&EPA
United States
Envirofunmlal Protection
Agnncy
Health Risk and Exposure Assessment for
Ozone
Final Report
Executive Summary
-------
This page left intentionally blank
-------
EPA-452/R-14-004f
August 2014
Health Risk and Exposure Assessment for Ozone
Final Report
Executive Summary
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
-------
DISCLAIMER
This final document has been prepared by staff from the Risk and Benefits Group, Health and
Environmental Impacts Division, Office of Air Quality Planning and Standards, U.S. Environmental
Protection Agency. Any findings and conclusions are those of the authors and do not necessarily reflect
the views of the Agency.
Questions related to this document should be addressed to Dr. Bryan Hubbell, U.S.
Environmental Protection Agency, Office of Air Quality Planning and Standards, C539-07, Research
Triangle Park, North Carolina 27711 (email: hubbell.bryan@epa.gov).
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
Introduction
A
s part of the review of the ozone (Os)
National Ambient Air Quality Standards
(NAAQS), EPA has prepared this Health
Risk and Exposure Assessment (HREA) to
provide estimates of exposures to Os and
resulting mortality and morbidity health risks.
The health effects evaluated in this HREA
are based on the findings of the Os ISA (U.S.
EPA, 2013) that short term Os exposures are
causally related to respiratory effects, and
likely causally related to cardiovascular
effects, and that long term Os exposures are
likely causally related to respiratory effects.
The assessment evaluated total exposures
and risks associated with the full range of
observed Os concentrations. In addition,
the HREA estimated the incremental
changes in exposures and risks associated
with ambient air quality adjusted to just
meeting the existing standard of 75 ppb
and just meeting potential alternative
standard levels of 70, 65, and 60 ppb using
the form and averaging time of the existing
standard, which is the annual 4th highest
daily maximum 8-hour Os concentration,
averaged over three consecutive years.
The results of the HREA are developed to
inform the Os Policy Assessment (PA) in
considering the adequacy of the existing Os
standards, and potential risk reductions
associated with potential alternative levels
of the standard. For added context
regarding existing Os air quality and the
potential impact to public health, initial
nonattainment area designations have
been made for 46 areas in the U.S. with
ambient Os concentrations exceeding the
existing standard (77 FR 30160). The figure
below provides the locations of
nonattainment areas and their respective
classifications and includes 227 counties
with an estimated 2010 population of just
over 123 million people.
As described in the conceptual framework
and scope in Chapters 2 and 3,
respectively, the HREA discusses air quality
considerations (Chapter 4) and evaluates
exposures and lung function risk in 15 urban
study areas (Chapters 5 and 6, respectively)
and risks based on application of results of
8-Hour Ozone Nonattainment Areas (2008 Standard)
Nonattainment areas are indicated by color.
When only 3 portion of a county is shown in
color.it indicates that only that part of the
county is within a nan attainment area boundary.
Nonattainment area ctossfficafions based on the existing Q? NAAQS.
(Source: http://www.epa .gov/airquaiily/greenboc*/}.
8-hour Ozone Classification
Ertreme
CH Severe 15
I I Serious
C^ Moderate
I IMardnal
ES-1
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
epidemiology studies in a subset of 12 urban
study areas (Chapter 7). In addition, to
place the urban study area analyses in a
broader context, the assessment estimated
the national burden of mortality associated
with recent Os levels, and evaluated the
representativeness of the urban areas in
characterizing Os exposures and risks across
the U.S. (Chapters). To further facilitate
interpretation of the results of the exposure
and risk assessment, Chapter? provides a
synthesis of the various results, focusing on
comparing and contrasting those results to
identify common patterns, or important
differences. It also includes an overall
integrated characterization of exposure and
risk in the context of key policy relevant
questions.
Conceptual Framework
and Scope
The HREA provides information to answer
key policy-relevant risk questions with
regards to evaluation of the adequacy
of the existing standards and evaluation of
potential alternative standards such as:
"To whaf extent do risk and/or exposure
analyses suggest that exposures of concern
for Os-re/afed health effects are likely to
occur with existing ambient levels of Os or
with levels that just meet the Os standard?
To what extent do alternative standards,
taking together levels, averaging times and
forms, reduce estimated exposures and risks
of concern attributable to Os and ofher
photochemical oxidants, and what are the
uncertainties associated with the estimated
exposure and risk reductions?"
In answering these key questions, the HREA
evaluates total exposures and risks
associated with the full range of observed
Os concentrations, as well as the
incremental changes in exposures and risks
for just meeting the existing standard and
just meeting several alternative standards.
With regard to selecting alternative levels for
the 8-hour Os standards for evaluation in the
quantitative risk assessment, we base the
range of selected levels on the evaluations
of the evidence provided in the first draft
PA, which received support from the CASAC
in their advisory letter on the first draft PA.
The first draft PA recommended evaluation
of 8-hour maximum concentrations in the
range of 60 to 70 ppb, with possible
consideration of levels somewhat below 60
ppb.
Ozone concentrations from 2006-2010 are
used in estimating exposures and risks for the
15 urban study areas. Because of the year-
to-year variability in Os concentrations, the
assessment evaluates air quality scenarios
for just meeting the existing and potential
alternative standards based on multiple
years of Os data to better capture the high
degree of variability in meteorological
conditions, as well as reflecting years with
higher and lower emissions of Os precursors.
The 15 urban study areas were selected to
be generally representative of U.S.
populations, geographic areas, climates,
and varying Os and co-pollutant levels.
These urban study areas include Atlanta,
GA; Baltimore, MD; Boston, MA; Chicago, IL;
Cleveland, OH; Dallas, TX; Denver, CO;
Detroit, Ml; Houston, TX; Los Angeles, CA;
New York, NY; Philadelphia, PA;
Sacramento, CA; St. Louis, MO; and
Washington, D.C.
We have identified the following goals for
the urban area exposure and risk
assessments: (1) to provide estimates of the
number and percent of people in the
general population and in at-risk
populations and lifestages with Os exposures
above health-based benchmark levels; (2)
to provide estimates of the number and
percent of people in the general population
and in at-risk populations and lifestages with
impaired lung function (defined based on
decrements in forced expiratory volume in
one second (FEVi) resulting from exposures
to Os; (3) to provide estimates of the
potential magnitude of premature mortality
ES-2
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
Policy Relevant Exposure and
Risk Questions
(Chapter 2)
Exposure Assessment
APEX
Urban Scale
Assessment of
Individual Exposure
(Chapter 5)
Urban Scale Risk
Analyses Based on
Application of Results
from Controlled
Human Exposure
Studies
(Chapter 6)
AQS, VNA,
CMAQ-HDDM
Air Quality Characterization
(Chapter 4)
Review of Health Evidence
(Chapter 2)
Risk Assessment
BenMAP
Urban Scale Risk
Analyses Based on
Application of Results
from Epidemiological
Studies
(Chapter 7)
Risk Characterization
(Chapter 9)
National Scale Risk
Burden Based on
Application of Results
from Epidemiological
Studies
(Chapter 8)
associated with both short-term and long-
term Os exposures, and selected morbidity
health effects associated with short-term Os
exposures; (4) to evaluate the influence of
various inputs and assumptions on risk
estimates to the extent possible given
available methods and data; (5) to gain
insights into the spatial and temporal
distribution of risks associated with Os
concentrations just meeting existing and
alternative standards, patterns of risk
reduction associated with meeting
alternative standards relative to the existing
standard, and uncertainties in the estimates
of risk and risk reductions.
In working towards these goals, we follow a
conceptual framework, shown in the figure
above , comprised of an air quality
characterization, a review of relevant
scientific evidence on health effects, the
1 The CMAQ model and associated
documentation is available for download at
https://www.cmascenter.org/cmaq/.
2 The APEX model and associated
documentation is available for download at
http://www.epa.gov/ttn/fera/human_apex.html
modeling of exposure, the modeling of risk,
and a risk characterization. As shown in this
framework, air quality is characterized
primarily by the combined use of ambient
monitoring data available in the EPA Air
Quality System (AQS), and a spatial
interpolation approach (Voronoi Neighbor
Averaging, VNA), along with Higher-Order
Decoupled Direct Method (HDDM)
capabilities in the Community Multi-scale Air
Quality (CMAQ)1 model. The modeling of
personal exposure and estimation of risks,
which rely on personal exposure estimates,
are implemented using the EPA's Air
Pollution Exposure model (APEX)2. Modeling
of population level risks for health endpoints
based on application of results of
epidemiological studies is implemented
using the environmental Benefits Mapping
and Analysis Program (BenMAP)3, a peer
reviewed software tool for estimating risks
3 The BenMAP model and associated
documentation is available for download at
http://www.epa.gov/air/benmap/
ES-:
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
and impacts associated with changes in
ambient air quality. The overall
characterization of risk draws from the results
of the exposure assessment and both types
of risk assessment.
Air Quality Characterization
In this analysis, we employed a
photochemical model-based adjustment
methodology (Simon et al., 2013) to
estimate the change in observed hourly Os
concentrations at a given set of monitoring
sites resulting from across-the-board
reductions in U.S. anthropogenic NOx and/or
VOC emissions. This information was then
used to adjust recent Os concentrations
(2006-2010) in the 15 urban study areas to
reflect just meeting the existing 8-hour Os
standard of 75 ppb and for just meeting
potential alternative standard levels of 70,
65, and 60 ppb. Because the form of the
existing Os standard is based on the 3-year
average of the 4th highest daily maximum 8-
hour average, we simulate just meeting the
potential alternative standards for two
periods, 2006-2008 and 2008-2010.
The use of the model-based adjustment
methodology is an example of how we
have brought improvements into this review
that better represent current scientific
understanding. The model-based
adjustment methodology represents a
substantial improvement over the quadratic
rollback method used to adjust Os
concentrations in past reviews. For
example, while the quadratic rollback was a
purely mathematical technique which
attempted to reproduce the distribution of
observed Os concentrations just meeting
various standards, the new methodology
uses photochemical modeling to simulate
the response in Os concentrations due to
changes in precursor emissions based on
current understanding of atmospheric
chemistry and pollutant transport. Second,
quadratic rollback used the same
mathematical formula to adjust
concentrations at all monitors within each
urban study area for all hours, while model-
based adjustment methodology allows the
adjustments to vary both spatially across
each study area and temporally across
hours of the day and across seasons. Finally,
quadratic rollback was designed to only
allow decreases in Os concentrations, while
the model-based adjustment methodology
allows both increases and decreases in Os
concentrations, which more accurately
reflects the scientific understanding that
increases in Os concentrations may occur in
response to reductions in NOx emissions in
some situations, such as in urban areas with
a large amount of NOx emissions.
Several general trends are evident in the
changes in Os patterns across the urban
study areas and across the different
standards under consideration. In all 15
urban study areas, peak Os concentrations
tended to decrease while the lowest Os
concentrations tended to increase as the
concentrations were adjusted to just meet
the existing and potential alternative
standards. In addition, high and mid-range
Os concentrations generally decreased in
rural and suburban portions of the case
study areas, while the Os response to NOx
reductions was more varied within urban
core areas. In particular, while the annual
4th highest daily maximum 8-hour
concentrations generally decreased in the
urban core of the study areas in response to
reductions in NOx emissions, the seasonal
mean of the daily maximum 8-hour Os
concentrations did not change significantly,
though it did exhibit some increases or
decreases in the various study areas as the
distribution of Os was further adjusted to
meet lower potential alternative standards.
The adjustments to Os to reflect just meeting
existing and potential alternative standards
are made by decreasing only U.S. emissions
of anthropogenic NOx primarily, and in a
few instance, both NOx and VOC
reductions. As such, the estimated changes
in exposure and risk, based on these air
quality changes, are solely attributable to
changes in U.S. emissions and are not meant
to reflect a specific air emission control
ES-4
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
mary
strategy that might be used by a state or
urban area to meet a standard.
Human Exposure Modeling
he population exposure assessment
evaluated exposures to Os using the
APEX exposure model which uses
time-activity diary and anthropometric
data coupled with local meteorology,
population demographics, and Os
concentrations to estimate the number
and percent of study group individuals
above exposure benchmarks. The
analyses examined exposure to Os for the
general population, all school-aged
children (ages 5-18), asthmatic school-
aged children, asthmatic adults (ages >
18), and older persons (ages 65 and
older), with a focus on when exposed
individuals were engaged in moderate or
greater exertion, for example, children
engaged in outdoor recreational
activities. Exposure is assessed in the 15
urban study areas for recent Os (2006-
2010) and for Os adjusted to just meet
existing and potential alternative
standards for two averaging periods
(2006-2008 and 2008-2010). The analysis
provided estimates of the number and
percent of several study groups of interest
exposed to concentrations above three
health-relevant 8-hour average Os
exposure benchmarks: 60, 70, and 80 ppb.
These benchmarks were selected to
provide perspective on the public health
impacts of Os-related health effects that
have been demonstrated in human
clinical and toxicological studies, but
cannot currently be evaluated in
quantitative risk assessments, such as lung
inflammation and increased airway
responsiveness. The Os ISA includes
studies showing significant effects at each
of these benchmark levels.
The analysis found that children are the
study group of greatest concern for Os
exposures due to the greater amount of
afternoon time they spend outdoors
engaged in moderate or higher exertion
activities and that they do so more
frequently of any of the at-risk study groups.
Based on this, we focus on the results for
children in this subsequent discussion. The
two figures below show the average across
2006-2010 of the percentage of school-
aged children experiencing 8-hour exposure
greater than 60 ppb at least once (top)
Altai
Eafimoi*
Eoston.
Chtago
Ckrretal
Dallas
Doirer
Detroi.
HewYoA
IMailphii
Saaametto
St Louis
1
1
1 1
1
1 1
1 1
1 1
1 1
1 1
1 1
1
1
1 1
1 1
1 1
OK 2K 4% 6% 8% 10K, 12K. 14% IfiK 18%
Beaxertof AUSchool-A^ ChJBrenwfiiatLsastQtieStirDaij'MaTt Ei5)ojii«>= 60ppb
stardirdtiTel(ppti) i 160 i IDJ i iTO i 1 75
Altai
Chicago
Ctrretal
Dallas
Domer
Los Angles
IfevYoA
Sunmato
SL Louis
, 1% ?/, ?/. 4% 5*/» 6K, 7% 8%
Percotof AUSchool-Ag ChiJravwih at LsistT™ 8-hr Daily Max Exposure >= SJppt
startfani tirel (ppt) i 160 i ifij i iTO i i7J
Average percent of all school-age children exposed of or
above 60 ppb at least once (tap) and at least twice
fboffornj per O3season for each urban study area across all
study years (2006-2070,1 considering each standard level.
Note: A stafKJord level oS 60 ppt> for me New Yorfc sfudy areo was nof
modeted. We do nof krrow whtf) fhe percen) risk would be lor NV under (he
60 ppb atfemah've sfondord but it woa/d nof necessar^x be zero OS
indicated by the figure.
ES-5
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
and at least twice (bottom) per Os season.
Based on this information, no more than 17
percent of any study group in any study
area, on average, was exposed at least
once at or above the 60 ppb benchmark,
when meeting the existing standard. When
meeting a standard level of 70 ppb, less
than 11 percent of any study group in any
study area, on average, was exposed at
least once at or above the 60 ppb exposure
benchmark. Adjusting ambient Os to just
meet a standard level of 65 ppb is
estimated to reduce the percent of persons
at or above an exposure benchmark of 60
ppb to 4 percent or less of any study group
and study area, on average.
For the exposure benchmark of 70 ppb, on
average less than 4 percent of any study
group, including all school-age children, in
any study area, was exposed at least once
at or above the exposure benchmark when
meeting the existing standard (not shown).
For the highest exposure benchmark of 80
ppb, on average less than 1 percent of any
study group in any study area was exposed
at least once at or above the exposure
benchmark when meeting the existing
standard (not shown). As expected, with
the lower ambient Os levels associated with
just meeting lower alternative standard
levels, the percentages of at-risk study
groups experiencing exposures above the
benchmark levels are smaller than when just
meeting the existing standard.
In considering two or more exceedances of
the 60 ppb benchmark, on average less
than 8 percent of any study group in any
study area experience such 8-hour
exposures when air quality is adjusted to
meet the existing standard. There were no
persons estimated to experience any multi-
day exposures above the exposure
benchmark of 80 ppb for any study group in
any study area, while less than 1 percent of
persons were estimated to experience two
or more 8-hour exposures at or above 70
ppb, when meeting the existing standard or
any of the alternative standard levels (not
shown).
In addition, the exposure assessment also
identified the specific microenvironments
and activities that contribute most to
exposure and evaluated at what times and
for how long individuals were in key
microenvironments and were engaged in
key activities, with a focus on persons
experiencing the highest daily maximum 8-
hour exposure within each study area. That
analysis indicated that children are an
important exposure study group, largely as a
result of the combination of large amounts
of afternoon time spent outdoors and their
engagement in moderate or high exertion
level activities. Highly exposed children, on
average, spend half of their outdoor time
engaged in moderate or greater exertion
levels, such as participating in sporting
activities. In addition, any people spending
a large portion of their time outdoors during
afternoon hours experienced the highest 8-
hour Os exposure concentrations given that
ambient Os concentrations are typically
highest during this time of day and other
microenvironments evaluated, particularly
indoor microenvironments, have much
lower Os concentrations than ambient
concentrations. Simulations of highly
exposed adults indicated that they also
spent large amounts of afternoon time
outdoors engaged in moderate or greater
exertion level activities though on average,
not participating in these events as
frequently as children.
Health Risks Based on
Controlled Human Exposure
Studies
T
his analysis uses the estimated Os
exposures from APEX, combined with
results from controlled human exposure
studies, to estimate the number and
percent of at-risk study groups (all children,
children with asthma, adults aged 18-35,
adults aged 36-55, and outdoor workers)
experiencing selected decrements in lung
function. The analysis focuses on estimates
of the percent of each at-risk population
ES-6
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
mary
experiencing a reduction in lung function for
three different health effect levels: 10, 15,
and 20 percent decrements in FEVi. These
health effect levels were selected based on
the published literature and conclusions
drawn regarding the adversity associated
with increasing lung function decrements
(O3 ISA, Section 6.2.1.1; Henderson, 2006).
Lung function decrements of 10 percent
and 15 percent in FEVi are considered
moderate decrements; 10 percent is
considered potentially adverse for people
with lung disease, while a 15 percent is
potentially adverse for active healthy
people. A 20 percent decrement in FEVi is
considered a large decrement that is
potentially adverse for healthy people and
can potentially cause more serious effects in
people with lung disease.
Two types of FEVi risk were used to estimate
lung function risks. One model was based
on application of a population level
exposure-response (E-R) function consistent
with the approach used in the previous Os
review. The second model was based on
application of an individual level risk
function (the McDonnell-Stewart-Smith
(MSS) model), newly introduced for this
review. The main difference between the
two models is that the MSS model includes
responses associated with a wider range of
exposure protocols used in the original
controlled human exposure studies (i.e.,
variable levels of exertion, lengths of
exposures, and patterns of exposure
concentrations) than compared to the
exposure-response model of previous
reviews. The models are similar in that both
models have a logistic form and are less
sensitive to changes at very low
concentrations of Os than to higher Os
concentrations. As a result, the models
estimate very few FEVi responses > 10%
when ambient concentrations are below
20 ppb and very few FEVi responses > 15%
when ambient concentrations are below
40 ppb. Because the individual level E-R
function approach allows for a more
complete estimate of risk, we focus on the
results of the MSS model for this discussion.
Lung function risks were estimated for each
of the 15 urban case study areas for recent
air quality conditions (2006-2010) and for air
quality adjusted to just meet the existing
and alternative standards for two design
value periods (2006-2008 and 2008-2010). As
with the exposure assessment, we focus on
lung function decrements in children as they
are the study group likely to have the
greatest percentage of that group at risk
due to higher levels of exposure and greater
levels of exertion. The figure below shows
the lung function risks associated with just
meeting the existing and potential
alternative standard levels, where risk is
taken to be the average value for each
study area (across all years considered) of
the percent of school-aged children with
FEVi decrement of 10 percent or greater.
This figure shows that there are decreases in
incremental risk for all 15 cities in the
progression from the level of the existing
standard, 75 ppb to the alternative
standard levels of 70, 65, and 60 ppb. The
estimated risks in this figure for Washington,
DC, for example, are about 16 percent for
Alarta
EdtiHOre
Boston
Chicago
Ctislmi
Mis
DOUBT
Detroit
Hjuston
Los Aigplss
HnvYori
HnUatfchtt
Ssirnittto
StLcui
Ufr'-.ViVi^riVi
1 1 1
1 1
1 1
1 1 1
1 1
1 1 1
1 III
1 1 1
1 1 1
1 1 1 1
1 1
1 1
1 1 1
1 1 1
1 1
0% 2% 4% (5% 8*/> 10% 12% 14% 16% 18% 20%
percaitof sdiool-a^dclvildmwlh IEV1 'icremat > 10%
staiiiirdl™el(jiib) i igi i i6J i 170 i i 75
Average percent of all school-age children with at least one
FEV\ decrement £ 10 percent per Os season in each urban
study area across at! study years (2006-20 W) considering
each standard level.
Note; A standard level ot 60 ppb for tfiewew York sfudy area was nof
modeted. We do not know whc?> (he percent risk wouW be tot NY under fhe
60 ppb alternative standard, but it wog/d not necessar/V be zero OS
ES-7
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
the existing standard level of 75 ppb and
aboutl 3 percent for the alternative
standard level of 70 ppb. The length of the
brown bar is the incremental reduction (3
percent) in the percent of persons at risk,
when adjusting air quality from the existing
standard of 75 ppb to the 70 ppb
alternative standard. The pattern of
reductions for lung function decrements
larger than 15 and 20 percent are similar to
those illustrated here (not shown).
Health Risks Based on
Application of Results of
Epidemiological Studies
The epidemiology-based risk assessment
evaluated mortality and morbidity risks
from short-term exposures, as well as
mortality risks from long-term exposures to
Os, by applying concentration-response (C-
R) functions derived from epidemiology
studies. Most of the endpoints evaluated in
epidemiology studies are for the entire study
population. Because most mortality and
hospitalizations occur in older persons, the
risk estimates for this portion of the analysis
are thus more focused in adults rather than
children, and thus differ in focus compared
to the human exposure and lung function
risk assessments. The analysis included both
a set of urban area study areas and a
national-scale assessment.
The urban study area analyses evaluated
mortality and morbidity risks, including
emergency department (ED) visits, hospital
admissions (HA), and respiratory symptoms
associated with recent Os concentrations
(2006-2010) and with O3 concentrations
adjusted to just meet the existing and
alternative Os standards. Mortality and
hospital admissions were evaluated in 12
urban study areas (a subset of the 15 urban
study areas evaluated in the exposure and
lung function risk assessments), while ED visits
and respiratory symptoms were evaluated in
a subset of study areas with supporting
epidemiology studies. The 12 urban areas
were: Atlanta, GA; Baltimore, MD; Boston,
MA; Cleveland, OH; Denver, CO; Detroit, Ml;
Houston, TX; Los Angeles, CA; New York, NY;
Philadelphia, PA; Sacramento, CA; and St.
Louis, MO. The urban study analyses focus
on risk estimates for the middle year of each
three-year ambient standard simulation
period (2006-2008 and 2008-2010) in order to
provide estimates of risk for a year with
generally higher Os levels (2007) and a year
with generally lower Os levels (2009).
In previous Os NAAQS reviews, health risks
were estimated for the portion of total Os
attributable to North American
anthropogenic sources (referred to in
previous Os reviews as "policy relevant
background"). In contrast, this assessment
provides risk estimates for the urban study
areas for Os concentrations down to zero,
reflecting the lack of evidence for a
detectable threshold in the C-R functions
(Os ISA), and the understanding that U.S.
populations may experience health risks
associated with Os resulting from emissions
from all sources, both natural and
anthropogenic, and within and outside the
U.S.
The figure below shows the results of the
mortality (top panel) and respiratory hospital
admissions (bottom panel) risk assessments
for all 12 urban areas associated with short-
term exposure to Os, showing the effect on
the incidence per 100,000 population just
meeting the existing 75 ppb standard and
potential alternative Os standards of 70, 65,
and 60 ppb in 2007. The overall trend across
urban areas is relatively small decreases in
mortality and morbidity risk as air quality is
adjusted to just meet incrementally lower
standard levels. In New York, there are
somewhat greater decreases, reflecting the
relatively large emission reductions used to
adjust air quality to just meet the 65 ppb
alternative standard, and the substantial
change in the distribution of Os
concentrations that resulted. Risks vary
substantially across urban study areas;
however, the general pattern of risk
reductions associated with air quality
adjusted to just meet alternative standards is
ES-8
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
similar between urban study areas. Risks are
generally lower in 2009 (not shown) relative
to 2007; though the patterns of reductions
are very similar between the two years. On
average, compared with air quality
adjusted to just meet the existing standard,
mortality and respiratory hospitalization risks
decrease by 5% or less for where ambient
concentrations are adjusted to meet a
standard level of 70 ppb, 10% or less for
meeting a level of 65 ppb, and 15% or less
for meeting a level of 60 ppb. Larger risk
Trend in ozone-related mortality across standard
levels (deaths per 100,000)
Trend in ozone-related HA across standard levels
(HA per 100,000)
reductions are estimated on days with
higher Os.
We also evaluated mortality risks in the 12
urban study areas associated with long-term
Os exposures (based on the April to
September average of the daily maximum
one-hour ambient Os concentrations). The
figure below shows the results of the long-
term mortality risk assessment for all 12 urban
study areas, showing the effect on the
incidence per 100,000 population
considering air quality adjusted to
just meeting the existing standard
and potential alternative Os
standard levels of 70, 65, and 60
ppb in 2007. Risks from long-term
exposures after adjusting air
quality to just meet the existing
standard are substantially greater
than risks from short-term
exposures, ranging from 15 to 30
percent of respiratory mortality
across urban areas. However, the
percent reductions in risks are
similar to those for mortality from
short-term exposures, e.g., less
than 10 percent reduction in risk
relative to just meeting the existing
standard in most areas when just
meeting the 70 ppb and 65 ppb
alternative standards, and less
than 20 percent reductions when
just meeting the 60 ppb
alternative standard level.
Shorf-ferm mortality risk per 100,000 population (top) and
respiratory hasp/fa) admissions risk per 100,000 population
{bottom) (OF 2007 considering each standard level,
Note: A standard level of 60 ppb for the New Voi* study area was nof modeled.
We do not know what the percent risk woufd be for NV under the 60 ppb
oftemafiVe standard, but it woutd nof necessarily be itero as indicated t>>' the
figure.
ES-9
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
Trend in ozone-related mortality across standard
levels (deaths per 100,000)
5
l J-iU. GA
rel, O-
• ;i
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
surrounding those areas, leading to risk
reductions that are not captured by the
urban-scale analysis.
Representativeness of
Exposure and Risk Results
and Associated
Uncertainties
As part of this assessment, we conducted
several analyses to determine the
extent to which our selected urban
areas represent: (1) the highest mortality
and morbidity risk areas in the U.S.; and (2)
the types of patterns of Os air quality
changes that we estimate would be
experienced by the overall U.S. population
in response to emissions reductions that
would decrease peak Os concentrations to
meet the existing standard or lower
alternative Os standard levels.
We selected urban study areas for the
exposure and risk analyses based on criteria
that included Os levels, at-risk study groups,
and related factors that were designed to
ensure we captured areas and persons likely
to experience high Os exposures and risks.
Based on the comparisons of distributions of
risk characteristics, the selected urban study
areas represent urban areas that are
among the most populated in the U.S., have
relatively high peak Os levels, and capture a
wide range of city-specific mortality risk
effect estimates. The analyses found that
the Os mortality risk for short-term Os
exposures in the 12 urban study areas are
representative of the full distribution of U.S.
Os-related mortality, representing both high
end and low end risk counties. For the long-
term exposure related mortality risk metric,
the 12 urban study areas are representative
of the central portion of the distribution of
risks across all U.S. counties, however, the
selected 12 urban areas do not capture the
very highest (greater than 98th percentile) or
lowest (less than 25th percentile) ends of the
national distribution of long-term exposure-
related Os-related risk.
While we selected urban study areas to
represent those populations likely to
experience elevated risks from Os exposure,
we did not include amongst the selection
criteria the responsiveness of Os in the urban
study area to decreases in Os precursor
emissions that would be needed to just
meet existing or potential alternative
standards. The additional analyses we
conducted suggest that many of the urban
study areas may show Os responses that are
typical of other large urban areas in the U.S.,
but may not represent the response of Os in
other populated areas of the U.S. These
other areas, including suburban areas,
smaller urban areas, and rural areas, would
be more likely than our urban study areas to
experience area-wide average decreases
in mean Os concentrations and, therefore,
associated decreases in mortality and
morbidity risks, as Os standards are met.
Even though large urban areas have high
population density, the majority of the U.S.
population lives outside of these types of
urban core areas, and thus, a large
proportion of the population is likely to
experience greater mortality and morbidity
risk reductions in response to reductions in 8-
hour Os concentrations than are predicted
by our modeling in the 12 selected urban
case study areas.
Because our selection strategy for risk
modeling was focused on identifying areas
with high risk, we tended to select large
urban population centers. This strategy was
largely successful in including urban areas in
the upper end of the Os risk distribution.
However, this also led to an
overrepresentation of the populations living
in locations where we estimate increasing
mean seasonal Os would occur in response
to decreases in Os precursor emissions that
would be needed to just meet existing or
alternative standards. The implication of this
is that our estimates of mortality and
morbidity risk reductions for the selected
urban areas should not be seen as
representative of potential risk reductions for
most of the U.S. population, and are likely to
understate the average risk reduction that
ES-11
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
would be experienced across the
population.
T While the best available science
information and methodologies are
used to estimate exposures and
associated health risk, there remain
significant uncertainties in each of the four
primary analytical areas of this HREA. For
example, a number of important
uncertainties are identified regarding the
modeling approaches used to characterize
air quality (i.e., the CMAQ modeling, the
HDDM method used simulate alternative air
quality scenarios, application of 2007
modeled sensitivities for months and years
not modeled), though results of our
uncertainty characterization show limited
instances of the potential for either under- or
over-estimating ambient concentrations
while also having a limited range of
potential bias, generally less than a few
ppb. Similar conclusions are drawn
regarding the most important uncertainties
in estimating exposures, in particular those
concerning concentrations at or above
exposure benchmark levels (i.e., the human
activity pattern data and afternoon time
spent outdoors). Extensive activity pattern
data evaluations considering numerous
influential factors (e.g., survey year,
geographic region, health condition)
combined with confidence in the
characterization of air quality used as input
to exposure calculations suggests a limited
potential for bias in our exposure estimates.
When considering the FEVi risk estimates,
the most important uncertainties are found
in the lung function risk model itself and the
moderate to high sensitivity of FEVi
responses to changes in values used for
certain input variables (i.e., inter- and infra-
individual variability in response). Important
uncertainties in ourepidemiological-based
risk estimates are associated with the C-R
functions (i.e., overall shape of function at
low Os concentrations and exposure
measurement error) and its application (i.e.,
the urban study area risk modeling domains
are extended beyond the original urban
area from which the functions were
derived).
Synthesis
o facilitate interpretation of the results of the
exposure and risk assessment, this
assessment provides a synthesis of the
various results, focusing on comparing and
contrasting those results to identify common
patterns, or important differences.
Consistent with the available evidence, we
estimated exposures relative to several
health-based exposure benchmarks, lung
function risks based on a threshold
exposure-response model of lung function
decrements, and mortality and morbidity
risks based on non-threshold C-R functions.
These three different analyses result in
differing sensitivities of results to changes in
ambient Os concentrations. Because the
three metrics are affected differently by
changes in Os at low concentration levels, it
is important to understand these changes in
Os at low concentrations in interpreting
differences in the results across metrics.
The exposure benchmark analysis is the least
sensitive to changes in Os in the lower part
of the Os concentration distribution,
because the lowest exposure benchmarks is
at 60 ppb, a level above the portion of the
overall Os concentrations distribution where
we observed increases when adjusting air
quality to just meet the existing and
alternative standards. Because the
modeled exposures will always be less than
or equal to the ambient monitor
concentrations, a benchmark of exposure
at 60 ppb is above the range of Os
concentrations where the model-based
adjustment approach estimates increases in
concentrations. Thus, this risk metric would
most reflect the decreases in Os at high
concentrations that are expected to result
from adjusting air quality to just meeting the
existing and potential alternative standards.
The lung function risk analysis is less sensitive
than the mortality and morbidity risk
assessments to changes at very low
concentrations of Os, because the risk
function is logistic and shows little response
ES-12
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
at lower Os doses that tend to occur when
ambient concentrations are lower
(generally less than 20 ppb for the 10
percent FEVi decrement and generally less
than 40 ppb for the 15 percent FEVi
decrement). However, because there are
still some ambient concentration increases
that occur in the 50 to 60 ppb range where
the estimated lung function risk model is
more responsive, there may be some
reduction in the net risk decrease when
adjusting air quality to just meet
progressively lower standard levels.
The mortality and morbidity risk assessment is
the analysis that is most sensitive to the
increases in Os in the lower part of the Os
concentration distribution that we estimated
would occur when adjusting air quality to
just meet the existing and alternative
standards some urban study areas. Mean Os
concentrations in the urban study areas
change little between air quality scenarios
of just meeting the existing standard and
progressively lower alternative standard
levels, because mean concentrations
reflect both the increases in Os at lower
concentrations and the decreases in Os
occurring on days with high Os
concentrations. This leads to small net
changes in mortality and morbidity risk
estimates for many of the urban study areas.
However, both the net change in risk and
the distribution of risk across the range of Os
concentrations may be relevant in
considering the degree of additional
protection provided by just meeting existing
and alternative standards.
In conclusion, we have estimated that
exposures and risks remain after just meeting
the existing Os standard and that that in
many cases, just meeting potential
alternative standard levels results in
reductions in those exposures and risks.
Meeting potential alternative standards has
larger impacts on metrics that are not
sensitive to changes in lower Os
concentrations. When meeting the 70, 65,
and 60 ppb alternative standards, the
percent of children experiencing exposures
above the 60 ppb health benchmark falls to
less than 20 percent, less than 10 percent,
and less than 3 percent in the worst Os year
for all 15 case study urban areas,
respectively. Lung function risk also drops
considerably as lower standards are met.
When meeting the 70, 65, and 60 ppb
alternative standards, the percent of
children with lung function decrements
greater than or equal to 10 percent in the
worst year falls to less than 21 percent, less
than 18 percent, and less than 14 percent in
the worst Os year for all 15 case study urban
areas, respectively. Mortality and
respiratory hospitalization risks decrease by
5% or less for a level of 70 ppb, 10% or less for
a level of 65 ppb, and 15% or less for a level
of 60 ppb. These smaller changes in the
mortality and morbidity risks, relative to the
exposures and lung function risk reductions,
reflect the impact of increasing Os on low
concentration days, and the non-threshold
nature of the C-R function. Larger mortality
and morbidity risk reductions are estimated
on days with higher baseline Os
concentrations.
While there remain significant uncertainties
identified in each of the analytical areas,
we have sufficient confidence in the overall
results for them to be useful in informing the
policy assessment. Our assessment suggests
that the highest confidence should be
placed in the results of the human exposure
and lung function risk results, largely
because they are based on results of
controlled human exposure studies and a
physiology-based risk model. Medium to
high confidence is placed in the results of
the assessment of epidemiology-based risks
associated with short-term Os exposures,
because while the large number of studies
supporting the C-R relationships provides
increased confidence, there still exists
uncertainties related to unexplained
heterogeneity between locations, exposure
measurement errors, and interpretation of
the shape of the C-R function at lower Os
concentrations. Lower confidence is
placed in the results of the assessment of
epidemiology-based mortality risks
associated with longer-term Os exposures,
primarily because that analysis is based on
ES-13
-------
Health Risk and Exposure Assessment
for Ozone, Final (July 2014)
only one well designed study, and because
of the uncertainty in that study about the
existence and location of a potential
threshold in the C-R function.
References
Henderson, R. (2006). Clean Air Scientific
Advisory Committee's (CASAC) Peer Review
of the Agency's 2nd Draft Ozone Staff
Paper. U.S. Environmental Protection
Agency Science Advisory Board. EPA-
C AS AC-0 7-001.
Simon, H.; K. R. Baker; F. Akhtar; S.L.
Napelenok; N. Possiel; B. Wells and B. Timin.
(2013). A direct sensitivity approach to
predict hourly ozone resulting from
compliance with the national ambient air
quality standard. Environmental Science
and Technology. (47):2304-2313.
U.S. EPA. (2012). Integrated Science
Assessment of Ozone and Related
Photochemical Oxidants (Final Report).
Research Triangle Park, NC: EPA Office of
Research and Development. (EPA
document number EPA/600/R-10/076F).
.
ES-14
-------
This page left intentionally blank
-------
United States Office of Air Quality Planning and Standards Publication No. EPA-452/R-14-004f
Environmental Protection Health and Environmental Impacts Division August 2014
Agency Research Triangle Park, NC
------- |