-------
1 samples "stack up" at the truncation points. Alternatively, new random values can be selected, in
2 which case the probability outside the limits is spread over the specified range, and thus the
3 probabilities inside the truncation limits will be higher than the theoretical untruncated
4 distribution.
5
6 3. Demographic Characteristics
7 The starting point for constructing a simulated individual is the population census database; this
8 contains population counts for each combination of age, gender, race, and sector. The user may
9 decide what spatial area is represented by a sector, but the default input file defines a sector as a
10 census tract. Census tracts are variable in both geographic size and population number, though
11 usually have between 1,500 and 8,000 persons. Currently, the default file contains population
12 counts from the 2000 census for every census tract in the United States, thus the default file
13 should be sufficient for most exposure modeling purposes. The combination of age, gender,
14 race, and sector are selected first. The sector becomes the home sector for the individual, and the
15 corresponding air district becomes the home district. The probabilistic selection of individuals is
16 based on the sector population and demographic composition, and taken collectively, the set of
17 simulated individuals constitutes a random sample from the study area.
18
19 The second step in constructing a simulated individual is to determine their employment status.
20 This is determined by a probability which is a function of age, gender, and home sector. An
21 input file is provided which contains employment probabilities from the 2000 census for every
22 combination of age (16 and over), gender, and census tract. APEX assumes that persons under
23 age 16 do not commute. For persons who are determined to be workers, APEX then randomly
24 selects a work sector, based on probabilities determined from the commuting matrix. The work
25 sector is used to assign a work district for the individual that may differ from the home district,
26 and thus different ambient air quality may be used when the individual is at work.
27
28 The commuting matrix contains data on flows (number of individuals) traveling from a given
29 home sector to a given work sector. Based on commuting data from the 2000 census, a
30 commuting data base for the entire United States has been prepared. This permits the entire list
31 of non-zero flows to be specified on one input file. Given a home sector, the number of
32 destinations to which people commute varies anywhere from one to several hundred other tracts.
5A-4
-------
1
2 4. Attributes of Individuals
3 In addition to the above demographic information, each individual is assigned status and
4 physiological attributes. The status variables are factors deemed important in estimating
5 microenvironmental concentrations, and are specified by the user. Status variables can include,
6 but are not limited to, people's housing type, whether their home has air conditioning, whether
7 they use a gas stove at home, whether the stove has a gas pilot light, and whether their car has air
8 conditioning. Physiological variables are important when estimating pollutant specific dose.
9 These variables could include height, weight, blood volume, pulmonary diffusion rate, resting
10 metabolic rate, energy conversion factor (liters of oxygen per kilocalorie energy expended),
11 hemoglobin density in blood, maximum limit on MET ratios (see below), and endogenous CO
12 production rate. All of these variables are treated probabilistically taking into account
13 interdependences, reflecting variability in the population.
14
15 5. Construction of Activity Diaries
16 The activity diary determines the sequence of microenvironments visited by the simulated
17 person. A longitudinal sequence of daily diaries must be constructed for each simulated
18 individual to cover the entire simulation period. The default activity diaries in APEX are derived
19 from those in the EPA's Consolidated Human Activity Database (CHAD), although the user
20 could provide area specific diaries if available. There are over 33,000 CHAD diaries, each
21 covering a 24 hour period, that have been compiled from several studies. CHAD is essentially a
22 cross-sectional database that, for the most part, only has one diary per person. Therefore, APEX
23 must assemble each longitudinal diary sequence for a simulated individual from many single-day
24 diaries selected from a pool of similar people.
25
26 APEX selects diaries from CHAD by matching gender and employment status, and by requiring
27 that age falls within a user-specified range on either side of the age of the simulated individual.
28 For example, if the user specifies plus or minus 20%, then for a 40 year old simulated individual,
29 the available CHAD diaries are those from persons aged 32 to 48. Each simulated individual
30 therefore has an age window of acceptable diaries; these windows can partially overlap those for
31 other simulated individuals. This differs from a cohort-based approach, where the age windows
32 are fixed and non-overlapping. The user may optionally request that APEX allow a decreased
5A-5
-------
1 probability for selecting diaries from ages outside the primary age window, and also for selecting
2 diaries from persons of missing gender, age, or employment status. These options allow the
3 model to continue the simulation when diaries are not available within the primary window.
4
5 The available CHAD diaries are classified into diary pools, based on the temperature and day of
6 the week. The model will select diaries from the appropriate pool for days in the simulation
7 having matching temperature and day type characteristics. The rules for defining these pools are
8 specified by the user. For example, the user could request that all diaries from Monday to Friday
9 be classified together, and Saturday and Sunday diaries in another class. Alternatively, the user
10 could instead create more than two classes of weekdays, combine all seven days into one class,
11 or split all seven days into separate classes.
12
13 The temperature classification can be based either on daily maximum temperature, daily average
14 temperature, or both. The user specifies both the ranges and numbers of temperatures classes.
15 For example, the user might wish to create four temperature classes and set their ranges to below
16 50, 50-69, 70-84, and above a daily maximum of 84°F. Then day type and temperature classes
17 are combined to create the diary pools. For example, if there are four temperature classes and
18 two day type classes, then there will be eight diary pools.
19
20 APEX then determines the day-type and the applicable temperature for each person's simulated
21 day. APEX allows multiple temperature stations to be used; the sectors are automatically
22 mapped to the nearest temperature station. This may be important for study areas such as the
23 greater Los Angeles area, where the inland desert sectors may have very different temperatures
24 from the coastal sectors. For selected diaries, the temperature in the home sector of the
25 simulated person is used. For each day of the simulation, the appropriate diary pool is identified
26 and a CHAD dairy is randomly drawn. When a diary for every day in the simulation period has
27 been selected, they are concatenated into a single longitudinal diary covering the entire
28 simulation for that individual. APEX contains three algorithms for stochastically selecting
29 diaries from the pools to create the longitudinal diary. The first method selects diaries at random
30 after stratification by age, gender, and diary pool; the second method selects diaries based on
31 metrics related to exposure (e.g., time spent outdoors) with the goal of creating longitudinal
5A-6
-------
1 diaries with variance properties designated by the user; and the third method uses a clustering
2 algorithm to obtain more realistic recurring behavioral patterns.
3
4 The final step in processing the activity diary is to map the CHAD location codes into the set of
5 APEX microenvironments, supplied by the user as an input file. The user may define the
6 number of microenvironments, from one up to the number of different CHAD location codes
7 (which is currently 115).
8
9 6. Microenvironmental Concentrations
10 The user provides rules for determining the pollutant concentration in each microenvironment.
11 There are two available models for calculating microenvironmental concentrations: mass balance
12 and regression factors. Any indoor microenvironment may use either model; for each
13 microenvironment, the user specifies whether the mass balance or factors model will be used.
14
15 6.1 Mass Balance Model
16 The mass balance method assumes that an enclosed microenvironment (e.g., a room in a
17 residence) is a single well-mixed volume in which the air concentration is approximately
18 spatially uniform. The concentration of an air pollutant in such a microenvironment is estimated
19 using the following four processes (as illustrated in Figure 1):
20 • Inflow of air into the microenvironment;
21 • Outflow of air from the microenvironment;
22 • Removal of a pollutant from the microenvironment due to deposition, filtration, and
23 chemical degradation; and
24 • Emissions from sources of a pollutant inside the microenvironment.
5A-7
-------
Microenvironment
Air
outflow
Indoorsources
Air
inflow
Removal due to:
•Chemical reactions
•Deposition
•Filtration
Figure 1. Components of the Mass Balance Model Used by APEX.
5
6
7
8
9
10
11
12
13
Considering the microenvironment as a well-mixed fixed volume of air, the mass balance
equation for a pollutant in the microenvironment can be written in terms of concentration:
dC(t)
where:
dt
in out removal source
(1)
C(7j = Concentration in the microenvironment at time t
C in = Rate of change in C(t) due to air entering the microenvironment
C out = Rate of change in C(t) due to air leaving the microenvironment
C removal = Rate of change in C(t) due to all internal removal processes
C source = Rate of change in C(t) due to all internal source terms
Concentrations are calculated in the same units as the ambient air quality data, e.g., ppm, ppb,
ppt, or |ig/m3. In the following equations concentration is shown only in |ig/m3 for brevity.
The change in microenvironmental concentration due to influx of air, C in, is given by:
5A-8
-------
1 O/n — (^outdoor x 'penetration x ^airexchange (2)
2 where:
3 Coutdoor = Ambient concentration at an outdoor microenvironment or
4 outside an indoor microenvironment (|ig/m3)
5 /penetration = Penetration factor (unitless)
6 Rear exchange = Air exchange rate (hr"1)
7 Since the air pressure is approximately constant in microenvironments that are modeled in
8 practice, the flow of outside air into the microenvironment is equal to that flowing out of the
9 microenvironment, and this flow rate is given by the air exchange rate. The air exchange rate
10 (hr"1) can be loosely interpreted as the number of times per hour the entire volume of air in the
11 microenvironment is replaced. For some pollutants (especially particulate matter), the process of
12 infiltration may remove a fraction of the pollutant from the outside air. The fraction that is
13 retained in the air is given by the penetration factor /penetration-
14
15 A proximity factor (fproximity) and a local outdoor source term are used to account for differences
16 in ambient concentrations between the geographic location represented by the ambient air quality
17 data (e.g., a regional fixed-site monitor) and the geographic location of the microenvironment.
18 That is, the outdoor air at a particular location may differ systematically from the concentration
19 input to the model representing the air quality district. For example, a playground or house
20 might be located next to a busy road in which case the air at the playground or outside the house
21 would have elevated levels for mobile source pollutants such as carbon monoxide and benzene.
22 The concentration in the air at an outdoor location or directly outside an indoor
23 microenvironment (Coutdoor) is calculated as:
outdoor proximity ambient LocalOutdoorSources \ /
25 where:
26 Cambient = Ambient air district concentration (|ig/m3)
27 /proximity = Proximity factor (unitless)
28 CiocaiOutdoorSources = The contribution to the concentration at this location from local
29 sources not represented by the ambient air district concentration
30 (|ig/m3)
5A-9
-------
1 During exploratory analyses, the user may examine how a microenvironment affects overall
2 exposure by setting the microenvironment' s proximity or penetration factor to zero, thus
3 effectively eliminating the specified microenvironment.
4 Change in microenvironmental concentration due to outflux of air is calculated as the
5 concentration in the microenvironment C(t) multiplied by the air exchange rate:
6 Cout = Rairexchange x C(f) (4)
7 The third term (C removal) in the mass balance calculation (1) represents removal processes within
8 the microenvironment. There are three such processes in general: chemical reaction, deposition,
9 and filtration. Chemical reactions are significant for O3, for example, but not for carbon
10 monoxide. The amount lost to chemical reactions will generally be proportional to the amount
1 1 present, which in the absence of any other factors would result in an exponential decay in the
12 concentration with time. Similarly, deposition rates are usually given by the product of a
13 (constant) deposition velocity and a (time-varying) concentration, also resulting in an
14 exponential decay. The third removal process is filtration, usually as part of a forced air
15 circulation or HVAC system. Filtration will normally be more effective at removing particles
16 than gases. In any case, filtration rates are also approximately proportional to concentration.
17 Change in concentration due to deposition, filtration, and chemical degradation in a
18 microenvironment is simulated based on the first-order equation:
removal \ deposition filtration + ^chemical ,<->.
20 where:
21 C removal = Change in microenvironmental concentration due to removal
22 processes (|ig/m3/hr)
23 ^deposition = Removal rate of a pollutant from a microenvironment due to
24 deposition (hr"1)
25 Rfiitration = Removal rate of a pollutant from a microenvironment due to
26 filtration (hr"1)
27 Rchemicai = Removal rate of a pollutant from a microenvironment due to
28 chemical degradation (hr"1)
29 Rremovai = Removal rate of a pollutant from a microenvironment due to the
30 combined effects of deposition, filtration, and chemical
3 1 degradati on (hr" * )
5 A-10
-------
1
2 The fourth term in the mass balance calculation represents pollutant sources within the
3 microenvironment. This is the most complicated term, in part because several sources may be
4 present. APEX allows two methods of specifying source strengths: emission sources and
5 concentration sources. Either may be used for mass balance microenvironments, and both can be
6 used within the same microenvironment. The source strength values are used to calculate the
7 term C source (|ig/m3/hr).
8 Emission sources are expressed as emission rates in units of |ig/hr, irrespective of the units of
9 concentration. To determine the rate of change of concentration associated with an emission
10 source SE, it is divided by the volume of the microenvironment:
11 Csource.SE = ~^~ (6)
12 where:
13 C Source,SE = Rate of change in C(t) due to the emission source SE (|ig/m3/hr)
14 SE = The emission rate (ng/hr)
15 V = The volume of the microenvironment (m3)
16 Concentration sources (Sc) however, are expressed in units of concentration. These must be the
17 same units as used for the ambient concentration (e.g., |ig/m3). Concentration sources are
18 normally used as additive terms for microenvironments using the factors model. Strictly
19 speaking, they are somewhat inconsistent with the mass balance method, since concentrations
20 should not be inputs but should be consequences of the dynamics of the system. Nevertheless, a
21 suitable meaning can be found by determining the rate of change of concentration ( C SOUrce) that
22 would result in a mean increase of Sc in the concentration, given constant parameters and
23 equilibrium conditions, in this way:
24 Assume that a microenvironment is always in contact with clean air (ambient = zero), and it
25 contains one constant concentration source. Then the mean concentration over time in this
26 microenvironment from this source should be equal to Sc. The mean source strength expressed
27 in ppm/hr or |ig/m3/hr is the rate of change in concentration ( C SOUrce,sc}- In equilibrium,
28 Cc C source, SC (7)
R -i- R
nair exchange ~t~rv removal
5 A-11
-------
1 where Cs is the mean increase in concentration over time in the microenvironment due to the
2 source C SOUrce,sc • C Source,sc can thus be written as
3 ^source, SC = ^S x ^mean (8)
4 where Rmean is the chemical removal rate. From Eq. 7, Rmean is equal to the sum of the air
5 exchange rate and the removal rate (Ratr exchange + Removal) under equilibrium conditions. In
6 general, however, the microenvironment will not be in equilibrium, but in such conditions there
7 is no clear meaning to attach to C S0urce,sc since there is no fixed emission rate that will lead to a
8 fixed increase in concentration. The simplest solution is to use Rmean = Ratr exchange + Removal-
9 However, the user is given the option of specifically specifying Rmean (see discussion of
10 parameters below). This may be used to generate a truly constant source strength C SOUrce,sc by
11 making Sc and Rmean both constant in time. If this is not done, then Rmean is simply set to the sum
12 of (Ratr exchange + Rremovai)- If these parameters change over time, then C S0urce,sc also changes.
13 Physically, the reason for this is that in order to maintain a fixed elevation of concentration over
14 the base conditions, then the source emission rate would have to rise if the air exchange rate were
15 to rise.
16 Multiple emission and concentration sources within a single microenvironment are combined
17 into the final total source term by combining equations 6 and 8:
1 "e nc
18 C source = ^source,SE +^source,SC =77 Zj S/ +^meanZj S/ (9)
V i=1 i=1
19 where:
20 SEI = Emission source strength for emission source / (|ig/hr,
21 irrespective of the concentration units)
22 Sa = Emission source strength for concentration source /' (|ig/m3)
23 ne = Number of emission sources in the microenvironment
24 nc = Number of concentration sources in the microenvironment
25 In equations 6 and 9, if the units of air quality are ppm rather than |ig/m3, 7/Fis replaced by f/V,
26 where/= ppm / |ig/m3 = gram molecular weight / 24.45. (24.45 is the volume (liters) of a mole
27 of the gas at 25 °C and 1 atmosphere pressure.)
5 A-12
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
Equations 2, 4, 5, and 9 can now be combined with Eq. 1 to form the differential equation for t
microenvironmental concentration C(t). Within the time period of a time step (at most 1 hour).
C source and C in are assumed to be constant. Using C combined = C source + C in leads to:
dc(f) -r P C(f} P C(f}
,, ~° combined '* air exchange ° V1 ) " removal ° A' /
combined mean \ /
Solving this differential equation leads to:
,x c ( , x c ^
P(f ) combined , p/v \ ^combined g-Rmean((-'o)
mean \ mean )
where:
C(to) = Concentration of a pollutant in a microenvironment at
beginning of a time step (|ig/m3)
C(t) = Concentration of a pollutant in a microenvironment at
within the time step (|ig/m3).
(10)
(11)
the
time t
Based on Eq. 1 1, the following three concentrations in a microenvironment are calculated:
P C(t > oo) combined ^source ~*~ ^in
r^mean ^ air exchange + ^removal
C(t0+T) = Cequil+(c(t0)-Cequil)e-R™*"T
] to+J / -.-1 f^RmeanT
C \C(t)dt C i \C(t ) C )
T R T
tf> mean
where:
(12)
(13)
(14)
Cequii = Concentration in a microenvironment (|ig/m3) if t -^ oo
(equilibrium state).
Cfto) = Concentration in a microenvironment at the beginning
time step (|ig/m3)
Cfto+T) = Concentration in a microenvironment at the end of the
(Hg/m3)
of the
time ste
Cmean = Mean concentration over the time step in a microenvironment
5 A-13
-------
1 7? = 7? -1-7? fhr \
J- J^mean *\-air exchange ~ -^removal \lii )
2 At each time step of the simulation period, APEX uses Eqs. 12, 13, and 14 to calculate the
3 equilibrium, ending, and mean concentrations, respectively. The calculation continues to the
4 next time step by using Cfto+T) for the previous hour as Cfto).
5 6.2 Factors Model
6 The factors model is simpler than the mass balance model. In this method, the value of the
1 concentration in a microenvironment is not dependent on the concentration during the previous
8 time step. Rather, this model uses the following equation to calculate the concentration in a
9 microenvironment from the user-provided hourly air quality data:
"c
10 P=Pff+"V9 HM
lu ^mean ^ambient 'proximity 'penetration ~"~/_j °C/ \L^J
11 where:
12 Cmean = Mean concentration over the time step in a microenvironment (|ig/m3)
13 Cambient = The concentration in the ambient (outdoor) environment (|ig/m3)
14 /proximity = Proximity factor (unitless)
15 /penetration = Penetration factor (unitless)
16 Sa = Mean air concentration resulting from source i (|ig/m3)
17 nc = Number of concentration sources in the microenvironment
18 The user may specify distributions for proximity, penetration, and any concentration source
19 terms. All of the parameters in the above equation are evaluated for each time step, although
20 these values might remain constant for several time steps or even for the entire simulation.
21
22 The ambient air quality data are supplied as time series over the simulation period at several
23 locations across the modeled region. The other variables in the factors and mass balance
24 equations are randomly drawn from user-specified distributions. The user also controls the
25 frequency and pattern of these random draws. Within a single day, the user selects the number
26 of random draws to be made and the hours to which they apply. Over the simulation, the same
27 set of 24 hourly values may either be reused on a regular basis (for example, each winter
28 weekday), or a new set of values may be drawn. The usage patterns may depend on day of the
29 week, on month, or both. It is also possible to define different distributions that apply if specific
30 conditions are met. The air exchange rate is typically modeled with one set of distributions for
5 A-14
-------
1 buildings with air conditioning and another set of distributions for those which do not. The
2 choice of a distribution within a set typically depends on the outdoor temperature and possibly
3 other variables. In total there are eleven such conditional variables which can be used to select
4 the appropriate distributions for the variables in the mass balance or factors equations.
5
6 For example, the hourly emissions of CO from a gas stove may be given by the product of three
7 random variables: a binary on/off variable that indicates if the stove is used at all during that
8 hour, a usage duration sampled from a continuous distribution, and an emission rate per minute
9 of usage. The binary on/off variable may have a probability for on that varies by time of day and
10 season of the year. The usage duration could be taken from a truncated normal or lognormal
11 distribution that is resampled for each cooking event, while the emission rate could be sampled
12 just once per stove.
13
14 7. Exposure time series and dose calculation
15 The activity diaries provide the time sequence of microenvironments visited by the simulated
16 individual and the activities performed by each individual. The pollutant concentration in the air
17 in each microenvironment is assumed to be spatially uniform throughout the microenvironment
18 and unchanging within each diary event and is calculated by either the factors or the mass
19 balance method, as specified by the user. The exposure of the individual is given by the time
20 sequence of airborne pollutant concentrations that are encountered in the microenvironments
21 visited. Figure 2 illustrates the exposures for one simulated 12-year old child over a 2-day
22 period. On both days the child travels to and from school in an automobile, goes outside to a
23 playground in the afternoon while at school, and spends time outside at home in the evening (H:
24 home, A: automobile, S: school, P: playground, O: outdoors at home).
25
5 A-15
-------
ppm
0.14:
0.12
O.t)i
0.08;
0.06
0.04:
0.02^
0.00;
o
o
HH
00:00 06:00 12:00 18:00 00:00 06:00 12:00 18:00 00:00
time of day
2 Figure 2. Microenvironmental and Exposure Concentrations for a Simulated Individual
3 over 48 Hours.
4
5 In addition to exposure, APEX models breathing rates based on the physiology of each
6 individual and the exertion levels associated with the activities performed. For each activity type
7 in CHAD, a distribution is provided for a corresponding normalized Metabolic Energy for a Task
8 (MET ratio). The MET ratio is a ratio of the metabolic energy requirements for the specific
9 activity as compared to the resting, or basal, metabolic rate. The MET ratios have less
10 interpersonal variation than do the absolute energy expenditures. Based on age and gender, the
11 resting metabolic rate, along with other physiological variables is determined for each individual
12 as part of their anthropometric characteristics. Because the MET ratios are sampled
13 independently from distributions for each diary event, it would be possible to produce time-series
14 of MET ratios that are physiologically unrealistic. APEX employs a MET adjustment algorithm
15 based on a modeled oxygen deficit to prevent such overestimation of MET and breathing rates.
16 The relationship between the oxygen deficit and the applied limits on MET ratios are nonlinear
17 and are derived from published data on work capacity and oxygen consumption. The resulting
18 combination of microenvironmental concentration and breathing ventilation rates provides a time
19 series of inhalation intake dose for most pollutants.
5 A-16
-------
1
2 APEX uses additional dose algorithms for the pollutants CO and PM2.5. For CO exposures,
3 APEX can calculate the time series of blood carboxyhemoglobin (COHb) levels. These are
4 determined by solving the non-linear Coburn, Forster, Kane equation using a fourth-order Taylor
5 series method. This algorithm is explicit (non-iterative), fast, and accurate, for any practical
6 COHb level (up to more than 50% COHb). PM2.5 dose is modeled as the mass of PM depositing
7 in the entire respiratory system, including the extrathoracic regions (mouth, nose, and
8 oropharynx) and the lungs. The PM dose algorithm was developed from the empirical lung
9 deposition equations of the International Commission on Radiological Protection's Human
10 Respiratory Tract Model for Radiological Protection. The empirical equations estimate
11 deposition by both aerodynamic and thermodynamic processes as a function of breathing rate,
12 lung physiology, and particle characteristics.
13 8. Model output
14 APEX calculates the exposure and dose time series based on the events as listed on the activity
15 diary with a minimum of one event per hour but usually more during waking hours. APEX can
16 aggregate the event level exposure and dose time series to output hourly, daily, monthly, and
17 annual averages . The types of output files are selected by the user, and can be as detailed as
18 event-level data for each simulated individual (note, Figure 2 was produced from the event
19 output file). A set of summary tables are produced for a variety of exposure and dose measures.
20 These include tables of person-minutes at various exposure levels, by microenvironment, a table
21 of person-days at or above each average daily exposure level, and tables describing the
22 distributions of exposures for different groups. An example of how APEX results can be
23 depicted is given in Figure 3, which shows the percent of children with at least one 8-hour
24 average exposure at or above different exposure levels, concomitant with moderate or greater
25 exertion. These are results from a simulation of Os exposures for the greater Washington, D.C.
26 metropolitan area for the year 2002. From this graph ones sees, for example, that APEX
27 estimates 30 percent of the children in this area experience exposures above 0.08 ppm-8hr while
28 exercising, at least once during the year.
29
5 A-17
-------
1
2
3
4
0.02
0.04 0.06
Ozone Exposure Level (ppm-8hr)
0.08
0.12
Figure 3. The Percent of Simulated Children (ages 5-18) at or above 8-hour Average
Exposure Levels While Exercising.
5 A-18
-------
Appendix 5B
Inputs to the APEX Exposure Model
Table of Contents
5B-1. POPULATION DEMOGRAPHICS 3
5B-2. POPULATION COMMUTING PATTERNS 3
5B-3. ASTHMA PREVALENCE RATES 5
5B-4. HUMAN ACTIVITY DATA 5
5B-5. PHYSIOLOGICAL DATA 10
5B-6. MICROENVIRONMENTS MODELED 10
5B-7. AIR EXCHANGE RATES FOR INDOOR RESIDENTIAL ENVIRONMENTS
12
5B-8. AIR CONDITIONING PREVALENCE 13
5B-9. AER DISTRIBUTIONS FOR OTHER INDOOR ENVIRONMENTS 15
5B-10. PROXIMITY AND PENETRATION FACTORS FOR OUTDOORS AND IN-
VEHICLE MICROENVIRONMENTS 17
5B-11. OZONE DECAY AND DEPOSITION RATES 19
5B-12. AMBIENT OZONE CONCENTRATIONS 19
5B-1. METEOROLOGICAL DATA 27
REFERENCES 29
5B-1
-------
List of Tables
Table 1. Studies in the Consoloidated Human Activity Database (CHAD) 8
Table 2. Microenvironments modeled 11
Table 3. AERs for Atlanta (Indoors - residences) 13
Table 4. AERs for Denver and Philadelphia (Indoors- residences) 13
Table 5. AERs for Los Angeles (Indoors - residences) 13
Table 6. American Housing Survey A/C prevalence from Current Housing Reports Table 1-4
For Selected Urban Areas (Total: seasonal, occupied, vacant) (housing units in
thousands) 16
Table 7. Distributions of penetration and proximity factors for in-vehicle microenvironments. 17
Table 8. VMT fractions of interstate, urban and local roads in the study areas 18
Table 9. Counties Modeled in Each Area 19
Table 10. Atlanta ozone monitoring sites 23
Table 11. Denver ozone monitoring sites 24
Table 12. Los Angeles ozone monitoring sites 24
Table 13. Philadelphia ozone monitoring sites 26
Table 14. Atlanta Meteorological Stations, Locations, and Hours of Missing Data 27
Table 15. Denver Meteorological Stations, Locations, and Hours of Missing Data 27
Table 16. Los Angeles Meteorological Stations, Locations, and Hours of Missing Data 28
Table 17. Philadelphia Meteorological Stations, Locations, and Hours of Missing Data 28
List of Figures
Figure 1. Air Conditioning Prevalence for Owner- and Renter-Occupied Housing Units in the
Los Angeles-Long Beach Area in 2003 14
Figure 2. Atlanta Ozone Monitors With 30 km Radii of Influence 20
Figure 3. Denver Ozone Monitors With 30 km Radii of Influence 21
Figure 4. Los Angeles Ozone Monitors With 30 km Radii of Influence 22
Figure 5. Philadelphia Ozone Monitors With 30 km Radii of Influence 23
5B-2
-------
1 The APEX model inputs require extensive analysis and preparation in order to ensure the
2 model run gives valid and relevant results. This Appendix describes preparation and the sources
3 of data for the APEX input files.
4 5B-1. POPULATION DEMOGRAPHICS
5 APEX takes population characteristics into account to develop accurate representations of
6 study area demographics. Population counts and employment probabilities by age and gender
7 are used to develop representative profiles of hypothetical individuals for the simulation. Tract-
8 level population counts by age in one-year increments, from birth to 99 years, come from the
9 2000 Census of Population and Housing Summary File 1. The Summary File 1 contains the 100-
10 percent data, which is the information compiled from the questions asked of all people and about
11 every housing unit.
12 In the 2000 U. S. Census, estimates of employment were developed by census tract.
13 Employment data from the 2000 census can be found on the U.S. census web site at the address
14 http://www.census.gov/population/www/cen2000/phc-t28.html (Employment Status: 2000-
15 Supplemental Tables). The file input to APEX is broken down by gender and age group, so that
16 each gender/age group combination is given an employment probability fraction (ranging from 0
17 to 1) within each census tract. The age groupings in this file are: 16-19, 20-21, 22-24, 25-29, 30-
18 34, 35-44, 45-54, 55-59, 60-61, 62-64, 65-69, 70-74, and >75. Children under 16 years of age
19 are assumed to be not employed.
20 5B-2. POPULATION COMMUTING PATTERNS
21 As part of the population demographics inputs, it is important to integrate working
22 patterns into the assessment. In addition to using estimates of employment by tract, APEX also
23 incorporates home-to-work commuting data.
24 Commuting data were originally derived from the 2000 Census and were collected as part
25 of the Census Transportation Planning Package (CTPP). These data are available from the U.S.
26 DOT Bureau of Transportation Statistics (BTS) at the web site http://transtats.bts.gov/. The data
27 used to generate APEX inputs were taken from the "Part 3-The Journey To Work" files. These
28 files contain counts of individuals commuting from home to work locations at a number of
29 geographic scales.
5B-3
-------
1 These data were processed to calculate fractions for each tract-to-tract flow to create the
2 national commuting data distributed with APEX. This database contains commuting data for
3 each of the 50 states and Washington, D.C.
4 Commuting within the Home Tract
5 The APEX data set does not differentiate people that work at home from those that
6 commute within their home tract.
7 Commuting Distance Cutoff
8 A preliminary data analysis of the home-work counts showed that a graph of log(flows)
9 versus log(distance) had a near-constant slope out to a distance of around 120 kilometers.
10 Beyond that distance, the relationship also had a fairly constant slope but it was flatter, meaning
11 that flows were not as sensitive to distance. A simple interpretation of this result is that up to
12 120 km, the majority of the flow was due to persons traveling back and forth daily, and the
13 numbers of such persons decrease fairly rapidly with increasing distance. Beyond 120 km, the
14 majority of the flow is made up of persons who stay at the workplace for extended times, in
15 which case the separation distance is not as crucial in determining the flow.
16 To apply the home-work data to commuting patterns in APEX, a simple rule was chosen.
17 It was assumed that all persons in home-work flows up to 120 km are daily commuters, and no
18 persons in more widely separated flows commute daily. This meant that the list of destinations
19 for each home tract was restricted to only those work tracts that are within 120 km of the home
20 tract. When the same cutoff was performed on the 1990 census data, it resulted in 4.75% of the
21 home-work pairs in the nationwide database being eliminated, representing 1.3% of the workers.
22 The assumption is that this 1.3% of workers do not commute from home to work on a daily
23 basis. It is expected that the cutoff reduced the 2000 data by similar amounts.
24 Eliminated Records
25 A number of tract-to-tract pairs were eliminated from the database for various reasons. A
26 fair number of tract-to-tract pairs represented workers who either worked outside of the U.S.
27 (9,631 tract pairs with 107,595 workers) or worked in an unknown location (120,830 tract pairs
28 with 8,940,163 workers). An additional 515 workers in the commuting database whose data
29 were missing from the original files, possibly due to privacy concerns or errors, were also
30 deleted.
5B-4
-------
1 APEX allows the user to specify how to handle individuals who commute to destinations
2 outside the study area. For this application, we do not simulate those individuals, since we have
3 not estimated ambient concentrations of O3 in counties outside of the modeled areas.
4 5B-3. ASTHMA PREVALENCE RATES
5 One of the important population subgroups for the exposure assessment is asthmatic
6 children. Evaluation of the exposure of this group with APEX requires the estimation of
7 children's asthma prevalence rates. The estimates are based on children's asthma prevalence
8 data from the National Health Interview Survey (NHIS). A detailed description of how the
9 NHIS data were processed for input to APEX is provided in Appendix 5C.
10 5B-4. HUMAN ACTIVITY DATA
11 Exposure models use human activity pattern data to predict and estimate exposure to
12 pollutants. Different human activities, such as outdoor exercise, indoor reading, or driving, have
13 different pollutant exposure characteristics. In addition, different human activities require
14 different metabolic rates, and higher rates lead to higher doses. To accurately model individuals
15 and their exposure to pollutants, it is critical to have a firm understanding of their daily activities.
16 The Consolidated Human Activity Database (CHAD) provides data on human activities
17 through a database system of collected human diaries, or daily activity logs (EPA, 2002). The
18 purpose of CHAD is to provide a basis for conducting multi-route, multi-media exposure
19 assessments (McCurdy et al., 2000).
20 The data contained within CHAD come from multiple surveys with varied structures.
21 Table 1 summarizes the studies in CHAD used in this modeling analysis, providing over 38,000
22 diary-days of activity data (over 13,000 diary-days for ages 5-18) collected between 1982 and
23 2009. In general, the surveys have a data foundation based on daily diaries of human activity.
24 This is the foundation from which CHAD was created. Individuals filled out diaries of their
25 daily activities and this information was input and stored in CHAD. Relevant data for these
26 individuals, such as age, are included as well. In addition, CHAD contains activity-specific
27 metabolic distributions developed from literature-derived data, which are used to provide an
28 estimate of metabolic rates of respondents through their various activities.
5B-5
-------
1 A key issue in this assessment is the development of an approach for creating (Vseason
2 or year-long activity sequences for individuals based on a cross-sectional activity data base of
3 24-hour records. The typical subject in the time/activity studies in CHAD provided less than two
4 days of diary data. For this reason, the construction of a season-long activity sequence for each
5 individual requires some combination of repeating the same data from one subject and using data
6 from multiple subjects. An appropriate approach should adequately account for the day-to-day
7 and week-to-week repetition of activities common to individuals while maintaining realistic
8 variability between individuals. The method in APEX for creating longitudinal diaries was
9 designed to capture the tendency of individuals to repeat activities, based on reproducing realistic
10 variation in a key diary variable, which is a user-selected function of diary variables. For this
11 analysis the key variable is set to the amount of time an individual spends outdoors each day,
12 which is one of the most important determinants of exposure to high levels of Os.
13 The actual diary construction method targets two statistics, a population diversity statistic
14 (D) and a within-person autocorrelation statistic (A). The D statistic reflects the relative
15 importance of within-person variance and between-person variance in the key variable. The A
16 statistic quantifies the lag-one (day-to-day) key variable autocorrelation. Desired D and A values
17 for the key variable are selected by the user and set in the APEX parameters file, and the method
18 algorithm constructs longitudinal diaries that preserve these parameters. Longitudinal diary data
19 from a field study of children ages 7-12 (Geyh et al., 2000; Xue et al., 2004) estimated values of
20 approximately 0.2 for D and 0.2 for A. In the absence of data for estimating these statistics for
21 younger children and others outside the study age range, and since APEX tends to underestimate
22 repeated activities, values of 0.5 for D and 0.2 for A are used for all ages.
23
24 CHAD Updates Since The Previous Ozone Review
25 Since the time of the prior Os NAAQS review conducted in 2007, there have been a
26 number new data sets incorporated into CHAD and used in our current exposure assessment,
27 most of which were from recently conducted studies. The data from these six additional studies
28 incorporated in CHAD have more than doubled the total activity pattern data used in the 2007 63
29 exposure modeling. The studies from which these new data were derived are briefly described
30 below.
31 • UMC and ISR. These diaries are from phase I (1997) and phase II (2002-03) of the
32 University of Michigan's Panel Study of Income Dynamics (PSID), respectively
5B-6
-------
1 (University of Michigan, 2012). Activity pattern data were collected from nearly 10,000
2 children ages 0-13 (phase I) and 5-19 (phase II) across the U.S. For each child, diary data
3 were collected on two nonconsecutive days in a single week, in no particular season,
4 though mostly occurring during the spring and fall (phase I), and winter (phase II)
5 months.
6 • NSA. The diaries were collected as part of the National Scale Activity Survey (NSAS),
7 an EPA-funded study of averting behavior related to air quality alerts (Knowledge
8 Networks, 2009). Data were collected from about 1,200 adults aged 35-92 in seven
9 metropolitan areas (Atlanta, St. Louis, Sacramento, Washington DC, Dallas, Houston,
10 and Philadelphia). Data were collected over 1-15 (partially consecutive) days across the
11 2009 ozone season, totaling approximately 7,000 person days of data.
12 • OAB. These diaries were collected in a study of children's activities on high and low
13 ozone days during the 2002 ozone season (Mansfield et al., 2009). Children from 35 U.S.
14 metropolitan areas having the worst O3 pollution households were studied, of whom
15 about half of the children were asthmatics. Activity data were collected on 6
16 nonconsecutive days from each subject, with some subjects providing fewer days,
17 totaling nearly 3,000 persons days of data.
18 • SEA. These diaries are from a PM exposure study of susceptible populations living in
19 Seattle, WA between 1999 to 2002 (Liu et al., 2003). Two cohorts were studied: an older
20 adult group with either chronic obstructive pulmonary disease (COPD) or coronary heart
21 disease and a child group with asthma. Activity data were collected on 10 consecutive
22 days from each subject, with some subjects providing fewer days. Over 1,300 daily
23 diaries were collected from the adult group and more than 300 from the children cohort.
24 • RTF. These diaries were collected in a panel study of PM exposure in the Research
25 Triangle Park, NC area (Williams et al., 2003a, b). Two older adult cohorts (ages 55-85)
26 were studied: a cohort having implanted cardiac defibrillators living in Chapel Hill, NC
27 and a second group of 30 people having controlled hypertension and residing in a low-to-
28 moderate SES neighborhood in Raleigh, NC. Data were collected on approximately 8
29 consecutive days in 4 consecutive seasons in 2000-2001. A total of 1000 diary-days are
30 included.
5B-7
-------
1 Table 1. Studies in the Consoloidated Human Activity Database (CHAD)
Study name
Baltimore
Retirement Home
Study (EPA)
California Youth
Activity Patterns
Study (CARS)
California Adults
Activity Patterns
Study (CARS)
California Children
Activity Patterns
Study (CARS)
Cincinnati Activity
Patterns Study
(EPRI)
Denver CO
Personal Exposure
Study (EPA)
Los Angeles Ozone
Exposure Study:
Elementary School
Los Angeles Ozone
Exposure Study:
High School
Geographic
coverage
One building
in Baltimore
California
California
California
Cincinnati
metro, area
Denver
metro, area
Los Angeles
Los Angeles
Study time
period
01/1997-02/1997,
07/1998-08/1998
10/1987-09/1988
10/1987-09/1988
04/1989- 02/1990
03/1985-04/1985,
08/1985
11/1982-02/1983
10/1989
09/1990-10/1990
Subject
ages
72-93
12- 17
18-94
<1 - 11
-------
National Human
Activity Pattern
Study (NHAPS):
Air
National Human
Activity Pattern
Study (NHAPS):
Water
National Study of
Avoidance of S
(NSAS)
Population Study of
Income Dynamics
PSID CDS I (Univ.
Michigan I)
Population Study of
Income Dynamics
PSID CDS II (Univ.
Michigan II)
RTI Ozone
Averting Behavior
RTF Panel (EPA)
Seattle
Washington, B.C.
(EPA)
Totals
National
National
7 U.S.
metropolitan
areas
National
National
35 U.S.
metropolitan
areas
RTF, NC
Seattle, WA
Wash., D.C.
metro, area
09/1992-10/1994
09/1992-10/1994
06/2009-09/2009
02/1997-12/1997
01/2002-12/2003
07/2002-08/2003
06/2000-05/2001
10/1999-03/2002
11/1982-02/1983
1982 - 2009
-------
1
2 5B-5. PHYSIOLOGICAL DATA
3 APEX requires values for several physiological parameters for subjects in order to
4 accurately model their metabolic processes that affect pollutant intake. This is because
5 physiological differences may cause people with the same exposure and activity scenarios to
6 have different pollutant intake levels. The physiological parameters file distributed with APEX
7 contains physiological data or distributions by age and gender for maximum ventilatory capacity
8 (in terms of age- and gender-specific maximum oxygen consumption potential), body mass,
9 resting metabolic rate, and oxygen consumption-to-ventilation rate relationships.
10 Also input to APEX are metabolic information for different activities listed in the diary file.
11 These metabolic activity levels are in the form of distributions. Some activities are specified as a
12 single point value (for instance, sleep), while others, such as athletic endeavors or manual labor,
13 are normally, lognormally, or otherwise statistically distributed. APEX samples from these
14 distributions and calculates values to simulate the variable nature of activity levels among
15 different people.
16 5B-6. MICROENVIRONMENTS MODELED
17 In APEX, microenvironments provide the exposure locations for modeled individuals.
18 For exposures to be accurately estimated, it is important to have realistic microenvironments that
19 are matched closely to where people are physically located on a daily and hourly basis. As
20 discussed in Appendix 5A, the two methods available in APEX for calculating pollutant
21 concentrations within microenvironments are a mass balance model and a transfer factor
22 approach. Table 2 lists the 28 microenvironments selected for this analysis and the exposure
23 calculation method for each. The parameters used in this analysis for modeling these
24 microenvironments are described in this section.
25
5B-10
-------
Table 2. Microenvironments modeled
1
2
O
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
Microenvironment
Indoor - Residence
Indoor - Community Center or Auditorium
Indoor - Restaurant
Indoor - Hotel, Motel
Indoor - Office building, Bank, Post office
Indoor - Bar, Night club, Cafe
Indoor - School
Indoor - Shopping mall, Non-grocery store
Indoor - Grocery store, Convenience store
Indoor - Metro- Sub way-Train station
Indoor - Hospital, Medical care facility
Indoor - Industrial, factory, warehouse
Indoor - Other indoor
Outdoor - Residential
Outdoor - Park or Golf course
Outdoor - Restaurant or Cafe
Outdoor - School grounds
Outdoor - Boat
Outdoor - Other outdoor non-residential
Near-road - Metro-Subway-Train stop
Near-road - Within 10 yards of street
Near-road - Parking garage (covered or below ground)
Near-road - Parking lot (open), Street parking
Near-road - Service station
Vehicle - Cars and Light Duty Trucks
Vehicle - Heavy Duty Trucks
Vehicle - Bus
Vehicle - Train, Subway
Calculation
Method
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Mass balance
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Factors
Parameters1
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
AER and DE
None
None
None
None
None
None
PR
PR
PR
PR
PR
PE and PR
PE and PR
PE and PR
PE and PR
1 AER=air exchange rate, DE=decay-deposition rate, PR=proximity factor, PE=penetration factor
5B-11
-------
1 5B-7. AIR EXCHANGE RATES FOR INDOOR RESIDENTIAL
2 ENVIRONMENTS
3 Distributions of AERs for the indoor microenvironments were developed using data from
4 several studies. The analysis of these data and the development of the distributions used in the
5 modeling are described in detail in EPA (2007) Appendix A. This analysis showed that the AER
6 distributions for the residential microenvironments depend on the type of air conditioning (A/C)
7 and on the outdoor temperature, as well as other variables for which we do not have sufficient
8 data to estimate. This analysis clearly demonstrates that the AER distributions vary greatly
9 across cities and A/C types and temperatures, so that the selected AER distributions for the
10 modeled cities should also depend upon the city, A/C type, and temperature. For example, the
11 mean AER for residences with A/C ranges from 0.39 for Los Angeles between 30 and 40 °C to
12 1.73 for New York between 20 and 25 °C. The mean AER for residences without A/C ranges
13 from 0.46 for San Francisco on days with temperature between 10 and 20 °C to 2.29 for New
14 York on days with temperature between 20 and 25 °C. The need to account for the city as well as
15 the A/C type and temperature is illustrated by the result that for residences with A/C on days
16 with temperature between 20 and 25 °C, the mean AER ranges from 0.52 for Research Triangle
17 Park to 1.73 for New York. For each combination of A/C type, city, and temperature with a
18 minimum of 11 AER values, exponential, lognormal, normal, and Weibull distributions were fit
19 to the AER values and compared. Generally, the lognormal distribution was the best-fitting of
20 the four distributions, and so, for consistency, the fitted lognormal distributions are used for all
21 the cases.
22 One limitation of this analysis was that distributions were available only for selected
23 cities, and yet the summary statistics and comparisons demonstrate that the AER distributions
24 depend upon the city as well as the temperature range and A/C type. Another important
25 limitation of the analysis was that distributions were not able to be fitted to all of the temperature
26 ranges due to limited data in these ranges. A description of how these limitations were addressed
27 can be found in EPA (2007) Appendix A.
28 City-specific AER distributions were used where possible; otherwise data for a similar
29 city were used. The AER distributions used for the exposure modeling are given in Table 3
30 (Atlanta), Table 4 (Denver and Philadelphia), and Table 5 (Los Angeles).
5B-12
-------
Table 3. AERs for Atlanta (Indoors - residences)
Microenvironment
Indoors - residences
Condi
°F
<50
50-67
68-76
>76
<50
50-67
>67
tions a
A/C
yes
yes
yes
yes
no
no
no
Distribution
(GM, GSD, min, max)
Lognormal(0.962, 1.809, 0.1, 10)
Lognormal(0.562, 1.906, 0.1, 10)
Lognormal(0.397, 1.889, 0.1, 10)
Lognormal(0.380, 1.709, 0.1, 10)
Lognormal(0.926, 2.804, 0.1, 10)
Lognormal(0.733, 2.330, 0.1, 10)
Lognormal(1.378, 2.276, 0.1, 10)
a Average daily temperature range (°F) and presence or absence of air conditioning
Table 4. AERs for Denver and Philadelphia (Indoors - residences)
Microenvironment
Indoors - residences
Condi
°F
<50
50-76
>76
<50
50-67
>67
tions a
A/C
yes
yes
yes
no
no
no
Distribution
(GM, GSD, min, max)
Lognormal(0.711, 2.018, 0.1, 10)
Lognormal(1.139, 2.677, 0.1, 10)
Lognormal( 1.244, 2.177, 0.1, 10)
Lognormal(1.016, 2.138, 0.1, 10)
Lognormal(0.791, 2.042, 0.1, 10)
Lognormal(1.606, 2.119, 0.1, 10)
a Average daily temperature range (°F) and presence or absence of air conditioning
Table 5. AERs for Los Angeles (Indoors - residences)
Microenvironment
Indoors - residences
Condi
°F
<68
68-76
>76
<68
68-76
>76
<68
68-76
>76
tions a
A/C
Central
Central
Central
Room
Room
Room
None
None
None
Distribution
(GM, GSD, min, max)
Lognormal(0.577, 1.897, 0.1, 10)
Lognormal(1.084, 2.336, 0.1, 10)
Lognormal(0.861, 2.344, 0.1, 10)
Lognormal(0.672, 1.863, 0.1, 10)
Lognormal(1.674, 2.223, 0.1, 10)
Lognormal(0.949, 1.644, 0.1, 10)
Lognormal(0.744, 2.057, 0.1, 10)
Lognormal( 1.448, 2.315, 0.1, 10)
Lognormal(0.856, 2.018, 0.1, 10)
a Average daily temperature range (°F) and type of air conditioning
5B-8. AIR CONDITIONING PREVALENCE
In previous applications of APEX, we obtained A/C prevalence from the American
Housing Survey (AHS), at the level of the metropolitan area. For this application, we take
5B-13
-------
1
2
3
4
5
6
7
9
10
11
12
13
14
15
16
17
18
advantage of A/C differentials between owner-occupied and rental housing to estimate A/C
prevalence at the Census tract level. In this first draft REA, we have done this additional
breakdown for Los Angeles only; in the next draft, this will be done for all cities. For example,
the AHS data for A/C prevalence in Los Angeles1 finds that owner-occupied housing units have
52% central A/C, while rental units have 26% central A/C. For housing units with no central and
one window A/C, the owner-occupied prevalence is 9% and the rentals 21% (Figure 1). The net
results of this is that owner-occupied housing tends to be much more airtight than rentals in Los
Angeles.
Figure 1. Air Conditioning Prevalence for Owner- and Renter-
Occupied Housing Units in the Los Angeles-Long Beach Area in 2003
Central
Room units
No A/C
Data from the American Housing Survey for the Los Angeles Metropolitan Area in 2003, Current Housing Reports,
Table 1-4
Since APEX is able to read in tract-level data, such as A/C prevalence, distance to
roadways, etc., and use these as conditional variables for microenvironmental distributions, we
use tract-level information on owner-occupied and rental housing units, together with the
1 Table 1-4. Selected Equipment and Plumbing - All Housing Units. American Housing Survey for the Los Angeles
Metropolitan Area in 2003, U.S. Department of Housing and Urban Development and U.S. Census Bureau.
5B-14
-------
1 corresponding AHS breakdown for each urban area (Table 6), and obtain tract-level variation in
2 A/C prevalence.
3 5B-9. AER DISTRIBUTIONS FOR OTHER INDOOR ENVIRONMENTS
4 To estimate AER distributions for non-residential, indoor environments (e.g., offices and
5 schools), we obtained and analyzed two AER data sets: "Turk" (Turk et al., 1989); and "Persily"
6 (Persily and Gorfain, 2004; Persily et al., 2005). The Turk data set includes 40 AER
7 measurements from offices (25 values), schools (7 values), libraries (3 values), and multi-
8 purpose buildings (5 values), each measured using an SFe tracer over two or four hours in
9 different seasons of the year. The Persily data were derived from the U.S. EPA Building
10 Assessment Survey and Evaluation (BASE) study, which was conducted to assess indoor air
11 quality, including ventilation, in a large number of randomly selected office buildings throughout
12 the U.S. This data base consists of a total of 390 AER measurements in 96 large, mechanically
13 ventilated offices. AERs were measured both by a volumetric method and by a CC>2 ratio
14 method, and included their uncertainty estimates. For these analyses, we used the recommended
15 "Best Estimates" defined by the values with the lower estimated uncertainty; in the vast majority
16 of cases the best estimate was from the volumetric method.
17 Due to the small sample size of the Turk data, the data were analyzed without
18 stratification by building type and/or season. For the Persily data, the AER values for each office
19 space were averaged, rather using the individual measurements, to account for the strong
20 dependence of the AER measurements for the same office space over a relatively short period.
21 The mean values are similar for the two studies, but the standard deviations are about twice as
22 high for the Persily data. We fitted exponential, lognormal, normal, and Weibull distributions to
23 the 96 office space average AER values from the more recent Persily data, and the best fitting of
24 these was the lognormal. The fitted parameters for this distribution are a geometric mean of
25 1.109 and a geometric standard deviation of 3.015. These are used for AER distributions for the
26 indoor non-residential microenvironments, except for restaurants, bars, night clubs, and cafes.
5B-15
-------
1 Table 6. American Housing Survey A/C prevalence from Current Housing Reports Table 1-4 For Selected Urban Areas
2 (Total: seasonal, occupied, vacant) (housing units in thousands)
Metropolitan
area
Atlanta
Boston
Chicago
Cleveland
Dallas
Ft. Worth -
Arlington
Denver
Detroit
Houston
Los Angeles-
Long Beach
Riverside-San
Bernardino-
Ontario
Anaheim - Santa
Ana
New York-
Nassau-Suffolk-
Orange
Northern NJ
Philadelphia
Sacramento
St. Louis
Seattle-Everett
Washington, DC
Baltimore
Area
MA
CMSA
PMSA
PMSA
PMSA
PMSA
MA
PMSA
PMSA
PMSA
PMSA
PMSA
PMSA
PMSA
PMSA
PMSA
MA
PMSA
MA
MSA
Years
2004
2007
2003
2009
2004
2002
2002
2004
2003
2009
2007
2003
2002
2002
2003
2009
2003
2009
2003
2009
2004
2004
2004
2009
2007
2007
Total
housing
units
1802.8
1151.0
3198.9
3010.7
856.1
1365.4
639.4
949.1
1900.6
1672.5
2160.1
3318.5
1229.5
995.6
4849.8
4493.3
2589.1
2681.7
2068.8
2122.2
727.5
1139.6
1075.6
1331.7
2133.5
1109.6
Central
A/C
1649.5
307.6
1919.6
2050.6
439.5
1256.9
556.0
469.7
1157.4
1194.3
1924.4
1284.8
866.5
472.1
794.6
872.4
1184.3
1334.4
1001.8
1169.4
581.4
974.4
77.9
172.7
1881.3
828.8
additional
central
265.9
20.3
87.6
116.2
14.8
185.3
70.5
18.6
39.4
46.5
167.8
84.6
68.2
25.8
50.2
38.2
70.2
106.7
54.6
56.1
32.4
53.7
1.6
6.7
150.8
46.2
1 room
unit
47.8
275.5
500.8
412.0
143.8
31.8
19.9
138.0
261.3
192.3
59.1
495.5
123.8
134.7
1401.5
1036.9
460.0
318.0
328.1
225.8
62.7
65.8
56.9
121.8
76.9
63.7
2 room
units
34.5
202.0
340.5
265.1
48.2
32.1
26.6
22.6
106.0
82.8
67.8
80.0
31.2
13.7
1155.7
1184.1
429.3
412.2
317.0
269.9
12.6
43.5
14.8
27.5
69.0
76.5
3+ room
units
18.9
157.8
102.8
124.4
17.6
29.6
24.4
4.1
39.8
29.2
62.9
43.7
5.0
4.7
690.3
812.6
324.5
375.1
241.1
275.2
2.4
16.6
6.4
8.6
66.8
66.3
Percent
central
A/C
91
27
60
68
51
92
87
49
61
71
89
39
70
47
16
19
46
50
48
55
80
86
7
13
88
75
Percent
window
units
6
55
30
27
24
7
11
17
21
18
9
19
13
15
67
68
47
41
43
36
11
11
7
12
10
19
Sum of
%central &
%window
97
82
90
95
76
99
98
67
82
90
98
57
83
63
83
87
93
91
91
91
91
97
15
25
98
93
MA - metropolitan area; CMSA - consolidated metropolitan statistical area; PMSA - primary metropolitan statistical area.
5B-16
-------
1 The AER distribution used for schools is a discrete distribution with values (0.8 1.3 1.8
2 2.19 2.2 2.21 3.0 0.6 0.1 0.6 0.2 1.8 1.3 1.2 2.9 0.9 0.9 0.9 0.9 0.4 0.4 0.4 0.4 0.9 0.9 0.9 0.9 0.3
3 0.3 0.3 0.3), taken from from Turk et al., 1989 and Shendell et al., 2004.
4 The AER distribution used for restaurants, bars, night clubs, and cafes is a discrete
5 distribution with values (1.46 2.64 5.09 9.07 4.25 3.46), from Bennett et al., 2012, who measured
6 these six values in restaurants. This distribution is also used for the Bar, Night club, and Cafe
7 microenvironments.
8 5B-10. PROXIMITY AND PENETRATION FACTORS FOR OUTDOORS AND
9 IN-VEHICLE MICROENVIRONMENTS
10 For the outdoors near-road, public garage/parking lot, and in-vehicle proximity factors,
11 and for the in-vehicle penetration factors, we use distributions developed from the Cincinnati
12 Ozone Study (American Petroleum Institute, 1997, Appendix B; Johnson et al., 1995). This field
13 study was conducted in the greater Cincinnati metropolitan area in August and September, 1994.
14 Vehicle tests were conducted according to an experimental design specifying the vehicle type,
15 road type, vehicle speed, and ventilation mode. Vehicle types were defined by the three study
16 vehicles: a minivan, a full-size car, and a compact car. Road types were interstate highways
17 (interstate), principal urban arterial roads (urban), and local roads (local). Nominal vehicle
18 speeds (typically met over one minute intervals within 5 mph) were at 35 mph, 45 mph, or 55
19 mph. Ozone concentrations were measured inside the vehicle, outside the vehicle, and at six
20 fixed-site monitors in the Cincinnati area. Table 7 lists the distributions developed for
21 penetration and proximity factors for in-vehicle microenvironments, which are used in this
22 modeling analysis.
23
24 Table 7. Distributions of penetration and proximity
25 factors for in-vehicle microenvironments
26
27
Gaussian
distributions
Penetration factors
Proximity factors
local roads
urban roads
interstate roads
Mean
0.300
0.755
0.754
0.364
Standard
deviation
0.232
0.203
0.243
0.165
5B-17
-------
1
2
3
4
5
6
7
8
9
10
11
12
The Vehicle Miles Of Travel (VMT) fractions (Table 8, summarized from the U.S.
Department of Transportation, Federal Highway Administration annual Highway Statistics
reports, Tables HM-71) are used as conditional variables, which determine selection of the
proximity factor distributions for in-vehicle microenvironments. For local and interstate road
types, the VMT for the same Department of Transportation (DOT) categories are used. For
urban roads, the VMT for all other DOT road types are summed (Other freeways/expressways,
Other principal arterial, Minor arterial, Collector). At the time of this writing, data were only
available for three of our modeled years, 2006-2008. We are assuming that 2009 and 2010
would be best represented by 2008. We plan to use the 2009 and 2010 statistics in the second
draft REA if they are available.
Table 8. VMT fractions of interstate, urban and local roads in the study areas
City
Atlanta
Baltimore
Boston
Chicago
Cleveland
Dallas
Denver-Aurora
Detroit
Houston
Los Angeles-
Long Beach-
Santa Ana
New York-
Newark
Philadelphia
Sacramento
Seattle
St. Louis
Washington, DC
2006
inter-
state urban local
0.34 0.46 0.20
0.34 0.59 0.07
0.32 0.55 0.13
0.30 0.58 0.12
0.40 0.44 0.16
0.30 0.66 0.04
0.23 0.67 0.10
0.25 0.65 0.10
0.24 0.72 0.04
0.29 0.66 0.05
0.19 0.66 0.15
0.23 0.65 0.12
0.25 0.72 0.03
0.29 0.60 0.11
0.36 0.45 0.19
0.30 0.62 0.08
2007
inter-
state urban local
0.34 0.47 0.19
0.34 0.59 0.07
0.32 0.55 0.13
0.30 0.58 0.12
0.40 0.44 0.16
0.30 0.66 0.04
0.24 0.66 0.10
0.25 0.65 0.10
0.24 0.72 0.04
0.29 0.67 0.04
0.19 0.65 0.16
0.24 0.65 0.11
0.24 0.70 0.06
0.29 0.60 0.11
0.37 0.45 0.18
0.31 0.61 0.08
2008
inter-
state urban local
0.32 0.45 0.23
0.34 0.59 0.07
0.32 0.54 0.14
0.31 0.57 0.12
0.39 0.45 0.16
0.30 0.65 0.05
0.25 0.65 0.10
0.24 0.66 0.10
0.24 0.73 0.03
0.28 0.67 0.05
0.19 0.66 0.15
0.24 0.65 0.11
0.24 0.69 0.08
0.29 0.60 0.11
0.37 0.45 0.18
0.30 0.62 0.08
13
14
15
U.S. Department of Transportation,
Urbanized Areas - Miles And Daily
fractions sum to 1.00.
Federal Highway Administration. Annual Highway Statistics, Table HM-71:
Vehicle Miles Of Travel. Some fractions have been adjusted so the three
5B-18
-------
2 5B-11. OZONE DECAY AND DEPOSITION RATES
3 A distribution for combined Os decay and deposition rates was obtained from the analysis
4 of measurements from a study by Lee et al. (1999). This study measured decay rates in the
5 living rooms of 43 residences in Southern California. Measurements of decay rates in a second
6 room were made in 24 of these residences. The 67 decay rates range from 0.95 to 8.05 hour"1. A
7 lognormal distribution was fit to the measurements from this study, yielding a geometric mean of
8 2.5 and a geometric standard deviation of 1.5. These values are constrained to lie between 0.95
9 and 8.05 hour"1. This distribution is used for all indoor microenvironments.
10
11 5B-12. AMBIENT OZONE CONCENTRATIONS
12 APEX requires hourly ambient 63 concentrations at a set of locations in the study area.
13 Data from EPA's AIRS Air Quality System (AQS) were used to prepare the ambient air quality
14 input files for 2006 to 2010 (see REA Section 4.3). The hourly O3 concentrations at the AIRS
15 sites in and around each urban area were used as input to APEX to represent the ambient
16 concentrations within each urban area. A 30 km radius of influence was used for each
17 monitoring site. This means that the ambient concentrations assigned to a Census tract are those
18 at the closest monitor, if that monitor is with 30 km of the center of the tract and the county is in
19 the list of modeled counties (Table 9); otherwise, the population in that county is not modeled.
20 Figures X to X show the monitoring sites with their 30 km radii of influence. The modeled area
21 is the interestion of the 30 km disks with the counties specified in Table 9.
22
23 Table 9. Counties Modeled in Each Area
Urban Area (List of Counties)
Atlanta area, GA (Barrow, Bartow, Bibb, Butts, Carroll Floyd, Cherokee, Clarke, Clayton,
Cobb, Coweta, Dawson, De Kalb, Douglas, Fayette, Forsyth, Fulton, Gwinnett, Hall, Haralson,
Heard, Henry, Jasper, Lamar, Meriwether, Gilmer, Newton, Paulding, Pickens, Pike, Polk,
Rockdale, Spalding, Troup, Upson, Walton, Chambers (AL))
Denver area, CO (Adams, Arapahoe, Boulder, Broomfield, Clear Creek, Denver, Douglas,
Elbert, Gilpin, Jefferson, Park, Larimer, Weld)
Los Angeles area, CA (Los Angeles, Orange, Riverside, San Bernardino, Ventura)
Philadelphia area (Kent, DE; New Castle, DE; Sussex, DE; Cecil, MD; Atlantic, NJ; Camden,
NJ; Cumberland, NJ; Gloucester, NJ; Mercer, NJ; Ocean, NJ; Berks, PA; Bucks, PA; Chester,
5B-19
-------
PA; Delaware, PA; Montgomery, PA; Philadelphia, PA)
1
2
Figure 2. Atlanta Ozone Monitors With 30 km Radii of Influence
Jackson Mad/son. E/berl
Bibb Macon >* \ Wilkinson
Twiggs
Laurens
4
5
5B-20
-------
1 Figure 3. Denver Ozone Monitors With 30 km Radii of Influence
2
3
4
5
6
Keystone^
Breckenridge r
BlueR
5B-21
-------
1 Figure 4. Los Angeles Ozone Monitors With 30 km Radii of Influence
i.os Angejes
Santa%lari
Sat\ Buer|averiTura (%ntura)JM Valfey -
OxnardCamarillo/ Los"n
San Diego
San Diego santee
El Cajon
City
5B-22
-------
1 Figure 5. Philadelphia Ozone Monitors With 30 km Radii of Influence
2
3
4
5
6
Easjurange
t">/°*,Eliz,beth NewY°rk
Ozone Monitoring Sites
Tables 9 to 12 list the ozone monitoring sites that were used in this analysis.
Table 10. Atlanta ozone monitoring sites
Monitor id
13021-0012-1
13021-0013-1
13055-0001-1
13059-0002-1
13067-0003-1
13077-0002-1
County
Bibb, GA
Bibb, GA
Floyd, GA
Clarke, GA
Cobb, GA
Coweta, GA
5B-23
-------
13085-0001-2 Dawson, GA
13089-0002-1 DeKalb, GA
13089-3001-1 DeKalb, GA
13097-0004-1 Douglas, GA
13113-0001-1 Fayette, GA
13121-0055-1 Fulton, GA
13135-0002-1 Gwinnett, GA
13151-0002-1 Henry, GA
13213-0003-1 Gilmer, GA
13223-0003-1 Paulding, GA
13247-0001-1 Rockdale, GA
Table 11. Denver ozone monitoring sites
Monitor id
08001-3001-2
08005-0002-1
08005-0006-1
08013-0011-1
08013-7001-1
08013-7002-1
08031-0002-5
08031-0014-2
08031-0025-1
08035-0004-1
08059-0002-1
08059-0005-1
08059-0006-1
08059-0011-1
08059-0013-1
08069-0007-1
08069-0011-1
08069-0012-1
08069-1004-1
08123-0009-1
County
Adams, CO
Arapahoe, CO
Arapahoe, CO
Boulder, CO
Boulder, CO
Boulder, CO
Denver, CO
Denver, CO
Denver, CO
Douglas, CO
Jefferson, CO
Jefferson, CO
Jefferson, CO
Jefferson, CO
Jefferson, CO
Larimer, CO
Larimer, CO
Larimer, CO
Larimer, CO
Weld, CO
Table 12. Los Angeles ozone monitoring sites
Monitor id
06037-0002-1
06037-0016-1
06037-0113-1
06037-1002-1
06037-1103-1
06037-1201-1
06037-1301-1
06037-1302-1
06037-1602-1
06037-1701-1
06037-2005-1
County
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
Los Angeles, CA
5B-24
-------
06037-4002-1 Los Angeles, CA
06037-4006-1 Los Angeles, CA
06037-5005-1 Los Angeles, CA
06037-6012-1 Los Angeles, CA
06037-9033-1 Los Angeles, CA
06037-9034-1 Los Angeles, CA
06059-0007-1 Orange, CA
06059-1003-1 Orange, CA
06059-2022-1 Orange, CA
06059-5001-1 Orange, CA
06065-0004-1 Riverside, CA
06065-0008-1 Riverside, CA
06065-0009-1 Riverside, CA
06065 -0012-1 Riverside, CA
06065 -1004-1 Riverside, CA
06065 -1010-1 Riverside, CA
06065 -1016-1 Riverside, CA
06065-1999-1 Riverside, CA
06065-2002-1 Riverside, CA
06065 -5 001 -1 Riverside, CA
06065 -6001 -1 Riverside, CA
06065-8001-1 Riverside, CA
06065-8005-1 Riverside, CA
06065 -9001 -1 Riverside, CA
06065-9003-1 Riverside, CA
06071 -0001 -1 San Bernardino, CA
06071-0005-1 San Bernardino, CA
06071-0012-1 San Bernardino, CA
06071-0306-1 San Bernardino, CA
06071-1001-1 San Bernardino, CA
06071 -1004-2 San Bernardino, CA
06071-1234-1 San Bernardino, CA
06071 -2002-1 San Bernardino, CA
06071 -4001 -1 San Bernardino, CA
06071-4003-1 San Bernardino, CA
06071 -9002-1 San Bernardino, CA
06071 -9004-1 San Bernardino, CA
06073-0001-1 San Diego, CA
06073-0003-1 San Diego, CA
06073-0006-1 San Diego, CA
06073-1001-1 San Diego, CA
06073-1002-1 San Diego, CA
06073-1006-1 San Diego, CA
06073-1008-1 San Diego, CA
06073-1010-1 San Diego, CA
06073-1011-3 San Diego, CA
06073-1016-1 San Diego, CA
06073-1201-1 San Diego, CA
06073-2007-1 San Diego, CA
06111-0007-1 Ventura, CA
06111-0009-1 Ventura, CA
06111-1004-1 Ventura, CA
5B-25
-------
06111-2002-1
06111-2003-1
06111-3001-1
Ventura, CA
Ventura, CA
Ventura, CA
Table 13. Philadelphia ozone monitoring sites
Monitor id
10001-0002-1
10003-1007-1
10003-1010-1
10003-1013-1
10005-1002-1
10005-1003-1
24015-0003-1
34001-0005-1
34001-0006-1
34007-0003-1
34007-1001-1
34011-0007-1
34015-0002-1
34021-0005-1
34029-0006-1
42011-0006-1
42011-0009-1
42011-0010-1
42011-0011-1
42017-0012-1
42029-0100-1
42045-0002-1
42091-0013-1
42101-0004-1
42101-0014-1
42101-0024-1
42101-0136-1
County
Kent, DE
New Castle, DE
New Castle, DE
New Castle, DE
Sussex, DE
Sussex, DE
Cecil, MD
Atlantic, NJ
Atlantic, NJ
Camden, NJ
Camden, NJ
Cumberland, NJ
Gloucester, NJ
Mercer, NJ
Ocean, NJ
Berks, PA
Berks, PA
Berks, PA
Berks, PA
Bucks, PA
Chester, PA
Delaware, PA
Montgomery, PA
Philadelphia, PA
Philadelphia, PA
Philadelphia, PA
Philadelphia, PA
4
5
6
7
8
9
10
Estimation of Missing Data
Missing air quality data were estimated by the following procedure. Where there were
consecutive strings of missing values (data gaps) of 4 or fewer hours, missing values were
estimated by linear interpolation between the valid values at the ends of the gap. Remaining
missing values at a monitor were estimated by fitting linear regression models for each hour of
the day, with each of the other monitors, and choosing the model which maximizes R2, for each
hour of the day, subject to the constraints that R2 be greater than 0.50 and the number of
regression data values (days) is at least 60. If there were any remaining missing values at this
point, for gaps of 6 or fewer hours, missing values were estimated by linear interpolation
5B-26
-------
1 between the valid values at the ends of the gap. Any remaining missing values were replaced
2 with the value at the closest monitoring site for that hour.
3 Spatial Interpolation
4 The 63 concentration for each hour at each Census tract is set to the concentration at the
5 Os monitor closest to the center of the Census tract. If no monitors are within 30 km of the tract
6 center, then the persons living in that tract are not modeled. This method was used in the
7 previous O3 NAAQS review. In the second draft REA, we plan to perform a sensitivity analysis
8 and compare this approach with using the prediction of a photochemical grid model to augment
9 the monitored concentrations to create a smooth spatial surface of Os concentrations.
10 5B-1. METEOROLOGICAL DATA
11 Hourly surface temperature measurements were obtained from the National Weather
12 Service ISH data files.2 The weather stations used for each city are given in Tables 9 to 12.
13 Missing data are estimated using the same algorithm as for missing air quality data (Section
14 5B.12). APEX uses the data from the closest weather station to each Census tract. Temperatures
15 are used in APEX both in selecting human activity data and in estimating AERs for indoor
16 microenvironments.
17
18
Table 14. Atlanta Meteorological Stations, Locations, and Hours of Missing Data
Station"
722190-13874
722195-03888
722270-13864
723200-93801
Latitude
33.633
33.767
33.917
34.350
Longitude
-84.433
-84.517
-84.517
-85.167
2006
0
14
2506
14
2007
0
15
1647
30
2008
101
113
267
187
2009
41
103
93
59
2010
18
29
74
68
a USAF ID-WBAN ID
Table 15. Denver Meteorological Stations, Locations, and Hours of Missing Data
Station Latitude Longitude 2006 2007 2008 2009 2010
724660-93037 38.817 -104.717 2 2 108 110 71
724666-93067 39.567 -104.850 2 1 104 53 45
http ://www. ncdc. noaa. gov/oa/climate/surfaceinventories .html
5B-27
-------
Station
724695-23036
725650-03017
Latitude
39.717
39.833
Longitude
-104.750
-104.650
Table 16. Los Angeles Meteorological
Station
722860-23119
722880-23152
722950-23174
722970-23129
723816-03159
723926-23136
Latitude
33.900
34.200
33.933
33.833
34.733
34.217
Longitude
-117.250
-118.350
-118.400
-118.167
-118.217
-119.083
Table 17. Philadelphia Meteorological
Station
724070-93730
724075-13735
724080-13739
724085-94732
724089-13781
724096-14706
725170-14737
Latitude
39.450
39.367
39.867
40.083
39.667
40.017
40.650
Longitude
-74.567
-75.083
-75.233
-75.017
-75.600
-74.600
-75.450
2006
33
0
Stations,
2006
13
2
0
2
126
21
Stations,
2006
4
20
1
0
22
66
5
2007
42
2
Locations,
2007
25
12
0
4
11
47
Locations,
2007
o
5
84
0
10
1
63
4
2008
104
91
and Hours
2008
103
152
113
99
438
311
and Hours
2008
142
268
122
143
156
132
148
2009
53
44
of Missing
2009
48
86
44
173
176
218
2010
33
40
Data
2010
29
37
19
269
411
139
of Missing Data
2009
112
73
57
60
244
83
74
2010
161
74
21
38
89
122
51
5B-28
-------
REFERENCES
AHS (2003). U.S. Bureau of the Census and U.S. Department of Housing and Urban
Development. 2003 American Housing Survey (AHS): National Survey Data. Available
at: http://www.census.gov/hhes/www/housing/ahs/ahs.html, and
http://www.huduser.org/datasets/ahs.html
Akland, G. G., Hartwell, T. D., Johnson, T. R., Whitmore, R. W. (1985). Measuring human
exposure to carbon monoxide in Washington, D. C. and Denver, Colorado during the
winter of 1982-83. Environ Sci Technol. 19:911-918.
American Petroleum Institute. (1997). Sensitivity Testing of pNEM/Os Exposure to Changes in
the Model Algorithms. Health and Environmental Sciences Department.
Bennett, D. H, W. Fisk, M. G. Apte, X. Wu, A. Trout, D. Faulkner, D. Sulivan (2012).
Ventilation, Temperature, and HVAC Characteristics in Small and Medium Commercial
Buildings (SMCBs) in California. Indoor Air. 22(4):309-320.
Geyh, A. S., Xue, I, Ozkaynak, H., Spengler, J. D. (2000). The Harvard Southern California
chronic ozone exposure study: assessing ozone exposure of grade-school-age children in
two southern California communities. Environ Health Perspect. 108: 265-270.
Hartwell, T. D., Clayton, C. A., Ritchie, R. M., Whitmore, R. W., Zelon, H. S., Jones, S. M.,
Whitehurst, D. A. (1984). Study of Carbon Monoxide Exposure of Residents of
Washington, DC and Denver, Colorado. Research Triangle Park, NC: U.S. Environmental
Protection Agency, Office of Research and Development, Environmental Monitoring
Systems Laboratory. EPA-600/4-84-031.
Johnson, T. (1984). A Study of Personal Exposure to Carbon Monoxide in Denver, Colorado.
Research Triangle Park, NC: U.S. Environmental Protection Agency, Environmental
Monitoring Systems Laboratory. EPA-600/4-84-014.
Johnson, T. (1989). Human Activity Patterns in Cincinnati, Ohio. Palo Alto, CA: Electric
Power Research Institute. EPRI EN-6204.
Johnson, T., Pakrasi, A., Wisbeth, A., Meiners, G., Ollison, W. (1995). Ozone exposures Within
Motor Vehicles - Results of a Field Study in Cincinnati, Ohio. Proceedings 88th annual
meeting and exposition of the Air & Waste Management Association, San Antonio, TX.
June 18-23, 1995. Preprint paper 95-WA84A.02.
Klepeis, N. E., Tsang, A. M., Behar, J. V. (1996). Analysis of the National Human Activity
Pattern Survey (NHAPS) Respondents from a Standpoint of Exposure Assessment.
Washington, DC: U.S. Environmental Protection Agency, Office of Research and
Development. EPA/600/R-96/074.
Knowledge Networks. (2009). Field Report: National Scale Activity Survey (NSAS).
Conducted for Research Triangle Institute. Submitted to Carol Mansfield November 13,
2009.
5B-29
-------
Lee, K., Vallarino, J., Dumyahn, T., Ozkaynak, H., and Spengler, J. D. (1999). Ozone decay
rates in residences. J Air Waste Manag Assoc. 49: 1238-1244.
Liu L-JS, Box M, Kalman D, Kaufman J, Koenig J, Larson T, Lumley T, Sheppard L, Wallace
L. (2003). Exposure assessment of particulate matter for susceptible populations in
Seattle. Environ Health Persp. 111:909-918.
Mansfield C, Houtven GV, Johnson FR, Yang J-C. (2009). Environmental Risks and Behavior:
Do children spend less time outdoors when ozone pollution is high? ASSA annual
meeting, January 5, 2009. Update of Houtven et al. (2003), using the OAB CHAD data set,
and related to Mansfield et al. (2006).
Mansfield C, Johnson FR, Van Houtven G. (2006). The missing piece: averting behavior for
children's ozone exposures. Resource Energy Econ. 28:215-228.
McCurdy, T., Glen, G., Smith, L., Lakkadi, Y. (2000). The National Exposure Research
Laboratory's Consolidated Human Activity Database. J Expo Anal Environ Epidemiol. 10:
566-578.
Persily, A. and Gorfain, J. (2004). Analysis of Ventilation Data from the U.S. Environmental
Protection Agency Building Assessment Survey and Evaluation (BASE) Study. National
Institute of Standards and Technology, NISTIR 7145, December 2004.
Persily, A., J. Gorfain, G. Brunner (2005). Ventilation Design and Performance in U.S. Office
Buildings. ASHRAE Journal. April 2005, 30-35.
Robinson, J. P., Wiley, J. A., Piazza, T., Garrett, K., and Cirksena, K. (1989). Activity Patterns
of California Residents and their Implications for Potential Exposure to Pollution.
California Air Resources Board, Sacramento, CA. CARB-A6-177-33.
Shendell, D. G., A. M. Winer, R. Weker, S. D. Colome. (2004). Evidence of inadequate
ventilation in portable classrooms: results of a pilot study in Los Angeles County. Indoor
Air 2004; 14: 154-158.
Spier, C. E., Little, D. E., Trim, S. C., Johnson, T. R., Linn, W. S., Hackney, J. D. (1992).
Activity patterns in elementary and high school students exposed to oxidant pollution. J
Expo Anal Environ Epidemiol. 2: 277-293.
Tsang A. M., and Klepeis, N. E. (1996). Descriptive Statistics Tables from a Detailed Analysis
of the National Human Activity Pattern Survey (NHAPS) Data. U.S. Environmental
Protection Agency. EPA/600/R-96/148.
Turk, B. H., Grimsrud, D. T., Brown, J. T., Geisling-Sobotka, K. L., Harrison, J., Prill, R. J.
(1989). Commercial Building Ventilation Rates and Particle Concentrations. ASHRAE
No. 3248.
U.S. EPA (2002). Consolidated Human Activities Database (CHAD) Users Guide. Database
and documentation available at: http://www.epa.gov/chadnetl/
5B-30
-------
U.S. EPA (2007). Ozone Population Exposure Analysis for Selected Urban Areas. Office of Air
Quality Planning and Standards, U.S. Environmental Protection Agency, Research Triangle
Park, NC. Available at: http://www.epa.gov/ttn/naaqs/standards/ozone/s_o3_cr_td.html
University of Michigan. (2012). The Panel Study of Income Dynamics (PSID). Data and
documentation available at http://psidonline.isr.umich.edu/.
Wiley, J. A., Robinson, J. P., Piazza, T., Garrett, K., Cirksena, K., Cheng, Y.-T., Martin, G.
(1991a). Activity Patterns of California Residents: Final Report. California Air Resources
Board, Sacramento, CA. ARB/R93/487. Available from: NTIS, Springfield, VA., PB94-
108719.
Wiley, J. A., Robinson, J. P., Cheng, Y.-T., Piazza, T., Stork, L., Pladsen, K. (1991b). Study of
Children's Activity Patterns: Final Report. California Air Resources Board, Sacramento,
CA. ARB-R-93/489.
Williams R, Suggs J, Rea A, Leovic K, et al. (2003a). The Research Triangle paniculate panel
study: PM mass concentrations relationships. AtmosEnviron. 37:5349-5363.
Williams R, Suggs J, Rea A, Sheldon L, et al. (2003b). The Research Triangle paniculate panel
study: modeling ambient source contributions to personal and residential PM mass
concentrations. Atmos Environ. 37:5365-5378.
Williams, R., Suggs, J., Creason, J., Rodes, C., Lawless, P., Kwok, R., Zweidinger, R., Sheldon,
L. (2000). The 1998 Baltimore paniculate matter epidemiology-exposure study: Part 2.
Personal exposure associated with an elderly population. J Expo Anal Environ Epidemiol.
10(6): 533-543.
Xue, J., McCurdy, T., Spengler, J., Ozkaynak, H. (2004). Understanding variability in time
spent in selected locations for 7-12-year old children. J Expo Anal Environ Epidemiol.
14(3): 222-33.
5B-31
-------
1 APPENDIX 5C: GENERATION OF ADULT AND CHILD CENSUS-
2 TRACT LEVEL ASTHMA PREVALENCE USING NHIS (2006-2010)
3 AND US CENSUS (2000) DATA
4 5C-1. OVERVIEW
5 This describes the generation of our census tract level children and adult asthma
6 prevalence data developed from the 2006-2010 National Health Interview Survey (NHIS) and
7 census tract level poverty information from the 2000 US Census. The approach is, for the most
8 part, a reapplication of work performed by Cohen and Rosenbaum (2005), though here we
9 incorporated a few modifications as described below. Details regarding the earlier asthma
10 prevalence work are documented in Appendix G of US EPA (2007).
11 Briefly in the earlier development work, Cohen and Rosenbaum (2005) calculated asthma
12 prevalence for children aged 0 to 17 years for each age, gender, and four US regions using 2003
13 NHIS survey data. The four regions defined by NHIS were 'Midwest', 'Northeast', 'South', and
14 'West'. The asthma prevalence was defined as the probability of a 'Yes' response to the
15 question "EVER been told that [the child] had asthma?"1 among those persons that responded
16 either 'Yes' or 'No' to this question.2 The responses were weighted to take into account the
17 complex survey design of the NHIS.3 Standard errors and confidence intervals for the
18 prevalence were calculated using a logistic model (PROC SURVEY LOGISTIC; SAS, 2012). A
19 scatter-plot technique (LOESS SMOOTHER; SAS, 2012) was applied to smooth the prevalence
20 curves and compute the standard errors and confidence intervals for the smoothed prevalence
21 estimates. Logistic analysis of the raw and smoothed prevalence curves showed statistically
22 significant differences in prevalence by gender and region, supporting their use as stratification
23 variables in the final data set. These smoothed prevalence estimates were used as an input to
24 EPA's Air Pollution Exposure Model (APEX) to estimate air pollutant exposure in asthmatic
25 children (US EPA, 2007; 2008; 2009).
26 For the current asthma prevalence data set development, several years of recent NHIS
27 survey data (2006-2010) were combined and used to calculate asthma prevalence. The current
28 approach estimates asthma prevalence for children (by age in years) as was done previously by
29 Cohen and Rosenbaum (2005) but now includes an estimate of adult asthma prevalence (by age
30 groups). In addition, two sets of asthma prevalence for each adults and children were estimated
1 The response was recorded as variable "CASHMEV" in the downloaded dataset. Data and documentation
are available at http://www.cdc.gov/nchs/nhis/quest data related 1997 forward.htm.
2 If there were another response to this variable other than "yes" or "no" (i.e., refused, not ascertained,
don't know, and missing), the surveyed individual was excluded from the analysis data set.
3 In the SURVEY LOGISTIC procedure, the variable "WTF_SC" was used for weighting, "PSU" was used
for clustering, and "STRATUM" was used to define the stratum.
5C-1
-------
31 here. The first data set, as was done previously, was based on responses to the question "EVER
32 been told that [the child] had asthma". The second data set was developed using the probability
33 of a 'Yes' response to a question that followed those that answered 'Yes' to the first question
34 regarding ever having asthma, specifically, do those persons "STILL have asthma?"4 And
35 finally, in addition to the nominal variables region and gender (and age and age groups), the
36 asthma prevalence in this new analysis were further stratified by a family income/poverty ratio
37 (i.e., whether the family income was considered below or at/above the US Census estimate of
38 poverty level for the given year).
39 These new asthma prevalence data sets were linked to the US census tract level poverty
40 ratios probabilities (US Census, 2007), also stratified by age and age groups. Given 1) the
41 significant differences in asthma prevalence by age, gender, region, and poverty status, 2) the
42 variability in the spatial distribution of poverty status across census tracts, stratified by age, and
43 3) the spatial variability in local scale ambient concentrations of many air pollutants, it is hoped
44 that the variability in population exposures is now better represented when accounting for and
45 modeling these newly refined attributes of this susceptible population.
46 5C-2. RAW ASTHMA PREVALENCE DATA SET DESCRIPTION
47 In this section we describe the asthma prevalence data sets used and identify the variables
48 retained for our final data set. First, raw data and associated documentation were downloaded
49 from the Center for Disease Control (CDC) and Prevention's National Health Interview Survey
50 (NHIS) website.5 The 'Sample Child' and 'Sample Adult' files were selected because of the
51 availability of person-level attributes of interest within these files, i.e., age in years ('age_p'),
52 gender ('sex'), US geographic region ('region'), coupled with the response to questions of
53 whether or not the surveyed individual ever had and still has asthma. In total, five years of
54 recent survey data were obtained, comprising over 50,000 children and 120,000 children for
55 years 2006-2010 (Table 5C-1).
56 Information regarding personal and family income and poverty ranking are also provided
57 by the NHIS in separate files. Five files ('INCIMPx.dat') are available for each survey year,
58 each containing either the actual responses (where recorded or provided by survey participant) or
59 imputed values for the desired financial variable.6 For this current analysis, the ratio of income
60 to poverty was used to develop a nominal variable: either the survey participant was below or
4 While we estimated two separate sets of prevalence using the "STILL" and "EVER" variables, only the
"STILL" data were used as input to our exposure model.
5 See http://www.cdc.gov/nchs/nhis.htm (accessed October 4, 2011).
6 Financial information was not collected from all persons; therefore the NHIS provides imputed data.
Details into the available variables and imputation method are provided with each year's data set. For example see
"Multiple Imputation of Family Income and Personal Earnings in the National Health Interview Survey: Methods
and Examples" at http://www.cdc.gov/nchs/data/nhis/tecdoc 2010.pdf.
5C-2
-------
61 at/above a selected poverty threshold. This was done in this manner to be consistent with data
62 generated as part of a companion data set, i.e., census tract level poverty ratio probabilities
63 stratified by age (see section 5C-5 below).
64 Given the changes in how income data were collected over the five year period of interest
65 and the presence of imputed data, a data processing methodology was needed to conform each of
66 the year's data sets to a compatible nominal variable. Briefly, for survey years 2006-2008,
67 poverty ratios ('RAT_CATF) are provided for each person as a categorical variable, ranging
68 from <0.5 to 5.0 by increments of either 0.25 (for poverty ratios categories between <0.5 - 2.0)
69 and 0.50 (for poverty ratios >5.0). For 2009 and 2010 data, the poverty ratio was provided as a
70 continuous variable ('POVRATI3') rather than a categorical variable.7
71 When considering the number of stratification variables, the level of asthma prevalence,
72 and poverty distribution among the survey population, sample size was an important issue. For
73 the adult data, there were insufficient numbers of persons available to stratify the data by single
74 ages (for some years of age there were no survey persons). Therefore, the adult survey data were
75 grouped as follows: ages 18-24, 25-34, 35-44, 45-54, 55-64, 65-74, and, >75.8 To increase the
76 number of persons within the age, gender, and four region groupings of our characterization of
77 'below poverty' asthmatics persons, the poverty ratio threshold was selected as <1.5, therefore
78 including persons that were within 50% above the poverty threshold. As there were five data
79 sets containing variable imputed poverty ratios (as well as a non varying values for where
80 income information was reported) for each year, the method for determining whether a person
81 was below or above the poverty threshold was as follows. If three or more of the five
82 imputed/recorded values were <1.5, the person's family income was categorized 'below' the
83 poverty threshold, if three or more of the 5 values were >1.5, the person's family income was
84 categorized 'above' the poverty threshold. The person-level income files were then merged with
85 the sample adult and child files using the 'FFHX' (a household identifier), 'FMX' (a family
86 identifier), and 'FPX' (an individual identifier) variables. Note, all persons within the sample
87 adult and child files had corresponding financial survey data.
88 Two asthma survey response variables were of interest in this analysis and were used to
89 develop the two separate prevalence data sets for each children and adults. The response to the
90 first question "Have you EVER been told by a doctor or other health professional that you [or
7 Actually, the 2009 data had continuous values for the poverty ratios ('POVRATI2') but the quality was
determined by us to be questionable: the value varied among family members by orders of magnitude - however, it
should be a constant. The income data ('FAMINCI2') provided were constant among family members, therefore we
combined these data with poverty thresholds obtained from the US Census (available at:
http://www.census.gov/hhes/www/povertv/data/threshld/thresh08.html') for year 2008 by family size (note, income
is the annual salary from the prior year) and calculated an appropriate poverty ratio for each family member.
8 These same age groupings were used to create the companion file containing the census tract level
poverty ratio probabilities (section 5C-5).
5C-3
-------
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
your child] had asthma?" was recorded as variable name 'CASHMEV for children and
'AASMEV for adults. Only persons having responses of either 'Yes' or 'No' to this question
were retained to estimate the asthma prevalence. This assumes that the exclusion of those
responding otherwise, i.e., those that 'refused' to answer, instances where it was "not
ascertained', or the person 'does not know', does not affect the estimated prevalence rate if either
'Yes' or 'No' answers could actually be given by these persons. There were very few persons
(<0.3%) that did provide an unusable response (Table 5C-1), thus the above assumption is
reasonable. A second question was asked as a follow to persons responding "Yes" to the first
question, specifically, "Do you STILL have asthma?" and noted as variables 'CASSTILL' and
' AASSTILL' for children and adults, respectively. Again, while only persons responding 'Yes'
and 'No' were retained for further analysis, the representativeness of the screened data set is
assumed unchanged from the raw survey data given the few persons having unusable data
Table 5C-1. Number of total surveyed persons from NHIS (2006-2010) sample adult and
child files and the number of those responding to asthma survey questions.
CHILDREN
All Persons
Yes/No Asthma
Yes/No to Still Have + No Asthma
ADULTS
All Persons
Yes/No Asthma
Yes/No to Still Have + No Asthma
2010
11,277
11,256
11,253
2010
27,157
27,157
27,113
2009
11,156
11,142
11,129
2009
27,731
27,715
27,686
2008
8,815
8,800
8,793
2008
21,781
21,766
21,726
2007
9,417
9,404
9,394
2007
23,393
23,372
23,349
2006
9,837
9,815
9,797
2006
24,275
24,242
24,208
TOTAL
50,502
50,417
50,366
TOTAL
124,337
124,252
124,082
107
108
109
110
111
112
113
114
115
116
117
118
119
5C-3. ASTHMA PREVALENCE: LOGISTIC MODELING
As described in the previous section, four person-level analytical data sets were created
from the raw NHIS data files, generally containing similar variables: a 'Yes' or 'No' asthma
response variable (either 'EVER' or 'STILL'), an age (or age group for adults), their gender
('male' or 'female'), US geographic region ('Midwest', 'Northeast', 'South', and 'West'), and
poverty status ('below' or above'). One approach to calculate prevalence rates and their
uncertainties for a given gender, region, poverty status, and age is to calculate the proportion of
'Yes' responses among the 'Yes' and 'No' responses for that demographic group, appropriately
weighting each response by the survey weight. This simplified approach was initially used to
develop 'raw' asthma prevalence rates however this approach may not be completely
appropriate. The two main issues with such a simplified approach are that the distributions of
the estimated prevalence rates would not be well approximated by normal distributions and that
5C-4
-------
120 the estimated confidence intervals based on a normal approximation would often extend outside
121 the [0, 1] interval. A better approach for such survey data is to use a logistic transformation and
122 fit the model:
123
124 Prob(asthma) = exp(beta) / (1 + exp(beta)),
125
126 where beta may depend on the explanatory variables for age, gender, poverty status, or
127 region. This is equivalent to the model:
128
129 Beta = logit (prob(asthma)} = log { prob(asthma) / [1 - prob(asthma)] }.
130
131 The distribution of the estimated values of beta is more closely approximated by a normal
132 distribution than the distribution of the corresponding estimates of prob(asthma). By applying a
133 logit transformation to the confidence intervals for beta, the corresponding confidence intervals
134 for prob(asthma) will always be inside [0, 1]. Another advantage of the logistic modeling is that
135 it can be used to compare alternative statistical models, such as models where the prevalence
136 probability depends upon age, region, poverty status, and gender, or on age, region, poverty
137 statu s but not gender.
138 A variety of logistic models were fit and compared to use in estimating asthma
139 prevalence, where the transformed probability variable beta is a given function of age, gender,
140 poverty status, and region. I used the SAS procedure SURVEYLOGISTIC to fit the various
141 logistic models, taking into account the NHIS survey weights and survey design (using both
142 stratification and clustering options), as well as considering various combinations of the selected
143 explanatory variables.
144 As an example, Table 5C-2 lists the models fit and their log-likelihood goodness-of-fit
145 measures using the sample child data and for the "EVER" asthma response variable. A total of
146 32 models were fit, depending on the inclusion of selected explanatory variables and how age
147 was considered in the model. The 'Strata' column lists the eight possible stratifications: no
148 stratification, stratified by gender, by region, by poverty status, by region and gender, by region
149 and poverty status, by gender and poverty status, and by region, gender and poverty status. For
150 example, "5. region, gender" indicates that separate prevalence estimates were made for each
151 combination of region and gender. As another example, "2. gender" means that separate
152 prevalence estimates were made for each gender, so that for each gender, the prevalence is
153 assumed to be the same for each region. Note the prevalence estimates are independently
154 calculated for each stratum.
155
5C-5
-------
156
157 The 'Description' column of Table 5C-2 indicates how beta depends upon the age:
158
159 Linear in age Beta = a + |3 * age, where a and |3 vary with strata.
160 Quadratic in age Beta = a + |3 x age + y x age2, where a |3 and y vary with strata.
161 Cubic in age Beta = a + |3 x age + y x age2 + 5 x age3, where a, |3, y, and 5 vary
162 with the strata.
163 f(age) Beta = arbitrary function of age, with different functions for
164 different strata
165
166 The category f(age) is equivalent to making age one of the stratification variables, and is
167 also equivalent to making beta a polynomial of degree 16 in age (since the maximum age for
168 children is 17), with coefficients that may vary with the strata.
169 The fitted models are listed in order of complexity, where the simplest model (1) is an
170 unstratified linear model in age and the most complex model (model 32) has a prevalence that is
171 an arbitrary function of age, gender, poverty status, and region. Model 32 is equivalent to
172 calculating independent prevalence estimates for each of the 288 combinations of age, gender,
173 poverty status, and region.
174
175
176
5C-6
-------
177
178
Table 5C-2. Example of alternative logistic models evaluated to estimate child asthma
prevalence using the "EVER" asthma response variable and goodness of fit test results.
Model
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
Description
. logit(prob) = linear in age
. logit(prob) = linear in age
. logit(prob) = linear in age
. logit(prob) = linear in age
. logit(prob) = linear in age
. logit(prob) = linear in age
. logit(prob) = linear in age
. logit(prob) = linear in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
Strata
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
- 2 Log Likelihood
288740115.1
287062346.4
288120804.1
287385013.1
286367652.6
286283543.6
285696164.7
284477928.1
286862135.1
285098650.6
286207721.5
285352164
284330346.1
284182547.5
283587631.7
282241318.6
286227019.6
284470413
285546716.1
284688169.9
283662673.5
283404487.5
282890785.3
281407414.3
285821686.2
283843266.2
284761522.8
284045849.2
282099156.1
281929968.5
281963915.7
278655423.1
DF
2
4
8
4
16
16
8
32
3
6
12
6
24
24
12
48
4
8
16
8
32
32
16
64
18
36
72
36
144
144
72
288
179
180
5C-7
-------
181 Table 5C-2 also includes the -2 Log Likelihood statistic, a goodness-of-fit measure, and
182 the associated degrees of freedom (DF), which is the total number of estimated parameters. Any
183 two models can be compared using their -2 Log Likelihood values: models having lower values
184 are preferred. If the first model is a special case of the second model, then the approximate
185 statistical significance of the first model is estimated by comparing the difference in the -2 Log
186 Likelihood values with a chi-squared random variable having r degrees of freedom, where r is
187 the difference in the DF (hence a likelihood ratio test). For all pairs of models from Table 5C-2,
188 all the differences in the -2 Log Likelihood statistic are at least 600,000 and thus significant at p-
189 values well below 1 percent. Based on its having the lowest -2 Log Likelihood value, the last
190 model fit (model 32: retaining all explanatory variables and usingf(age)) was preferred and used
191 to estimate the asthma prevalence.9
192 The SURVEYLOGISTIC procedure produces estimates of the beta values and their 95%
193 confidence intervals for each combination of age, region, poverty status, and gender. By
194 applying the inverse logit transformation,
195
196 Prob(asthma) = exp( beta) / (1 + exp(beta)),
197
198 one can convert the beta values and associated 95% confidence intervals into predictions
199 and 95% confidence intervals for the prevalence. The standard error for the prevalence was
200 estimated as
201
202 Std Error (Prob(asthma)} = Std Error (beta) x exp(- beta) / (1 + exp(beta) )2,
203
204 which follows from the delta method (i.e., a first order Taylor series approximation).
205 Estimated asthma prevalence using this approach and termed here as 'unsmoothed' are provided
206 in Attachment A. Results for children are given in Attachment A Tables 1 ('EVER' had
207 Asthma) and 2 (' STILL' have asthma) while adults are provided in Attachment A Tables 3
208 ('EVER' had Asthma) and 4 ('STILL' have asthma). Graphical representation is also provided
209 in a series of plots within Attachment A Figures 1-4. The variables provided in the tabular
210 presentation are:
211
212 • Region
213 • Gender
9 Similar results were obtained when estimating prevalence using the 'STILL' have asthma variable as well
as when investigating model fit using the adult data sets. Note that because age was a categorical variable in the
adult data sets it could only be evaluated using f(age_group). See Attachment B Tables 1 - 4 for all model fit
results.
5C-8
-------
214 • Age (in years) or Age_group (age categories)
215 • Poverty Status
216 • Prevalence = predicted prevalence
217 • SE = standard error of predicted prevalence
218 • LowerCI = lower bound of 95 % confidence interval for predicted prevalence
219 • UpperCI = upper bound of 95 % confidence interval for predicted prevalence
220
221 5C-4. ASTHMA PREVALENCE: APPLICATION OF LOESS SMOOTHER
222 The estimated prevalence curves shows that the prevalence is not necessarily a smooth
223 function of age. The linear, quadratic, and cubic functions of age modeled by
224 SURVEYLOGISTIC were identified as a potential method for smoothing the curves, but they
225 did not provide the best fit to the data. One reason for this might be due to the attempt to fit a
226 global regression curve to all the age groups, which means that the predictions for age A are
227 affected by data for very different ages. A local regression approach that separately fits a
228 regression curve to each age A and its neighboring ages was used, giving a regression weight of
229 1 to the age A, and lower weights to the neighboring ages using a tri-weight function:
230
231 Weight = {1 - [ |age - A / q ]3}, where | age - A <= q.
232
233 The parameter q defines the number of points in the neighborhood of the age A. Instead
234 of calling q the smoothing parameter, SAS defines the smoothing parameter as the proportion of
235 points in each neighborhood. A quadratic function of age to each age neighborhood was fit
236 separately for each gender and region combination. These local regression curves were fit to the
237 beta values, the logits of the asthma prevalence estimates, and then converted them back to
238 estimated prevalence rates by applying the inverse logit function exp(beta) / (1 + exp(beta)). In
239 addition to the tri-weight variable, each beta value was assigned a weight of
240 1 / [std error (beta)]2, to account for their uncertainties.
241 In this application of LOESS, weights of 1 / [std error (beta)]2 were used such that a2 =
242 1. The LOESS procedure estimates a2 from the weighted sum of squares. Because it is assumed
243 a2 = 1, the estimated standard errors are multiplied by 1 / estimated a and adjusted the widths of
244 the confidence intervals by the same factor.
245 One data issue was an overly influential point that needed to be adjusted to avoid
246 imposing wild variation in the "smoothed" curves: for the West region, males, age 0, above
247 poverty threshold, there were 249 children surveyed that all gave 'No' answers to the asthma
248 question, leading to an estimated value of -14.203 for beta with a standard error of 0.09. In this
249 case the raw probability of asthma equals zero, so the corresponding estimated beta would be
5C-9
-------
250 negative infinity, but SAS's software gives -14.203 instead. To reduce the excessive impact of
251 this single data point, we replaced the estimated standard error by 4, which is approximately four
252 times the maximum standard error for all other region, gender, poverty status, and age
253 combinations.
254 There are several potential values that can be selected for the smoothing parameter; the
255 optimum value was determined by evaluating three regression diagnostics: the residual standard
256 error, normal probability plots, and studentized residuals. To generate these statistics, the
257 LOESS procedure was applied to estimated smoothed curves for beta, the logit of the prevalence,
258 as a function of age, separately for each region, gender, and poverty classification. For the
259 children data sets, curves were fit using the choices of 0.4, 0.5, 0.6, 0.7, 0.8, 0.9, and 1.0 for the
260 smoothing parameter. This selected range of values was bounded using the following
261 observations. With only 18 points (i.e., the number of ages), a smoothing parameter of 0.2
262 cannot be used because the weight function assigns zero weights to all ages except age A, and a
263 quadratic model cannot be uniquely fit to a single value. A smoothing parameter of 0.3 also
264 cannot be used because that choice assigns a neighborhood of 5 points only (0.3 x 18 = 5,
265 rounded down), of which the two outside ages have assigned weight zero, making the local
266 quadratic model fit exactly at every point except for the end points (ages 0, 1, 16 and 17).
267 Usually one uses a smoothing parameter below 1 so that not all the data are used for the local
268 regression at a given x value. Note also that a smoothing parameter of 0 can be used to generate
269 the unsmoothed prevalence. The selection of the smoothing parameter used for the adult curves
270 would follow a similar logic, although the lower bound could effectively be extended only to 0.9
271 given the number of age groups. This limits the selection of smoothing parameter applied to the
272 two adult data sets to a value of 0.9, though values of 0.8 - 1.0 were nevertheless compared for
273 good measure.
274 The first regression diagnostic used was the residual standard error, which is the LOESS
275 estimate of a. As discussed above, the true value of a equals 1, so the best choice of smoothing
276 parameter should have residual standard errors as close to 1 as possible. Attachment B, Tables 5
277 - 8 contain the residual standard errors output from the LOESS procedure, considering region,
278 gender, poverty status and each data set examined. For children 'EVER' having asthma and
279 when considering the best 20 models (of the 112 possible) using this criterion (note also within
280 0.06 RSE units of 1), the best choice varies with gender, region, and poverty status between
281 smoothing parameters of 0.6, 0.7, and 0.8 (Table 5C-3). Similar results were observed for the
282 'STILL' data set, though a value of 0.6 would be slightly preferred. Either adult data set could
283 be smoothed using a value of 0.8 or 0.9 given the limited selection of smoothing values, though
284 0.9 appears a better value for the ' STILL' data set.
285
5C-10
-------
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
303
Table 5C-3. Top 20 model smoothing fits where residual standard error at or a value of
1.0.
Data Set
Children
Adults
Asthma
EVER
STILL
EVER
STILL
Smoothing Parameter
0.4
2
2
n/a
n/a
0.5
2
3
n/a
n/a
0.6
5
4
n/a
n/a
0.7
5
2
n/a
n/a
0.8
4
3
6
5
0.9
1
3
6
7
1.0
1
3
8
8
The second regression diagnostic was developed from an approximate studentized
residual. The residual errors from the LOESS model were divided by standard error (beta) to
make their variances approximately constant. These approximately studentized residuals should
be approximately normally distributed with a mean of zero and a variance of a2 = 1. To test this
assumption, normal probability plots of the residuals were created for each smoothing parameter,
combining all the studentized residuals across genders, regions, poverty status, and ages. These
normal probability plots are provided in Attachment B, Figures 1-4. The results for the
children data indicate little distinction or affect by the selection of a particular smoothing
parameter (e.g., see Figure 5C-1 below), although linearity in the plotted curve is best expressed
with smoothing parameters at or above values of 0.6. When considering the adult data sets,
again the appropriate value would be 0.9, as Attachment B Figures 3 and 4 support this
conclusion.
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.7
1 -
-1 -
0.1
10
25
I
50
I
75
90
I
95
99
99.9
Normal Percentiles
Figure 5C-1. Normal probability plot of studentized residuals generated using logistic
model, smoothing set to 0.7, and the children 'EVER' asthmatic data set.
5C-11
-------
304
305
306
307
308
309
310
311
312
313
314
315
316
317
318
319
320
321
322
The third regression diagnostic, presented in Attachment B Figures 5 - 8 are plots of the
studentized residuals against the smoothed beta values. All the studentized residuals for a given
smoothing parameter are plotted together within the same graph. Also plotted is a LOESS
smoothed curve fit to the same set of points, with SAS's optimal smoothing parameter choice, to
indicate the typical pattern. Ideally there should be no obvious pattern and an average
studentized residual close to zero with no regression slope (e.g., see Figure 5C-2). For the
children data sets, these plots generally indicate no unusual patterns, and the results for
smoothing parameters 0.4 through 0.6 indicate a fit LOESS curve closest to the studentized
residual equals zero line. When considering the adult data sets, again the appropriate value
would be 0.9, as Attachment B Figures7 and 8 support this conclusion.
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
SmoothingParameter=0.6
student
5-
4-
3-
2-
1-
o-
-1:
-2-
-3-
-4:
-5 -
^P00^00000^^yw(^^'^yi^^ -
s£-J •"• sx
+
I I I
-5.00000 -4.00000 -3.00000
Predicted
reggendpov ° ° ° AU: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
^ Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
t~ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
8i
1 1 1 1
1
1
ni
-:Jx
^
1 '
-2.00000
|r^>°Q O
^••oepA^
^D
i i
-1.00000 0
logitprev
0
o
4-
X
X
0
o
+
X
X
0
0
4-
X
X
Midwest-Female-Abo vePovertyLev
Midwest-Male-Abo vePovertyLevel
Northeast-Female-Abo vePovertyL
Northeast-Male-Abo vePovertyLev
South-Female- Abo vePovertyLevel
South-Male- Abo vePovertyLevel
West-Female-Abo vePovertyLevel
West-Male-Abo vePovertyLevel
Figure 5C-2. Studentized residuals versus model predicted betas generated using a logistic
model and using the children 'EVER' asthmatic data set, with smoothing set to 0.6.
When considering both children asthma prevalence responses evaluated, the residual
standard error (estimated values for sigma) suggests the choice of smoothing parameter as 0.6 to
0.8. The normal probability plots of the studentized residuals suggest preference for smoothing
at or above 0.6. The plots of residuals against smoothed predictions suggest the choices of 0.4
5C-12
-------
323 through 0.6. We therefore chose the final value of 0.6 to use for smoothing the children's asthma
324 prevalence. For the adults, 0.9 was selected for smoothing.
325 Smoothed asthma prevalence and associated graphical presentation are provided in
326 Attachment C, following a similar format as the unsmoothed data provided in Attachment A.
327 5C-5. CENSUS TRACT LEVEL POVERTY RATIO DATA SET DESCRIPTION AND
328 PROCESSING
329 This section describes the approach used to generate census tract level poverty ratios for
330 all US census tracts, stratified by age and age groups where available. The data set generation
331 involved primarily two types of data downloaded from the 2000 US Census, each are described
332 below.
333 First, individual state level SF3 geographic data ("geo") .uf3 files and associated
334 documentation were downloaded10 and, following import by SAS (SAS, 2012), were screened
335 for tract level information using the "sumlev" variable equal to ' 140'. For quality control
336 purposes and ease of matching with the poverty level data, our geo data set retained the
337 following variables: stusab, sumlev, logrecno, state, county, tract, name, latitude, and longitude.
338 Second, the individual state level SF3 files ("30";) were downloaded, retaining the
339 number of persons across the variable "PCT50" for all state "logrecno". n The data provided by
340 the PCT50 variable is stratified by age or age groups (ages <5, 5, 6-11, 12-14, 15, 16-17, 18-24,
341 25-34, 35-44, 45-54, 55-64, 65-74, and >75) and income/poverty ratios, given in increments of
342 0.25. We calculated two new variables for each state logrecno using the number of persons from
343 the PCT50 stratifications; the fraction of those persons having poverty ratios < 1.5 and > 1.5 by
344 summing the appropriate PCT50 variable and dividing by the total number of persons in that
345 age/age group. Finally the poverty ratio data were combined with the above described census
346 tract level geographic data using the "stusab" and "logrecno" variables. The final output was a
347 single file containing relevant tract level poverty probabilities by age groups for all US census
348 tracts (where available).
10 Geographic data were obtained from http://www2.census.gov/census 2000/datasets/Summary File 3/.
Information regarding variable names is given in Figure 2-5 of US Census (2007).
11 Poverty ratio data were obtained from http://www2.census.gov/census 2000/datasets/Summary File 3/.
Information regarding poverty ratio names variable names is given in chapter 6 of US Census Bureau (2007). We
used the variable "PCT50", an income to poverty ratio variable stratified by various ages and age groups and
described in chapter 7 of US Census Bureau (2007).
5C-13
-------
349 5C-6. COMBINED CENSUS TRACT LEVEL POVERTY RATIO AND ASTHMA
350 PREVALENCE DATA
351 Because the prevalence data are stratified by standard US Census defined regions,12 we
352 first mapped the tract level poverty level data to an appropriate region based on the State.
353 Further, as APEX requires the input data files to be complete, additional processing of the
354 poverty probability file was needed. For where there was missing tract level poverty
355 information13, we substituted an age-specific value using the average for the particular county the
356 tract was located within. The frequency of missing data substitution comprised 1.7% of the total
357 poverty probability data set. The two data sets were merged and the final asthma prevalence was
358 calculated using the following weighting scheme:
359
360 prevalence=round((pov_prob*prev_poor)+((l-pov_prob)*prev_notpoor),0,0001);
361
362 whereas each US census tract value now expresses a tract specific poverty-weighted
363 prevalence, stratified by ages (children 0-17), age groups (adults), and two genders. These final
364 prevalence data are found within the APEX asthmaprevalence. txt file.
365
366 5C-7. REFERENCES
367 Cohen J and Rosenbaum A. (2005). Analysis of NHIS Asthma Prevalence Data. Memorandum to John Langstaff
368 by ICF Incorporated. For US EPA Work Assignment 3-08 under EPA contract 68D01052.
369 SAS. (2012). SAS/STAT 9.2 User's Guide, Second Edition. Available at:
370 http://support.sas.com/documentation/cdl/en/statug/63033/PDF/default/statug.pdf.
371 US Census Bureau. (2007). 2000 Census of Population and Housing. Summary File 3 (SF3) Technical
372 Documentation, available at: http://www.census.gov/prod/cen2000/doc/sf3.pdf. Individual SF3 files '30' (for
373 income/poverty variables pct50) for each state were downloaded from:
374 http://www2.census.gov/census 2000/datasets/Summary File 3/.
375 US EPA. (2007). Ozone Population Exposure Analysis for Selected Urban Areas (July 2007). Office of Air
376 Quality Planning and Standards, Research Triangle Park, NC. EPA-452/R-07-010. Available at:
377 http://epa.gOv/ttn/naaqs/standards/ozone/s o3crtd.html.
378 US EPA. (2008). Risk and Exposure Assessment to Support the Review of the NOa Primary National Ambient Air
379 Quality Standard. Report no. EPA-452/R-08-008a. November 2008. Available at:
380 http://www.epa.gov/ttn/naaqs/standards/nox/data/2008112 l_NO2_REA_final.pdf.
381 US EPA. (2009). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
382 Quality Standard. Report no. EPA-452/R-09-007. August 2009. Available at:
383 http://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
12 For example, see http://www.cdc.gov/std/stats 10/census.htm.
13 Whether there were no data collected by the Census or whether there were simply no persons in that age
group is relatively inconsequential to estimating the asthmatic persons exposed, particularly considering latter case
as no persons in that age group would be modeled.
5C-14
-------
384 APPENDIX 5C, ATTACHMENT A: UNSMOOTHED ASTHMA
385 PREVALENCE TABLES AND FIGURES.
Appendix 5C, Attachment A, Table-1. Unsmoothed prevalence for children "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
Prevalence
0.0018
0.0387
0.0367
0.0395
0.0815
0.0885
0.0438
0.1374
0.0820
0.1027
0.0995
0.1129
0.1752
0.1331
0.1944
0.1383
0.1731
0.1311
0.0564
0.0585
0.1256
0.1127
0.1746
0.1584
0.1229
0.0867
0.1523
0.2070
0.2293
0.1359
0.1501
0.1527
0.1197
0.2103
0.2054
0.1844
0.0061
0.0258
0.0848
0.0996
0.0876
0.1593
0.0977
0.1793
0.1503
0.1418
0.1569
0.1717
0.2054
0.1846
0.1671
0.1454
0.1557
0.1320
0.0293
0.1051
0.1786
0.2066
SE
0.0018
0.0233
0.0148
0.0186
0.0298
0.0207
0.0200
0.0277
0.0246
0.0220
0.0193
0.0277
0.0391
0.0256
0.0477
0.0302
0.0341
0.0256
0.0353
0.0197
0.0487
0.0419
0.0395
0.0447
0.0417
0.0353
0.0392
0.0486
0.1109
0.0470
0.0484
0.0380
0.0462
0.0760
0.0597
0.1134
0.0044
0.0178
0.0231
0.0261
0.0223
0.0313
0.0229
0.0313
0.0356
0.0265
0.0322
0.0371
0.0338
0.0358
0.0291
0.0356
0.0278
0.0233
0.0176
0.0376
0.0652
0.0513
LowerCI
0.0002
0.0117
0.0165
0.0155
0.0390
0.0556
0.0176
0.0916
0.0450
0.0669
0.0675
0.0688
0.1112
0.0905
0.1173
0.0890
0.1160
0.0885
0.0160
0.0299
0.0567
0.0529
0.1100
0.0888
0.0616
0.0381
0.0902
0.1275
0.0800
0.0670
0.0774
0.0921
0.0544
0.0980
0.1121
0.0491
0.0015
0.0066
0.0491
0.0588
0.0527
0.1069
0.0611
0.1259
0.0930
0.0973
0.1035
0.1106
0.1470
0.1244
0.1175
0.0885
0.1087
0.0926
0.0089
0.0509
0.0835
0.1236
UpperCI
0.0129
0.1208
0.0797
0.0972
0.1624
0.1382
0.1046
0.2010
0.1450
0.1545
0.1442
0.1797
0.2652
0.1916
0.3049
0.2086
0.2502
0.1898
0.1799
0.1112
0.2552
0.2240
0.2658
0.2664
0.2301
0.1851
0.2456
0.3182
0.5043
0.2562
0.2710
0.2427
0.2431
0.3949
0.3462
0.4976
0.0247
0.0957
0.1426
0.1636
0.1423
0.2306
0.1527
0.2489
0.2340
0.2021
0.2306
0.2568
0.2795
0.2650
0.2322
0.2297
0.2182
0.1848
0.0922
0.2047
0.3418
0.3247
5C-15
-------
Appendix 5C, Attachment A, Table-1. Unsmoothed prevalence for children "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Age
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
Prevalence
0.2770
0.2504
0.2186
0.2192
0.2902
0.1242
0.2897
0.2669
0.2589
0.2429
0.1470
0.1965
0.1855
0.3740
0.0055
0.0296
0.0697
0.0723
0.1142
0.1058
0.0933
0.1084
0.0780
0.1362
0.0979
0.1697
0.0535
0.0910
0.1500
0.1733
0.1884
0.1694
0.0315
0.1230
0.0703
0.1860
0.1666
0.2347
0.0682
0.0972
0.2049
0.1695
0.0988
0.2622
0.1377
0.3506
0.1869
0.1965
0.1986
0.1625
0.0256
0.0542
0.0635
0.0835
0.1378
0.1444
0.2175
0.2019
0.1878
0.1286
0.1879
0.2532
0.1801
0.1581
SE
0.0638
0.0499
0.0447
0.0456
0.0649
0.0437
0.0639
0.0613
0.1050
0.0693
0.0490
0.0509
0.0611
0.1042
0.0054
0.0164
0.0252
0.0250
0.0254
0.0296
0.0254
0.0251
0.0221
0.0374
0.0298
0.0382
0.0229
0.0273
0.0207
0.0355
0.0510
0.0395
0.0251
0.0576
0.0277
0.0555
0.0598
0.0636
0.0250
0.0362
0.0604
0.0698
0.0440
0.0734
0.0525
0.0762
0.0537
0.0534
0.0470
0.0602
0.0130
0.0231
0.0220
0.0232
0.0329
0.0357
0.0482
0.0343
0.0373
0.0342
0.0278
0.0420
0.0233
0.0340
LowerCI
0.1703
0.1656
0.1436
0.1428
0.1806
0.0607
0.1815
0.1646
0.1068
0.1329
0.0742
0.1150
0.0935
0.1998
0.0008
0.0099
0.0337
0.0362
0.0731
0.0602
0.0541
0.0681
0.0442
0.0780
0.0530
0.1073
0.0228
0.0499
0.1138
0.1142
0.1077
0.1052
0.0064
0.0469
0.0319
0.1002
0.0791
0.1329
0.0327
0.0458
0.1107
0.0717
0.0400
0.1445
0.0629
0.2188
0.1031
0.1120
0.1221
0.0754
0.0094
0.0231
0.0318
0.0478
0.0849
0.0875
0.1376
0.1429
0.1252
0.0751
0.1394
0.1799
0.1388
0.1022
UpperCI
0.4170
0.3600
0.3184
0.3211
0.4312
0.2374
0.4285
0.4021
0.5051
0.4017
0.2703
0.3151
0.3345
0.5884
0.0368
0.0854
0.1384
0.1394
0.1741
0.1793
0.1563
0.1682
0.1339
0.2272
0.1738
0.2578
0.1204
0.1604
0.1953
0.2541
0.3085
0.2613
0.1404
0.2852
0.1479
0.3193
0.3175
0.3802
0.1366
0.1944
0.3478
0.3505
0.2240
0.4277
0.2752
0.5100
0.3148
0.3217
0.3065
0.3158
0.0679
0.1218
0.1228
0.1418
0.2158
0.2291
0.3263
0.2774
0.2719
0.2115
0.2485
0.3439
0.2303
0.2366
5C-16
-------
Appendix 5C, Attachment A, Table-1. Unsmoothed prevalence for children "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Age
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
Prevalence
0.2043
0.1752
0.1798
0.1836
0.0375
0.1649
0.2200
0.1124
0.2651
0.2398
0.3209
0.2651
0.2905
0.3810
0.3382
0.2485
0.2819
0.2961
0.2876
0.2632
0.2407
0.3123
0.0129
0.0191
0.0558
0.0793
0.0834
0.0932
0.1446
0.1439
0.1111
0.1258
0.0626
0.1288
0.1064
0.1387
0.1621
0.1399
0.1362
0.1299
0.0495
0.0734
0.0828
0.0973
0.1578
0.1409
0.1536
0.1658
0.1428
0.2123
0.1408
0.2249
0.1741
0.1463
0.2428
0.1947
0.1285
0.1322
0.0135
0.0782
0.1134
0.1063
0.1679
0.1644
SE
0.0447
0.0287
0.0360
0.0282
0.0275
0.0506
0.0503
0.0445
0.0909
0.0651
0.0432
0.0572
0.0969
0.0773
0.1019
0.0708
0.0705
0.0685
0.0713
0.0661
0.0559
0.0734
0.0080
0.0084
0.0147
0.0200
0.0184
0.0222
0.0226
0.0248
0.0194
0.0222
0.0154
0.0210
0.0182
0.0222
0.0243
0.0169
0.0253
0.0197
0.0216
0.0210
0.0207
0.0271
0.0372
0.0300
0.0381
0.0332
0.0302
0.0413
0.0347
0.0466
0.0519
0.0296
0.0437
0.0399
0.0344
0.0323
0.0065
0.0162
0.0190
0.0211
0.0303
0.0226
LowerCI
0.1303
0.1257
0.1195
0.1346
0.0087
0.0877
0.1371
0.0501
0.1262
0.1355
0.2427
0.1686
0.1401
0.2446
0.1732
0.1359
0.1656
0.1808
0.1695
0.1548
0.1483
0.1885
0.0038
0.0080
0.0330
0.0479
0.0537
0.0579
0.1057
0.1017
0.0784
0.0883
0.0383
0.0928
0.0756
0.1006
0.1198
0.1100
0.0938
0.0959
0.0207
0.0415
0.0503
0.0556
0.0976
0.0917
0.0927
0.1104
0.0931
0.1425
0.0855
0.1467
0.0941
0.0972
0.1675
0.1280
0.0747
0.0807
0.0052
0.0517
0.0811
0.0714
0.1165
0.1247
UpperCI
0.3056
0.2387
0.2614
0.2454
0.1477
0.2887
0.3337
0.2330
0.4738
0.3885
0.4107
0.3908
0.5070
0.5392
0.5551
0.4102
0.4371
0.4448
0.4440
0.4107
0.3660
0.4701
0.0427
0.0447
0.0928
0.1286
0.1273
0.1467
0.1948
0.1996
0.1550
0.1762
0.1005
0.1759
0.1478
0.1881
0.2156
0.1763
0.1938
0.1737
0.1137
0.1268
0.1336
0.1649
0.2450
0.2103
0.2439
0.2414
0.2126
0.3042
0.2233
0.3288
0.2997
0.2142
0.3382
0.2847
0.2122
0.2092
0.0342
0.1165
0.1563
0.1554
0.2360
0.2136
5C-17
-------
Appendix 5C, Attachment A, Table-1. Unsmoothed prevalence for children "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Poverty Status
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Age
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Prevalence
0.1328
0.1542
0.1502
0.1522
0.1485
0.1767
0.1915
0.1939
0.1381
0.1579
0.1698
0.1530
0.0610
0.1005
0.1102
0.1699
0.1642
0.2510
0.2064
0.1588
0.2518
0.2246
0.2022
0.1890
0.2322
0.2345
0.2265
0.1801
0.1286
0.1916
0.0049
0.0390
0.0269
0.0439
0.0232
0.0988
0.0829
0.1065
0.0960
0.1124
0.0978
0.1186
0.1655
0.0855
0.1258
0.1482
0.1394
0.2285
0.0064
0.0443
0.0523
0.0403
0.0346
0.0887
0.1351
0.1364
0.1106
0.1254
0.0585
0.0747
0.0720
0.1898
0.1431
0.1168
SE
0.0212
0.0270
0.0224
0.0232
0.0240
0.0255
0.0236
0.0255
0.0196
0.0246
0.0193
0.0240
0.0181
0.0206
0.0225
0.0324
0.0288
0.0485
0.0339
0.0309
0.0503
0.0381
0.0368
0.0344
0.0383
0.0454
0.0489
0.0371
0.0303
0.0297
0.0037
0.0202
0.0097
0.0153
0.0079
0.0294
0.0223
0.0281
0.0280
0.0296
0.0285
0.0188
0.0352
0.0196
0.0278
0.0213
0.0254
0.0375
0.0064
0.0195
0.0220
0.0140
0.0177
0.0372
0.0432
0.0360
0.0244
0.0405
0.0204
0.0264
0.0279
0.0591
0.0431
0.0304
LowerCI
0.0964
0.1083
0.1114
0.1121
0.1073
0.1322
0.1495
0.1487
0.1039
0.1154
0.1352
0.1117
0.0338
0.0667
0.0732
0.1154
0.1152
0.1682
0.1477
0.1072
0.1663
0.1588
0.1394
0.1305
0.1656
0.1573
0.1448
0.1183
0.0799
0.1399
0.0011
0.0139
0.0132
0.0219
0.0118
0.0544
0.0484
0.0627
0.0534
0.0662
0.0545
0.0864
0.1074
0.0542
0.0806
0.1111
0.0967
0.1632
0.0009
0.0185
0.0226
0.0202
0.0126
0.0380
0.0703
0.0798
0.0711
0.0650
0.0292
0.0368
0.0331
0.0993
0.0773
0.0692
UpperCI
0.1802
0.2148
0.1994
0.2033
0.2018
0.2323
0.2419
0.2487
0.1813
0.2122
0.2110
0.2061
0.1076
0.1488
0.1626
0.2431
0.2285
0.3572
0.2808
0.2290
0.3622
0.3078
0.2839
0.2658
0.3153
0.3345
0.3361
0.2645
0.2005
0.2566
0.0216
0.1048
0.0541
0.0858
0.0450
0.1730
0.1384
0.1752
0.1666
0.1846
0.1695
0.1606
0.2463
0.1324
0.1911
0.1949
0.1969
0.3101
0.0441
0.1025
0.1166
0.0788
0.0919
0.1934
0.2439
0.2234
0.1682
0.2283
0.1137
0.1460
0.1496
0.3323
0.2495
0.1906
5C-18
-------
Appendix 5C, Attachment A, Table-1. Unsmoothed prevalence for children "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Age
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Prevalence
0.0814
0.0637
0.0000
0.0244
0.0517
0.0601
0.1698
0.1236
0.1376
0.1288
0.1018
0.1884
0.1604
0.2121
0.1833
0.2105
0.1475
0.1641
0.1958
0.2113
0.0135
0.0812
0.0417
0.1182
0.1349
0.1562
0.1853
0.1484
0.1549
0.1275
0.1742
0.1909
0.1678
0.1793
0.1919
0.1410
0.1863
0.2030
SE
0.0290
0.0235
0.0000
0.0121
0.0155
0.0172
0.0275
0.0288
0.0264
0.0354
0.0223
0.0315
0.0273
0.0298
0.0349
0.0397
0.0309
0.0263
0.0282
0.0289
0.0128
0.0317
0.0131
0.0351
0.0329
0.0401
0.0444
0.0343
0.0343
0.0418
0.0431
0.0554
0.0599
0.0491
0.0454
0.0577
0.0384
0.0493
LowerCI
0.0398
0.0305
0.0000
0.0092
0.0285
0.0339
0.1224
0.0772
0.0934
0.0738
0.0657
0.1342
0.1138
0.1596
0.1244
0.1431
0.0966
0.1188
0.1463
0.1602
0.0020
0.0370
0.0224
0.0647
0.0823
0.0926
0.1133
0.0928
0.0988
0.0654
0.1049
0.1046
0.0800
0.1021
0.1180
0.0606
0.1223
0.1229
UpperCI
0.1593
0.1285
0.0000
0.0635
0.0920
0.1041
0.2307
0.1918
0.1980
0.2152
0.1547
0.2579
0.2215
0.2762
0.2618
0.2987
0.2187
0.2224
0.2569
0.2733
0.0832
0.1691
0.0765
0.2061
0.2131
0.2514
0.2883
0.2288
0.2346
0.2338
0.2751
0.3227
0.3185
0.2959
0.2966
0.2946
0.2734
0.3165
5C-19
-------
Appendix 5C, Attachment A, Table 2. Unsmoothed prevalence for children "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
Prevalence
0.0018
0.0387
0.0302
0.0395
0.0531
0.0617
0.0386
0.0801
0.0492
0.0789
0.0625
0.0856
0.1269
0.1089
0.1580
0.0863
0.1300
0.0989
0.0564
0.0486
0.0959
0.0697
0.1697
0.0819
0.0809
0.0680
0.1257
0.1394
0.1871
0.0726
0.1101
0.1258
0.0999
0.1648
0.1647
0.1747
0.0061
0.0214
0.0752
0.0692
0.0527
0.1293
0.0710
0.1369
0.1047
0.1096
0.1004
0.1340
0.1093
0.1029
0.1230
0.1007
0.1141
0.0644
0.0274
0.0892
0.1786
0.1620
0.2557
0.1914
0.1432
0.1788
0.2414
0.1114
SE
0.0018
0.0233
0.0135
0.0186
0.0214
0.0173
0.0192
0.0239
0.0151
0.0200
0.0162
0.0232
0.0357
0.0264
0.0478
0.0213
0.0319
0.0236
0.0353
0.0183
0.0434
0.0338
0.0387
0.0265
0.0357
0.0325
0.0346
0.0398
0.1071
0.0266
0.0452
0.0354
0.0435
0.0745
0.0576
0.1141
0.0044
0.0175
0.0222
0.0203
0.0201
0.0303
0.0193
0.0301
0.0299
0.0269
0.0281
0.0348
0.0242
0.0210
0.0236
0.0305
0.0268
0.0193
0.0175
0.0369
0.0652
0.0475
0.0634
0.0400
0.0333
0.0378
0.0604
0.0404
LowerCI
0.0002
0.0117
0.0125
0.0155
0.0238
0.0354
0.0143
0.0442
0.0267
0.0476
0.0373
0.0498
0.0717
0.0669
0.0849
0.0526
0.0792
0.0613
0.0160
0.0229
0.0383
0.0263
0.1065
0.0428
0.0332
0.0261
0.0719
0.0779
0.0548
0.0349
0.0477
0.0711
0.0413
0.0640
0.0799
0.0429
0.0015
0.0042
0.0417
0.0385
0.0247
0.0805
0.0413
0.0878
0.0589
0.0669
0.0571
0.0791
0.0700
0.0684
0.0837
0.0548
0.0711
0.0354
0.0077
0.0386
0.0835
0.0888
0.1517
0.1248
0.0894
0.1162
0.1429
0.0533
UpperCI
0.0129
0.1208
0.0715
0.0972
0.1142
0.1055
0.0999
0.1411
0.0888
0.1280
0.1029
0.1433
0.2145
0.1724
0.2751
0.1382
0.2062
0.1556
0.1799
0.1000
0.2206
0.1723
0.2594
0.1512
0.1840
0.1661
0.2105
0.2369
0.4777
0.1451
0.2340
0.2130
0.2226
0.3629
0.3094
0.4997
0.0247
0.1008
0.1319
0.1213
0.1090
0.2011
0.1193
0.2072
0.1793
0.1745
0.1704
0.2179
0.1665
0.1520
0.1771
0.1780
0.1780
0.1143
0.0925
0.1927
0.3418
0.2772
0.3974
0.2821
0.2215
0.2649
0.3780
0.2180
5C-20
-------
Appendix 5C, Attachment A, Table 2. Unsmoothed prevalence for children "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Age
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
Prevalence
0.2022
0.1731
0.2271
0.1627
0.0967
0.1509
0.1167
0.3301
0.0055
0.0296
0.0697
0.0470
0.0717
0.0642
0.0709
0.0697
0.0609
0.0996
0.0740
0.1028
0.0386
0.0187
0.0907
0.1270
0.0974
0.1239
0.0078
0.1230
0.0658
0.1700
0.1139
0.2219
0.0583
0.0495
0.0850
0.0652
0.0988
0.2587
0.0882
0.3162
0.1293
0.1798
0.1429
0.1133
0.0131
0.0505
0.0635
0.0582
0.1007
0.1245
0.1990
0.1240
0.1482
0.0980
0.0999
0.1805
0.1204
0.0855
0.1243
0.1249
0.1198
0.0690
0.0375
0.1649
SE
0.0624
0.0406
0.1064
0.0591
0.0413
0.0506
0.0490
0.1005
0.0054
0.0164
0.0252
0.0158
0.0199
0.0196
0.0254
0.0180
0.0209
0.0334
0.0260
0.0305
0.0187
0.0095
0.0181
0.0344
0.0267
0.0375
0.0078
0.0576
0.0272
0.0576
0.0456
0.0583
0.0290
0.0252
0.0368
0.0294
0.0440
0.0734
0.0426
0.0739
0.0372
0.0479
0.0381
0.0426
0.0101
0.0227
0.0220
0.0216
0.0281
0.0318
0.0511
0.0274
0.0321
0.0321
0.0216
0.0342
0.0211
0.0237
0.0351
0.0247
0.0283
0.0173
0.0275
0.0506
LowerCI
0.1061
0.1072
0.0822
0.0767
0.0406
0.0757
0.0495
0.1683
0.0008
0.0099
0.0337
0.0240
0.0413
0.0349
0.0346
0.0416
0.0307
0.0507
0.0366
0.0565
0.0147
0.0069
0.0609
0.0733
0.0562
0.0671
0.0011
0.0469
0.0287
0.0842
0.0503
0.1282
0.0215
0.0179
0.0354
0.0264
0.0400
0.1416
0.0332
0.1913
0.0722
0.1039
0.0831
0.0527
0.0029
0.0206
0.0318
0.0277
0.0574
0.0742
0.1171
0.0795
0.0956
0.0506
0.0648
0.1229
0.0848
0.0491
0.0702
0.0839
0.0744
0.0418
0.0087
0.0877
UpperCI
0.3511
0.2675
0.4908
0.3125
0.2129
0.2781
0.2512
0.5456
0.0368
0.0854
0.1384
0.0897
0.1218
0.1151
0.1398
0.1143
0.1171
0.1865
0.1439
0.1797
0.0975
0.0500
0.1330
0.2108
0.1636
0.2177
0.0541
0.2852
0.1436
0.3133
0.2376
0.3561
0.1484
0.1294
0.1903
0.1521
0.2240
0.4249
0.2146
0.4746
0.2209
0.2930
0.2348
0.2269
0.0574
0.1185
0.1228
0.1181
0.1705
0.2013
0.3177
0.1885
0.2227
0.1813
0.1509
0.2573
0.1682
0.1449
0.2108
0.1819
0.1872
0.1117
0.1477
0.2887
5C-21
-------
Appendix 5C, Attachment A, Table 2. Unsmoothed prevalence for children "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Age
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
Prevalence
0.1621
0.1015
0.2486
0.1479
0.2630
0.1707
0.2056
0.3343
0.2276
0.1643
0.1117
0.1931
0.1714
0.2043
0.1684
0.2140
0.0129
0.0144
0.0452
0.0675
0.0540
0.0572
0.1002
0.0894
0.0762
0.0969
0.0473
0.0847
0.0768
0.0700
0.1059
0.0930
0.0702
0.0867
0.0404
0.0613
0.0704
0.0812
0.1404
0.1276
0.0792
0.1262
0.1185
0.1147
0.1038
0.1461
0.1299
0.1013
0.1699
0.1591
0.0633
0.0975
0.0044
0.0700
0.0911
0.0962
0.1230
0.1321
0.0999
0.1114
0.0946
0.1108
0.1010
0.0946
SE
0.0496
0.0440
0.0909
0.0487
0.0391
0.0507
0.0966
0.0680
0.0786
0.0600
0.0389
0.0430
0.0664
0.0555
0.0501
0.0526
0.0080
0.0076
0.0169
0.0196
0.0150
0.0138
0.0186
0.0191
0.0160
0.0210
0.0135
0.0165
0.0152
0.0158
0.0211
0.0186
0.0156
0.0162
0.0203
0.0183
0.0193
0.0254
0.0367
0.0304
0.0288
0.0305
0.0290
0.0286
0.0301
0.0366
0.0490
0.0262
0.0385
0.0365
0.0273
0.0299
0.0025
0.0162
0.0195
0.0206
0.0259
0.0204
0.0192
0.0214
0.0168
0.0202
0.0186
0.0175
LowerCI
0.0864
0.0420
0.1131
0.0753
0.1939
0.0926
0.0751
0.2162
0.1093
0.0770
0.0552
0.1223
0.0764
0.1162
0.0912
0.1286
0.0038
0.0051
0.0215
0.0379
0.0311
0.0354
0.0692
0.0584
0.0502
0.0627
0.0269
0.0576
0.0518
0.0447
0.0711
0.0624
0.0451
0.0597
0.0149
0.0338
0.0408
0.0434
0.0826
0.0789
0.0381
0.0775
0.0724
0.0694
0.0579
0.0879
0.0600
0.0602
0.1071
0.0998
0.0267
0.0526
0.0014
0.0442
0.0595
0.0627
0.0805
0.0970
0.0681
0.0758
0.0664
0.0770
0.0699
0.0655
UpperCI
0.2835
0.2255
0.4621
0.2701
0.3463
0.2935
0.4521
0.4776
0.4145
0.3164
0.2132
0.2914
0.3410
0.3338
0.2901
0.3345
0.0427
0.0402
0.0926
0.1175
0.0920
0.0911
0.1431
0.1346
0.1141
0.1466
0.0819
0.1231
0.1124
0.1080
0.1550
0.1364
0.1077
0.1242
0.1050
0.1085
0.1189
0.1471
0.2286
0.1997
0.1573
0.1989
0.1881
0.1836
0.1792
0.2331
0.2589
0.1655
0.2590
0.2441
0.1427
0.1737
0.0135
0.1092
0.1373
0.1449
0.1833
0.1774
0.1443
0.1608
0.1330
0.1569
0.1438
0.1348
5C-22
-------
Appendix 5C, Attachment A, Table 2. Unsmoothed prevalence for children "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Poverty Status
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Age
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
Prevalence
0.1340
0.1122
0.0713
0.0899
0.0871
0.0700
0.0477
0.0859
0.0820
0.1434
0.1320
0.2314
0.1395
0.1207
0.2064
0.1364
0.1473
0.1390
0.1673
0.1684
0.0936
0.1379
0.0816
0.1057
0.0013
0.0353
0.0159
0.0284
0.0183
0.0689
0.0477
0.0469
0.0756
0.0686
0.0791
0.0763
0.1023
0.0571
0.1012
0.0923
0.0787
0.1303
0.0064
0.0443
0.0249
0.0372
0.0114
0.0491
0.1016
0.0908
0.0874
0.0839
0.0275
0.0339
0.0551
0.1028
0.1312
0.0630
0.0758
0.0328
0.0000
0.0039
0.0305
0.0384
SE
0.0207
0.0226
0.0153
0.0158
0.0147
0.0178
0.0162
0.0197
0.0201
0.0319
0.0265
0.0486
0.0302
0.0269
0.0474
0.0279
0.0315
0.0286
0.0339
0.0449
0.0305
0.0353
0.0275
0.0289
0.0013
0.0202
0.0076
0.0132
0.0071
0.0276
0.0166
0.0144
0.0263
0.0196
0.0250
0.0124
0.0260
0.0163
0.0251
0.0207
0.0214
0.0294
0.0064
0.0195
0.0153
0.0137
0.0102
0.0294
0.0419
0.0302
0.0258
0.0267
0.0137
0.0160
0.0254
0.0393
0.0440
0.0247
0.0287
0.0163
0.0000
0.0040
0.0113
0.0129
LowerCI
0.0983
0.0750
0.0466
0.0635
0.0623
0.0421
0.0242
0.0544
0.0503
0.0914
0.0881
0.1498
0.0902
0.0771
0.1285
0.0903
0.0956
0.0917
0.1109
0.0975
0.0485
0.0820
0.0415
0.0609
0.0002
0.0113
0.0062
0.0113
0.0085
0.0308
0.0239
0.0255
0.0376
0.0388
0.0420
0.0553
0.0614
0.0323
0.0615
0.0590
0.0458
0.0827
0.0009
0.0185
0.0074
0.0179
0.0020
0.0148
0.0440
0.0464
0.0484
0.0443
0.0103
0.0133
0.0219
0.0474
0.0662
0.0288
0.0354
0.0122
0.0000
0.0005
0.0147
0.0197
UpperCI
0.1801
0.1646
0.1077
0.1260
0.1206
0.1141
0.0916
0.1330
0.1309
0.2178
0.1931
0.3397
0.2097
0.1840
0.3145
0.2009
0.2203
0.2051
0.2445
0.2752
0.1729
0.2226
0.1544
0.1772
0.0095
0.1045
0.0401
0.0695
0.0389
0.1468
0.0928
0.0846
0.1459
0.1185
0.1440
0.1043
0.1655
0.0989
0.1622
0.1416
0.1322
0.1993
0.0441
0.1025
0.0805
0.0756
0.0638
0.1506
0.2174
0.1698
0.1529
0.1532
0.0715
0.0839
0.1315
0.2089
0.2435
0.1324
0.1546
0.0850
0.0000
0.0289
0.0623
0.0735
5C-23
-------
Appendix 5C, Attachment A, Table 2. Unsmoothed prevalence for children "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Above Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Below Poverty
Age
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Prevalence
0.1363
0.0933
0.0803
0.1014
0.0537
0.1120
0.1202
0.1333
0.1258
0.1039
0.0873
0.0881
0.1066
0.1364
0.0135
0.0812
0.0308
0.0944
0.1056
0.0856
0.1277
0.0943
0.1282
0.0883
0.0697
0.0954
0.0759
0.0600
0.1457
0.1099
0.0957
0.1136
SE
0.0261
0.0268
0.0208
0.0320
0.0182
0.0242
0.0253
0.0271
0.0286
0.0328
0.0217
0.0222
0.0230
0.0284
0.0128
0.0317
0.0080
0.0311
0.0306
0.0256
0.0356
0.0353
0.0343
0.0287
0.0228
0.0365
0.0316
0.0276
0.0391
0.0551
0.0350
0.0421
LowerCI
0.0927
0.0523
0.0478
0.0537
0.0273
0.0726
0.0788
0.0885
0.0796
0.0549
0.0531
0.0532
0.0692
0.0897
0.0020
0.0370
0.0185
0.0486
0.0588
0.0471
0.0726
0.0443
0.0746
0.0459
0.0363
0.0440
0.0329
0.0239
0.0844
0.0394
0.0458
0.0534
UpperCI
0.1960
0.1608
0.1317
0.1834
0.1029
0.1689
0.1791
0.1959
0.1934
0.1879
0.1404
0.1425
0.1607
0.2021
0.0832
0.1691
0.0510
0.1755
0.1822
0.1508
0.2149
0.1897
0.2115
0.1632
0.1298
0.1947
0.1655
0.1427
0.2398
0.2713
0.1894
0.2254
5C-24
-------
Appendix 5C, Attachment A, Table 3. Unsmoothed prevalence for adults "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Age grp
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
Prevalence
0.1633
0.1347
0.1214
0.1157
0.1360
0.1104
0.0990
0.1990
0.1896
0.1789
0.1903
0.2760
0.1459
0.1295
0.1658
0.1254
0.0934
0.0659
0.0856
0.0884
0.0808
0.1672
0.1103
0.0945
0.1445
0.1623
0.1474
0.0830
0.1834
0.1375
0.1297
0.1209
0.1306
0.1244
0.0844
0.1642
0.1726
0.1771
0.2140
0.2174
0.1752
0.0941
0.1658
0.1262
0.0773
0.0976
0.0911
0.0926
0.0689
0.1753
0.1255
0.1317
0.1189
0.1681
0.1383
0.0943
0.1501
0.1290
0.1050
0.1163
0.1279
0.1231
0.0939
0.1511
SE
0.0154
0.0096
0.0084
0.0072
0.0103
0.0107
0.0095
0.0156
0.0177
0.0209
0.0180
0.0255
0.0205
0.0202
0.0158
0.0092
0.0083
0.0057
0.0086
0.0106
0.0110
0.0182
0.0156
0.0191
0.0204
0.0203
0.0307
0.0217
0.0199
0.0107
0.0109
0.0095
0.0106
0.0130
0.0101
0.0194
0.0170
0.0172
0.0204
0.0232
0.0186
0.0132
0.0223
0.0126
0.0094
0.0086
0.0096
0.0128
0.0127
0.0200
0.0178
0.0244
0.0162
0.0490
0.0313
0.0265
0.0121
0.0084
0.0074
0.0060
0.0087
0.0102
0.0092
0.0133
LowerCI
0.1353
0.1169
0.1059
0.1022
0.1171
0.0910
0.0819
0.1701
0.1573
0.1415
0.1576
0.2289
0.1101
0.0948
0.1371
0.1085
0.0784
0.0555
0.0701
0.0697
0.0617
0.1345
0.0832
0.0632
0.1089
0.1263
0.0968
0.0492
0.1476
0.1178
0.1097
0.1034
0.1113
0.1010
0.0666
0.1296
0.1418
0.1459
0.1767
0.1753
0.1417
0.0712
0.1265
0.1034
0.0607
0.0820
0.0740
0.0704
0.0478
0.1395
0.0945
0.0909
0.0906
0.0923
0.0875
0.0536
0.1279
0.1134
0.0914
0.1051
0.1119
0.1044
0.0773
0.1269
UpperCI
0.1958
0.1547
0.1389
0.1306
0.1575
0.1332
0.1193
0.2314
0.2268
0.2237
0.2281
0.3285
0.1908
0.1744
0.1990
0.1446
0.1109
0.0779
0.1040
0.1114
0.1050
0.2060
0.1447
0.1391
0.1893
0.2061
0.2182
0.1367
0.2256
0.1598
0.1527
0.1409
0.1528
0.1523
0.1064
0.2059
0.2084
0.2132
0.2567
0.2664
0.2147
0.1234
0.2142
0.1531
0.0980
0.1158
0.1117
0.1209
0.0982
0.2179
0.1648
0.1872
0.1545
0.2865
0.2118
0.1606
0.1754
0.1464
0.1205
0.1285
0.1459
0.1446
0.1136
0.1790
5C-25
-------
Appendix 5C, Attachment A, Table 3. Unsmoothed prevalence for adults "EVER" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age grp
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
Prevalence
0.1336
0.1452
0.1622
0.2039
0.1616
0.1127
0.1438
0.1095
0.0890
0.0704
0.0782
0.0789
0.0893
0.1473
0.0914
0.0972
0.1062
0.1068
0.0966
0.0702
0.1595
0.1387
0.1368
0.1431
0.1478
0.1541
0.1231
0.1522
0.1191
0.1466
0.1874
0.1747
0.1318
0.1370
0.1499
0.1304
0.0984
0.0944
0.0917
0.1168
0.1208
0.1589
0.0846
0.0760
0.1422
0.0979
0.1349
0.0937
SE
0.0087
0.0125
0.0128
0.0179
0.0163
0.0133
0.0100
0.0078
0.0066
0.0051
0.0071
0.0078
0.0111
0.0152
0.0122
0.0139
0.0138
0.0156
0.0149
0.0130
0.0150
0.0096
0.0109
0.0092
0.0094
0.0130
0.0117
0.0184
0.0118
0.0182
0.0219
0.0181
0.0179
0.0198
0.0188
0.0107
0.0080
0.0081
0.0075
0.0126
0.0160
0.0222
0.0128
0.0135
0.0214
0.0176
0.0323
0.0194
LowerCI
0.1175
0.1224
0.1386
0.1711
0.1321
0.0891
0.1253
0.0952
0.0769
0.0610
0.0654
0.0649
0.0698
0.1199
0.0701
0.0732
0.0821
0.0799
0.0710
0.0486
0.1323
0.1209
0.1168
0.1261
0.1303
0.1302
0.1020
0.1195
0.0978
0.1145
0.1483
0.1419
0.1005
0.1027
0.1167
0.1108
0.0837
0.0796
0.0780
0.0943
0.0928
0.1201
0.0626
0.0535
0.1052
0.0684
0.0831
0.0620
UpperCI
0.1515
0.1714
0.1889
0.2413
0.1962
0.1415
0.1645
0.1258
0.1027
0.0811
0.0932
0.0956
0.1135
0.1797
0.1184
0.1280
0.1363
0.1414
0.1301
0.1004
0.1911
0.1586
0.1595
0.1621
0.1671
0.1813
0.1479
0.1920
0.1441
0.1859
0.2341
0.2131
0.1709
0.1806
0.1905
0.1527
0.1153
0.1116
0.1076
0.1438
0.1558
0.2073
0.1133
0.1069
0.1894
0.1381
0.2116
0.1393
5C-26
-------
Appendix 5C, Attachment A, Table 4. Unsmoothed prevalence for adults "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Age grp
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
Prevalence
0.1062
0.0859
0.0859
0.0858
0.0996
0.0755
0.0643
0.1306
0.1329
0.1354
0.1398
0.2110
0.1190
0.1029
0.0790
0.0599
0.0486
0.0447
0.0555
0.0524
0.0477
0.0938
0.0572
0.0731
0.0969
0.1350
0.1349
0.0643
0.1123
0.0917
0.0944
0.0858
0.0945
0.0898
0.0706
0.1232
0.1180
0.1265
0.1745
0.1744
0.1388
0.0488
0.0888
0.0655
0.0409
0.0564
0.0469
0.0641
0.0527
0.0780
0.0847
0.0795
0.0798
0.1322
0.1055
0.0758
0.0893
0.0731
0.0689
0.0716
0.0865
0.0914
0.0599
0.0996
SE
0.0133
0.0090
0.0081
0.0061
0.0090
0.0083
0.0073
0.0144
0.0143
0.0187
0.0166
0.0221
0.0180
0.0183
0.0125
0.0066
0.0063
0.0049
0.0059
0.0076
0.0088
0.0143
0.0137
0.0162
0.0208
0.0205
0.0294
0.0213
0.0148
0.0102
0.0092
0.0080
0.0086
0.0106
0.0098
0.0182
0.0147
0.0138
0.0185
0.0211
0.0148
0.0088
0.0161
0.0093
0.0061
0.0078
0.0085
0.0105
0.0110
0.0129
0.0171
0.0212
0.0196
0.0492
0.0296
0.0247
0.0090
0.0064
0.0051
0.0049
0.0064
0.0090
0.0072
0.0119
LowerCI
0.0828
0.0699
0.0713
0.0746
0.0832
0.0608
0.0514
0.1049
0.1073
0.1027
0.1102
0.1709
0.0879
0.0722
0.0577
0.0482
0.0377
0.0360
0.0450
0.0394
0.0331
0.0693
0.0355
0.0470
0.0630
0.0997
0.0869
0.0332
0.0864
0.0735
0.0778
0.0714
0.0790
0.0711
0.0537
0.0918
0.0921
0.1018
0.1412
0.1369
0.1123
0.0341
0.0620
0.0495
0.0304
0.0429
0.0328
0.0463
0.0348
0.0562
0.0566
0.0467
0.0489
0.0617
0.0600
0.0395
0.0732
0.0615
0.0595
0.0626
0.0747
0.0753
0.0473
0.0786
UpperCI
0.1354
0.1052
0.1031
0.0986
0.1188
0.0934
0.0802
0.1614
0.1634
0.1764
0.1757
0.2575
0.1590
0.1448
0.1071
0.0743
0.0625
0.0554
0.0683
0.0694
0.0682
0.1258
0.0908
0.1119
0.1461
0.1804
0.2035
0.1208
0.1447
0.1138
0.1141
0.1029
0.1127
0.1128
0.0924
0.1634
0.1499
0.1560
0.2137
0.2196
0.1704
0.0693
0.1257
0.0862
0.0547
0.0738
0.0667
0.0880
0.0789
0.1075
0.1248
0.1322
0.1275
0.2608
0.1789
0.1406
0.1086
0.0866
0.0797
0.0818
0.1000
0.1105
0.0756
0.1254
5C-27
-------
Appendix 5C, Attachment A, Table 4. Unsmoothed prevalence for adults "STILL" having asthma.
Smoothed
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
No
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age grp
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
Prevalence
0.0867
0.1152
0.1369
0.1780
0.1303
0.0895
0.0608
0.0471
0.0451
0.0359
0.0413
0.0441
0.0636
0.0617
0.0344
0.0488
0.0800
0.0676
0.0687
0.0331
0.0908
0.0819
0.0994
0.0937
0.1013
0.1103
0.0783
0.0901
0.0861
0.1081
0.1391
0.1293
0.1053
0.1061
0.0620
0.0528
0.0582
0.0499
0.0542
0.0756
0.0711
0.0741
0.0457
0.0344
0.1119
0.0528
0.1159
0.0442
SE
0.0079
0.0113
0.0123
0.0173
0.0152
0.0118
0.0079
0.0053
0.0048
0.0040
0.0055
0.0057
0.0097
0.0086
0.0064
0.0109
0.0131
0.0122
0.0129
0.0083
0.0143
0.0070
0.0090
0.0095
0.0087
0.0114
0.0092
0.0135
0.0111
0.0143
0.0179
0.0164
0.0166
0.0162
0.0104
0.0068
0.0061
0.0065
0.0072
0.0102
0.0133
0.0132
0.0097
0.0089
0.0198
0.0137
0.0336
0.0131
LowerCI
0.0725
0.0948
0.1144
0.1467
0.1033
0.0689
0.0471
0.0377
0.0365
0.0288
0.0317
0.0342
0.0470
0.0468
0.0239
0.0314
0.0579
0.0473
0.0473
0.0202
0.0663
0.0691
0.0830
0.0766
0.0854
0.0898
0.0621
0.0669
0.0667
0.0831
0.1075
0.1005
0.0770
0.0782
0.0445
0.0410
0.0473
0.0386
0.0416
0.0579
0.0491
0.0520
0.0301
0.0207
0.0786
0.0316
0.0644
0.0246
UpperCI
0.1035
0.1393
0.1629
0.2144
0.1631
0.1154
0.0782
0.0587
0.0556
0.0446
0.0535
0.0567
0.0855
0.0810
0.0494
0.0751
0.1097
0.0957
0.0987
0.0539
0.1231
0.0968
0.1186
0.1141
0.1197
0.1347
0.0982
0.1202
0.1105
0.1394
0.1781
0.1648
0.1425
0.1424
0.0858
0.0679
0.0715
0.0642
0.0702
0.0982
0.1019
0.1046
0.0689
0.0568
0.1570
0.0870
0.1996
0.0781
5C-28
-------
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Above Poverty Level
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
prev
0 1
4 5
9 10 11 12 13 14 15 16 17
345
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
gender
Female
Male
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
prev
0.6-
0 1
' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i ' ' ' ' i '
3 4 5 6 7 8 9 10 11 12 13 14 15
' ' i ' ' ' ' i
16 17
prev
0.6
0.4-
± \
X
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
age
Female ^~^~ Male
gender
Female
Male
Appendix 5C, Attachment A, Figure 1. Unsmoothed prevalence and confidence intervals for children 'EVER' having asthma.
5C-29
-------
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat= Above Poverty Level
prev
1 1 ' ' • ' i ' ' ' ' i • ' ' ' i • ' ' • i • ' ' • i ' ' ' • i '
• i ' • ' ' i ' • ' ' i '
4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
1 i' ' ' ' i '' ' ' i '' ' ' i ' '' ' i ' '' ' i '
01234567
1 l ' ' ' ' l ' ' ' ' l ' ' ' ' l ' ' ' ' l ' ' ' ' l ' ' ' ' l ' ' ' ' l '
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=West pov_rat= Above Poverty Level
1 i' ' ' • i' • ' ' i '
345678
10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 1. Raw asthma 'EVER' prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
0 1
34567
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Appendix 5C, Attachment A, Figure 1, cont. Unsmoothed prevalence and confidence intervals for children 'EVER' having
asthma.
5C-30
-------
Figure 2. Raw asthma 'STILL1 prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
Figure 2. Raw asthma 'STILL1 prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
34567
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Raw asthma 'STILL' prevalence rates and confidence intervals
region=Midwest pov_rat=Above Poverty Level
1 I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
age
Female
Male
3456
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Raw asthma 'STILL' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
prev
345
gender
7 8
10 11 12 13 14 15 16 17
age
Appendix 5C, Attachment A, Figure 2. Unsmoothed prevalence and confidence intervals for children 'STILL' having asthma.
5C-31
-------
Figure 2. Raw asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat=Above Poverty Level
prev
I ' ' ' ' I' ' ' ' I' ' ' ' I '' ' ' I '' ' ' I ' '' ' I ' '' ' I ' ' '' I '
012345678
1 r ''' i '''' i'''' i'''' i '''' i '''' i '''
10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Raw asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
prev
0.6-
0.4-
0.3-
0.2:
o.i-
0.0-
mmi
4567
1 I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' '
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Raw asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Above Poverty Level
prev
0.6-
4567
1 I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' ' ' I ' ' '
9 10 11 12 13 14 15 16 17
gender
age
Fern ale
Male
Figure 2. Raw asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
prev
I ' ' ' ' I' ' ' ' I'
01234567
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Appendix 5C, Attachment A, Figure 2, cont. Unsmoothed prevalence and confidence intervals for children 'STILL' having
asthma.
5C-32
-------
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Above Poverty Level
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
18-24
25-34
gender
age_grp
Female
Male
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
75+
prev
0.4-
35-44
gender
45-54
age_grp
Female
55-64
Male
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
75+
18-24
gender
Female
Male
18-24
gender
Female
Male
Appendix 5C, Attachment A, Figure 3. Unsmoothed prevalence and confidence intervals for adults 'EVER' having asthma.
5C-33
-------
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat=Above Poverty Level
18-24
25-34
35-44
gender
45-54
age_grp
Female
55-64
65-74
Male
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
18-24
gender
age_grp
Female Male
75+
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=West pov_rat=Above Poverty Level
prev
0.4:
0.2-
18-24
45-54 55-64
65-74
gender
age_grp
Female
Male
Figure 3. Raw adult asthma 'EVER' prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
45-54
55-64
gender
age_grp
Female Male
Appendix 5C, Attachment A, Figure 3, cont. Unsmoothed prevalence and confidence intervals for adults 'EVER' having
asthma.
75+
5C-34
-------
Figure 4. Raw adult asthma 'STILL1 prevalence rates and confidence intervals
region=Midwest pov rat=Above Poverty Level
Figure 4. Raw adult asthma 'STILL1 prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
prev
0.4
gender
45-54
age_grp
Female
Figure 4. Raw adult asthma 'STILL' prevalence rates and confidence intervals
region=Midwest pov rat=Below Poverty Level
gender
age_grp
Female
Figure 4. Raw adult asthma 'STILL' prevalence rates and confidence intervals
region=Northeast pov rat=Below Poverty Level
prev
0.4-
18-24
25-34
35-44
45-54
55-64
65-74
gender
age_grp
Female
gender
age_grp
Female
Appendix 5C, Attachment A, Figure 4. Unsmoothed prevalence and confidence intervals for adults 'STILL' having asthma.
5C-35
-------
Figure 4. Raw adult asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat=Above Poverty Level
Figure 4. Raw adult asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Above Poverty Level
prev
0.4:
0.3:
18-24
25-34
35-44
gender
age_grp
Female
Male
Figure 4. Raw adult asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
prev
0.4-
gender
45-54
age_grp
Female
Male
Figure 4. Raw adult asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
18-24
35-44
gender
45-54
age_grp
Female
55-64
Male
65-74
75+
18-24
25-34
35-44
gender
45-54
age_grp
Female
Male
Appendix 5C, Attachment A, Figure 4, cont. Unsmoothed prevalence and confidence intervals for adults 'STILL' having
asthma.
5C-36
-------
386 APPENDIX 5C, ATTACHMENT B: LOGISTIC MODEL FIT TABLES
387 AND FIGURES.
Appendix 5C, Attachment B, Table 1. Alternative logistic models for estimating child asthma prevalence using the "EVER" asthma
response variable and goodness of fit test results.
Description
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
Stratification Variable
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1 . none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
-2 log likelihood
288740115.1
287062346.4
288120804.1
287385013.1
286367652.6
286283543.6
285696164.7
284477928.1
286862135.1
285098650.6
286207721.5
285352164
284330346.1
284182547.5
283587631.7
282241318.6
286227019.6
284470413
285546716.1
284688169.9
283662673.5
283404487.5
282890785.3
281407414.3
285821686.2
283843266.2
284761522.8
284045849.2
282099156.1
281929968.5
281963915.7
278655423.1
DF
2
4
8
4
16
16
8
32
3
6
12
6
24
24
12
48
4
8
16
8
32
32
16
64
18
36
72
36
144
144
72
288
Appendix 5C, Attachment B, Table 2. Alternative logistic models for estimating child asthma prevalence using the "STILL" asthma
response variable and goodness of fit test results.
Description
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
1. logit(prob) = linear in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
2. logit(prob) = quadratic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
Stratification Variable
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1 . none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
-2 log likelihood
181557347.7
180677544.6
180947344.2
180502490.5
179996184.8
179517528
179637601.4
178567573.9
180752073.1
179771977.6
180088080.5
179611530.4
179004935.6
178519078.1
178640744.8
177414967.2
180247874.1
179235170
179583725.1
179067549.2
178407915.7
DF
2
4
8
4
16
16
8
32
3
6
12
6
24
24
12
48
4
8
16
8
32
5C-37
-------
Appendix 5C, Attachment B, Table 2. Alternative logistic models for estimating child asthma prevalence using the "STILL" asthma
response variable and goodness of fit test results.
Description
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
3. logit(prob) = cubic in age
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
4. logit(prob) = f(age)
Stratification Variable
6. region, poverty
7. gender, poverty
8. region, gender, poverty
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
-2 log likelihood
177897359.3
178029240
176642073.7
179972765.3
178918713.8
178852704.9
178599743.4
177075815.4
176418872.7
177422457.4
173888684.9
DF
32
16
64
18
36
72
36
144
144
72
288
Appendix 5C, Attachment B, Table 3. Alternative logistic models for estimating adult asthma prevalence using the "EVER" asthma
response variable and goodness of fit test results.
Description
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
Stratification Variable
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
-2 log likelihood
825494282
821614711.2
824598583.4
823443004.3
820520390.7
821958349.1
819560679.9
817723710
DF
7
14
28
14
56
56
28
112
Appendix 5C, Attachment B, Table 4. Alternative logistic models for estimating adult asthma prevalence using the "STILL" asthma
response variable and goodness of fit test results.
Description
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
4. logit(prob) = f(age grp)
Stratification Variable
1. none
2. gender
3. region
4. poverty
5. region, gender
6. region, poverty
7. gender, poverty
8. region, gender, poverty
-2 log likelihood
600538044.1
594277797.3
599561222.3
597511872.6
593112157.6
596008068.6
591394271.8
589398969.5
DF
7
14
28
14
56
56
28
112
5C-38
-------
Appendix 5C, Attachment B, Table 5. Effect on residual standard error by varying LOESS smoothing parameter
while fitting children "EVER" having asthma data set.
Region
South
Northeast
South
Midwest
Midwest
South
South
Midwest
West
South
Midwest
Midwest
Midwest
Midwest
Northeast
South
South
Midwest
West
Northeast
South
Northeast
Midwest
Northeast
Midwest
West
West
South
South
Midwest
Midwest
South
Northeast
Midwest
South
South
Midwest
Midwest
Midwest
Midwest
South
South
Northeast
South
Midwest
West
South
South
South
West
South
West
South
Northeast
Midwest
Midwest
South
Northeast
West
West
Gender
Female
Female
Male
Male
Male
Female
Male
Male
Female
Female
Female
Male
Male
Female
Female
Male
Female
Female
Male
Female
Female
Female
Male
Female
Male
Female
Male
Female
Female
Male
Female
Male
Female
Male
Male
Female
Male
Male
Male
Female
Male
Male
Female
Male
Female
Male
Male
Female
Male
Female
Male
Male
Female
Male
Male
Male
Female
Male
Female
Female
Poverty Ratio
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Smoothing Parameter
0.5
0.7
0.6
0.9
0.8
0.8
0.5
1
0.7
0.7
0.7
0.4
0.8
0.6
0.4
0.7
0.6
0.8
0.6
0.6
0.5
0.8
0.7
0.5
0.9
0.8
0.5
0.9
0.4
0.7
0.9
0.6
0.9
0.5
0.7
0.4
0.6
0.4
1
0.5
0.8
0.5
1
1
1
0.4
0.8
0.6
0.9
0.9
0.4
0.4
0.7
0.6
0.6
0.5
0.8
0.5
0.6
1
Residual Standard Error
0.999919
1.00088
1.003839
1.00548
1.010889
1.012178
0.982885
1.023284
0.973279
0.97298
1.028007
0.970948
0.965591
1.038233
0.961444
1.040867
0.954946
1.045107
1.052418
0.946315
0.945525
1.054556
0.940657
0.940383
1.063971
1.066819
1.067075
1.067923
0.930104
0.929292
1.072631
0.927161
1.074984
0.917969
0.912266
1.089646
0.90827
0.906073
1.094737
1.096459
1.099725
0.898228
1.101884
0.896985
1.103976
0.894137
0.893364
0.891551
0.890138
1.111538
0.885511
1.115223
0.86999
0.86934
0.86245
0.857982
0.857778
0.857592
0.852664
1.147894
5C-39
-------
Appendix 5C, Attachment B, Table 5. Effect on residual standard error by varying LOESS smoothing parameter
while fitting children "EVER" having asthma data set.
Region
South
South
Northeast
West
West
West
South
West
West
South
Northeast
South
West
Northeast
West
Northeast
West
Midwest
Northeast
Northeast
Midwest
West
Midwest
Northeast
South
Northeast
West
Northeast
West
Northeast
West
West
Midwest
West
West
Northeast
West
Northeast
Northeast
Northeast
Northeast
Midwest
Northeast
West
Northeast
West
Northeast
Northeast
Midwest
West
Midwest
Midwest
Gender
Female
Female
Male
Male
Female
Female
Male
Female
Female
Male
Male
Female
Female
Male
Female
Male
Male
Female
Male
Male
Female
Male
Female
Male
Male
Male
Male
Male
Female
Male
Male
Female
Female
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Male
Female
Male
Female
Male
Female
Female
Poverty Ratio
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Smoothing Parameter
1
0.9
0.7
0.7
0.9
0.8
0.4
0.7
1
0.9
0.8
1
0.6
1
0.5
0.9
1
1
1
0.7
0.4
0.5
0.9
0.6
1
0.8
0.9
0.9
0.5
0.4
0.6
0.4
0.8
0.8
0.7
0.5
0.8
0.8
0.7
0.9
0.6
0.7
0.5
0.4
1
0.9
0.4
0.4
0.6
1
0.5
0.4
Residual Standard Error
0.849143
0.847567
0.844668
1.163749
1.163943
1.166005
0.826195
1.174564
1.178045
1.178803
0.820245
1.182254
1.187757
0.811815
0.808706
0.805685
0.804743
0.799988
0.799128
0.798212
1.20612
0.793132
0.788082
0.78547
1.216423
0.78144
0.780843
0.779772
1.224495
0.769037
0.763027
0.762134
0.758775
0.756848
0.752592
0.729776
1.284153
1.292845
1.296274
1.308752
1.309671
0.688366
1.314991
1.31595
1.327129
1.35931
1.37577
0.618785
0.607758
1.395061
0.541466
0.522325
5C-40
-------
Appendix 5C, Attachment B, Table 6. Effect on residual standard error by varying LOESS smoothing parameter
while fitting children "STILL" having asthma data set.
Region
South
Northeast
Northeast
Northeast
Northeast
Midwest
Midwest
Midwest
Midwest
Northeast
South
Midwest
Northeast
South
Northeast
Northeast
Northeast
Midwest
South
Northeast
Northeast
Midwest
Midwest
Midwest
Northeast
Midwest
South
Northeast
South
West
South
South
South
South
West
South
West
West
South
West
Midwest
West
West
Midwest
Northeast
West
Midwest
West
South
Midwest
West
Midwest
Northeast
West
Midwest
Midwest
Midwest
South
Gender
Female
Male
Male
Male
Male
Male
Male
Female
Male
Male
Male
Female
Male
Female
Male
Male
Male
Male
Female
Male
Male
Male
Female
Male
Male
Male
Female
Male
Male
Male
Male
Male
Female
Female
Male
Male
Female
Female
Male
Female
Male
Male
Male
Female
Female
Female
Female
Male
Female
Female
Male
Male
Female
Female
Male
Female
Female
Male
Poverty Ratio
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Smoothing Parameter
1
0.9
0.7
0.9
0.4
0.7
0.8
1
0.6
0.8
0.6
0.5
0.5
0.4
0.6
0.5
0.8
0.9
0.6
1
1
0.5
0.9
1
0.7
0.7
0.7
0.6
1
0.9
0.5
0.7
0.8
0.9
0.8
0.9
0.7
1
0.8
0.9
0.8
0.7
1
0.8
0.4
0.8
0.6
0.7
0.5
0.7
0.6
0.4
0.5
0.7
0.9
0.9
1
0.4
Residual Standard Error
1.000117
1.000909
1.000993
0.997502
0.997275
0.996943
0.996544
1.003498
0.995815
0.995723
1.007198
0.99235
1.008536
0.99041
1.009859
1.01048
1.011028
1.011038
1.013156
1.01445
1.016505
1.01692
0.979917
1.020707
1.021388
0.977074
0.976479
1.024042
0.975784
1.025093
1.026184
0.971057
0.965833
0.965238
1.03481
0.964953
1.036384
1.040924
0.957162
1.044522
1.04601
1.04802
1.050309
0.946142
0.94543
1.055218
0.938888
1.063545
1.063816
0.931681
1.079146
1.080605
1.083479
1.084472
1.084476
0.914962
0.913089
1.087093
5C-41
-------
Appendix 5C, Attachment B, Table 6. Effect on residual standard error by varying LOESS smoothing parameter
while fitting children "STILL" having asthma data set.
Region
Midwest
West
Midwest
Midwest
Northeast
South
Midwest
West
Midwest
Northeast
West
West
South
Midwest
Midwest
West
South
Northeast
South
Northeast
Northeast
West
Northeast
West
South
West
South
West
South
West
West
West
Midwest
West
Midwest
South
South
South
South
West
West
Midwest
South
Northeast
Northeast
West
South
Northeast
West
South
Northeast
Northeast
Northeast
Northeast
Gender
Female
Female
Male
Male
Female
Male
Male
Female
Female
Male
Female
Male
Female
Male
Female
Female
Female
Female
Female
Female
Female
Male
Female
Male
Female
Male
Female
Female
Female
Female
Female
Male
Female
Female
Female
Male
Male
Female
Male
Female
Male
Female
Male
Female
Female
Male
Male
Female
Male
Male
Female
Female
Female
Female
Poverty Ratio
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Smoothing Parameter
0.8
0.6
0.6
1
0.6
0.4
0.5
0.6
0.7
0.4
0.5
0.5
0.6
0.4
0.4
0.8
1
0.7
0.5
0.9
1
0.4
0.8
0.4
0.7
0.6
0.9
0.5
0.8
0.4
0.9
0.5
0.6
1
0.4
0.5
0.6
0.4
0.7
0.4
0.8
0.5
0.8
0.7
0.6
0.9
0.9
0.8
1
1
0.5
0.9
1
0.4
Residual Standard Error
0.912722
0.912605
0.907737
1.103127
1.103286
1.112998
0.878223
1.124127
0.875579
0.874469
0.873529
1.127032
0.87206
0.869726
1.135372
1.136048
0.863066
1.140006
0.858107
1.147352
1.148471
1.152015
1.153553
0.845979
0.842335
0.8413
0.841106
1.166931
0.830955
0.826586
1.183444
0.815615
0.802622
1.20757
0.78769
1.214019
1.216661
0.781555
1.242272
1.252141
1.254244
0.742493
1.294055
1.32003
1.355219
1.356792
1.365737
1.39015
1.405599
1.408469
1.431367
1.503674
1.574778
1.605
5C-42
-------
Appendix 5C, Attachment B, Table 7. Effect on residual standard error by varying LOESS smoothing parameter
while fitting adults "EVER" having asthma data set.
Region
Midwest
South
West
West
South
Midwest
West
West
West
Northeast
West
Midwest
Northeast
South
Midwest
Midwest
South
Northeast
South
South
South
Northeast
West
South
South
Northeast
Northeast
Northeast
West
Northeast
Northeast
South
Midwest
West
South
West
Northeast
Midwest
West
West
South
Northeast
Northeast
Midwest
Midwest
Midwest
Midwest
Midwest
Gender
Female
Female
Female
Male
Female
Female
Male
Female
Male
Female
Male
Female
Male
Male
Male
Male
Female
Male
Female
Female
Female
Male
Female
Male
Male
Female
Male
Male
Female
Female
Male
Male
Male
Female
Male
Female
Female
Male
Male
Male
Male
Female
Female
Male
Male
Female
Female
Female
Poverty Ratio
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Smoothing Parameter
1
1
0.9
0.8
1
0.9
0.8
0.8
0.9
1
1
0.8
0.8
1
0.9
1
0.9
1
0.8
0.9
0.8
1
1
0.9
1
0.9
0.9
0.9
1
1
0.8
0.9
1
0.8
0.8
0.9
0.8
0.9
0.9
1
0.8
0.9
0.8
0.8
0.8
1
0.9
0.8
Residual Standard Error
0.983356
1.040607
1.044712
0.937658
1.06598
0.911278
1.095844
0.893319
0.886119
0.875056
0.858542
0.843191
1.177547
0.813689
1.190978
0.785268
0.77381
1.241548
0.751726
0.747912
0.740577
0.732859
1.275049
0.708509
0.706944
0.699107
1.301543
0.677309
0.669638
0.662619
0.646318
0.64328
1.395026
0.597305
0.58427
0.567466
0.528031
0.49517
1.523816
1.537805
0.400237
0.394894
0.362058
0.306085
0.169594
1.910643
1.920542
2.249162
5C-43
-------
Appendix 5C, Attachment B, Table 8. Effect on residual standard error by varying LOESS smoothing parameter
while fitting adults "STILL" having asthma data set.
Region
South
West
West
West
West
West
West
South
Midwest
Northeast
Midwest
South
Midwest
South
South
West
South
Midwest
Northeast
Northeast
South
South
South
West
South
Northeast
Northeast
Midwest
Northeast
Northeast
Northeast
Midwest
South
Northeast
South
Midwest
Midwest
Northeast
West
Northeast
Northeast
Midwest
Midwest
Midwest
Midwest
West
West
West
Gender
Male
Female
Female
Female
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Female
Male
Female
Male
Male
Female
Male
Female
Female
Male
Female
Male
Male
Male
Male
Male
Female
Male
Male
Female
Female
Male
Male
Female
Female
Male
Female
Male
Male
Male
Poverty Ratio
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Smoothing Parameter
0.8
0.8
0.9
1
1
0.8
0.9
0.9
1
1
0.9
0.8
0.8
1
0.9
1
0.9
1
0.9
1
1
0.9
0.8
0.9
1
0.8
1
1
0.9
0.9
0.8
0.9
1
1
0.8
0.9
0.8
0.8
0.8
0.9
0.8
1
0.9
0.8
0.8
0.8
1
0.9
Residual Standard Error
1.015193
1.045714
1.051807
1.061488
0.92928
0.925921
0.915895
1.097531
0.89825
1.102905
0.876146
1.128781
0.870507
1.130393
0.835583
0.825684
1.192655
0.788217
0.786205
1.21537
1.23752
0.748499
0.717121
0.670751
0.664236
0.65848
0.653985
0.650735
0.630298
1.370134
1.375365
0.620174
1.400273
0.581032
0.568428
0.508247
0.503315
0.478186
0.464598
0.453855
0.396203
1.616706
1.636938
0.295923
1.883863
2.16547
2.200364
2.396381
5C-44
-------
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.4
25 50 75
Normal Percentiles
90 95
99
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.6
I I
5 50 75
Normal Percentiles
90
I
95
I
99
r
99.9
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.5
5 10
I I
5 50 75
Normal Percentiles
90 95
I
99
99.9
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.7
25 50 75
Normal Percentiles
90 95
99
99.9
Appendix 5C, Attachment B, Figure 1. Normal probability plots of studentized residuals generated using logistic model and children
'EVER' asthmatic data set.
5C-45
-------
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.8
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.9
\
\
\ I I
5 10 25 50 75 90 95
Normal Percentiles
99
O
99.9
O
I I
5 50 75
Normal Percentiles
90 95
I
99
r
99.9
$
•9 0
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter= 1
5 10 25 50 75
Normal Percentiles
90 95
99
99.9
Appendix 5C, Attachment B, Figure 1, cont. Normal probability plots of studentized residuals generated using logistic model and children
'EVER' asthmatic data set.
5C-46
-------
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
Smo othingParameter=0.4
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
Smoothing? arameter=0.6
Normal Percentiles
Normal Percentiles
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
Smo othingParameter=0.5
75
90
99
99.9
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
SmoothingParameter=0.7
0.1
5 10
75
90
99
99.9
Normal Percentiles
Normal Percentiles
Appendix 5C, Attachment B, Figure 2. Normal probability plots of studentized residuals generated using logistic model and children
'STILL' asthmatic data set.
5C-47
-------
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
Smo othingParameter=0.8
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
Smoothing? arameter= 1
O
Normal Percentiles
Normal Percentiles
Normal probability plot of studentized residuals by smoothing parameter
All genders, regions, poverty ratios combined
Smo othingParameter=0.9
4 -
-4 -
5 10 25
75
90 95
Normal Percentiles
99
99.9
Appendix 5C, Attachment B, Figure 2, cont Normal probability plots of studentized residuals generated using logistic model and children
'STILL' asthmatic data set.
5C-48
-------
Normal probability plot of studentized residuals by smoothing parameter
Adults: All genders, regions, poverty ratios combined
SmoothingParameter=0.8
n i
5 10
\ I I
25 50 75
Normal Percentiles
I
90
I
95
00
I
99
99.9
Normal probability plot of studentized residuals by smoothing parameter
Adults: All genders, regions, poverty ratios combined
SmoothingParameter= 1
o
o
i
10
I
75
I
90
99
Normal Percentiles
Normal probability plot of studentized residuals by smoothing parameter
Adults: All genders, regions, poverty ratios combined
SmoothingParameter=0.9
O
o
25 50 75
Normal Percentiles
90 95
99
99.9
Appendix 5C, Attachment B, Figure 3. Normal probability plots of studentized residuals generated using logistic model and adult 'EVER'
asthmatic data set.
5C-49
-------
Normal probability plot of studentized residuals by smoothing parameter
Adults Still: All genders, regions, poverty ratios combined
SmoothingParameter=0.8
O
\ I
5 10
\ I I
25 50 75
Normal Percentiles
I
90
\
95
O
I
99
1 -
Normal probability plot of studentized residuals by smoothing parameter
Adults Still: All genders, regions, poverty ratios combined
SmoothingParameter= 1
O 00
O
99.9
I
75
I
90
99
Normal Percentiles
Normal probability plot of studentized residuals by smoothing parameter
Adults Still: All genders, regions, poverty ratios combined
SmoothingParameter=0.9
O
10 25 50 75
Normal Percentiles
90 95
99
99.9
Appendix 5C, Attachment B, Figure 4. Normal probability plots of studentized residuals generated using logistic model and adult 'STILL'
asthmatic data set.
5C-50
-------
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
SmoothingParameter=0.4
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
Smo othingParameter=0.6
student
student
reggendpov
O X
-4.00000
-3.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O C Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
SmoothingParameter=0.5
-4.00000
-3.00000
-1.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
Northeast-Female-BelowPovertyL
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
u u Northeast-Female-BelowPovertyL
O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
X X South-Male-BelowPovertyLevel
O O O Northeast-Fern
o ~ -- -
XXX South-Male-BelowPovertyLevci
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
Smo othingParameter=0.7
/s /s W est-r emale-AbovePovertyLe\
X X West-Male-AbovePovertyLevel
student
student
Predicted logitprev
Predicted logitprev
reggendpov 0 O 0 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
"t" South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
reggendpov O O O All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
v v v "^st-Male-BelowPovertyLevel
XXX
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Appendix 5C, Attachment B, Figure 5. Studentized residuals generated using logistic model versus model predicted betas and the child
'EVER' asthmatic data set.
5C-51
-------
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
SmoothingParameter=0.8
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
Smoothing? arameter= 1
student
reggendpov
*o
-4.00000
-3.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ + 4 South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O C Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of prevalence rates by smoothing parameter
SmoothingParameter=0.9
student
O X
student
5-
4:
3-
2-
1 -
O
-4.00000
-3.00000
-2.00000
Predicted logitprev
reggendpov 000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
Midwest-Male-L^.... , v „.. „
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O ( Northeast-Male-AbovePovertyLev
~*" ~*" ~*" South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Predicted logitprev
reggendpov 0 O O All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
"t" South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLei
O O O Midwest-Male-AbovePovertyLevel
~r bouth-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Appendix 5C, Attachment B, Figure 5, cont. Studentized residuals generated using logistic model versus model predicted betas and the child
'EVER' asthmatic data set.
5C-52
-------
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
SmoothingParameter=0.4
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
Smo othingParameter=0.6
student
reggendpov
O
-4.00000
-3.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O ( Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
SmoothingParameter=0.5
student
O
-4.00000
-3.00000
-1.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-remate-BetowPovertyLev
Midwest-Male-BelowPovertyLevel
Northeast-Female-BelowPovertyL
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLeve]
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
Smo othingParameter=0.7
student
O
student
Predicted logitprev
Predicted logitprev
reggendpov 0 O 0 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
"t" South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
reggendpov O O 0 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
~t~ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
v v v "r,st-Male-BelowPovertyLevel
XXX
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Appendix 5C, Attachment B, Figure 6. Studentized residuals generated using logistic model versus model predicted betas and the child
'STILL' asthmatic data set.
5C-53
-------
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
SmoothingParameter=0.8
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
SmoothingParameter= 1
student
student
reggendpov
-4.00000
-3.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O ( Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of still prevalence rates by smoothing parameter
SmoothingParameter=0.9
-4.00000
-3.00000
-1.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-remale-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
Northeast-Female-BelowPovertyL
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLeve]
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
student
O
O
reggendpov
Predicted logitprev
All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
"t" South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Appendix 5C, Attachment B, Figure 6, cont. Studentized residuals generated using logistic model versus model predicted betas using child
'STILL' asthmatic data set.
5C-54
-------
Studentized residual versus smoothed logits of adult prevalence rates by smoothing parameter
SmoothingParameter=0.8
Studentized residual versus smoothed logits of adult prevalence rates by smoothing parameter
SmoothingParameter= 1
student
3-
-3.00000
reggendpov
-2.00000
Predicted logitprev
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ + 4 South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O C Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of adult prevalence rates by smoothing parameter
SmoothingParameter=0.9
student
3-
2-
1 -
o-
-1 -
2 -
-3-
O xO
+ OgjO X >t|3*pxX, ~
0 ^QXX+X x x Q
O
§P
°
-3.00000
Predicted logitprev
reggendpov 000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
Midwest-Male-L^.... . v „.. „
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O ( Northeast-Male-AbovePovertyLev
~*" ~*" ~*" South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
student
3-
2-
1 -
o-
-1 •
-2-
-3-
®+<7-^%
0 X*
X °
%$!J$$%Qfe<#b
x + ^ x x o
O
reggendpov
Predicted logitprev
0 O O All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
"t" South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLei
O O O Midwest-Male-AbovePovertyLevel
~r bouth-Femate-AbovePovertyLevet
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Appendix 5C, Attachment B, Figure 7. Studentized residuals generated using logistic model versus model predicted betas using adult
'EVER' asthmatic data set.
5C-55
-------
Studentized residual versus smoothed logits of adult still prevalence rates by smoothing parameter
SmoothingParameter=0.8
Studentized residual versus smoothed logits of adult still prevalence rates by smoothing parameter
Smoothing? arameter= 1
student
3-
1 •
-1 -
-2 "
-3-
X
vv\ ••?;aja^^#:^^^'ii8^({fe)./vnsrvvn
O
-4.00000
-3.00000
Predicted logitprev
reggendpov
000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ + 4 South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O C Northeast-Male-AbovePovertyLev
+ + + South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Studentized residual versus smoothed logits of adult still prevalence rates by smoothing parameter
SmoothingParameter=0.9
student
3:
2-.
I -
-4.00000
-3.00000
Predicted logitprev
reggendpov 000 All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
Midwest-Male-L^.... , v „.. „
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
+ South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLev
O O O Midwest-Male-AbovePovertyLevel
Northeast-Female-AbovePovertyL
O O ( Northeast-Male-AbovePovertyLev
~*" ~*" ~*" South-Female-AbovePovertyLevel
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
student
3-
1 -
o-
-1 •
-3-
0 x x><+ +°
^OOC^^^v^'^^P^'-^'^'^^^o
x"6p 0°
X
reggendpov
Predicted logitprev
0 O O All: LOESS Smoothed
O O O Midwest-Female-BelowPovertyLev
Midwest-Male-BelowPovertyLevel
O O O Northeast-Female-BelowPovertyL
O O O Northeast-Male-BelowPovertyLev
"t" South-Female-BelowPovertyLevel
XXX South-Male-BelowPovertyLevel
XXX West-Female-BelowPovertyLevel
XXX West-Male-BelowPovertyLevel
O O O Midwest-Female-AbovePovertyLei
O O O Midwest-Male-AbovePovertyLevel
~r bouth-Femate-AbovePovertyLevet
South-Male-AbovePovertyLevel
XXX West-Female-AbovePovertyLevel
XXX West-Male-AbovePovertyLevel
Appendix 5C, Attachment B, Figure 8. Studentized residuals generated using logistic model versus model predicted betas using adult
'STILL' asthmatic data set.
5C-56
-------
APPENDIX 5C, ATTACHMENT C: SMOOTHED ASTHMA
PREVALENCE TABLES AND FIGURES.
Appendix 5C, Attachment C, Table 1. Smoothed prevalence for children "EVER" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
Prevalence
0.0083
0.0179
0.0327
0.0509
0.0671
0.0854
0.0995
0.1041
0.1024
0.1020
0.1055
0.1192
0.1390
0.1529
0.1603
0.1597
0.1517
0.1374
0.0413
0.0706
0.1047
0.1356
0.1553
0.1488
0.1327
0.1341
0.1535
0.1729
0.1861
0.1691
0.1470
0.1439
0.1541
0.1707
0.1962
0.2323
0.0133
0.0313
0.0585
0.0898
0.1111
0.1256
0.1411
0.1496
0.1502
0.1542
0.1627
0.1760
0.1876
0.1847
0.1764
0.1641
0.1487
0.1318
0.0429
0.0908
0.1530
0.2110
SE
0.0050
0.0066
0.0076
0.0096
0.0122
0.0134
0.0141
0.0145
0.0132
0.0121
0.0127
0.0137
0.0163
0.0176
0.0176
0.0160
0.0161
0.0229
0.0168
0.0168
0.0173
0.0208
0.0237
0.0229
0.0228
0.0224
0.0239
0.0270
0.0311
0.0300
0.0247
0.0239
0.0244
0.0275
0.0427
0.0813
0.0066
0.0091
0.0102
0.0121
0.0145
0.0149
0.0158
0.0164
0.0161
0.0166
0.0173
0.0181
0.0186
0.0181
0.0170
0.0149
0.0144
0.0201
0.0176
0.0214
0.0235
0.0277
LowerCI
0.0022
0.0079
0.0195
0.0336
0.0448
0.0602
0.0725
0.0765
0.0769
0.0784
0.0806
0.0922
0.1070
0.1182
0.1254
0.1277
0.1197
0.0945
0.0167
0.0416
0.0724
0.0962
0.1100
0.1053
0.0902
0.0920
0.1080
0.1215
0.1272
0.1131
0.1006
0.0990
0.1078
0.1186
0.1187
0.1002
0.0045
0.0164
0.0398
0.0666
0.0831
0.0964
0.1100
0.1171
0.1182
0.1211
0.1283
0.1397
0.1501
0.1483
0.1422
0.1341
0.1198
0.0937
0.0173
0.0536
0.1084
0.1566
UpperCI
0.0310
0.0397
0.0541
0.0766
0.0993
0.1198
0.1351
0.1403
0.1352
0.1317
0.1369
0.1527
0.1787
0.1956
0.2026
0.1979
0.1903
0.1956
0.0985
0.1174
0.1491
0.1879
0.2146
0.2062
0.1910
0.1912
0.2136
0.2401
0.2640
0.2451
0.2097
0.2045
0.2156
0.2395
0.3065
0.4512
0.0391
0.0588
0.0851
0.1200
0.1471
0.1621
0.1793
0.1892
0.1891
0.1942
0.2041
0.2193
0.2319
0.2277
0.2167
0.1994
0.1833
0.1823
0.1026
0.1498
0.2118
0.2780
5C-57
-------
Appendix 5C, Attachment C, Table 1. Smoothed prevalence for children "EVER" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Age
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
Prevalence
0.2428
0.2458
0.2393
0.2261
0.2225
0.2354
0.2499
0.2553
0.2512
0.2149
0.1941
0.2027
0.2364
0.3045
0.0115
0.0278
0.0533
0.0823
0.1027
0.1066
0.1023
0.0979
0.1010
0.1146
0.1179
0.1170
0.1154
0.1246
0.1405
0.1551
0.1714
0.1883
0.0394
0.0754
0.1188
0.1539
0.1684
0.1503
0.1355
0.1263
0.1322
0.1583
0.1818
0.2030
0.2293
0.2437
0.2368
0.2188
0.1906
0.1572
0.0279
0.0444
0.0668
0.0948
0.1269
0.1665
0.1891
0.1901
0.1858
0.1873
0.1908
0.1926
0.1934
0.1847
SE
0.0303
0.0285
0.0270
0.0268
0.0290
0.0311
0.0339
0.0357
0.0377
0.0355
0.0308
0.0292
0.0390
0.0768
0.0066
0.0095
0.0108
0.0127
0.0152
0.0150
0.0143
0.0137
0.0144
0.0166
0.0171
0.0175
0.0164
0.0148
0.0148
0.0152
0.0209
0.0376
0.0211
0.0229
0.0229
0.0265
0.0295
0.0269
0.0245
0.0231
0.0257
0.0301
0.0342
0.0358
0.0359
0.0366
0.0335
0.0286
0.0298
0.0443
0.0107
0.0103
0.0106
0.0134
0.0174
0.0209
0.0207
0.0204
0.0189
0.0189
0.0180
0.0163
0.0168
0.0172
LowerCI
0.1828
0.1888
0.1853
0.1729
0.1655
0.1741
0.1831
0.1852
0.1779
0.1473
0.1353
0.1462
0.1617
0.1652
0.0032
0.0131
0.0340
0.0584
0.0737
0.0777
0.0749
0.0715
0.0734
0.0828
0.0852
0.0836
0.0838
0.0955
0.1109
0.1245
0.1302
0.1189
0.0119
0.0383
0.0770
0.1043
0.1131
0.1003
0.0902
0.0836
0.0853
0.1029
0.1183
0.1355
0.1600
0.1726
0.1713
0.1625
0.1335
0.0822
0.0119
0.0265
0.0470
0.0692
0.0933
0.1257
0.1478
0.1494
0.1479
0.1494
0.1545
0.1595
0.1592
0.1499
UpperCI
0.3150
0.3133
0.3033
0.2900
0.2924
0.3101
0.3311
0.3409
0.3423
0.3025
0.2703
0.2741
0.3320
0.4921
0.0402
0.0583
0.0827
0.1150
0.1413
0.1445
0.1383
0.1325
0.1375
0.1566
0.1611
0.1615
0.1568
0.1611
0.1765
0.1916
0.2223
0.2851
0.1222
0.1433
0.1789
0.2214
0.2432
0.2193
0.1987
0.1862
0.1993
0.2358
0.2689
0.2926
0.3172
0.3323
0.3179
0.2879
0.2645
0.2796
0.0639
0.0733
0.0940
0.1284
0.1702
0.2173
0.2387
0.2389
0.2307
0.2322
0.2333
0.2307
0.2329
0.2253
5C-58
-------
Appendix 5C, Attachment C, Table 1. Smoothed prevalence for children "EVER" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Age
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
Prevalence
0.1797
0.1781
0.1795
0.1838
0.0946
0.1345
0.1759
0.2132
0.2353
0.2638
0.2909
0.3169
0.3272
0.3238
0.3163
0.3022
0.2846
0.2779
0.2702
0.2698
0.2745
0.2843
0.0137
0.0266
0.0453
0.0687
0.0928
0.1142
0.1298
0.1333
0.1231
0.1095
0.1033
0.1086
0.1212
0.1368
0.1437
0.1448
0.1395
0.1283
0.0496
0.0682
0.0893
0.1111
0.1319
0.1473
0.1553
0.1592
0.1650
0.1766
0.1825
0.1805
0.1837
0.1932
0.1891
0.1760
0.1560
0.1298
0.0335
0.0629
0.0985
0.1306
0.1472
0.1523
SE
0.0168
0.0156
0.0162
0.0251
0.0396
0.0296
0.0264
0.0326
0.0361
0.0316
0.0305
0.0339
0.0405
0.0439
0.0429
0.0412
0.0388
0.0367
0.0343
0.0316
0.0349
0.0575
0.0056
0.0064
0.0068
0.0086
0.0112
0.0123
0.0128
0.0123
0.0117
0.0109
0.0102
0.0103
0.0110
0.0113
0.0111
0.0104
0.0113
0.0172
0.0153
0.0123
0.0116
0.0141
0.0171
0.0181
0.0183
0.0183
0.0188
0.0198
0.0216
0.0219
0.0221
0.0218
0.0202
0.0181
0.0195
0.0271
0.0089
0.0093
0.0094
0.0116
0.0133
0.0130
LowerCI
0.1458
0.1465
0.1467
0.1350
0.0365
0.0817
0.1251
0.1503
0.1653
0.2004
0.2287
0.2475
0.2451
0.2356
0.2304
0.2199
0.2074
0.2048
0.2016
0.2062
0.2048
0.1760
0.0056
0.0156
0.0325
0.0522
0.0710
0.0900
0.1042
0.1085
0.0996
0.0877
0.0830
0.0881
0.0991
0.1138
0.1210
0.1235
0.1166
0.0952
0.0250
0.0458
0.0670
0.0838
0.0987
0.1120
0.1193
0.1231
0.1277
0.1374
0.1398
0.1373
0.1401
0.1499
0.1487
0.1398
0.1178
0.0810
0.0186
0.0453
0.0797
0.1073
0.1204
0.1259
UpperCI
0.2195
0.2149
0.2178
0.2452
0.2240
0.2134
0.2416
0.2932
0.3236
0.3388
0.3621
0.3954
0.4214
0.4265
0.4169
0.3995
0.3769
0.3651
0.3518
0.3445
0.3573
0.4250
0.0334
0.0450
0.0629
0.0901
0.1203
0.1439
0.1605
0.1627
0.1512
0.1359
0.1279
0.1332
0.1475
0.1635
0.1699
0.1690
0.1661
0.1709
0.0962
0.1004
0.1181
0.1459
0.1740
0.1914
0.1997
0.2035
0.2104
0.2241
0.2347
0.2336
0.2371
0.2453
0.2374
0.2192
0.2037
0.2015
0.0596
0.0867
0.1212
0.1581
0.1787
0.1831
5C-59
-------
Appendix 5C, Attachment C, Table 1. Smoothed prevalence for children "EVER" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
Prevalence
0.1539
0.1485
0.1461
0.1517
0.1639
0.1772
0.1794
0.1752
0.1705
0.1652
0.1600
0.1562
0.0629
0.0922
0.1253
0.1578
0.1852
0.1975
0.2038
0.2087
0.2078
0.2080
0.2122
0.2137
0.2192
0.2199
0.2059
0.1946
0.1827
0.1709
0.0131
0.0188
0.0264
0.0361
0.0469
0.0647
0.0857
0.1008
0.1032
0.1063
0.1166
0.1181
0.1196
0.1202
0.1241
0.1389
0.1665
0.2118
0.0250
0.0309
0.0387
0.0488
0.0602
0.0843
0.1143
0.1295
0.1195
0.0950
0.0786
0.0812
0.0979
0.1278
0.1324
0.1188
SE
0.0128
0.0125
0.0123
0.0124
0.0129
0.0134
0.0128
0.0127
0.0120
0.0108
0.0118
0.0190
0.0140
0.0118
0.0123
0.0156
0.0186
0.0190
0.0198
0.0204
0.0203
0.0206
0.0203
0.0202
0.0214
0.0220
0.0209
0.0186
0.0177
0.0246
0.0067
0.0057
0.0053
0.0064
0.0083
0.0105
0.0130
0.0144
0.0151
0.0144
0.0140
0.0129
0.0131
0.0130
0.0127
0.0125
0.0152
0.0305
0.0138
0.0099
0.0082
0.0099
0.0129
0.0169
0.0197
0.0191
0.0175
0.0151
0.0139
0.0150
0.0179
0.0221
0.0211
0.0176
LowerCI
0.1278
0.1231
0.1212
0.1265
0.1375
0.1496
0.1530
0.1491
0.1458
0.1428
0.1358
0.1189
0.0383
0.0694
0.1008
0.1265
0.1479
0.1592
0.1639
0.1675
0.1669
0.1664
0.1711
0.1727
0.1759
0.1755
0.1639
0.1571
0.1471
0.1235
0.0042
0.0096
0.0171
0.0245
0.0317
0.0451
0.0611
0.0733
0.0746
0.0786
0.0893
0.0927
0.0938
0.0945
0.0987
0.1136
0.1358
0.1525
0.0073
0.0152
0.0243
0.0312
0.0374
0.0538
0.0776
0.0930
0.0861
0.0666
0.0530
0.0537
0.0651
0.0866
0.0925
0.0853
UpperCI
0.1842
0.1782
0.1752
0.1810
0.1943
0.2085
0.2093
0.2049
0.1984
0.1902
0.1876
0.2026
0.1016
0.1215
0.1547
0.1951
0.2294
0.2424
0.2506
0.2570
0.2558
0.2567
0.2601
0.2612
0.2698
0.2718
0.2554
0.2385
0.2246
0.2317
0.0400
0.0365
0.0407
0.0531
0.0689
0.0919
0.1189
0.1372
0.1412
0.1424
0.1509
0.1494
0.1513
0.1519
0.1548
0.1687
0.2025
0.2864
0.0819
0.0618
0.0612
0.0757
0.0955
0.1296
0.1652
0.1775
0.1636
0.1338
0.1150
0.1209
0.1447
0.1848
0.1859
0.1631
5C-60
-------
Appendix 5C, Attachment C, Table 1. Smoothed prevalence for children "EVER" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Prevalence
0.0917
0.0600
0.0057
0.0191
0.0479
0.0903
0.1300
0.1437
0.1374
0.1290
0.1365
0.1560
0.1794
0.1980
0.1948
0.1818
0.1771
0.1801
0.1897
0.2081
0.0258
0.0442
0.0700
0.1005
0.1323
0.1609
0.1663
0.1582
0.1536
0.1543
0.1630
0.1746
0.1828
0.1809
0.1800
0.1828
0.1881
0.1964
SE
0.0164
0.0186
0.0035
0.0067
0.0092
0.0114
0.0149
0.0158
0.0157
0.0148
0.0148
0.0154
0.0160
0.0175
0.0180
0.0175
0.0164
0.0148
0.0149
0.0248
0.0126
0.0124
0.0119
0.0144
0.0190
0.0218
0.0213
0.0205
0.0204
0.0214
0.0240
0.0270
0.0270
0.0276
0.0259
0.0233
0.0242
0.0396
LowerCI
0.0615
0.0300
0.0014
0.0084
0.0306
0.0673
0.0993
0.1110
0.1050
0.0985
0.1058
0.1236
0.1454
0.1608
0.1566
0.1449
0.1423
0.1484
0.1577
0.1567
0.0087
0.0237
0.0479
0.0729
0.0959
0.1186
0.1247
0.1182
0.1140
0.1128
0.1168
0.1230
0.1306
0.1280
0.1298
0.1371
0.1405
0.1234
UpperCI
0.1347
0.1163
0.0229
0.0428
0.0743
0.1201
0.1685
0.1842
0.1779
0.1671
0.1743
0.1950
0.2193
0.2413
0.2396
0.2256
0.2183
0.2167
0.2264
0.2709
0.0738
0.0812
0.1013
0.1370
0.1799
0.2147
0.2184
0.2086
0.2040
0.2075
0.2228
0.2420
0.2498
0.2495
0.2440
0.2396
0.2471
0.2978
5C-61
-------
Appendix 5C, Attachment C, Table 2. Smoothed prevalence for children "STILL" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
Prevalence
0.0082
0.0168
0.0289
0.0420
0.0509
0.0573
0.0611
0.0624
0.0629
0.0663
0.0737
0.0889
0.1056
0.1157
0.1191
0.1177
0.1107
0.0999
0.0381
0.0620
0.0875
0.1079
0.1187
0.1117
0.0940
0.0974
0.1144
0.1237
0.1196
0.1074
0.1025
0.1096
0.1236
0.1412
0.1633
0.1906
0.0122
0.0268
0.0480
0.0710
0.0842
0.0934
0.1056
0.1117
0.1111
0.1138
0.1126
0.1108
0.1129
0.1139
0.1128
0.1054
0.0935
0.0782
0.0402
0.0824
0.1338
0.1774
0.1949
0.1867
0.1807
0.1734
0.1739
0.1814
SE
0.0051
0.0064
0.0070
0.0086
0.0103
0.0108
0.0109
0.0107
0.0100
0.0096
0.0108
0.0126
0.0151
0.0163
0.0160
0.0144
0.0143
0.0205
0.0164
0.0160
0.0160
0.0183
0.0202
0.0194
0.0188
0.0187
0.0205
0.0220
0.0237
0.0225
0.0199
0.0211
0.0229
0.0266
0.0413
0.0779
0.0064
0.0083
0.0091
0.0113
0.0134
0.0138
0.0144
0.0149
0.0152
0.0155
0.0153
0.0146
0.0137
0.0132
0.0127
0.0118
0.0133
0.0184
0.0177
0.0213
0.0225
0.0255
0.0267
0.0237
0.0222
0.0221
0.0248
0.0269
LowerCI
0.0021
0.0073
0.0169
0.0267
0.0326
0.0378
0.0412
0.0427
0.0443
0.0481
0.0533
0.0649
0.0768
0.0845
0.0882
0.0896
0.0831
0.0632
0.0146
0.0349
0.0581
0.0738
0.0811
0.0758
0.0602
0.0634
0.0765
0.0830
0.0766
0.0672
0.0664
0.0712
0.0815
0.0924
0.0914
0.0722
0.0038
0.0135
0.0315
0.0500
0.0591
0.0673
0.0779
0.0829
0.0820
0.0840
0.0831
0.0826
0.0861
0.0880
0.0878
0.0822
0.0682
0.0462
0.0151
0.0463
0.0917
0.1282
0.1429
0.1402
0.1371
0.1301
0.1260
0.1297
UpperCI
0.0319
0.0382
0.0490
0.0655
0.0788
0.0859
0.0897
0.0902
0.0886
0.0907
0.1012
0.1206
0.1435
0.1565
0.1588
0.1530
0.1461
0.1544
0.0956
0.1076
0.1295
0.1550
0.1704
0.1616
0.1439
0.1469
0.1676
0.1805
0.1821
0.1673
0.1551
0.1649
0.1830
0.2099
0.2746
0.4158
0.0384
0.0525
0.0725
0.1001
0.1187
0.1282
0.1416
0.1489
0.1489
0.1525
0.1507
0.1472
0.1466
0.1462
0.1438
0.1343
0.1269
0.1292
0.1028
0.1425
0.1911
0.2401
0.2601
0.2443
0.2344
0.2273
0.2350
0.2478
5C-62
-------
Appendix 5C, Attachment C, Table 2. Smoothed prevalence for children "STILL" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Age
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
Prevalence
0.1813
0.1749
0.1702
0.1499
0.1366
0.1484
0.1846
0.2590
0.0153
0.0281
0.0437
0.0584
0.0657
0.0668
0.0678
0.0696
0.0737
0.0840
0.0807
0.0710
0.0629
0.0680
0.0786
0.0913
0.1095
0.1328
0.0234
0.0564
0.1040
0.1466
0.1618
0.1441
0.1124
0.0751
0.0633
0.0838
0.1288
0.1778
0.2073
0.2063
0.1929
0.1703
0.1414
0.1108
0.0225
0.0368
0.0562
0.0797
0.1035
0.1289
0.1472
0.1423
0.1290
0.1251
0.1288
0.1262
0.1246
0.1230
0.1207
0.1114
0.0983
0.0823
0.0930
0.1202
SE
0.0282
0.0282
0.0298
0.0296
0.0269
0.0268
0.0359
0.0740
0.0089
0.0096
0.0090
0.0098
0.0112
0.0111
0.0111
0.0114
0.0124
0.0147
0.0144
0.0134
0.0116
0.0113
0.0117
0.0120
0.0165
0.0330
0.0142
0.0190
0.0219
0.0272
0.0304
0.0280
0.0238
0.0174
0.0157
0.0188
0.0270
0.0336
0.0349
0.0328
0.0287
0.0235
0.0234
0.0327
0.0108
0.0105
0.0104
0.0127
0.0162
0.0187
0.0190
0.0181
0.0163
0.0159
0.0155
0.0139
0.0139
0.0149
0.0144
0.0126
0.0124
0.0171
0.0402
0.0280
LowerCI
0.1275
0.1214
0.1143
0.0959
0.0876
0.0987
0.1185
0.1306
0.0042
0.0132
0.0276
0.0402
0.0449
0.0461
0.0471
0.0482
0.0506
0.0569
0.0541
0.0466
0.0416
0.0469
0.0564
0.0681
0.0781
0.0753
0.0061
0.0266
0.0648
0.0964
0.1056
0.0928
0.0698
0.0447
0.0364
0.0507
0.0802
0.1154
0.1410
0.1435
0.1375
0.1248
0.0974
0.0567
0.0078
0.0195
0.0373
0.0559
0.0730
0.0931
0.1102
0.1070
0.0973
0.0943
0.0985
0.0989
0.0971
0.0939
0.0925
0.0868
0.0743
0.0518
0.0347
0.0710
UpperCI
0.2514
0.2454
0.2457
0.2268
0.2066
0.2169
0.2761
0.4484
0.0537
0.0589
0.0683
0.0840
0.0950
0.0958
0.0967
0.0993
0.1062
0.1224
0.1187
0.1068
0.0938
0.0976
0.1085
0.1214
0.1513
0.2234
0.0856
0.1157
0.1627
0.2167
0.2400
0.2168
0.1761
0.1234
0.1078
0.1355
0.2004
0.2638
0.2941
0.2873
0.2637
0.2281
0.2009
0.2051
0.0633
0.0682
0.0838
0.1123
0.1449
0.1757
0.1938
0.1868
0.1690
0.1641
0.1668
0.1598
0.1584
0.1594
0.1560
0.1420
0.1291
0.1285
0.2262
0.1964
5C-63
-------
Appendix 5C, Attachment C, Table 2. Smoothed prevalence for children "STILL" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Age
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
Prevalence
0.1475
0.1714
0.1860
0.2060
0.2256
0.2496
0.2727
0.2579
0.2318
0.1902
0.1624
0.1641
0.1699
0.1797
0.1933
0.2097
0.0131
0.0228
0.0352
0.0495
0.0633
0.0740
0.0826
0.0888
0.0860
0.0791
0.0747
0.0736
0.0776
0.0851
0.0871
0.0876
0.0859
0.0819
0.0396
0.0573
0.0772
0.0963
0.1120
0.1206
0.1219
0.1152
0.1131
0.1190
0.1208
0.1195
0.1275
0.1405
0.1394
0.1296
0.1136
0.0923
0.0228
0.0476
0.0793
0.1076
0.1193
0.1194
0.1145
0.1071
0.1011
0.1000
0.1059
0.1122
SE
0.0256
0.0311
0.0335
0.0276
0.0276
0.0317
0.0387
0.0395
0.0366
0.0310
0.0268
0.0254
0.0251
0.0244
0.0276
0.0451
0.0059
0.0063
0.0064
0.0074
0.0089
0.0092
0.0096
0.0099
0.0100
0.0095
0.0088
0.0085
0.0087
0.0093
0.0093
0.0087
0.0091
0.0136
0.0135
0.0113
0.0109
0.0136
0.0165
0.0174
0.0173
0.0162
0.0157
0.0161
0.0175
0.0178
0.0192
0.0197
0.0184
0.0166
0.0184
0.0249
0.0070
0.0082
0.0089
0.0109
0.0123
0.0117
0.0111
0.0105
0.0099
0.0098
0.0102
0.0106
LowerCI
0.0997
0.1134
0.1232
0.1519
0.1708
0.1866
0.1964
0.1810
0.1611
0.1311
0.1116
0.1155
0.1216
0.1321
0.1397
0.1274
0.0048
0.0124
0.0236
0.0355
0.0464
0.0561
0.0638
0.0695
0.0666
0.0606
0.0576
0.0570
0.0606
0.0669
0.0688
0.0702
0.0681
0.0567
0.0186
0.0371
0.0564
0.0704
0.0805
0.0874
0.0888
0.0842
0.0829
0.0880
0.0874
0.0857
0.0910
0.1026
0.1037
0.0973
0.0791
0.0503
0.0116
0.0325
0.0619
0.0859
0.0949
0.0960
0.0924
0.0861
0.0813
0.0806
0.0855
0.0910
UpperCI
0.2130
0.2508
0.2708
0.2732
0.2919
0.3255
0.3653
0.3535
0.3216
0.2678
0.2302
0.2278
0.2323
0.2396
0.2612
0.3253
0.0349
0.0415
0.0522
0.0685
0.0857
0.0969
0.1063
0.1129
0.1105
0.1025
0.0963
0.0944
0.0989
0.1078
0.1099
0.1087
0.1080
0.1169
0.0823
0.0876
0.1048
0.1306
0.1536
0.1641
0.1652
0.1556
0.1524
0.1591
0.1646
0.1642
0.1757
0.1893
0.1848
0.1706
0.1605
0.1634
0.0443
0.0693
0.1011
0.1341
0.1490
0.1475
0.1411
0.1323
0.1251
0.1236
0.1305
0.1376
5C-64
-------
Appendix 5C, Attachment C, Table 2. Smoothed prevalence for children "STILL" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Age
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
Prevalence
0.1103
0.1052
0.0983
0.0899
0.0811
0.0727
0.0499
0.0749
0.1033
0.1305
0.1519
0.1595
0.1598
0.1540
0.1466
0.1457
0.1504
0.1508
0.1506
0.1470
0.1345
0.1215
0.1080
0.0948
0.0077
0.0122
0.0181
0.0248
0.0305
0.0382
0.0482
0.0573
0.0628
0.0697
0.0768
0.0786
0.0808
0.0829
0.0845
0.0908
0.1016
0.1180
0.0244
0.0270
0.0306
0.0354
0.0407
0.0577
0.0807
0.0954
0.0876
0.0648
0.0495
0.0473
0.0606
0.0845
0.0931
0.0846
0.0629
0.0376
0.0007
0.0052
0.0225
0.0596
SE
0.0105
0.0105
0.0094
0.0081
0.0089
0.0136
0.0126
0.0110
0.0116
0.0149
0.0177
0.0180
0.0185
0.0180
0.0170
0.0170
0.0171
0.0171
0.0184
0.0192
0.0179
0.0159
0.0164
0.0227
0.0049
0.0046
0.0045
0.0055
0.0068
0.0077
0.0091
0.0098
0.0106
0.0106
0.0099
0.0094
0.0100
0.0108
0.0111
0.0110
0.0129
0.0236
0.0144
0.0091
0.0074
0.0090
0.0112
0.0146
0.0185
0.0181
0.0159
0.0127
0.0107
0.0110
0.0137
0.0179
0.0180
0.0154
0.0143
0.0146
0.0007
0.0027
0.0063
0.0095
LowerCI
0.0893
0.0843
0.0795
0.0737
0.0636
0.0479
0.0285
0.0542
0.0805
0.1012
0.1171
0.1240
0.1234
0.1186
0.1130
0.1122
0.1167
0.1171
0.1146
0.1097
0.0999
0.0907
0.0770
0.0555
0.0019
0.0053
0.0105
0.0153
0.0186
0.0245
0.0318
0.0393
0.0432
0.0497
0.0577
0.0603
0.0615
0.0621
0.0632
0.0694
0.0766
0.0753
0.0066
0.0128
0.0179
0.0201
0.0221
0.0328
0.0483
0.0624
0.0583
0.0419
0.0306
0.0282
0.0366
0.0526
0.0603
0.0562
0.0379
0.0158
0.0001
0.0014
0.0112
0.0398
UpperCI
0.1356
0.1305
0.1210
0.1093
0.1028
0.1089
0.0860
0.1027
0.1316
0.1666
0.1948
0.2029
0.2045
0.1977
0.1879
0.1870
0.1917
0.1921
0.1955
0.1943
0.1788
0.1607
0.1494
0.1573
0.0306
0.0278
0.0310
0.0401
0.0494
0.0590
0.0724
0.0829
0.0904
0.0970
0.1016
0.1018
0.1056
0.1100
0.1121
0.1179
0.1337
0.1803
0.0862
0.0561
0.0518
0.0615
0.0738
0.0996
0.1319
0.1434
0.1296
0.0989
0.0792
0.0781
0.0988
0.1329
0.1411
0.1253
0.1026
0.0868
0.0067
0.0192
0.0447
0.0884
5C-65
-------
Appendix 5C, Attachment C, Table 2. Smoothed prevalence for children "STILL" having asthma.
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age
4
5
6
7
8
9
10
11
12
13
14
15
16
17
0
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
Prevalence
0.0989
0.1070
0.0959
0.0830
0.0877
0.1029
0.1189
0.1292
0.1214
0.1050
0.0981
0.0997
0.1091
0.1290
0.0263
0.0374
0.0518
0.0681
0.0871
0.1074
0.1167
0.1138
0.1073
0.0964
0.0830
0.0745
0.0825
0.1000
0.1074
0.1120
0.1127
0.1084
SE
0.0140
0.0147
0.0141
0.0126
0.0124
0.0135
0.0140
0.0153
0.0154
0.0139
0.0127
0.0116
0.0128
0.0231
0.0130
0.0101
0.0086
0.0105
0.0143
0.0173
0.0183
0.0186
0.0177
0.0164
0.0149
0.0151
0.0165
0.0197
0.0200
0.0193
0.0222
0.0340
LowerCI
0.0691
0.0754
0.0660
0.0565
0.0613
0.0737
0.0883
0.0955
0.0879
0.0749
0.0707
0.0742
0.0810
0.0814
0.0088
0.0204
0.0358
0.0483
0.0604
0.0749
0.0820
0.0789
0.0741
0.0659
0.0557
0.0474
0.0527
0.0643
0.0707
0.0760
0.0724
0.0531
UpperCI
0.1397
0.1496
0.1372
0.1203
0.1239
0.1419
0.1584
0.1724
0.1653
0.1452
0.1346
0.1327
0.1454
0.1984
0.0761
0.0673
0.0742
0.0952
0.1240
0.1517
0.1635
0.1615
0.1529
0.1389
0.1221
0.1152
0.1268
0.1524
0.1600
0.1620
0.1714
0.2088
5C-66
-------
Appendix 5C, Attachment C, Table 3. Smoothed prevalence for adults "EVER" having asthma
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Age group
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
Prevalence
0.1642
0.1341
0.1193
0.1204
0.1246
0.1165
0.0980
0.2014
0.1812
0.1782
0.2104
0.2295
0.1892
0.1176
0.1705
0.1209
0.0886
0.0727
0.0770
0.0828
0.0847
0.1654
0.1143
0.1066
0.1376
0.1643
0.1396
0.0853
0.1791
0.1423
0.1256
0.1246
0.1281
0.1151
0.0879
0.1646
0.1705
0.1842
0.2084
0.2180
0.1695
0.0960
0.1728
0.1163
0.0932
0.0901
0.0963
0.0874
0.0708
0.1734
0.1323
0.1182
0.1254
0.1361
0.1305
0.0988
0.1533
0.1235
0.1114
0.1149
0.1261
0.1188
0.0959
0.1491
SE
0.0141
0.0063
0.0058
0.0057
0.0066
0.0062
0.0089
0.0153
0.0114
0.0130
0.0146
0.0164
0.0145
0.0173
0.0149
0.0063
0.0053
0.0046
0.0054
0.0058
0.0106
0.0175
0.0109
0.0122
0.0146
0.0164
0.0160
0.0205
0.0176
0.0076
0.0072
0.0071
0.0076
0.0070
0.0098
0.0182
0.0110
0.0126
0.0143
0.0156
0.0118
0.0125
0.0210
0.0081
0.0070
0.0063
0.0072
0.0073
0.0118
0.0193
0.0138
0.0135
0.0144
0.0198
0.0195
0.0255
0.0114
0.0054
0.0050
0.0047
0.0058
0.0058
0.0087
0.0122
LowerCI
0.1219
0.1142
0.1012
0.1025
0.1040
0.0971
0.0719
0.1531
0.1445
0.1370
0.1638
0.1770
0.1435
0.0690
0.1249
0.1008
0.0719
0.0583
0.0602
0.0647
0.0545
0.1122
0.0808
0.0703
0.0936
0.1141
0.0918
0.0353
0.1265
0.1183
0.1029
0.1024
0.1043
0.0934
0.0598
0.1104
0.1356
0.1442
0.1629
0.1684
0.1321
0.0603
0.1126
0.0914
0.0721
0.0710
0.0744
0.0656
0.0398
0.1138
0.0896
0.0772
0.0816
0.0786
0.0743
0.0373
0.1185
0.1065
0.0956
0.0998
0.1077
0.1004
0.0701
0.1107
UpperCI
0.2176
0.1568
0.1402
0.1409
0.1486
0.1392
0.1322
0.2603
0.2248
0.2284
0.2662
0.2920
0.2453
0.1933
0.2284
0.1444
0.1087
0.0904
0.0980
0.1053
0.1292
0.2370
0.1593
0.1585
0.1979
0.2309
0.2068
0.1920
0.2474
0.1701
0.1525
0.1509
0.1565
0.1412
0.1273
0.2383
0.2123
0.2323
0.2627
0.2773
0.2149
0.1495
0.2560
0.1469
0.1197
0.1139
0.1237
0.1155
0.1229
0.2552
0.1911
0.1768
0.1879
0.2253
0.2191
0.2366
0.1959
0.1429
0.1295
0.1320
0.1472
0.1400
0.1297
0.1978
5C-67
-------
Appendix 5C, Attachment C, Table 3. Smoothed prevalence for adults "EVER" having asthma
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age group
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
18-24
25-34
35-44
45-54
55-64
65-74
75+
Prevalence
0.1365
0.1414
0.1686
0.1881
0.1651
0.1125
0.1445
0.1086
0.0860
0.0742
0.0733
0.0790
0.0900
0.1433
0.1031
0.0934
0.1055
0.1072
0.0942
0.0712
0.1571
0.1415
0.1373
0.1423
0.1497
0.1445
0.1266
0.1434
0.1318
0.1440
0.1806
0.1713
0.1511
0.1292
0.1566
0.1233
0.1025
0.0908
0.0955
0.1067
0.1265
0.1521
0.0942
0.0885
0.1133
0.1237
0.1134
0.0961
SE
0.0066
0.0078
0.0097
0.0115
0.0101
0.0124
0.0095
0.0050
0.0044
0.0040
0.0045
0.0048
0.0102
0.0144
0.0087
0.0090
0.0101
0.0108
0.0092
0.0123
0.0135
0.0067
0.0070
0.0067
0.0071
0.0070
0.0112
0.0164
0.0092
0.0117
0.0144
0.0136
0.0117
0.0177
0.0173
0.0069
0.0060
0.0054
0.0059
0.0068
0.0152
0.0204
0.0095
0.0102
0.0130
0.0156
0.0142
0.0190
LowerCI
0.1149
0.1159
0.1369
0.1505
0.1325
0.0755
0.1147
0.0926
0.0720
0.0616
0.0594
0.0639
0.0606
0.1000
0.0766
0.0664
0.0751
0.0750
0.0666
0.0385
0.1163
0.1201
0.1150
0.1207
0.1268
0.1220
0.0929
0.0945
0.1026
0.1074
0.1350
0.1284
0.1141
0.0785
0.1067
0.1019
0.0839
0.0741
0.0774
0.0860
0.0834
0.0938
0.0660
0.0590
0.0753
0.0789
0.0726
0.0474
UpperCI
0.1614
0.1714
0.2059
0.2324
0.2039
0.1644
0.1805
0.1269
0.1025
0.0891
0.0902
0.0974
0.1316
0.2013
0.1376
0.1300
0.1462
0.1510
0.1314
0.1279
0.2089
0.1660
0.1631
0.1670
0.1758
0.1704
0.1702
0.2117
0.1678
0.1903
0.2374
0.2248
0.1974
0.2054
0.2240
0.1485
0.1247
0.1107
0.1174
0.1318
0.1871
0.2373
0.1327
0.1308
0.1670
0.1888
0.1727
0.1849
5C-68
-------
Appendix 5C, Attachment C, Table 4. Smoothed prevalence for adults "STILL" having asthma
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Midwest
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
Northeast
South
South
South
South
South
South
South
South
Gender
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Poverty Status
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Age group
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
Prevalence
0.1046
0.0888
0.0835
0.0893
0.0909
0.0811
0.0630
0.1327
0.1280
0.1315
0.1600
0.1777
0.1488
0.0940
0.0807
0.0584
0.0479
0.0472
0.0522
0.0528
0.0481
0.0912
0.0683
0.0694
0.1015
0.1338
0.1202
0.0709
0.1098
0.0965
0.0899
0.0901
0.0917
0.0862
0.0726
0.1212
0.1199
0.1338
0.1655
0.1824
0.1273
0.0529
0.0922
0.0600
0.0488
0.0483
0.0563
0.0576
0.0554
0.0791
0.0800
0.0805
0.0857
0.1064
0.1040
0.0771
0.0891
0.0735
0.0684
0.0732
0.0846
0.0817
0.0641
0.0948
SE
0.0121
0.0057
0.0052
0.0050
0.0057
0.0051
0.0067
0.0139
0.0095
0.0114
0.0134
0.0146
0.0128
0.0157
0.0115
0.0045
0.0040
0.0038
0.0042
0.0045
0.0081
0.0136
0.0091
0.0109
0.0141
0.0165
0.0161
0.0210
0.0134
0.0065
0.0063
0.0060
0.0062
0.0059
0.0093
0.0166
0.0093
0.0106
0.0127
0.0143
0.0098
0.0086
0.0154
0.0058
0.0050
0.0051
0.0065
0.0063
0.0106
0.0128
0.0119
0.0135
0.0162
0.0224
0.0200
0.0236
0.0083
0.0039
0.0036
0.0037
0.0046
0.0047
0.0070
0.0105
LowerCI
0.0703
0.0714
0.0675
0.0738
0.0736
0.0654
0.0438
0.0907
0.0980
0.0961
0.1181
0.1318
0.1091
0.0513
0.0491
0.0448
0.0359
0.0358
0.0395
0.0393
0.0268
0.0542
0.0430
0.0402
0.0624
0.0866
0.0751
0.0250
0.0721
0.0765
0.0708
0.0718
0.0727
0.0681
0.0467
0.0744
0.0914
0.1013
0.1260
0.1381
0.0972
0.0300
0.0509
0.0428
0.0340
0.0334
0.0376
0.0393
0.0281
0.0430
0.0459
0.0427
0.0419
0.0475
0.0501
0.0241
0.0649
0.0615
0.0571
0.0617
0.0705
0.0674
0.0443
0.0641
UpperCI
0.1528
0.1100
0.1030
0.1077
0.1118
0.1002
0.0898
0.1899
0.1656
0.1772
0.2132
0.2352
0.1998
0.1659
0.1299
0.0758
0.0637
0.0620
0.0687
0.0706
0.0847
0.1496
0.1067
0.1173
0.1610
0.2010
0.1869
0.1850
0.1638
0.1210
0.1136
0.1124
0.1151
0.1085
0.1110
0.1915
0.1559
0.1747
0.2143
0.2370
0.1650
0.0917
0.1616
0.0836
0.0696
0.0693
0.0834
0.0837
0.1062
0.1409
0.1360
0.1465
0.1672
0.2211
0.2035
0.2203
0.1212
0.0876
0.0817
0.0866
0.1012
0.0987
0.0920
0.1380
5C-69
-------
Appendix 5C, Attachment C, Table 4. Smoothed prevalence for adults "STILL" having asthma
Smoothed
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Yes
Region
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
South
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
West
Gender
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Female
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Male
Poverty Status
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Above Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Below Poverty Level
Age group
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
18-24
25-34
35-44
45-44
55-64
65-74
75+
Prevalence
0.0942
0.1086
0.1446
0.1618
0.1379
0.0881
0.0600
0.0490
0.0421
0.0386
0.0384
0.0457
0.0627
0.0583
0.0443
0.0492
0.0720
0.0771
0.0608
0.0353
0.0842
0.0876
0.0931
0.0981
0.1028
0.0984
0.0825
0.0863
0.0934
0.1091
0.1332
0.1292
0.1169
0.1021
0.0597
0.0569
0.0549
0.0525
0.0562
0.0660
0.0783
0.0720
0.0484
0.0539
0.0784
0.0936
0.0758
0.0489
SE
0.0059
0.0073
0.0095
0.0112
0.0095
0.0109
0.0073
0.0035
0.0033
0.0031
0.0034
0.0038
0.0089
0.0080
0.0053
0.0067
0.0090
0.0096
0.0075
0.0082
0.0115
0.0054
0.0062
0.0065
0.0067
0.0061
0.0090
0.0121
0.0078
0.0100
0.0120
0.0120
0.0104
0.0148
0.0092
0.0046
0.0045
0.0046
0.0053
0.0058
0.0131
0.0125
0.0068
0.0084
0.0115
0.0155
0.0129
0.0136
LowerCI
0.0758
0.0859
0.1149
0.1267
0.1082
0.0570
0.0392
0.0381
0.0322
0.0292
0.0282
0.0343
0.0382
0.0358
0.0290
0.0303
0.0460
0.0492
0.0390
0.0154
0.0522
0.0708
0.0742
0.0781
0.0820
0.0795
0.0565
0.0524
0.0695
0.0789
0.0967
0.0929
0.0854
0.0609
0.0351
0.0432
0.0414
0.0389
0.0407
0.0487
0.0437
0.0389
0.0294
0.0311
0.0465
0.0517
0.0413
0.0182
UpperCI
0.1166
0.1365
0.1806
0.2043
0.1742
0.1337
0.0907
0.0629
0.0550
0.0510
0.0520
0.0607
0.1013
0.0937
0.0672
0.0790
0.1112
0.1188
0.0937
0.0787
0.1328
0.1080
0.1163
0.1226
0.1281
0.1213
0.1189
0.1387
0.1243
0.1489
0.1806
0.1770
0.1580
0.1662
0.0998
0.0745
0.0723
0.0704
0.0770
0.0889
0.1364
0.1295
0.0787
0.0919
0.1293
0.1635
0.1350
0.1250
5C-70
-------
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Above Poverty Level
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
prev
0.6-
345
gender
9 10 11 12 13 14 15 16 17
age
Female
Male
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
345
gender
9 10 11 12 13 14 15 16 17
Female
Male
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
prev
0 1
3 4 5 6 7
' I ' ' ' ' I ' ' ' ' I ' ' ' ' I '
9 10 11 12 13 14 15
' I ' ' ' ' I
16 17
345
9 10 11 12 13 14 15 16 17
gender
Female
Male
gender
Female
Male
Appendix 5C, Attachment C, Figure 1. Smoothed prevalence and confidence intervals for children 'EVER' having asthma.
5C-71
-------
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat=Above Poverty Level
1 I' ' ' ' I' ' ' ' I '' ' ' I '' ' ' I ' '' ' I ' '' ' I ' ' '' I ' ' '' I ' ' ' ' I ' ' ' ' I ' ' ' ' I' ' ' ' I' ' ' ' I '' ' ' I '' ' ' I '
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
Female
Male
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
4567
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
prev
0.6-
prev
0.6
0.3-
Figure 1. Smoothed asthma 'EVER' prevalence rates and confidence intervals
region=West pov_rat= Above Poverty Level
3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 1. Smoothed asthma 'EVER1 prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
E-*
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
age
gender Female Male
Appendix 5C, Attachment C, Figure 1, cont. Smoothed prevalence and confidence intervals for children 'EVER' having
asthma.
5C-72
-------
Figure 2. Smoothed asthma 'STILL1 prevalence rates and confidence intervals
region=Midwest pov_rat=Above Poverty Level
Figure 2. Smoothed asthma 'STILL1 prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
prev
0.6-
0.4
0.3
prev
123456
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Smoothed asthma 'STILL' prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
age
Female
Male
prev
3456
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Smoothed asthma 'STILL' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
gender
age
Female
Male
Appendix 5C, Attachment C, Figure 2. Smoothed prevalence and confidence intervals for children 'STILL' having asthma.
5C-73
-------
Figure 2. Smoothed asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat= Above Poverty Level
prev
0.6-
0123456
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Smoothed asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
prev
34567
gender
Female
Male
Figure 2. Smoothed asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Above Poverty Level
prev
0.6-1
0123456
9 10 11 12 13 14 15 16 17
gender
age
Female
Male
Figure 2. Smoothed asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
prev
0.6-
01 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
gender
age
Female
Appendix 5C, Attachment C, Figure 2, cont. Smoothed prevalence and confidence intervals for children 'STILL' having
asthma.
5C-74
-------
Figure 3. Smoothed adult asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Above Poverty Level
Figure 3. Smoothed adult asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Above Poverty Level
prev
0.4:
25-34
35-44
45-54
55-64
65-74
75+
prev
0.4:
18-24
25-34
35-44
45-54
65-74
75+
gender
age_grp
Fern ale
Male
Figure 3. Smoothed adult asthma 'EVER' prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
gender
age_grp
Female
Male
Figure 3. Smoothed adult asthma 'EVER' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
prev
0.4:
25-34
35-44 45-54
55-64 65-74
75+
prev
0.4:
25-34 35-44
45-54
gender
age_grp
Female
Male
gender
age_grp
Female
Male
Appendix 5C, Attachment C, Figure 3. Smoothed prevalence and confidence intervals for Adults 'EVER' having asthma.
5C-75
-------
Figure 3. Smoothed adult asthma 'EVER1 prevalence rates and confidence intervals
region=South pov_rat=Above Poverty Level
Figure 3. Smoothed adult asthma 'EVER1 prevalence rates and confidence intervals
region=West pov_rat=Above Poverty Level
prev
0.4-
35-44
gender
45-54
age_grp
Female
55-64
65-74
75+
Male
Figure 3. Smoothed adult asthma 'EVER' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
35-44
gender
45-54
age_grp
Female
55-64
65-74
75+
prev
0.4-
18-24 25-34 35-44 45-54 55-64 65-74
gender
age_grp
Female
Male
Figure 3. Smoothed adult asthma 'EVER' prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
prev
0.4:
0.3:
0.2:
18-24
35-44
gender
45-54
age_grp
Female
55-64
65-74
Male
Appendix 5C, Attachment C, Figure 3, cont. Smoothed prevalence and confidence intervals for Adults 'EVER' having
asthma.
75+
5C-76
-------
Figure 4. Smoothed adult asthma 'STILL' prevalence rates and confidence intervals
region=Midwest pov_rat= Above Poverty Level
Figure 4. Smoothed adult asthma 'STILL' prevalence rates and confidence intervals
region=Northeast pov_rat= Above Poverty Level
prev
0.4:
0.3-
18-24
18-24
25-34
35-44
45-54
55-64
65-74
age_grp
Fern ale
Male
Figure 4. Smoothed adult asthma 'STILL' prevalence rates and confidence intervals
region=Midwest pov_rat=Below Poverty Level
gender
age_grp
Female
Male
Figure 4. Smoothed adult asthma 'STILL' prevalence rates and confidence intervals
region=Northeast pov_rat=Below Poverty Level
prev
0.4:
18-24
25-34
35-44
45-54
55-64
65-74
45-54
gender
age_grp
Female Male
gender
age_grp
Female
Male
Appendix 5C, Attachment C, Figure 4. Smoothed prevalence and confidence intervals for Adults 'STILL' having asthma.
5C-77
-------
Figure 4. Smoothed adult asthma 'STILL1 prevalence rates and confidence intervals
region=South pov_rat= Above Poverty Level
Figure 4. Smoothed adult asthma 'STILL' prevalence rates and confidence intervals
region=West pov_rat=Above Poverty Level
prev
0.4:
0.2-
i
4
gender
45-54
age_grp
Female
Male
Figure 4. Smoothed adult asthma 'STILL' prevalence rates and confidence intervals
region=South pov_rat=Below Poverty Level
18-24
35-44
gender
45-54
age_grp
Female
55-64
65-74
Male
prev
0.4-
I
1
gender
45-54
age_grp
Female
Male
Figure 4. Smoothed adult asthma 'STILL1 prevalence rates and confidence intervals
region=West pov_rat=Below Poverty Level
gender
45-54
age_grp
Female
Male
Appendix 5C, Attachment C, Figure 4, cont. Smoothed prevalence and confidence intervals for Adults 'STILL' having
asthma.
5C-78
-------
1 APPENDIX 5D: VARIABILITY ANALYSIS AND UNCERTAINTY
2 CHARACTERIZATION
3 5D-1. OVERVIEW
4 An important issue associated with any population exposure or risk assessment is the
5 characterization of variability and uncertainty. Variability refers to the inherent heterogeneity in
6 a population or variable of interest (e.g., residential air exchange rates). The degree of variability
7 cannot be reduced through further research, only better characterized with additional
8 measurement. Uncertainty refers to the lack of knowledge regarding the values of model input
9 variables (i.e., parameter uncertainty), the physical systems or relationships used (i.e., use of
10 input variables to estimate exposure or risk or model uncertainty), and in specifying the scenario
11 that is consistent with purpose of the assessment (i.e., scenario uncertainty). Uncertainty is,
12 ideally, reduced to the maximum extent possible through improved measurement of key
13 parameters and iterative model refinement. The approaches used to assess variability and to
14 characterize uncertainty in this REA are discussed in the following two sections. The primary
15 purpose of this characterization is to provide a summary of variability and uncertainty
16 evaluations conducted to date regarding our Os exposure assessments and APEX exposure
17 modeling and to identify the most important elements of uncertainty in need of further
18 characterization. Each section contains a concise tabular summary of the identified components
19 and how, for elements of uncertainty, each source may affect the estimated exposures.
20 5D-2. TREATMENT OF VARIABILITY AND CO-VARIABILITY
21 The purpose for addressing variability in this REA is to ensure that the estimates of
22 exposure and risk reflect the variability of ambient 63 concentrations, population characteristics,
23 associated O3 exposure and intake dose, and potential health risk across the study area and for
24 the simulated at-risk populations. In this REA, there are several algorithms that account for
25 variability of input data when generating the number of estimated benchmark exceedances or
26 health risk outputs. For example, variability may arise from differences in the population
27 residing within census tracts (e.g., age distribution) and the activities that may affect population
28 exposure to 63 and the resulting intake dose estimate (e.g., time spent outdoors, performing
29 moderate or greater exertion level activities outdoors). A complete range of potential exposure
30 levels and associated risk estimates can be generated when appropriately addressing variability in
5D-1
-------
31 exposure and risk assessments; note however that the range of values obtained would be within
32 the constraints of the input parameters, algorithms, or modeling system used, not necessarily the
33 complete range of the true exposure or risk values.
34 Where possible, staff identified and incorporated the observed variability in input data
35 sets rather than employing standard default assumptions and/or using point estimates to describe
36 model inputs. The details regarding variability distributions used in data inputs are described in
37 Appendix 5B, while details regarding the variability addressed within its algorithms and
38 processes are found in the APEX TSD (US EPA, 2012).
39 Briefly, APEX has been designed to account for variability in most of the input data,
40 including the physiological variables that are important inputs to determining exertion levels and
41 associated ventilation rates. APEX simulates individuals and then calculates O3 exposures for
42 each of these simulated individuals. The individuals are selected to represent a random sample
43 from a defined population. The collection of individuals represents the variability of the target
44 population, and accounts for several types of variability, including demographic, physiological,
45 and human behavior. In this assessment, we simulated 200,000 individuals to reasonably capture
46 the variability expected in the population exposure distribution for each study area. APEX
47 incorporates stochastic processes representing the natural variability of personal profile
48 characteristics, activity patterns, and microenvironment parameters. In this way, APEX is able
49 to represent much of the variability in the exposure estimates resulting from the variability of the
50 factors effecting human exposure.
51 We note also that correlations and non-linear relationships between variables input to the
52 model can result in the model producing incorrect results if the inherent relationships between
53 these variables are not preserved. That is why APEX is also designed to account for co-
54 variability, or linear and nonlinear correlation among the model inputs, provided that enough is
55 known about these relationships to specify them. This is accomplished by providing inputs that
56 enable the correlation to be modeled explicitly within APEX. For example, there is a non-linear
57 relationship between the outdoor temperature and air exchange rate in homes. One factor that
58 contributes to this non-linear relationship is that windows tend to be closed more often when
59 temperatures are at either low or high extremes than when temperatures are moderate. This
60 relationship is explicitly modeled in APEX by specifying different probability distributions of air
5D-2
-------
61 exchange rates for different ambient temperatures. In any event, APEX models variability and
62 co-variability in two ways:
63 • Stochastically. The user provides APEX with probability distributions
64 characterizing the variability of many input parameters. These are treated
65 stochastically in the model and the estimated exposure distributions reflect this
66 variability. For example, the rate of Os removal in houses can depend on a
67 number of factors which we are not able to explicitly model at this time, due to a
68 lack of data. However, we can specify a distribution of removal rates which
69 reflects observed variations in Os decay. APEX randomly samples from this
70 distribution to obtain values which are used in the mass balance model. Further,
71 co-variability can be modeled stochastically through the use of conditional
72 distributions. If two or more parameters are related, conditional distributions that
73 depend on the values of the related parameters are input to APEX. For example,
74 the distribution of air exchange rates (AERs) in a house depends on the outdoor
75 temperature and whether or not air conditioning (A/C) is in use. In this case, a set
76 of AER distributions is provided to APEX for different ranges of temperatures
77 and A/C use, and the selection of the distribution in APEX is driven by the
78 temperature and A/C status at that time. The spatial variability of A/C prevalence
79 is modeled by supplying APEX with A/C prevalence for each Census tract in the
80 modeled area.
81 • Explicitly. For some variables used in modeling exposure, APEX models
82 variability and co-variability explicitly and not stochastically. For example,
83 hourly-average ambient O3 concentrations and temperatures are used in model
84 calculations. These are input to the model for every hour in the time period
85 modeled at different spatial locations, and in this way the variability and co-
86 variability of hourly concentrations and temperatures are modeled explicitly.
87 Important sources of the variability and co-variability accounted for by APEX and used
88 for this exposure analysis are summarized in Tables 5D-1 and 5D-2 below, respectively.
89
90
5D-3
-------
91 Table 5D-1. Components of exposure variability modeled by APEX.
Component
Variability Source
Comment
Population data
Individuals are randomly sampled from US census tracts
used in each model study area, stratified by age (single
years), gender, and employment status probability
distributions (US Census Bureau, 2007a).
Commuting data
Employed individuals are probabilistically assigned ambient
concentrations originating from either their home or work tract
based on US Census derived commuter data (US Census
Bureau, 2007a).
Simulated
Individuals
Activity patterns
Data diaries are randomly selected from CHAD master
(>38,000 diaries) using six diary pools stratified by two day-
types (weekday, weekend) and three temperature ranges (<
55.0 >, between 55.0 and 83.9°F, and >84.0 >). The CHAD
diaries capture real locations that people visit and the
activities they perform, ranging from 1 minute to 1 hour in
duration (US EPA, 2002).
Longitudinal profiles
A sequence of diaries is linked together for each individual
that preserves both the inter- and intra-personal variability in
human activities (Glen et al., 2008).
Asthma prevalence
Asthma prevalence is stratified by two genders, single age
years (0-17), seven age groups, (18-24, 25-34, 35-44, 45-54,
55-64, 65-74, and, >75), four regions (Midwest, Northeast,
South, and West), and US census tract level poverty ratios
(CDC, 2011; US Census Bureau, 2007b).
Measured ambient O3
concentrations
Ambient Input
Temporal: 1-hour concentrations for an entire O3 season or
year predicted using ambient monitoring data.
Spatial: Several monitors are used to represent ambient
conditions within each study area; each monitor was assigned
a 30 km zone of influence, though value from closest monitor
is used for each tract. Four US study areas assess regional
differences in ambient conditions.
Meteorological data
Spatial: Values from closest available local surface National
Weather Service (NWS) station were used.
Temporal: 1-hour temperature data input for each year; daily
values calculated by APEX.
Microenvironmental
Approach
Microenvironments:
General
Twenty-eight total microenvironments are represented,
including those expected to be associated with high exposure
concentrations (i.e., outdoors and outdoor near-road). Where
this type of variability is incorporated within particular
microenvironmental algorithm inputs, this results in
differential exposure estimates for each individual (and event)
as persons spend varying time frequency within each
microenvironment and ambient concentrations vary spatially
within and between study areas.
Microenvironments:
Spatial Variability
Ambient concentrations used in microenvironmental
algorithms vary spatially within (where more than one site
available) and among study areas. Concentrations near
roadways are adjusted to account fortitration by NO.
5D-4
-------
Component
Physiological
Factors and
Algorithms
Variability Source
Microenvironments:
Temporal Variability
Air exchange rates
Proximity factors for
on- and near roads
Resting metabolic
rate (RMR)
Maximum normalized
oxygen consumption
rate (NVO2)
Maximum oxygen
debt (MOXD)
Recovery time
METS by activity
Oxygen uptake per
unit of energy
expended (UCF)
Body mass
Height
Body surface area
Comment
All exposure calculations are performed at the event-level
when using either factors or mass balance approach
(durations can be as short as one minute). In addition, for the
indoor microenvironments, using a mass balance model
accounts for O3 concentrations occurring during a previous
hour (and of ambient origin) to calculate a current event's
indoor O3 concentrations.
Several lognormal distributions are sampled based on five
daily mean temperature ranges, study area, and study-area
specific A/C prevalence rates.
Three distributions are used, stratified by road-type (urban,
interstate, and rural), selected based on VMTto address
expected ozone titration by NO near roads.
Regression equations for three age-group (18-29, 30-59, and
60+) and two genders were used with body mass as the
independent variable (see Johnson et al. (2000) and section
5.3 of APEX TSD).
Single year age- and gender-specific normal distributions are
randomly sampled for each person (Isaacs and Smith, 2005
and section 7.2 of APEX TSD). This variable is used to
calculate maximum metabolic equivalents (METS).
Normal distributions for maximum obtainable oxygen,
stratified by 3 age groups (ages 0-11, 12-18, 19-100) and two
genders (Isaacs and Smith, 2007 and section 7.2 of APEX
TSD). Used when adjusting METS to address fatigue and
EPOC.
One uniform distribution randomly sampled to estimate the
time required to recover a maximum oxygen deficit (Isaacs
and Smith, 2007 and section 7.2 of APEX TSD).
Values randomly sampled from distributions developed for
specific activities (a few are age-group specific) (McCurdy,
2000; US EPA, 2002).
Values randomly sampled from a uniform distribution to
convert energy expenditure to oxygen consumption (Johnson
et al., 2000 and section 5.3 of APEX TSD).
Randomly selected from population-weighted lognormal
distributions with age- and gender-specific geometric mean
(GM) and geometric standard deviation (GSD) derived from
the National Health and Nutrition Examination Survey
(NHANES) for the years 1999-2004 (Isaacs and Smith (2005)
and section 5.3 of APEX TSD).
Values randomly sampled from distributions used are based
on equations developed for each gender by Johnson (1 998)
using height and weight data from Brainard and Burmaster
(1 992) (also see Appendix B of 201 0 CO REA).
Point estimates of exponential parameters used for
calculating body surface area as a function of body mass
(Burmaster, 1998)
5D-5
-------
Component
Variability Source
Comment
Ventilation rate
Event-level activity-specific regression equations stratified by
four age groups, using age, gender, body mass normalized
oxygen consumption rate as independent variables, and
accounting for intra and interpersonal variability (Graham and
McCurdy, 2005).
Fatigue and EPOC
APEX approximates the onset of fatigue, controlling for
unrealistic or excessive exercise events in each persons
activity time-series while also estimating excess post-
exercise oxygen consumption (EPOC) that may occur
following vigorous exertion activities (Isaacs et al., 2007 and
section 7.2 of APEX TSD).
92
93
Table 5D-2. Important components of co-variability.
Type of Co-variability
Within-person correlations 1
Between-person correlations
Correlations between profile variables
and microenvironment parameters
Correlations between demographic
variables (e.g., age, gender) and activities
Correlations between activities and
microenvironment parameters
Correlations among microenvironment
parameters in the same
microenvironment
Correlations between demographic
variables and air quality
Correlations between meteorological
variables and activities
Correlations between meteorological
variables and microenvironment
parameters
Correlations between drive times in
CHAD and commute distances traveled
Consistency of occupation/school
microenvironmental time and time spent
commuting/busing for individuals from
one working/school day to the next.
Modeled
by
APEX?
Yes
No
Yes
Yes
No
Yes
Yes
Yes
Yes
Yes
No
Treatment in APEX / Comments
Sequence of activities performed,
microenvironments visited, and general
physiological parameters (body mass, height,
ventilation rates).
Judged as not important.
Profiles are assigned microenvironment
parameters.
Age and gender are used in activity diary selection.
Perhaps important, but do not have data. For
example, frequency of opening windows when
cooking or smoking tobacco products.
Modeled with joint conditional variables.
Modeled with the spatially varying demographic
variables and air quality input to APEX.
Temperature is used in activity diary selection.
The distributions of microenvironment parameters
can be functions of temperature.
CHAD diary selection is weighted by commute
times for employed persons during weekdays.
Simulated individuals are assigned activity diaries
longitudinally without regard to occupation or
school schedule (note though, longitudinal
variable used to develop annual profile is time
spent outdoors).
1 The term correlation is used to represent linear and nonlinear relationships.
5D-6
-------
94 5D-3. CHARACTERIZATION OF UNCERTAINTY
95 While it may be possible to capture a range of exposure or risk values by accounting for
96 variability inherent to influential factors, the true exposure or risk for any given individual within
97 a study area is largely unknown. To characterize health risks, exposure and risk assessors
98 commonly use an iterative process of gathering data, developing models, and estimating
99 exposures and risks, given the goals of the assessment, scale of the assessment performed, and
100 limitations of the input data available. However, significant uncertainty often remains and
101 emphasis is then placed on characterizing the nature of that uncertainty and its impact on
102 exposure and risk estimates.
103 In the final 2008 Oj NAAQS rule,1 EPA staff performed such a characterization and at
104 that time, identified the most important uncertainties affecting the exposure estimates. The key
105 elements of uncertainty were 1) the modeling of human activity patterns over an Os season, 2)
106 the modeling of variations in ambient Os concentrations near roadways, 3) the modeling of air
107 exchange rates that affect the amount of Os that penetrates indoors, and 4) the characterization of
108 energy expenditure (and related ventilation rate estimates) for children engaged in various
109 activities. Further, the primary findings of a quantitative Monte Carlo analysis also performed at
110 that time indicated that the overall uncertainty of the APEX estimated exposure distributions was
111 relatively small: the percent of children or asthmatic children with exposures above 0.06, 0.07, or
112 0.08 ppm-8hr under moderate exertion have 95% were estimated by APEX to have uncertainty
113 intervals of at most ±6 percentage points. Details for these previously identified uncertainties are
114 discussed in the 2007 Os Staff Paper (section 4.6) and in a technical memorandum describing the
115 2007 Os exposure modeling uncertainty analysis (Langstaff, 2007).
116 The REA' s conducted for the most recent NO2 (US EPA, 2008), SO2 (US EPA, 2009),
117 and CO (US EPA, 2010) NAAQS reviews also presented characterizations of the uncertainties
118 associated with APEX exposure modeling (among other pollutant specific issues), albeit mainly
119 qualitative evaluations. Conclusions drawn from all of these assessments regarding exposure
120 modeling uncertainty have been integrated here, following the standard approach used by EPA
121 staff since 2008 and outlined by WHO (2008) to identify, evaluate, and prioritize the most
122 important uncertainties relevant to the estimated potential health effect endpoints used in this Os
1 Federal Register Vol. 73, No. 60. Available at: http://www.epa.gov/ttn/naaqs/standards/ozone/fr/20080327.pdf
5D-7
-------
123 REA. Staff selected the qualitative approach used for this first draft 63 REA as a step towards
124 developing an appropriate probabilistic uncertainty analysis, perhaps similar to that performed at
125 the time of the 2007 O3 REA by Langstaff (2007).
126 The qualitative approach used in this first draft Os REA varies from that described by
127 WHO (2008) in that a greater focus was placed on evaluating the direction and the magnitude2 of
128 the uncertainty; that is, qualitatively rating how the source of uncertainty, in the presence of
129 alternative information, may affect the estimated exposures and health risk results. In addition
130 and consistent with the WHO (2008) guidance, staff discuss the uncertainty in the knowledge
131 base (e.g., the accuracy of the data used, acknowledgement of data gaps) and decisions made
132 where possible (e.g., selection of particular model forms), although qualitative ratings were
133 assigned only to uncertainty regarding the knowledge base.
134 First, staff identified the key aspects of the assessment approach that may contribute to
135 uncertainty in the exposure and risk estimates and provided the rationale for their inclusion.
136 Then, staff characterized the magnitude and direction of the influence on the assessment results
137 for each of these identified sources of uncertainty. Consistent with the WHO (2008) guidance,
138 staff subjectively scaled the overall impact of the uncertainty by considering the degree of
139 uncertainty as implied by the relationship between the source of uncertainty and the exposure
140 concentrations.
141 Where the magnitude of uncertainty was rated low, it was judged that changes within the
142 source of uncertainty would have only a small effect on the exposure results. For example, we
143 have commonly employed statistical procedure to substitute missing concentration values to
144 complete the APEX ambient input data sets. Staff has consistently compared the air quality
145 distributions and found negligible differences between the substituted data set and the one with
146 missing values (e.g., Tables 5-13 through 5-16 of US EPA, 2010), primarily because of the
147 infrequency of missing value substitutions needed to complete a data set. There is still
148 uncertainty in the approach used, and there may be alternative, and possibly better, methods
149 available to perform such a task. However, in this instance, staff judged that the quantitative
150 comparison of the ambient concentration data sets indicates that there would likely be little
151 influence on exposure estimates by the data substitution procedure used.
!This is synonymous with the "level of uncertainty" discussed in WHO (2008), section 5.1.2.2.
5D-8
-------
152 A magnitude designation of moderate implies that a change within the source of
153 uncertainty would likely have a moderate (or proportional) effect on the results. For example,
154 the magnitude of uncertainty associated with using the quadratic approach to represent a
155 hypothetical future air quality scenario was rated as low-moderate. While we do not have
156 information regarding how the ambient Os concentration distribution might look in the future, we
157 do know however what the distribution might look like based on historical trends and the
158 emission sources. These historical data and trends serve to generate algorithms used to adjust air
159 quality. If these trends in observed concentrations and emissions were to remain constant in the
160 future, then the magnitude of the impact to estimated exposures in this assessment would be
161 judged as likely low or having negligible impact on the estimated exposures. However, if there
162 are entirely new emission sources in the future or if the approach developed is not equally
163 appropriate across the range of assessed study areas, the magnitude of influence might be judged
164 as greater. For example, when comparing exposure estimates for one year that used three
165 different 3-year periods to adjust that year's air quality levels to just meet the current standard,
166 staff observed mainly proportional differences (e.g., a factor of two or three) in the estimated
167 number of persons exposed in more than half of the twelve study areas (Langstaff, 2007).
168 Assuming that these types of ambient concentration adjustments could reflect the addition of a
169 new or unaccounted for emission source in a particular study area, staff also judged the
170 magnitude of influence in using the quadratic approach to adjust air quality data to represent a
171 hypothetical future scenario as moderate. A characterization of high implies that a small change
172 in the source would have a large affect on results, potentially an order of magnitude or more.
173 This rating would be used where the model estimates were extremely sensitive to the identified
174 source of uncertainty.
175 In addition to characterizing the magnitude of uncertainty, staff also included the
176 direction of influence, indicating how the source of uncertainty was judged to affect estimated
177 exposures or risk estimates; either the estimated values were possibly over- or under-estimated.
178 In the instance where the component of uncertainty can affect the assessment endpoint in either
179 direction, the influence was judged as both. Staff characterized the direction of influence as
180 unknown when there was no evidence available to judge the directional nature of uncertainty
181 associated with the particular source. Staff also subj ectively scaled the knowledge-base
182 uncertainty associated with each identified source using a three-level scale: low indicated
5D-9
-------
183 significant confidence in the data used and its applicability to the assessment endpoints,
184 moderate implied that there were some limitations regarding consistency and completeness of
185 the data used or scientific evidence presented, and high indicated the extent of the knowledge-
186 base was extremely limited.
187 The output of the uncertainty characterization is a summary describing, for each
188 identified source of uncertainty, the magnitude of the impact and the direction of influence the
189 uncertainty may have on the exposure and risk characterization results. At this point we have
190 identified a total of 28 sources of uncertainty associated with our approach to model 63
191 population exposure, each broadly summarized in Table 5D-3, including newly identified
192 elements. We then judged whether these results from our historical characterizations were an
193 appropriate characterization of the elements within our current exposure assessment, while also,
194 considering our new analysis of the attributes contributing to those persons highly exposed. The
195 most influential elements of uncertainty in need of further investigation are:
196 • Activity Patterns
197 o In general, with a focus on representation of time spent outdoors
198 o Longitudinal Activity Profiles (e.g., investigation of alternative
199 approaches and assignment of more rigid schedules)
200 • Spatial Variability in Os Concentrations (as the outdoor microenvironment is the
201 most important determinant for 8-hour exposure benchmark exceedances, most
202 elements should be systematically re-evaluated)
203 • Physiological Processes
204 o Metabolic equivalents (METs) distributions (updated information
205 availability, short-term activity evaluations)
206 o Ventilation rate equations
207 Newly identified elements would also be a part of this new uncertainty characterization in
208 future drafts. These include:
209 • The new modeling approach used to simulate ambient air quality that just meets
210 the current standard (if done for future next drafts)
211 • Poverty Status (US Census) Weighted Asthma Prevalence (CDC)
212 • Commuting (CHAD drive times linked with Census commute distances)
213 • Resting Metabolic Rate (RMR) equations
5D-10
-------
214 • At-risk population (effect of averting behavior on activity pattern data)
5D-11
-------
1 Table 5D-3. Characterization of key uncertainties in historical and current APEX exposure assessments.
Sources of Uncertainty
Category
Ambient Monitoring
Concentrations
Element
Database Quality
Instrument Measurement
Error
Missing Data Substitution
Method
Temporal Representation
Spatial Representation:
Large Scale
Spatial Representation:
Neighborhood Scale (1)
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Over
Over
Both
Both
Both
Both
Magnitude
Low
Low
Low
Low
Low
Low
Knowledge-
base
Uncertainty
Low
Low
Low
Low
Low
Low
Comments
All ambient pollutant measurements
available from AQS are both
comprehensive and subject to quality
control.
Mean bias estimated as 1 .2% (CV of
4.4%). See Table 2 and Figure 6 of
Langstaff(2007).
Overall completeness of data yield
negligible mean bias (~0) along with
an estimated standard deviation of 4
ppb when replacing missing values.
See Table 3 of Langstaff (2007).
Appropriately uses 1 -hour time-series
of O3 concentrations for 5 years.
Tens of monitors used in each study
area.
Spatial interpolation using jackknife
method (removal of a single monitor)
yielded generally unbiased
observed/predicted ratios (mean
1.06), having an estimated standard
deviation of 0.2. Langstaff (2007).
Is rating
appropriate for
current APEX O3
exposure
assessment?
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
Yes. For the
uncertainties
characterized, the
historical rating is
appropriate.
However local-
5D-12
-------
Sources of Uncertainty
Category
Adjustment of Air
Quality to Simulate
Just Meeting the
Current Standard
APEX' General
Input Databases
Element
Spatial Representation:
Neighborhood Scale (2)
Spatial Representation:
Vertical Profile
Quadratic Approach
New Model Simulation
Approach
Population Demographics
and Commuting (US
Census)
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Over
Both
Both
Under
Magnitude
Low
Moderate
Low -
Moderate
Low
Knowledge-
base
Uncertainty
Low
Moderate
Moderate
Low
Comments
When reducing the APEX radius
setting from an unlimited value
(actual value used) to 10 km (i.e., the
tendency would be to more
accurately represent exposure), a
smaller fraction (1-3 percentage
points) of population exceeds
benchmark levels. See Figures 7-9
of Langstaff (2007).
Differences between ground-level (0-
3 meters) and building rooftop sited
(25 meters) monitor concentrations
can be significant. Most importantly,
use of higher elevation monitors
would tend to overestimate ground-
level exposures (i.e., persons
outdoors).
Variable differences (e.g., none to a
factor of two or three) in the
estimated number of persons
exposed across study areas when
using differing 3-year roll-back
periods for a single year of air quality
(Langstaff, 2007).
New approach developed for this
REA, newly identified, not evaluated.
Comprehensive and subject to quality
control. Differences in 2000 versus
modeled years (2006-10) likely small
when estimating percent of
population exposed.
Is rating
appropriate for
current APEX O3
exposure
assessment?
scale spatial
representation
(not
characterized)
may result in a
different
characterization.
Yes. Given
judged impact to
exposure,
additional
characterization is
needed.
Yes. Uncertainty
in the approach
has resulted in
plans to use
alternative
approach.
New. Needs
characterization.
Yes. No further
characterization
needed.
5D-13
-------
Sources of Uncertainty
Category
APEX:
Microenvironmental
Concentrations
Element
Activity Patterns (CHAD)
Meteorological (NWS)
Poverty Status (US
Census) Weighted
Asthma Prevalence
(CDC)
Outdoor Near-Road and
Vehicular: Proximity
Factors
Indoor: Near-Road
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Unknown
Both
nc
Both
Over
Magnitude
Low -
Moderate
Low
nc
Low
Low
Knowledge-
base
Uncertainty
Moderate
Low
nc
Low-
Moderate
Low
Comments
Comprehensive and subject to quality
control. However, comprised of
multiple studies, varying survey
techniques, historical data, broad
location/activity code assignments,
among other issues, add to difficulties
in assessing uncertainties.
Comprehensive and subject to quality
control, few missing values. Limited
application in selecting CHAD diaries
and AERs.
New data set generated for this REA,
newly identified, not evaluated.
Uncertainty in mean value used
approximated as 15 percentage
points. See Figure 10 and Table 7 of
Langstaff (2007). May be of greater
importance in certain study areas.
Expected reduction in O3 for persons
residing near roads not modeled
here, but when included, there is a
small reduction (~3%) in the number
of persons experiencing exposure
above benchmark levels (Langstaff,
2007).
Is rating
appropriate for
current APEX O3
exposure
assessment?
Yes. Given
judged impact to
exposure,
additional
characterization is
needed.
Yes. No further
characterization
needed.
New. Needs
characterization.
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
5D-14
-------
Sources of Uncertainty
Category
Element
Indoor: Air Exchange
Rates
Indoor: A/C Prevalence
(AHS)
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Both
Both
Magnitude
Low
Low
Knowledge-
base
Uncertainty
Moderate
Low
Comments
Uncertainty due to random sampling
variation via bootstrap distribution
analysis indicated the AER GM and
GSD uncertainty for a given study
area tends range to at most from
fitted ±1.0 GM and ± 0.5 GSD hr"1.
Non-representativeness remains an
important issue as city-to-city
variability can be wide ranging
(GM/GSD pairs can vary by factors of
2-3) and data available for city-
specific evaluation are limited (US
EPA, 2007). Also, indoor exposures
are estimated as not important to 8-
hour average daily maximum O3
exposure.
Comprehensive and subject to quality
control, estimated 95th percentile
confidence bounds range from a few
to just over ten percentage points,
though some cities use older year
data (Table 9 of Langstaff, 2007).
Note, variable indicates
presence/absence not actual use.
Also, indoor exposures are estimated
here as limited in importance to 8-
hour average daily maximum
exposures and sensitivity analyses in
NO2 REA (in-vehicle was most
influential exposure ME) concluded
prevalence variable was of limited
importance.
Is rating
appropriate for
current APEX O3
exposure
assessment?
Yes. No further
characterization
nppHpH
l l^djdj .
Yes. No further
characterization
needed.
5D-15
-------
Sources of Uncertainty
Category
APEX: Simulated
Activity Profiles
Element
Indoor: Removal Rate
Vehicular: Penetration
Factors
Longitudinal Profiles
Commuting
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Both
Both
Under
nc
Magnitude
Low
Low
Low -
Moderate
nc
Knowledge-
base
Uncertainty
Low
Moderate
Moderate
nc
Comments
Greatest uncertainty in the input
distribution regarded
representativeness, though estimated
as unbiased but correct to within
10%.
Input distribution is from an older
measurement study though
consistent with recent, albeit limited
data.
Depending on the longitudinal profile
method selected, the number of
persons experiencing multiple
exposure events at or above a
selected level could differ by about 15
to 50% (see Appendix B, Attachment
4 of NO2 REA). Long-term diary
profiles (i.e., monthly, annual) do not
exist for a population, limiting the
evaluation. Modeling does not assign
rigid schedules for workers or
children attending school.
New method used in this assessment
designed to link Census commute
distances with CHAD vehicle drive
times, newly identified, not evaluated.
Note while vehicle time accounted for
through diary selection, not rigidly
scheduled.
Is rating
appropriate for
current APEX O3
exposure
assessment?
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
Yes. Given
judged impact to
exposure,
additional
characterization is
needed.
NPW Npprl^
INCVV. INC/C/UO
pwoli lation
C? VC1 ILICIlll_/l 1
5D-16
-------
Sources of Uncertainty
Category
APEX:
Physiological
Processes
Element
At-Risk Population
Body Mass (NHANES)
NVO2max
RMR
METS distributions
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Both
Unknown
Unknown
nc
Over
Magnitude
Low
Low
Low
nc
Low -
Moderate
Knowledge-
base
Uncertainty
Low -
Moderate
Low
Low
nc
Low-
M ode rate
Comments
Asthmatics activity patterns are
similar to that of non-asthmatics (both
types of diaries are used in our
simulations, regardless of health
status). See discussion in SO2 REA
(section 8. 11. 2.2. 5).
Comprehensive and subject to quality
control, though older (1999-2004)
than current simulated population,
possible small regional variation is
not represented by national data.
Upper bound control for unrealistic
activity levels rarely used by model,
thus likely not very influential.
Approach from older literature
(Schofield, 1985), linked to estimated
ventilation rates, not previously
evaluated.
APEX estimated daily mean METs
range from about 0.1 to 0.2 units
(between about 5-10%) higher than
independent literature reported
values (Table 15 of Langstaff, 2007).
Shorter-term values are of greater
importance in this assessment.
Is rating
appropriate for
current APEX O3
exposure
assessment?
Yes. For the
uncertainties
characterized, the
historical rating is
appropriate.
However, averting
behavior (where
present in input
data and currently
undesignated)
may result in a
different
characterization.
Yes. No further
characterization
needed.
Yes. No further
characterization
needed.
New. Needs
characterization.
Vpc C^iwpn
judged impact to
exposure,
additional
characterization is
needed
5D-17
-------
Sources of Uncertainty
Category
Element
Ventilation rates
Historical Uncertainty Characterization
Influence of Uncertainty
on Exposure/Intake
Dose Estimates
Direction
Over
Magnitude
Low -
Moderate
Knowledge-
base
Uncertainty
Low -
Moderate
Comments
APEX estimated daily ventilation
rates can be greater (2-3 m3/day)
than literature reported measurement
values (Table 25 of Langstaff, 2007),
though accounting for measurement
bias minimizes the discrepancy
(Graham and McCurdy, 2005). Also,
a shorter-term comparison (for hours
rather than daily), while more
informative, is lacking due to limited
data.
Is rating
appropriate for
current APEX O3
exposure
assessment?
Yes Given
judged impact to
exposure,
additional
characterization is
needed
5D-18
-------
5D-4. REFERENCES
Brainard J and Burmaster D. (1992). Bivariate distributions for height and weight of men and women in the United
States. RiskAnalysis. 12(2):267-275.
Burmaster DE. (1998). Lognormal distributions for skin area as a function of body weight. RiskAnalysis. 18(1):27-
32.
CDC. (2011). Summary Health Statistics for U.S. Adults: National Health Interview Survey, years 2006-10. U.S.
Department of Health and Human Services, Hyattsville, MD. Data and documentation available at:
http://www.cdc.gov/nchs/nhis.htm (accessed October 4, 2011).
Glen G, Smith L, Isaacs K, McCurdy T, Langstaff J. (2008). A new method of longitudinal diary assembly for
human exposure modeling. JExpos Sci Environ Epidem. 18:299-311.
Graham SE and T McCurdy. (2005). Revised ventilation rate (VE) equations for use in inhalation-oriented
exposure models. Report no. EPA/600/X-05/008. Report is found within Appendix A of US EPA (2009).
Metabolically Derived Human Ventilation Rates: A Revised Approach Based Upon Oxygen Consumption
Rates (Final Report). Report no. EPA/600/R-06/129F. Appendix D contains "Response to peer-review
comments on Appendix A", prepared by S. Graham (US EPA). Available at:
http://cfpub.epa.gov/ncea/cfm/recordisplav.cfm?deid=202543
Issacs K and Smith L. (2005). New Values for Physiological Parameters for the Exposure Model Input File
Physiology.txt. Technical memorandum to Tom McCurdy, NERL WA10. December 20, 2005. Provided
in Appendix A of the CO REA (US EPA, 2010).
Isaacs K, Glen G, McCurdy T., and Smith L. 2007. Modeling energy expenditure and oxygen consumption in
human exposure models: Accounting for fatigue and EPOC. J Expos Sci Environ Epidemiol. 18(3):289-98.
Johnson T. (1998). Memo No. 5: Equations for Converting Weight to Height Proposed for the 1998 Version of
pNEM/CO. Memorandum Submitted to U.S. Environmental Protection Agency. TRJ Environmental, Inc.,
713 Shadylawn Road, Chapel Hill, North Carolina 27514.
Johnson T, Mihlan G, LaPointe J, Fletcher K, Capel J, Rosenbaum A, Cohen J, Stiefer P. (2000). Estimation of
carbon monoxide exposures and associated carboxyhemoglobin levels for residents of Denver and Los
Angeles using pNEM/CO. Appendices. EPA constract 68-D6-0064.
Langstaff JE. (2007). OAQPS Staff Memorandum to Ozone NAAQS Review Docket (OAR-2005-0172). Subject:
Analysis of Uncertainty in Ozone Population Exposure Modeling. [January 31, 2007]. Available at:
http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3crtd.html
McCurdy T. (2000). Conceptual basis for multi-route intake dose modeling using an energy expenditure approach.
J Expos Anal Environ Epidemiol. 10:1-12.
SchofieldWN. (1985). Predicting basal metabolic rate, new standards, and review of previous work. Hum Nutr
ClinNutr. 39C(S1):5-41.
US Census Bureau. (2007a). Employment Status: 2000- Supplemental Tables. Available at:
http://www.census.gov/population/www/cen2000/phc-t28.html.
US Census Bureau. (2007b). 2000 Census of Population and Housing. Summary File 3 (SF3) Technical
Documentation, available at: http://www.census.gov/prod/cen2000/doc/sf3.pdf. Individual SF3 files '30'
19
-------
(for income/poverty variables pct49-pct51) for each state were downloaded from:
http://www2.census.gov/census 2000/datasets/Summary File 3/.
US EPA. (2002). EPA's Consolidated Human Activities Database. Available at: http://www.epa.gov/chad/.
US EPA. (2007). Ozone Population Exposure Analysis for Selected Urban Areas. Office of Air Quality Planning
and Standards, U.S. Environmental Protection Agency, Research Triangle Park, NC. Available at:
http://www.epa.gOv/ttn/naaqs/standards/ozone/s o3crtd.html
US EPA. (2008). Risk and Exposure Assessment to Support the Review of the NOa Primary National Ambient Air
Quality Standard. Report no. EPA-452/R-08-008a. November 2008. Available at:
http://www.epa.gov/ttn/naaqs/standards/nox/data/2008 1 12 l_NO2_REA_final.pdf .
US EPA. (2009). Risk and Exposure Assessment to Support the Review of the SO2 Primary National Ambient Air
Quality Standard. Report no. EPA-452/R-09-007. August 2009. Available
athttp://www.epa.gov/ttn/naaqs/standards/so2/data/200908SO2REAFinalReport.pdf.
US EPA. (2010). Quantitative Risk and Exposure Assessment for Carbon Monoxide - Amended. EPA Office of
Air Quality Planning and Standards. EPA-452/R-10-009. July 2010. Available at:
http://www.epa.gov/ttn/naaqs/standards/co/data/CO-REA-Amended-Julv2010.pdf
US EPA. (2012). Total Risk Integrated Methodology (TRIM) - Air Pollutants Exposure Model Documentation
(TRIM.Expo / APEX, Version 4.4) Volume I: User's Guide. Office of Air Quality Planning and Standards,
U.S. Environmental Protection Agency, Research Triangle Park, NC. EPA-452/B-12-001a. Available at:
http://www.epa.gov/ttn/fera/human _apex.html
WHO. (2008). Harmonization Project Document No. 6. Part 1: Guidance document on characterizing and
communicating uncertainty in exposure assessment. Available at:
http://www.who.int/ipcs/methods/harmonization/areas/exposure/en/.
20
-------
1 APPENDIX 7A.
2 INPUT DATA USED IN MODELING RISK FOR THE 12 URBAN STUDY AREAS
3 This appendix presents data used in modeling risk for the 12 urban study areas (in Table 7A-
4 1). In some cases (as noted below) data are presented in an aggregated fashion. If the reader
5 would like the dis-aggregated data, they can consult the original data sources cited in the
6 relevant sections of Chapter 7. Table 7A-1 is organized by health endpoint. The specific
7 types of data provided for each endpoint are described below (note, only those data fields
8 requiring additional clarification are described here, many are self explanatory).
9 • Study information (C-R function): these fields provide information on the C-R
10 functions used in modeling endpoints covered in the risk assessment including (a)
11 ozone metric and risk modeling period, (b) age range of the population modeled, the
12 effect estimate (including statistical fit information), the model form and additional
13 details related to the model (e.g., lag structure, copollutants control if relevant) (see
14 section 7.3.2).
15 • Baseline incidence: annual incidence per 100,000 general population for the specific
16 risk period modeled for that health endpoint in the risk assessment (i.e., these are not
17 annual values, but rather incidence rates for the risk modeling period). Incidence rates
18 are provided for both simulation years (2007 and 2009) (see section 7.3.4).
19 • Population: count of individuals matched to the population being modeled for the
20 particular health endpoint (provided for 2007 and 2009 - see section 7.3.5).
21 • Surrogate LMLs: These are the LMLs obtained from the composite monitor
22 distributions used to model each health endpoint (for a given urban study area and
23 simulation year) (see section 7.3.3).
7A-1
-------
Table 7A-1 Selected Model Inputs Used in Generating Risk Estimates for the First Draft REA
Endfioifrt
Mortality. All Cause
Mortality, All Cause
Mortality, All Cause
Mortality, All Cause
Mortality. All Cause
Mortality, All Cause
Mortality. All Cause
Mortality, All Cause
Mortality, All Cause
Mortality, All Cause
Mortality, All Cause
Mortality, All Cause
Mortality.. Ncn-
Accidental
Mortality, Non-
fee id ental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality, Ncn-
Accidental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality, Non-
Accidental
Mortality. Nort-
Accidental
Mortality, Nan-
Accidental
Study
Zanabstti and
Schwartz \b\, 2CCS
Zanobetti and
Schwartsjb*., 200S
Zanobetti and
SchwartzlbJ, 2C03
Zanobetti and
Schwartz (bj, 2CCS
Zanobetti and
Schwartzlb}, 200 B
Zancbstti and
Sclrwarujb), 2GC3
Zanabetti and
S-chwartzjb', 20CS
Zanobetti and
S-chwartzfb*., 20CS
Zanobetti and
Schwartzlb}, 20CB
Zanobetti an-d
Schwartz Sbj, 2CCS
SchwartzfbJ, 2COS
Zancbetti and
Schwartz |bj, 2CCS
Belletal., 2CG-4
Beiletal., 2CG4
Belletal., 200^
Belletal., 2004
Belletal., 2GQ-&
Belletal., 200*
Belletal., 2CCi
Belletal., 200*
Belletal.. 2004
Belletal., 2004
Belletal., 2CQ£
Belletal., 20C4
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland. CH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CC^
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Study information (C-& function)
Air metric
DSHourMean
BSHcurMean
DBHourMean
DSHourMean
DSHourMean
DSHourMean
DSHourMean
DSHourMean
DSHourMean
DBHourMean
DSHourMean
DSHourMean
DSHourMax
DSHourMax
DS Hou rM ax
DSHourMss
DSHourMax
DBHourMax
DSHourMax
DSHourMax
DSHourMax
DSHourMax
DSHourMax
iRisic assessment
modeling
period
June-August
June-August
June-August
June-August
June-August
June-August
June-August
J u n e-August
June-August
June-August
June-August
June-August
March-October
isy
April-October
I? I-
April-
September JS)
April-October
M a rr h-
April-
September J6)
J a n u a ry~
December (12)
J 3 n u a ry-
December |12;
April-October
April-October
J a n u a ry-
Decernber jl2}
April-October
Age range
0-99
C-93
C-99
C-99
G-99
0-99
0-99
0-99
C-99
'C-99
0-99
0-39
C-99
C-99
0-99
C-99
C-99
C-99
0-99
C-99
0-99
0-99
0-99
C-99
Lag
distributed
lag 0-3 d
distributed
lagO-3 d
distributed
lag 0-3 d
distributed
!agO-3d
distributed
Iag0-3d
distributed
lagC-3d
distributed
lag 0-3 d
distributed
leg 0-3 d
distributed
laf.0-3d
distributed
Iag0-3d
distributed
Iag0-3d
distributed
Iag0-3d
distributed
Iag0-6d
distributed
lagO-Sd
distributed
IsgC-Sd
distributed
Iag0-ed
distributed
lag 0-6 d
distributed
lagO-Sd
distributed
lag 0-6 d
distributed
lagO-Sd
distributed
lag 0-6 d
distributed
iagC-€d
distributed
Iag0-ed
distributed
Iag0-6d
AdiJition»l
study details
Statistica
1 Model
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
Bert
estimate
(Bet.)
0 OCC2§5*
O.CC0515
0.000 6-S16
0.0005962
O.OC0351B
0.0010*59
0.0001523
0.0002737
C.OC10325
O.OOC-62'6
0.0005691
0.0005«i
0.0007053
0.0003*28
0.0006.3-87
O.OCC-4353
0.0003862
0.0003307
O.COO"22
0.000*168
0.000528'
0.0003736
0.000^2*2
0.000*113
S£ (effect
estimatef
0.00028S6
0.00031^
fl.COC32Si
0.00035*6
G .GCO-^C SB
0.0003*41
0.000262S
0.0002134
0.0002357
0.00031*6
0.0003S35
0.000333*
0.0002252
0.0002533
O.OO02S7
O.OS030S
0.0003105
0.000259
0.0003013
C.COC23S3
0.0003191
0.0003213
0.0002935
0.00031*1
Baseline
incideraeb
2007
133
219
190
2S7
133
230
1*0
152
131
26*
132
213
332
17 i
360
530
380
42*
500
5E2
401
562
661
*S3
2009
134
211
1S3
256
173
218
135
1*3
172
250
175
20*
32*
*56
3*7
564
355
*02
*S*
5*3
385
530
636
*4*
Population
2007
3,356,357
2,1*6,632
',021,878
1,323,261
553.317
1,986,360
3.730.576
9,331,639
11,0*3,330
1,456,148
1,405,7"
2,193,2*2
3,356,357
2,1*6,632
4.021,378
1,323,261
558,817
1,986,360
3,730,575
9,931,6*0
11,0*3,332
1,456.148
1,405,7**
2,198,2*2
2009
3,972,395
2,19*,116
4,0*8,879
1,317,923
559,791
1,969,826
3,350,328
10,0*2,327
11,108,750
1,***,16*
1,*53,703
2,211,259
3,972,395
2,19*. 116
4,0*8,879
1,317,928
559,791
1,969,326
3,350,325
10,0*2,327
11,108,750
1,4*4,15*
1,453,703
2,211,259
Surrogate LMLs
jppb)
20O7
24
13
19
6
21
19
10
31
10
12
30
22
17
13
12
12
4
13
6
9
10
13
13
8
2009
21
2*
17
16
~>7
11
15
22
12
1*
30
-j-j
5
9
12
15
16
1*
,
a
8
9
5
7
7A-2
-------
Endpoint
Mortality,
Cardiovascular
Mortality,
Cardicvascular
Mortality,
Cardiovascular
Me reality,
Cardicvascular
Mortality,
Cardiovascular
Mortality,
Cardie-vascular
Mortality,
Cardiovascular
Mortality,
Cardiovascular
Mortality,
Cardicvascular
Mortality,
Cardiovascular
Mortality,
Cardiovascular
Mortality,
Cardicvascular
Mortality,
Bespiratcry
Mortality,
PesciraMrv
Mortality,
Respiratory
Mortality,
Respiratory
Mortality,
Respiratory
M orta 1 ity,
Respiratory
Mortality,
Rsspirstcry
Mortality,
Res-pi ratcry
Mortality,
Respi rater/
Mortality,
Respiratory
Mortality,
Respiratcrv
Mortality,
Respiratory
Study
Zanobetti and
Schwartz (b), 2O08
Zanobetti and
Schwartz (b), 2QQS
Zanobetti and
Schwartz |b), 200B
ZanDbetti and
Schwartz (b), 2O3S
Zanobetti and
Schwartz Ib), 200S
Zanobetti and
Schwartz IbJ. 2008-
Zanc-hetti and
Schwartz (b), 2OOS
Zanabettl and
Schwartz (b), 2QO&
Zanobetti and
Schwartz (b). 20QS
Zanobetti and
Schwartz (b), 2008
ZanDbetti and
Schwartz lb), 2OOS
Zanobetti and
Schwartz (b), 2QQ5
Zanobetti and
Schwartz (b), 2OOS
Zancbetti and
Sch\vartz|b), 2GQS
Zanobetti and
Schwartz lb;. 2008
Zanobetti and
Schwartz Ib), 2OOS
Zanobetti and
Schwartz (b), 2OOS
Zancbetti and
Schwartz |b), 20OS
Zanobetti and
Schwartz |b/, 2008
Zanobetti and
Schwartz (b), 2O3S
Zanobetti and
Schwartz (b«, 2OOB
Zanobetti and
Schwartz !>', 2OQS
Zanobetti and
Schwartz (b), 2O3S
Zanobetti and
Schwartz Ib), 2008
Urban study area
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Atlanta, GA
Baltimore, MD
Boston. MA
Cleveland, OH
Denver, CO
Detroit, Ml
Houston, "TX
Los Angeles, CA
New York, NY
Philadelphia, PA
St. Louis, MO
Study information |C-R function)
Air metric
DSHourMean
DSHourMean
DBHourMean
DSHourMear
DSHourMean
DSHourMean
DSHourMean
DBHourMean
DSHourMean
DSHourMean
DSHourMean
DSHourMean
DSHcurMean
DSHourMean
DBHourMean
DSHourMean
DBHourMean
DSHourMean
DBHourMean
DSHourMean
DSHourMean
DSHourMean
DSHourMean
DSHcurMean
Risk assessment
modeling
period
June- August
June-August
June-August
Jure-AugUSt
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
J u n e-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
Age range
0-99
0-99
•0-99
•0-93
•0-99
O-S9
O-99
•0-99
•0-99
•0-99
•0-99
0-99
0-S9
0-99
C-99
•0-99
O-S9
•0-99
•0-99
0-99
•0-99
•0-99
•0-99
-0-99
La£
distributed
leg -3-3 d
distributed
lag 0-3 d
distributed
lag '3-3 d
distributed
las 0-3 d
distributed
lasC-3d
distributed
lag -3-3 d
distributed
lag '3-3 d
distributed
lag 0-3 d
distributed
lag '3-3 d
distributed
Iag3-3d
distributed
lag '3-3 d
distributed
lag 0-3 d
distributed
la|C-3 d
distributed
lag '3-3 d
distributed
lag '3-3 d
distributed
lagO-3d
distributed
lag -3-3 d
distributed
lag -3-3 d
distributed
lagO-3d
distributed
lagO-3d
distributed
lag 0-3 d
distributed
lag '3-3 d
distributed
lag -3-3 d
distributed
§O-3d
Additional
study details
Statist! da
1 Model
log-
linear
log-
linear
log-
linear
log-
liresr
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linssr
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
Icg-
lircar
log-
linear
log-
linear
log-
linear
leg-
linear
Effect
estimate
(tea)
0.0006124
C."C3665
C.::C7E6E
O.OO07935
O.OOOS25S
c.::i-2i3
O.O3O6319
O.OO03071
O.O012475
C.CCC9SSS
O.OO0754
O.O007753
•:.::i:t-:
O.OOOS325
O.O010BS7
O.OO094S1
0.0008799
O.O01O421
:.:cci7::
O.OOO4179
O.OOOS568
O.O007B69
C.C3C733
O.OOOS74I'1
SE (effect
estimate}"
O.OOO335S
O.OO0353
:.:::36i7
O.O303711
O.OO04O33
:.;::3663
O.O333116
O.O302432
c.:c::eis
0.0003563
O.Q3O39S1
c.:c:3e-6
O.OCX33746
O.OO037S3
:.:c:37E5
c.:"3s:5
:.:3c-:32
o.oooaas
c.:::357s
O.O303101
O.OM3505
c.:::37S7
O.OOO397S
:.:::3S3
Baseline
incidence1
2O07
37
32
53
99
49
93
41
56
..,
34
&3
34
11
2=
23
21
13
IB
10
14
16
21
20
13
2OO9
37
B2
5S
99
49
93
41
56
,-,
34
BO
34
11
19
19
20
IS
17
w
14
15
23
19
13
Population
2O07
3,356,357
2,146,632
4,021,S7S
1,323,261
553,317
1.936,360
3,730,576
9,331,639
11.O43.330
1,456,14S
1,405,744
2,193,242
3,356,357
2,146,632
4.021.B7B
1,323,261
553,317
1.936,360
3,733,576
9,931,639
Il,0i3,330
1,456,143
1,405,744
2,193,242
2OO9
3,972,395
2,194,116
4,043,879
1,317,923
559,791
1,969,326
3,350,328
10/342,327
11,103,750
1,444,164
1,453,703
2,211,259
3,972,395
2,194,116
4,048,879
1,317,923
559,791
1.969,826
3,850,328
1O,O42,327
11,103,750
1.---.K-
1,453,703
2,211,259
Surrogate LMLs
(ppbl
2O07
24
13
19
6
21
19
10
31
10
12
30
22
24
13
19
6
21
19
10
31
10
12
30
n
2O09
21
24
17
16
22
11
15
22
12
14
3O
22
21
24
17
16
22
11
15
22
12
14
3O
22
7A-2
-------
End point
Asthma
Exacerbation, Chest
Tightness
.Asthma
Exacerbation, Chest
Tightresi
Asthma
Exacerbation, Chest
Ti|i".rezi
.Asthma
Exacerbation^ Chest
Tightness
Asthma
Exacerbation,
Shortness cf Breath
Asthma
Exacerbation,
Shortness of Breath
Asthma
Exacerbation,
Wheeze
Emergency Room
Visits. .Asthma
Emergency Room
Visits, .Asthma
Emergency Room
Visits, .Asthma
Emergency Room
Visits, Asthma
Emergency Room
Visits, .Asthma
Emergency Room
Visits, Respiratory
Emergency Room
Visits, Respiratory
Emergency Room
Visits, Pespiratcry
Emergency Room
Visits, Respiratory
Emergency Room
Visits, Respiratory
Emergency Room-
Visits, Respiratory
Em e rge n cy Room
Visits, Respiratory
Emergency Room
Visits, Respiratory
- A All Respiratory
Study
Gentetal., 2O03
Gentetal., 2OC3
Gentetal., 2003
Gent etal., 2003
Gentetal., 2003
Gentetal., 2003
Gentetal., 2CO3
Itoetal., 2O07
Itnetal., 2CO7
Itcetal., 2007
Itoetal., 20O7
Itoetal., 2007
Darrowet al., 2O11
Strickland etal.,
2010
Strickland etal.,
2010
Tolbertet =1., 2CC7
Tolbertetal., 2-0-07
Tolbertetal., 2007
Tolbertetal.. 2007
Tolbertetal., 2O07
Katsouyanni etal.,
''.ZZ3
Urban study area
Boston, MA
Boston, MA
Boston. MA
icston, MA
Boston, MA
3-oston, MA
Boston, MA
New York, NY
New York, NY
New York, NY
New York, NY
New York, NY
Atlanta, G A
Atlanta, GA
Atlanta, GA
Atlanta, GA
Atlanta. GA
Atlanta. GA
Atlsrts, GA
Atlanta, GA
Cetrcit, Ml
Study information |C-R function}
Air metric
DlHourMax
DBHourMax
DlHourMax
DlHourMax
DlHcurMax
DSHourMax
DlHcurMax
DSHourMax
DSHaurMax
DSHcurMax
DSHcurMax
DSHourMax
DBHourMax
DBHourMax
DSHourMax
DBHcurMax
DSHaurMax
DBHourMax
DSHcurMax
DSHourMax
DlHourMax
Risk assessment
modeling
period
April-
September [&)
.April-
September |6)
April-
September |6)
April-
September |6)
April-
September |E)
April-
September (6)
April-
September 16)
April-October
P>
April-October
P)
April-October
0
April-October
(7)
April-October
0
March-October
IB)
March-October
IS)
March-October
g
March-October
H
March-October
la)
March-October
IS!
March-October
is;
March-October
IS)
June-August
Age range
0-12
•3-12
O-12
•3-12
0-12
O-12
O-12
0-99
•3-99
033
•3-99
•3-99
0-99
5-17
5-17
•3-99
3-99
•3-99
0-99
0-99
65-33
lag
Lag Id
La- Id
L=sld
Lag Id
Lagld
Ligld
LagOd
average cf lag
0 and lag 1
average cf lag
0 and lag 1
average cf lag
0 and lag 1
average of lag
•3 and lag 1
average cf lag
0 and lag 1
Lag Id
distributed
lag 0-7 d
average of
lags -3-2
average of
lags '3-2
average of
lags 0-2
average of
lags '3-2
average of
1 =523-2
average of
lags 0-2
average cf lag
C and lag 1
Additional
study details
PM2.5
PM2.5
PM2.5
PM2.5
NO2
CO
sc:
CO
NO2
PM1C
PM10, NO2
penalized
splines
Statistics
1 Model
logistic
Icgistic
logistic
Icgistic
logistic
Icgistic
logistic
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
leg-
linear
log-
linear
log-
linear
log-
linear
log-
linear
log-
linear
Effect
estimate
(Beta)
C.C;C76C3
C.C:57C3E
O.O077052
O.O070131
C.CC3B77
0.0052473
:.c;e:::i
O.O052134
0.0039757
0.0032337
0.0055437
0.004115
O.O006S52
O.O047S64
O.OQ2699
O.OQ12S6
0.00114OS
0.0010287
O.OOOS032
:.::c77-3
0.00056
SE (effect
estimate}'
O.O02OO02
C.CC2C:i7
O.O0226&6
O.SX522734
:.CC173-7
O.O021SOB
:.c:::::s
O.O3O9OS7
•3.&3397S9
0.0009359
O.O33S939
c.cc:3::e
0.00013S5
0.0007602
0.0006456
O.OOO2OS2
O.OO022S3
0.0002506
c.c:c:e7
c.;:::67:
:.:::3s:
Baseline
incidence1
2O07
19,541
19,541
19,541
19,541
2-.-2c
24,426
45,595
6S6-
6S6
SS6
ES6
636
2.SB9
5,464
5,464
2.S-S9
2.SS9
2.SB9
1.SS3
2.&B3
1.34B
2O09
19,546
13,546
19,548
19,546
24,432
24,432
-5.6C7
686
686
ES6
6S6
6S6
2,9-32
5,464
5/6-
2,902
2.902
2,9Q2
:.9:2
2,902
1.336
Popu
20X)7
662,064
662/364
662,064
662,064
EE:.ce;
662,064
662,064
11,043,332
11,043,332
11,043,332
11/343,332
11/343,332
3,556,358
697,630
697,690
3,S56.35S
3,S56,35S
3,856,358
3. BSE, 353
3,356,35S
221,636
ation
2009
669,219
669,219
669,219
669,219
669,219
669,219
669,219
11,1OB,75O
11,108,750
11,108,750
11,108,750
11,108,750
3,972,395
714,368
714.36S
3,972,395
3.972.395
3,972.395
3.572.335
3,972,395
218,112
Surrogate LMLs
(ppb)
2007
14
12
14
14
14
12
14
10
10
10
10
10
17
17
17
17
17
17
a
17
23
2009
15
12
15
15
15
12
15
8
3
8
a
8
5
5
5
5
5
5
S
5
17
7A-4
-------
Endpoint
HA, All Respiratory
HA, All Respiratory
HA, Asthma
HA. Asthma
HA, Chronic Lung
-:::•::
HA, Chronic Lung
Disease [less
Asthma)
HA. Chronic Lung
Disease (less
Asth m a )
HA, Chronic Lung
Disease [less
Asthma)
HA, Chronic Lung
Disease [less
Astn m a J
HA, Chronic Lung
Asthma)
HA, Chronic Lung
Disease (less
Asth m a )
HA, Chronic Lung
Disease (less
Asth m a )
HA, Chronic Lung
Disease Hess
Asth m a )
HA, Chronic Lung
Disease (less
As t h rr 3 '•
HA, Chronic Lung
Asthma)
HA, Chronic Lung
Disease [less
.Asth m a )
HA, Chronic Lung
.Asth m a )
Study
Katsouyanni et al.,
20B9
Linn etal., 2000
Silverman and Ito,
2010
Silverman and Itc,
2010
Lin etal. (a), 2OOS
Medina-Ramon et
al, 2-306
Medina-Ramon et
a I, 2O36
Medina-Ramon et
al, 2006
Medina-Ramon et
al, 2006
Medina-Ramcn et
al, 2006
Medina-Ramon et
=i. ::C-E
Medina-Ramon et
al, 2-306
Medina-Ramon et
al, 2006
Medina-Ramon et
al, 200S
Medina-Ramon et
si. ::ce
Medina-Ramon et
al, 2006
Medina-Ramon et
al, 2036
Urban study area
Detroit, Ml
IDS Angeles., CA
New York, NY
New York, NY
New York, NY
Atlanta, GA
Baltimore, MD
Boston, MA
Cleveland. OH
Denver, CC
Detroit, Ml
Houston, TX
Los Angeles, CA
New York, NY
Philadelphia, PA
Sacramento, CA
St. Louis, MO
Study
Air metric
DIHourMax
D24HourMean
C 3 H c L rM a>:
DSHcurMax
DlHcurMa*
CBHcurMear
DSHourMean
DSHourMean
DSHourMean
DSHourMean
DS H c u rM e a n
DSHourMean
DSHc-urMean
DSHourMean
DSHourMean
DSHcurMean
DSHourMean
Risk assessment
modeling
period
June-August
June-August
April-October
17)
April-October
April-October
(7)
June-August
J u n e-Augu st
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
June-August
Age range
65-99
3-0-99
£-12
6-15
•2-17
65-99
65-99
65-99
65-99
G5-99
65-99
€5-99
65-99
65-99
£5-99
65-99
65-99
nformation JC-R function \
i^e
average cf lag
0 a n d 1 a g 1
LagOd
average of lag
0 and lag 1
average cf lag
0 and lag 1
Lag2d
distributed
lagC-ld
distributed
lagO-ld
distributed
lag '0-1 d
distributed
lag 0-1 d
lag -0-1 d
distributed
lag 0-1 d
distributed
la = C-ld
distributed
lag -0-1 d
distributed
lag -0-1 d
distributed
lagC-ld
distributed
lag 0-1 d
distributed
lag 0-1 d
Additional
study details
natural
splines
PM2.5
Statistics
1 Model
log-
linear
log-
linear
log-
linear
linear
loS-
linear
legist!:
logistic
logistic
logistic
logistic
logistic
Icgistic
logistic
logistic
logistic
logistic
logistic
Effect
estimate
(Bon)
C.CCC5-
3.0036
C.CC73C7
2.C-C55553
0.0007609
C.CCCB-
C.-33C5-
0.00354
0.00054
0.00054
O.O3054
•3/53354
0.00054
0.00354
C.CCC5-
C.CCC5-
. . . .
5E {effect
estimate!"
0.'3O33571
3.0037
C.CC37SE:
3.-3036926
0/300163
C.:C:IBB
O.CXX5199
0.000199
C.CCC153
0.000199
O.OO3199
O.O03199
c.:::i53
0/3O3199
C.CCCI5S
O.OO3199
Baseline
incidence'
2007
1,343
269
192
13;
234
3E7
314
25E
320
204
375
261
187
19E
265
161
.
2009
1,336
273
IB:
IB:
233
365
312
254
317
204
372
259
187
195
262
162
TT3
Population
2007
221,636
5640233.5
1.352.727
1.S52.727
:,S93,SS7
331. S12
2E3.211
509,362
195,957
55,913
221, €3E
2B1.-77
1.-315/399
1,3E4,875
133. 7E5
152.374
.
2009
218,112
5730434.5
1,363,523
1,863,528
2,593,341
325. 37B
272,392
519,354
192.596
53.947
213,112
3:7.353
1,048,772
1,375,434
17B.3S-
158.266
Surrogate LMU
(PPM
2O07
23
IS
1C
1C
16
24
13
19
f.
:i
19
1C
31
13
12
3O
22
20O9
17
IS
8
8
11
21
24
17
16
22
11
15
22
12
14
30
2 a-all Beta distributions assumed to be normal
3 b-Gent et al., 2003 also uses the following prevalence rates: 0.028 (wheeze), 0.015 (shortness of breath), 0.012 (chest tightness) (from study)
7A-5
-------
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
APPENDIX 8A.
SUPPLEMENT TO THE REPRESENTATIVENESS ANALYSIS OF
THE 12 URBAN STUDY AREAS
Following the analysis discussed in Chapter 8, this appendix provides graphical
comparisons of the empirical distributions of components of the risk function, and additional
variables that have been identified as potentially influencing the risk associated with ozone
exposures. In each graph, the blue line represents the cumulative distribution function (CDF) for
the complete set of data available for the variable. In some cases, this many encompass all
counties in the U.S., while in others it may be based on a subset of the U.S., usually for large
urban areas. The black squares at the bottom of each graph represent the specific value of the
variable for one of the case study locations, with the line showing where that value intersect the
CDF of the nationwide data.
8-A.l. ELEMENTS OF THE RISK EQUATION
Comparison of Urban Case Study Area with U.S. Distribution (3143 U.S.
Counties) - Population
Urban case study
areas are all above
the 90th percentile of
county populations
100
1000
10000 100000
Population, 2008
1000000
10000000
16
•All Counties CDF
Case Study Counties
8A-1
-------
Figure 8-0.1 Comparison of distributions for key elements of the risk equation: Total
population
4
5
6
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Percent Younger than 15 Years Old
14
16
18 20 22 24 26
% Younger than 15 Years Old, 2005
•All Counties CDF
Case Study Counties
28
30
Figure 8-0.2 Comparison of distributions for key elements of the risk equation: Percent of
population younger than 15 years old
8A-2
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Percent 65 Years and Older
Urban case study
counties are all below
the 60th percentile of
county % of population
65 years and older
11 13 15 17 19
% 65 Years and Older, 2005
21 23
•All Counties CDF
Case Study Counties
25 27
Figure 8-0.3 Comparison of distributions for key elements of the risk equation: Percent of
population 65 and older
8A-3
-------
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Percent 85 Years and Older
100%
90%
80%
S 70%
§ 60%
o
u
^ 50%
^ 40%
* 30%
20%
10%
0%
0.5
Urban case study
areas are all below
the 75th percentile
of % of population
85 years and older
1.5
2 2.5 3 3.5
% 85 Years and Older, 2005
4.5
5.5
•All Counties CDF
Case Study Counties
2 Figure 8-0.4 Comparison of distributions for key elements of the risk equation: Percent of
3 population 85 and older
8A-4
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (671 U.S.
Counties with Ozone Monitors) -
100%
90%
80%
I 70%
o 60%
"S 50%
o
'E 40%
o
J 30%
S? 20%
10%
^^^^^ __i
>e
/
Ul
a:
/
i— i
•on
/
1 1
ai i\
/
ii — i
n
Ul
eai
• — i
i o-
y
/
i 1
nr u
f
\ 1
ai
• — i
y
u
ivia
^
f
\ — i
X
ii
uzone
^, — '
S*
\ , 1
^^^^^^^^
1 .
30 40 50 60
Seasonal Mean 8-hr Daily Max Ozone Concentration, Average 2006-2008
(ppb)
70
•All Counties CDF
Case Study Counties
Figure 8-0.5 Comparison of distributions for key elements of the risk equation: Seasonal
mean 8-hr daily maximum ozone concentration
8A-5
-------
Comparison of Urban Case Study Area with U.S. Distribution (725 U.S.
Counties with Ozone Monitors) -
4th High 8-hr Daily Maximum Ozone
40
50 60 70 80 90 100
4th High 8-hr Daily Maximum Ozone, 2007 (ppb)
110
•All Counties CDF
Case Study Counties
2 Figure 8-0.6 Comparison of distributions for key elements of the risk equation: 4 highest
3 8-hr daily maximum ozone concentration
8A-6
-------
Comparison of Urban Case Study Area with U.S. Distribution (671 U.S.
Counties with Ozone Monitors) -
Seasonal Mean 1-hr Daily Max Ozone
1/1
Q)
C
3
O
u
1
o
'E
o
°
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
^
1
x
1 1
II II
40 50 60 70 80
Seasonal Mean 1-hr Daily Max Ozone Concentration, Average 2006-2008
(ppb)
•All Counties CDF
Case Study Counties
2 Figure 8-0.7 Comparison of distributions for key elements of the risk equation: Seasonal
3 mean 1-hr daily maximum ozone concentration
8A-7
-------
Comparison of Urban Case Study Area with U.S. Distribution (671 U.S.
Counties with Ozone Monitors) -
Seasonal Mean Ozone
20 30 40 50
Seasonal Mean Ozone Concentration, Average 2006-2008 (ppb)
60
•All Counties CDF
Case Study Counties
2 Figure 8-0.8 Comparison of distributions for key elements of the risk equation: Seasonal
3 mean ozone concentration
8A-8
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (3137 U.S.
Counties) - All Cause Mortality
Urban case
study areas are
all below the
85th percentile
of county all
cause mortality
500
600 700 800 900 1000 1100 1200 1300 1400
All Cause Mortality per 100,000 Population, 1999-2005
1500
•All Counties CDF
Case Study Counties
Figure 8-0.9 Comparison of distributions for key elements of the risk equation: Baseline
all-cause mortality
8A-9
-------
1
2
3
Comparison of Urban Case Study Area with U.S. Distribution (3135 U.S.
Counties) - Non Accidental Mortality
300
Urban case study
counties are all
below the 80th
percentile of non
accidental
mortality
500 700 900 1100 1300
Non Accidental Mortality per 100,000 Population, 1999-2005
1500
•All Counties CDF
Case Study Counties
Figure 8-0.10
Comparison of distributions for key elements of the risk equation: Baseline
non-accidental mortality
8 A-10
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (3110 U.S.
Counties) - Cardiovascular Mortality
100
200
300
400
500
600
Cardiovascular Mortality per 100,000 Population, 1999-2005
•All Counties CDF
Case Study Counties
Figure 8-0.11 Comparison of distributions for key elements of the risk equation: Baseline
cardiovascular mortality
8 A-11
-------
1
2
3
4
5
6
Comparison of Urban Case Study Area with U.S. Distribution (2993 U.S.
Counties) - Respiratory Mortality
100%
90%
80%
S 70%
§ 60%
o
!x 50%
^ 40%
* 30%
20%
10%
0%
«-
20
^^
if* • • • • Jp
Urban case study
counties are all below
the 50th percentile of
respiratory mortality
40 60 80 100 120 140
Respiratory Mortality per 100,000 Population, 1999-2005
•All Counties CDF
Case Study Counties
160
Figure 8-0.12 Comparison of distributions for key elements of the risk equation: Baseline
respiratory mortality
8 A-12
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (95
NMMAPS Cities) -
Non Accidental Mortality Risk ((J)
0%
0.0002
0.0004 0.0006 0.0008 0.001
Non Accidental Mortality Risk Coefficient ((J)
0.0012
•All Cities CDF
Case Study Cities
Figure 8-0.13 Comparison of distributions for key elements of the risk equation: Non-
accidental mortality risk coefficient from Bell et al. (2004)
8 A-13
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (48 Z&S
Cities) - All Cause Mortality Risk (p)
0)
.a
o
c
TO
M
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
0.0002
0.0004 0.0006 0.0008 0.001
All Cause Mortality Risk Coefficient (3)
0.0012
All cities
•All Cities CDF
Case Study Cities
Figure 8-0.14 Comparison of distributions for key elements of the risk equation: All-cause
mortality risk coefficient from Zanobetti and Schwartz (2008)
8 A-14
-------
1
2
3
4
Comparison of Urban Case Study Area with U.S. Distribution (48 Z&S
Cities) - Cardiovascular Mortality Risk Coefficient (3)
1/1
0)
s
u
co"
8
2.
N
CD
s
£
ro
•E
0)
.a
o
c
(D
M
'o
*
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
*
0.0002
0.0005 0.0008 0.0011
Cardiovascular Mortality Risk Coefficient
0.0014
0.0017
All Cities
•All Cities CDF
Case Study Cities
Figure 8-0.15 Comparison of distributions for key elements of the risk equation:
Cardiovascular mortality risk coefficient from Zanobetti and Schwartz
(2008)
8 A-15
-------
Comparison of Urban Case Study Area with U.S. Distribution (48 Z&S
Cities) - Respiratory Mortality Risk ((J)
0%
0.0003
0.0005 0.0007 0.0009 0.0011
Respiratory Mortality Risk Coefficient ((J)
0.0013
All Cities
•All Cities CDF
Case Study Cities
2 Figure 8-0.16 Comparison of distributions for key elements of the risk equation:
3 Respiratory mortality risk coefficient from Zanobetti and Schwartz (2008)
8 A-16
-------
1
2
4
5
6
8-A.2. VARIABLES EXPECTED TO INFLUENCE THE RELATIVE RISK
FROM OZONE
i. Demographic Variables
Comparison of Urban Case Study Area with U.S. Distribution (3143 U.S.
Counties) - Population Density
10 100 1000 10000
Population Per Square Mile, 2008
100000
1000000
•All Counties CDF
Case Study Counties
Figure 8-0.17 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Population density
8 A-17
-------
100%
90%
80%
I 70%
c
o 60%
3 50%
"S 40%
S?
30%
20%
10%
0%
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Median Age
32 34 36 38
40 42 44 46
Median Age, 2005
48 50
52 54
•All Counties CDF
Case Study Counties
2 Figure 8-0.18 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Median age
8 A-18
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) -
Percent Less than High School Education
J.UU/O
90%
80%
.2 70%
o 60% -J
u
"! 50%
'o 40%
30% -
20%
10%
^^
I
1 — ^-1 1
/
• 1 1
/
U II
>
f
• Ul
r
• — ~-i
/
in
/
i — j
/
r
i 1
/
r
i— i
Ul
X
x^
i , 1
i , , , , , ,
10 15 20 25 30 35 40 45 50 55 60 65
% Less than High School Education, 2000
•All Counties CDF
Case Study Counties
Figure 8-0.19 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Percent less than high school education
8 A-19
-------
1
2
3
4
5
6
7
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Unemployment rate
6 8
Unemployment rate, 2005
10
•All Counties CDF
Case Study Counties
12
Figure 8-0.20 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Unemployment rate
8A-20
-------
1/1
Q)
C
3
O
u
100%
90%
80%
70%
60%
50%
40%
30%
20%
10%
0%
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Percent Non-White
Urban
study
areas are
all above
*—%
20 30 40
Percent Non-White, 2005
50
60
70
•All Counties CDF
Case Study Counties
2 Figure 8-0.21 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Percent non-white
8A-21
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (3141 U.S.
Counties) - Urbanicity
4 5
Urbanicity, 2003
•All Counties CDF
Case Study Counties
Figure 8-0.22 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Urbanicity
8A-22
-------
Comparison of Urban Case Study Area with U.S. Distribution (76 Cities) -
Air Conditioning Prevalence
10
20
Urban study areas
are all below the
90th percentile of
percent of
residences with no
air conditioning
30 40 50 60 70
No air conditioning, 2004 (%)
80
•All Cities CDF
Case Study Cities
90 100
2 Figure 8-0.23 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Air conditioning prevalence
8A-23
-------
1
2
3
4
5
6
Comparison of Urban Case Study Area with U.S. Distribution (366 U.S.
Cities) - Public Transportation Use
10 15 20 25 30
% Commuting by Public Transportation, 2010
•All Counties CDF
Case Study Counties
35
40
Figure 8-0.24 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Percent commuting by public transportation
n.
Health Conditions
8A-24
-------
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) -
Acute Myocardial Infarction
234567
Acute Myocardial Infarction Prevalence, 2007 (%)
•All Counties CDF
Case Study Counties
2 Figure 8-0.25 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Acute myocardial infarction prevalence
8A-25
-------
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) - Diabetes
8 9 10 11
Diabetes Prevalence, 2007 (%)
12
•All Counties CDF
Case Study Counties
13
14
2 Figure 8-0.26 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Diabetes prevalence
8A-26
-------
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) - Stroke
8 60%
1.5
2.5 3 3.5
Stroke Prevalence, 2007 (%)
•All Counties CDF
Case Study Counties
4.5
2 Figure 8-0.27 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Stroke prevalence
8A-27
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) - Coronary Heart Disease
45678
Coronary Heart Disease Prevalence, 2007 (%)
10
•All Counties CDF
Case Study Counties
Figure 8-0.28 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Coronary heart disease prevalence
8A-28
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (182 BFRSS
Cities)- Obesity
15
20 25 30
Obesity Prevalence, 2007 (%)
35
40
•All Counties CDF
Case Study Counties
Figure 8-0.29 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Obesity prevalence
8A-29
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (183 BFRSS
Cities) - Vigorous Activity 20min
15
20 25 30 35
Vigorous Activity 20 minutes per day, 2007 (%)
40
•All Counties CDF
Case Study Counties
Figure 8-0.30 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Vigorous activity at least 20 minutes per day
8A-30
-------
1
2
3
4
5
6
7
Comparison of Urban Case Study Area with U.S. Distribution (182 BFRSS
Cities) - Moderate Activity SOmin or Vigorous Activity 20min
35 40 45 50 55 60 65
Moderate Activity 30 minutes per day or Vigorous Activity 20 minutes per
day, 2007 (%)
•All Counties CDF
Case Study Counties
Figure 8-0.31 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Moderate activity at least 30 minutes per day or
vigorous activity at least 20 minutes per day
8A-31
-------
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) - Asthma Prevalence
8 10
Asthma Prevaldence, 2007(%)
12
•All Counties CDF
Case Study Counties
14
2 Figure 8-0.32 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Asthma prevalence
8A-32
-------
Comparison of Urban Case Study Area with U.S. Distribution (184 BFRSS
Cities) - Ever Smoked
10
15
20 25
Ever Smoked, 2007(%)
30
•All Counties CDF
Case Study Counties
35
2 Figure 8-0.33 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Smoking prevalence
8A-33
-------
2
3
4
1/1
111.
Air Quality and Climate Variables
Comparison of Urban Case Study Area with U.S. Distribution (617
U.S.Counties with PM2 5 Monitors) - Annual Average PM2 5
10 12 14 16 18 20
Annual Average PM2 5, 2007 (u,g/m3)
•All Counties CDF
Case Study Counties
22
24
Figure 8-0.34 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Annual average PMi.s concentration
8A-34
-------
1
2
Comparison of Urban Case Study Area with U.S. Distribution (617 U.S.
Counties with PM2 5 Monitors) -
98th Percentile PM,,-
10
20 30 40 50 60
98th Percentile PM2 5, 2007 (u.g/m3)
•All Counties CDF
Case Study Counties
70
80
Figure 8-0.35 Comparison of distributions for selected variables expected to influence the
r>th
relative risk from ozone: 98 percentile PMi.s concentration
8A-35
-------
Comparison of Urban Case Study Area with U.S. Distribution (204 U.S.
Counties in MCAPS Database) -
Percent Days with PM2 5 Exceeding 35 u.g/m3
5 10 15
% days with PM2 5 exceeding 35 u,g/m3,1999-2002
•All Counties CDF
Case Study Counties
20
2 Figure 8-0.36 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Percent of days with PM2.s exceeding 35 ug/m3
8A-36
-------
1
2
3
4
5
Comparison of Urban Case Study Area with U.S. Distribution (202 U.S.
Counties in MCAPS Database) - Average Temperature
40
50 60
Average Temperature (°F)
70
80
•All Counties CDF
Case Study Counties
Figure 8-0.37 Comparison of distributions for selected variables expected to influence the
relative risk from ozone: Average temperature
8A-37
-------
Comparison of Urban Case Study Area with U.S. Distribution (All U.S.
Counties) - July Temperature
66
68
70 72 74 76 78 80 82
Mean Temperature for July, 1941-1970 (°F)
84
86
•All Counties CDF
Case Study Counties
2 Figure 8-0.38 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: July temperature
8A-38
-------
Comparison of Urban Case Study Area with U.S. Distribution (All U.S.
Counties) - July Humidity
20
30 40 50 60
Relative Humidity for July, 1941-1970
70
80
•All Counties CDF
Case Study Counties
2 Figure 8-0.39 Comparison of distributions for selected variables expected to influence the
3 relative risk from ozone: Relative humidity
8A-39
-------
United States Office of Air Quality Planning and Standards Publication No. EPA 452/P-12-001
Environmental Protection Air Quality Strategies and Standards Division July 2012
Agency Research Triangle Park, NC
------- |