United States
Environmental Protection Agency
Washington, DC 20460
July 2013
User's Manual
for the Co-Benefits
Risk Assessment
(COBRA)
Screening Model
Version: 2.61
Developed for
Climate Protection Partnerships Division
State and Local Climate and Energy Programs
State Climate and Energy Program
-------
Table of Contents
Table of Contents 1
ACKNOWLEDGEMENTS 2
INSTALLATION INSTRUCTIONS 3
System Requirements 3
Installation 3
Launching the Model 3
Technical Assistance 3
CHAPTER 1. Introduction 4
What is COBRA? 4
How is COBRA used? 5
Overview of Model 7
Caveats and Limitations 13
Additional Information 15
CHAPTER 2. Qui ck- Start Tutori al 16
Step 1. Open the model 16
Step 2. View the baseline emissions data 17
Step 3. Select the geography for emissions changes 17
Step 4. Define the emissions changes, select a discount rate, and run the scenario 18
Step 5. View the results 21
Step 6. Export and save your results 26
CHAPTER 3. Exploring Baseline Emissions Data 28
Baseline Emissions: Tables 28
Baseline Emissions: Maps 30
CHAPTER 4. Creating a New Emissions Scenario 32
Selecting Scenario Geography 32
Grouping Counties 33
Defining Scenario Emissions 35
CHAPTER 5. Viewing Results 37
Viewing Scenario Definition 37
Air Quality: Tables 37
Health Effects: Tables 39
Results: Maps 43
Saving Results 45
Glossary 47
Appendix A. Description of Source-Receptor Matrix and Emissions Data A-l
Appendix B. Derivation of Health Impact Functions B-l
Appendix C. COBRA Health Impact Functions C-l
Appendix D. Baseline Incidence Rates for Adverse Health Effects D-l
Appendix E. Population Forecasts E-l
Appendix F. Economic Value of Health Effects F-l
Appendix G. Additional Quick Start Tutorials on Sample COBRA Scenarios G-l
Appendix H. References H-l
1 July 2013
-------
ACKNOWLEDGEMENTS
COBRA was originally developed by Abt Associates Inc. in 2002 and updated in 2012
under contract with EPA's State and Local Climate and Energy Program. It is managed
by Denise Mulholland of EPA HQ.
EPA thanks the following individuals for assessing the technical and scientific aspects of
COBRA during a formal technical peer review of the original model: Dallas Burtraw,
Resources for the Future; Nino Kuenzli, Keck School of Medicine, University of
Southern California; and Jonathan Levy, Department of Environmental Health, Harvard
University. We also thank several individuals and organizations identified as likely users
of COBRA that served as an informal review group and provided comments on the
functionality, ease of use, and/or technical aspects of the model: Bryan Garcia,
Connecticut Clean Energy Fund; Lisa Herschberger, Minnesota Pollution Control
Agency; Marney Hoefer, Wisconsin Department of Natural Resources; Chris James,
Connecticut Department of Environmental Protection; Iyad Kheirbek, Northeast States
for Coordinated Air Use Management; Derek Murrow, Environment Northeast; and
Glenn Sappie, North Carolina Department of Environment and Natural Resources.
In addition, we thank Art Diem (EPA HQ), Doug Latimer (EPA Region 8), Bryan
Hubbell (EPA OAQPS), Neal Fann (EPA OAQPS), Alison Eyth (EPA OAQPS), and
Mark Houyoux (EPA OAQPS) for their extensive assistance and feedback during the
development of the original model and/or during the update.
Finally, EPA thanks the staff of Abt Associates Inc. that worked on COBRA, particularly
Donald McCubbin (now with US AID), Anna Belova, Jin Huang, Carleen Ghio, Andreas
Maier, Hardee Mahoney, Sue Greco, Frank Divita, and Jacqueline Haskell, for their
invaluable expertise and support in developing and updating this innovative screening
model.
2
July 2013
-------
INSTALLATION INSTRUCTIONS
System Requirements
Users must have at least 2 GB free hard drive space, 1 GB of RAM, and a CD-ROM
drive. COBRA was designed and tested on Windows XP. There are no known issues with
regards to running COBRA on Windows Vista or Windows 7.
Installation
COBRA can be downloaded directly to your computer or installed from a CD sent to you
in the mail. If you are downloading the COBRA model, note that the installer file is large
and the amount of time required to complete the download will depend on your
connection speed. Find the program 'setup.exe' in the location where the installer file
was saved. If you are installing the COBRA model from a CD, exit all programs and
insert the Installation disk into your CD-ROM drive. The installation program may start
automatically; if not, go to Start... Run... and then find the program 'setup.exe' in your
computer's CD-ROM drive.
During installation, follow the prompts on your screen. COBRA is a large program, and
depending on the speed of your computer, it may take five minutes to an hour to
complete the installation.
Launching the Model
To launch the model, go to Start... Programs... COBRA. To allow COBRA to run
efficiently, turn off any antivirus programs.
Technical Assistance
For more information, please contact Denise Mulholland at 202-343-9274 or
mulholland.denise@epa.gov.
3
July 2013
-------
CHAPTER 1.
Introduction
What is COBRA?
COBRA is a screening tool that provides preliminary estimates of the impact of air
pollution emission changes on ambient particulate matter (PM) air pollution
concentrations, translates this into health effect impacts, and then monetizes these
impacts,1 as illustrated below.
User-defined
changes in c
emissions
Changes in
ambient PM2 5 c
concentrations
. Changes in
^ health effects
Changes in
monetary
impacts
The model does not require expertise in air quality modeling, health effects assessment,
or economic valuation. Built into COBRA are emissions inventories, a simplified air
quality model, health impact equations, and economic valuations ready for use, based on
assumptions that EPA currently uses as reasonable best estimates. Analyses can be
performed at the state or county level and across the 14 major emissions categories (these
categories are called "tiers") included in the National Emissions Inventory.2'3
COBRA presents results in tabular as well as geographic form, and enables policy
analysts to obtain a first-order approximation of the benefits of different mitigation
scenarios under consideration and to quickly compare outcomes in terms of PM2.5 air
quality or health effects. However, COBRA is only a screening tool. More sophisticated,
albeit time-consuming, modeling approaches are currently available to obtain a more
refined picture of the health and economic impacts of changes in emissions.
1 In calculating health impacts, COBRA generates mean estimates of health impacts. This is in contrast to a risk
assessment, which typically builds in a margin of safety by presenting 95th percentile estimates.
2 The emissions inventory in COBRA includes fourteen broad tier 1 categories (e.g., on-road motor vehicles); within
each of these larger categories there are tier 2 (e.g., diesels), and tier 3 (e.g., heavy duty diesels) categories. The
fourteen tier 1 categories include: Chemical & Allied Product Manufacturing, Fuel Comb Electric Utilities, Fuel
Combustion Industrial, Fuel Combustion Other, Highway Vehicles, Metals Processing, Miscellaneous, Natural
Sources, Off-Highway, Other Industrial Processes, Petroleum & Related Industries, Solvent Utilization, Storage &
Transport, and Waste Disposal & Recycling.
3 Details on emissions categories are available at: http://www.epa.gov/ttnchiel/net/2008inventorv.html
4
July 2013
-------
Ch. 1. Introduction
How is COBRA used?
COBRA can be used to quickly identify
important emission sources and compare
the impacts of different types of control
options. Using the mapping capabilities in
COBRA, users can identify the locations
and types of emissions sources that
contribute to local air quality problems.
When considering different policy options,
COBRA can help identify those options
that are likely to maximize health benefits,
or that could be expected to achieve health
risk reductions in the most cost-effective
manner. Once state and local officials
narrow the set of most promising policy
options through COBRA, they can then
conduct analyses with more sophisticated
air quality models to finalize their policy
choices.
The model contains detailed emissions
estimates for the year 2017, developed for
the Mercury and Air Toxics Standards (MATS) Final Rule (77 FR 9304-9513), which
limits mercury and other toxic air pollution from coal- and oil-fired power plants.4 The
assumptions underlying these emissions data are detailed in the Emissions Modeling for
the Final MATS Technical Support Document (U.S. EPA, 201 lb). The air quality
modeling platform for MATS is based on emissions data, meteorology, initial conditions,
and boundary conditions from 2005 and uses 2017 as the future year of analysis. COBRA
uses the 2017 "control case" developed by EPA for the MATS Rule, which includes:
electrical generating unit emissions (reflecting the implementation of both MATS
and the Cross-State Air Pollution Rule),5
mobile emissions (reflecting the impacts of implementation of the Energy
Independence and Security Act of 2007 and the Energy Policy Act of 2005 on
mobile source fuels), and
average year fire data.
Who can use COBRA?
State and local officials who would
like to quickly identify important
emission sources and compare the
impacts of different control options;
Analysts looking to improve their
understanding of the air quality
improvements and health benefits
associated with clean energy policies
under consideration;
Environmental agencies trying to
inexpensively screen through many
options to identify those that
maximize the health benefits and to
quantify the economic value of health
improvements;
Energy officials looking to estimate
and promote the air quality, health,
and associated economic co-benefits
of their energy efficiency or
renewable energy policies; and
Transportation planners interested
in understanding the air quality and
health impacts of fuel switching or
reductions in vehicle miles traveled.
4 Read about the current status of MATS at http://www.epa.gov/airaualitv/powerplanttoxics/actions.html.
5 On August 21, 2012, the U.S. Court of Appeals for the D.C. Circuit issued an opinion that would vacate the Cross
State Air Pollution Rule. On October 5, 2012 the United States filed a petition seeking rehearing of that decision.
Further information about CSAPR (77 FR 34830) is available at: http://www.epa.gov/airtransport/.
5
July 2013
-------
Ch. 1. Introduction
COBRA users can create their own new scenarios by specifying increases or reductions
to the emissions estimates for the analysis year (i.e., 2017). Emissions changes can be
entered at the county, state, or national level.
COBRA then generates changes in PM2.5 concentrations between the baseline scenario
(the "business-as-usual" estimates for the analysis year) and the control scenario (the
analysis year modified by the user's emissions changes). A source-receptor matrix
translates the air pollution emissions changes into changes in ambient PM2.5 (for more
information about the emissions inventory and the source-receptor matrix, see Appendix
A). Using a range of health impact functions, COBRA then translates the ambient PM2.5
changes into changes in the incidence of human health effects (see Appendices B through
E). Finally, the model places a dollar value on these health effects (for more information,
see Appendix F).6 COBRA estimates the change in air pollution-related health impacts,
and estimates the economic value of these impacts, using an approach that is generally
consistent with EPA Regulatory Impact Analyses (U.S. EPA, 2012f; U.S. EPA, 2012g).
These analyses reflect the current state of the science regarding the relationship between
particulate matter and adverse human health.
Outcomes can be modeled nationwide or for smaller geographic areas. Results include
changes in ambient PM2.5 concentrations, and changes in the number of cases of a variety
of health endpoints that have been associated with PM2.5. These health endpoints include:
Adult and infant mortality;
Non-fatal heart attacks;
Respiratory-related and cardiovascular-related hospitalizations;
Acute bronchitis;
Upper and lower respiratory symptoms;
Asthma-related emergency room visits;
Asthma exacerbations;
Minor restricted activity days (i.e., days on which activity is reduced, but not
severely restricted); and
Work days lost due to illness.
6 There is a large literature regarding the health impacts of air pollution and approaches to value these impacts.
COBRA uses a subset that EPA deems most credible. More sophisticated users interested in using additional
approaches may want to use EPA's Enviromnental Benefits Mapping and Analysis Program (BenMAP), which is
available at: http://www.epa.gov/air/bemnap/.
6
July 2013
-------
Ch. 1. Introduction
Users can view the results in tabular or map form as well as export the data for use in
their own communications.
Overview of Model
The COBRA screening model is a stand-alone Windows application that contains all of
the data needed for the analysis of alternative emissions scenarios; the user is only
required to enter changes in emissions. Upon launching the model, you will see the
Overview screen, which explains that you may do one of two things: explore the baseline
emissions data, or create your own emissions scenario. Even if you are mainly interested
in creating your own scenario, you may want to look at the baseline data first, since new
scenarios are created by specifying increases or decreases to the baseline for one or more
tier categories.
File View Help
COBRA
Screening
Model
Analysis Year: 2017
Scenario Options
Run a new scenario:
( [nationwide}
C for individual states:
Alabama
~ Arizona
~ Arkansas
~ California
~ Colorado
~ Connecticut
~ Delaware
~ District of Columbia
~ Florida
~ Geofgia
DIdaho
Start I
Overview Emissions
Welcome to the Co-Benefits Risk Assessment
Screening Model (COBRA)
Tobeginusing COBRA you may:
1) Explore the analysis year 2017 emissions data
This data can be accessed in table and map form by clicking on the
"Emissions" button above. Viewing the baseline data first can help
you decide what changes you want to make in your own scenario.
2) Create your own scenario.
Vou can create a new scenario through the left panel of this page
or load in a previously saved scenario through 'File' -> 'Load'
Once you are ready to run a comparison, return to the Overview screen and use the
Scenario Options panel on the left of the screen to indicate the geographic level at which
you wish to make your emissions changes. 'Nationwide' means that any emissions
changes will be applied to all sources in that category throughout the entire U.S.
7
July 2013
-------
Ch. 1. Introduction
Alternatively, you can select 'for individual states' and select one or more states in the
window below. Any emissions changes you make will be applied only to those selected
states. On the following screens you can indicate if you want to make changes at the
county level of individual states; if so, you can combine some or all of the counties into
groups. Note that COBRA allows you to enter individual county changes for up to 10
counties per state.
AL AZ AR j CA
Define Alabama's emission increases/reductions:
C statewide
( tor individual counties:!
~ Autauga
U Choctaw
~ Baldwin
D Clarke
~ Barbour
~ Clay
~ Bibb
D Cleburne
~ Blount
~ Coffee
~ Bullock
~ Colbert
~ Butler
D Conecuh
~ Calhoun
~ Coosa
~ Chambers
~ Covington
~ Cherokee
O Crenshaw
~ Chilton
~ Cullman
< | fit
~
< Back |
Continue *> |
Once your geography is determined, you can enter your emissions changes for the nation,
or for each state, county, or group of counties, depending on your previous selection.
COBRA provides three levels of emissions sources (tiers) in a directory tree structure.
The choice of the category depends on the source that a policy or action is expected to
affect. For example, to assess the impacts of a renewable energy or energy efficiency
policy that is expected to affect utility-related emissions, you would select 'FUEL
COMB. ELEC. UTIL.' as the first tier. If you know the specific fuel source within the
utility category that would be reduced or displaced (e.g. coal or natural gas), you can
select the appropriate second tier. For a policy that involves fuel switching or reductions
8
July 2013
-------
Ch. 1. Introduction
in transportation through vehicle miles travelled, you would select 'HIGHWAY
VEHICLES' as the first tier.
Exhibit 1. Basic Tools and Data Sources for Determining Emission Reductions
Online Tool
Description
EPA's Emissions & Generation Resource Integrated
Database (eGrid)
(http://www.epa.gov/cleanenergv/energy-
resources/egrid/index.html)
EPA's Power Plant Emissions Calculator (P-PEC)
http://www.epa.gov/airaualitv/eere/auantifv.html
eCalc (http://ecalc.tamu.eduA
EPA's Motor Vehicle Emission Simulator
(MOVES)
(http://www.epa.gov/otaa/models/moves/index.htm)
National Emissions Inventory
(http://www.epa.gov/ttncliiel/net/2Q08inventorv.ht
M)
OTC Workbook (http://www.otcair.org)
Power Profiler (www.epa.gov/powerprofiler/)
Provides data on the enviromnental characteristics
of electric generation by power plants in the United
States.
Estimates potential emission reductions at power
plants within a county or nonattainment area due to
renewable energy or energy efficiency policies and
programs.
Uses both energy and emissions modeling to
determine emission reductions from energy
efficiency and renewable energy programs in the
Electric Reliability Council of Texas region.
Estimates emissions from mobile sources,
including emissions from cars, trucks, and
motorcycles.
Allows users to view emissions by sector (for 60
emissions inventory sectors) for specific pollutants
at varying levels of geographic aggregation.
Predicts emission reductions from energy portfolio
policies and energy efficiency programs and other
measures affecting renewable resources or multiple
pollutants.
Allows users to view the emissions that can be
attributed to electricity use in homes or businesses.
Note: For more details on these basic tools and on other methods, see: (1) Chapter 4 of EPA's "Assessing
the Multiple Benefits of Clean Energy: A Resource for States" report (U.S. EPA, 201 la), available at
http://www.epa.gov/statelocalclimate/resources/benefits.html; or (2) Appendix I of EPA's "Roadmap for
Incorporating Energy Efficiency/Renewable Energy Policies and Programs into State and Tribal
Implementation Plans" (U.S. EPA, 2012e), available at
http://www.epa.gov/airaualitv/eere/pdfs/appendixl.pdf.
Once you have determined the appropriate tier category, click on it and enter the
emission changes in tons or percentages, for one or more of the six included pollutants.
Absolute emission reductions in tons can be estimated using a variety of methods ranging
from basic to sophisticated. See Exhibit 1 above for a description of a few basic methods.
Percentage reductions can be used to assess the benefits of a goal that calls for reductions
in activity levels or emissions from a particular source, such as a renewable portfolio
9
July 2013
-------
standard, transportation policies requiring reductions in vehicle miles traveled, and
energy efficiency programs.
For example, EPA's Emissions & Generation Resource Integrated Database (eGrid)
website provides electric generation data and corresponding emissions rates for the
United States. On the eGrid website, click on 'eGRID2012 year 2009 Summary Tables
(PDF).' Using the Western Electricity Coordinating Council (WECC) Southwest Region
(which includes Arizona and New Mexico) as an example, you can obtain the following
information:
Emissions: The "Year 2009 eGRID Subregion Emissions - Criteria Pollutants"
table on page 2 summarizes emissions data in several regions. The annual sulfur
dioxide (SO2) emissions for the WECC Southwest Region in 2009 were
approximately 58,000 tons.
Electric Generation: The "Year 2009 eGRID Subregion Resource Mix" table on
page 5 summarizes electric generation data by region. The net generation for the
WECC Southwest Region in 2009 was approximately 186 million megawatt
hours (MWh).
Emissions Rates: The "Year 2009 eGRID Subregion Output Emission Rates -
Criteria Pollutants" table on page 4 provides the WECC Southwest Region's non-
baseload output emissions rate for SO2: 0.3913 lbs. per MWh.7
If a policy is expected to reduce electric generation by 20% in the WECC Southwest
Region, we can calculate the reduction in MWh: 20% x 186 million MWh = 37 million
MWh. We can then calculate the emission reductions as:
Emission Reduction = 37 million MWh x 0.3913 per MWh = 14 million lbs.
This reduction is equal to 7,000 tons (14 million lbs. ^ 2000 lbs. per ton) of SO2.
After you have calculated emissions changes, you can enter these changes for as many
tier categories as you wish, and different sets of changes for each state, county, or county
group.
7 Non-baseload emissions come from power plants that are brought online only when there is excess demand (U.S.
EPA, 2012b).
10
July 2013
-------
Ch. 1. Introduction
AL
AR
CA
All Counties
To change emissions estimates, click on a source category and enter your changes in the panel below. You MUST click the Apply
Edits button after editing each source category for your changes to be recorded.
Currently active category:
|(No selected category)
B CHEMICAL & ALLIED PRODUCT MFG
i FUEL COMB ELEC.UTIL.
3 FUEL COMB. INDUSTRIAL
i FUEL COMB.OTHER
a HIGHWAY VEHICLES
i METALS PROCESSING
MISCELLANEOUS
i NATURAL SOURCES
S3 OFF-HIGHWAY
SI OTHER INDUSTRIAL PROCESSES
i PETROLEUM & RELATED INDUSTRIES
B SOLVENT UTILIZATION
i STORAGE & TRANSPORT
6 WASTE DISPOSAL & RECYCLING
PM 2.5:
S02:
NOx:
NH3:
VOC:
( reduce by r; 77
_ . ' (enter amount)
< increase by 1
% ,educebv Ifenter amount)
( increase by 1
( reduce by r: r
_ . ' (enter amount)
( increase by 1
( reduce by r,: t
_ . , (enter amount)
( increase by 1
( reduce by r; r
_ . , (enter amount)
( increase by 1
Apply Edits
(* percent
C tons
( percent
C tons
f* percent
C tons
( percent
C tons
(* percent
C tons
<- Back
Summarize Edits
Run Scenario ->
With your emissions changes entered, click Apply Edits to ensure that the changes are
effective, then Run Scenario to run the comparison between the scenario you have just
created and the baseline scenario. The model will ask you to choose a discount rate (more
detail (described in more detail in the Chapter 2 Tutorial) and name the scenario before it
starts. When the model is done running, you can examine the results.
Regardless of the geographic level at which you made your emissions changes, you can
examine the results for every county in every state in the country. For each county,
COBRA calculates three types of results: the change in ambient PM2.5 concentration; the
change in health effects associated with that change; and the dollar value associated with
the change in health effects.
These results can be viewed in tables in the results section, or graphically on a map of the
U.S. You can select the specific states or effects of interest in the tables from the
dropdown tab under "View New Table By..." For the map, you can zoom and drag the
map to get to the scale of interest and then select the results you want to see.
11
July 2013
-------
Ch. 1. Introduction
COBRA
File View Help
COBRA
Screening
Model
Analysis Yeai: 2017
Air Quality
Table Options
Current table:
Scenario Name:
Test Run
View:
All States
View new table by:
| choose state
3]
View
View Scenario Definition
Export Scenario Definition
Overview Emissions Test Run
Air Quality: Tables j Health Effects: Tables | Results: Maps |
Export current data view
FIPS
County
State
Control PM 2.5
Base PM 2.5
Delta PM 2.5
~
18151
Steuben
IN
9.496
9.496
0
18153
Sullivan
IN
8.464
8.464
0
18155
Switzerland
IN
9 704
9.704
0
18157
Tippecanoe
IN
9.963
9.963
0
18159
Tipton
IN
12.081
12.081
0
18161
Union
IN
10.846
10.846
0
18163
Vanderburgh
IN
11.373
11.373
0
18165
Vermillion
IN
8.567
8.567
0
18167
Vigo
IN
8.899
8.899
0
18169
Wabash
IN
10.8
10.8
0
18171
Warren
IN
9.145
9.145
0
18173
Warrick
IN
9.397
9.397
0
18175
Washington
IN
10.049
10.049
0
18177
Wayne
IN
11.142
11.142
0
18179
Wells
IN
10.717
10 717
0
18181
White
IN
9.302
9.302
0
~
12
July 2013
-------
Ch. 1. Introduction
COBRA
Screening
Model
Analysis Yeai: 2017
Results Map Options
Current map view:
Scenario Name:
Test Run
Quantity:
Adult Mortality (low)
Change map quantity:
| Delta PM 2.5 (ug/'rn3]
View |
Change numeric ranges:
Change |
View Scenario Definition
Export Scenario Definition
Air Quality: Tables | Health Effects: Tables [ Results: Maps
Zoom tools: ฎN zoom in | Qx zoom out | C^C full exjent |
Export Map
(0.0000) 0.0000
0.0000 - 0.0000
0.0000 - 0.0000
0.0000 - 0.0000
0.0000 ฆ 0.0000
Results sets and subsets, including maps, can be exported for use in outside programs and
presentations. Scenario results can be saved and reloaded into COBRA at a future time.
To save your results, click on File... Save and then select your scenario from the list of
currently open scenarios. To reload results that you have previously saved, click on
File... Load and then browse to the location on your computer where the results are
saved. Up to five sets of results can be loaded at once; you can toggle back and forth
between them. At any time in the program you can view your scenario definition (the
summary of emissions changes you have made to the baseline), or export it for future
reference by clicking on Export Scenario Definition.
Caveats and Limitations
There are limitations to the COBR A screening model that make it inappropriate for
certain types of analyses:
Complicated analyses. Scenarios involving hundreds or perhaps thousands of
different changes in emissions from sources over a wide geographic area, for
13
July 2013
-------
example, would take a prohibitive amount of time, due to limitations in the user
interface.
Determination of attainment. Modeling the attainment of national ambient air
quality standards requires more sophisticated air quality modeling than that
currently built into COBRA.
Estimating rebound effects. Emissions in some states and regions are "capped"
and emission allowances are traded across entities within the designated area. In
these areas, a reduction in emissions in one location may result in an increase
(rebound) in emissions in another area under the cap, unless the emission
allowances are retired through the scenario evaluated. COBRA does not
automatically capture this potential effect so care should be exercised when using
COBRA to analyze the net impacts of a change in policy. COBRA is more suited
to an attributable risk assessment, which addresses the magnitude of an emission
source and the impact of controlling its emissions. That information can be used
to develop policies targeted to the appropriate sources.
Because COBRA is intended primarily as a screening tool, it uses a relatively simple air
quality model, which introduces additional uncertainty. While some comparative work to
test the performance of COBRA's air quality model is ongoing, it is not yet fully
validated.
As with more complex air pollution benefits models, there is substantial uncertainty
surrounding the values of key inputs to COBRA - in the air quality model, emissions
inventory, health impact functions, and economic values - and users should exercise
caution when interpreting the results of analyses.
Some of the uncertainty in COBRA reflects variability (for example, a health impact
function that is appropriate for one location may not be appropriate for another location if
the function actually varies across locations). Much of the uncertainty, however, reflects
the insufficient level of knowledge about the true values of model inputs.
The appendices discuss these issues and provide sources for additional information.
However, developing a quantified confidence interval for the results is beyond the scope
of this model. As an alternative, users should consider using sensitivity analyses to
determine how their conclusions might change with differences in the location and
amount of emissions. When more detailed analyses are required, users should be
cognizant of the model's limitations, and consider using more sophisticated modeling
approaches.
14
July 2013
-------
Chapter 2 of this User's Manual provides a quick tutorial for the new user. Baseline data
are examined, and a simple new scenario is defined and run, and the results are displayed
in tables and maps. The following chapters provide more detailed information on each
step, and describe additional options you can use for more complicated analyses.
Chapter 3 describes the baseline emissions data you can examine using the Emissions
button at the top of the Overview screen.
Chapter 4 provides details on different ways to define your new scenario, and run the
comparison between it (the control scenario) and the baseline scenario.
Chapter 5 describes the different ways to view, export, and save your results, and to
reload previously saved results.
A Glossary is provided at the end of the manual.
Additional Information
The Appendices to this manual provide additional information on the methods and
assumptions used in the model.
Appendix A: Description of Source-Receptor Matrix and Emissions Data.
Describes the source-receptor matrix embedded within the model that translates
the air pollution emissions changes into changes in ambient particulate matter.
Appendix B: Derivation of Health Impact Functions. Explains the derivation
of the types of health impact functions used in COBRA.
Appendix C: COBRA Health Impact Functions. Provides an overview of all
the health impact functions used in COBRA to convert changes in ambient PM2.5
into health effects.
Appendix D: Baseline Incidence Rates for Adverse Health Effects. Lists the
baseline incidence rates for each of the types of adverse health effects.
Appendix E: Population Forecasts. Describes the population forecasting
procedure.
Appendix F: Economic Value of Health Effects. Lists the equations and sources
of the values used to monetize the health effects.
Appendix G: Additional Quick Start Tutorials on Sample COBRA Scenarios
Appendix H: References. Provides all of the sources referenced in the
Appendices or used in the model.
15
July 2013
-------
CHAPTER 2.
Quick-Start Tutorial
This tutorial will give you a quick introduction to how COBRA works, and how to work
through the steps of a simple analysis. At each step, only the basic functions are
described; for more advanced options, see the subsequent chapters.
COBRA allows you to estimate the impact of a change in air pollution resulting from a
new policy or other type of change. In this example, we will consider changes in one state
(we have arbitrarily selected Pennsylvania) that result in a decrease in emissions from
electricity generating plants. If a statewide plan to switch 25 percent of electricity
generation to renewable sources were put into effect, what would be the difference in
ambient particulate matter levels and health effects, compared to doing nothing? This
tutorial will show you how to use COBRA to examine this type of scenario through the
following steps:
Step 1. Open the model.
Step 2. View the baseline emissions data.
Step 3. Select the geography for emissions changes.
Step 4. Define the emissions changes, select a discount rate, and run the comparison.
Step 5. View the results.
Step 6. Save or export your results.
Additional examples for assessing the impacts of a transportation policy, renewable
energy supply standards, and energy efficiency programs can be found in Appendix G.
Step 1. Open the model.
To open COBRA, click Start... All Programs... COBRA... COBRA. The model will
open. COBRA will load estimates of emissions, population, and other data for the
analysis year (i.e., 2017) for you to view and modify. The results of your analysis will be
expressed in terms of the difference between the baseline scenario (estimates for 2017 in
absence of the policy) and the control scenario (estimates for 2017 reflecting effects of
changes in emissions due to the policy). Note that the final results will be for 2017 only.
16
July 2013
-------
Ch. 2. Quick-Start Tutorial
Step 2. View the baseline emissions data.
COBRA will display the main screen. You will see two tabs at the very top: Overview
and Emissions. Click Emissions and you will see displayed all of the estimated
emissions data for the country in 2017, by state, county, source category, and pollutant.
To see emissions data for Pennsylvania only, change the selection in the drop down box
under 'View new table by' from 'All States" to 'Pennsylvania,' and click View. Now you
can scroll through the table to see the estimated emissions for 2017. In Step 4, you can
choose to enter emissions as percentage changes; such emissions changes will be based
on these estimates.
File View Help
COBRA
Screening
Model
Analysis Year: 2017
Base Emissions
Table Options
Current table:
Data for
Pennsylvania
All Counties
View new table by:
| Pennsylvania
| All Counties
[frievd
Overview Emissions
Base Emissions: lables | Base Emissions: Maps |
Export current data view
~
State
County
Tierl
Tier2
PA
Adams
FUEL COMB. INDUSTRIAL
COAL
PA
Adams
FUEL COMB. INDUSTRIAL
OIL
PA
Adams
FUEL COMB. INDUSTRIAL
GAS
PA
Adams
FUEL COMB. INDUSTRIAL
OTHER
PA
Adams
FUEL COMB. OTHER
COMMERCIAL/INSTITUTION;
PA
Adams
FUEL COMB. OTHER
COMMERCIAL/INSTITUTION,
PA
Adams
FUEL COMB. OTHER
COMMERCIAL/INSTITUTION,
PA
Adams
FUEL COMB. OTHER
MISC. FUEL COMB. (EXCEPT
PA
Adams
FUEL COMB. OTHER
RESIDENTIAL WOOD
PA
Adams
FUEL COMB. OTHER
RESIDENTIAL WOOD
PA
Adams
FUEL COMB. OTHER
RESIDENTIAL OTHER
PA
Adams
FUEL COMB. OTHER
RESIDENTIAL OTHER
PA
Adams
FUEL COMB. OTHER
RESIDENTIAL OTHER
PA
Adams
OTHER INDUSTRIAL PROCESSES
AGRICULTURE. FOOD, a KIN
F1
l
-------
entire country (although only a subset of states usually experience PM2.5 reductions and
health benefits).
Click on the Overview tab. This is where you begin to define your scenario. In the left
hand panel you can select your geography. Since we are only looking at changes
statewide in Pennsylvania, click 'for individual states' under 'Run a new scenario' and
check the box next to 'Pennsylvania' in the list. Click Start.
COBRA will ask if you want define Pennsylvania's emission increases/reductions
statewide or for individual counties. If you wanted to vary the emissions changes across
counties, or only make changes in some counties, you would select the second option.
For instance, if you know the counties in which sources that are likely to be affected
(such as power plants) are located, you can enter emissions changes in those counties
only. However, in this example we are looking at a statewide change, so select
'statewide' and click Continue.
Step 4. Define the emissions changes, select a discount rate, and run
the scenario.
This page allows you to specify exactly how emissions will change in your control
scenario. The bottom left window contains a directory tree with all of the tier 1, 2, and 3
source categories (see Appendix A for a list of source categories and their emissions).
You can define emissions changes at any level, but each level always includes all the
levels indented underneath it. In this example, we are only interested in electrical utilities,
so click on the plus sign to the left of 'FUEL COMB. ELEC. UTIL.' This will open the
tier 2 and 3 categories underneath. You can roll your mouse over any category and a
yellow 'Tier Information' box will appear with the number of tons of each pollutant in
the baseline emissions data for that source category, including all of the categories
indented under it.
We want to change all of the source categories under 'FUEL COMB. ELEC. UTIL,' so
click on that entry in the directory tree to highlight it. The boxes on the right-hand side of
the screen list each pollutant included in the model. Since our scenario reduces all
baseline emissions by 25 percent, type in '25' in the box next to each pollutant. The
default selections are 'reduce by' and 'percent'; leave them as they are. Click Apply
Edits to save your changes, COBRA will highlight the source categories that you have
changed. Your screen should look like the one below. Click Run Scenario.
18
July 2013
-------
Ch. 2. Quick-Start Tutorial
Defir
Define scenario
PA
All Counties
To change emissions estimates, click on a source category and enter your changes in the panel below. You MUST click the Apply
Edits button after editing each source category for your changes to be recorded.
Currently active category:
Ifuel comb. elec. util.
B CHEMICAL & ALLIED PRODUCT MFG
G-D FUEL COMB ELEC. UTIL
: E COAL
a GAS
I a INTERNAL COMBUSTION
| a oil
j E OTHER
FUEL COMB INDUSTRIAL
ffl FUEL COMB OTHER
a HIGHWAY VEHICLES
i METALS PROCESSING
3 MISCELLANEOUS
i NATURALSOURCES
6 OFF-HIGHWAY
< I i" I >
PM 2.5:
S02:
NOx:
NH3:
VOC:
( reduce by
C increase by '
< reduce by
r increase by
( reduce by r^~~
C increase by '
( reduce by
C increase by '
C* reduce by
C increase by '
' Apply Edits |
(* percent
r tons
(* percent
C tons
(* percent
r tons
percent
C tons
(* percent
C tons
<- Back
Summarize Edits
Run Scenario >
A pop-up box will open, asking you to choose a discount rate for the COBRA session.
The discount rate you select is used to express future economic values in present terms.
Not all health effects and associated economic values occur in the year of analysis (as
explained in Step 5 below), and people are generally willing to pay more for something
now than for the same thing later. Therefore, COBRA accounts for this time preference
(i.e., a general preference for receiving benefits now rather than later) by discounting
benefits received later. There is an ongoing discussion within the federal government
about the most appropriate discount rate in this context; typically either 3% or 7%
discount rates are used. Based on EPA's Guidelines for Preparing Economic Analyses
(U.S. EPA, 2010a), it is recommended that COBRA users calculate monetized health
benefits using both discount rates and then evaluate whether the overall outcome of the
analysis is affected by the choice of discount rate. For more details on discount rates, see
Appendix F.
In this scenario, we will use a 3% discount rate. Select 3% in the pop-up box and click
Continue
19
July 2013
-------
Ch. 2. Quick-Start Tutorial
Select a Discount Rate for the Scenario
COBRA estimates the economic value of current and future avoided deaths and
illnesses expected based on emissions reductions in the year 2017. Emission reductions
require investments and, like all investments, there are trade offs, or opportunity costs, of
picking one investment over another, each with their own set and schedule of expected
benefits. T o reflect the opportunity costs of the investments foregone by investing in
emission reductions and to figure out how much future benefits are worth today, COBRA
users must select a discount rate.
Rather than using just a single rate, EPA's Guidelines for Economic Analysis recommend that analysts use
a bounding approach to discounting, developing an upper and lower bound for their estimates. They
advise use of both:
a 3% rate, reflecting the interest rate consumers might earn on Government backed securities, and
a IX rate, reflecting the opportunity cost of private capital, based on estimates from the Office of
Management and Budget.
NOTE: A higher discount rate favors those investments with immediate benefits and reduces the value of
future benefits more than a lower discount rate, which places a greater value on future benefits to society.
For more information on discount rates and how EPA uses them in monetizing health benefits, see the
User Manual.
In order to run the COBRA model, please select a discount rate to use in this COBRA session.
Continue
After selecting a discount rate, COBRA will ask you to supply a name for your scenario.
You may want to include your choice of discount rate in the scenario name. Enter 'Penn
Utility Reduction - 3%' (or whatever representative name you prefer) and click OK. You
will see a message indicating that the scenario run is processing and may take a few
minutes. The run time depends on your computer.
Scenario name
Enter a name for your scenario:
Penn Utility Reduction 3%|
OK
Cancel
20
July 2013
-------
Step 5. View the results.
Once your run is complete, COBRA will display your results. You will see a screen with
three tabs at the top: 'Air Quality: Tables,' 'Health Effects: Tables,' and 'Results: Maps.'
Click on the first tab, 'Air Quality: Tables.' This tab shows the change in air quality
(PM2.5) between the baseline scenario and your new scenario (the control scenario) for
2017. The default view shows the whole country, but since we expect the majority of the
air quality changes to be in Pennsylvania, change the selection under 'View new table by'
to Pennsylvania, and click View. The table will now display all the counties in
Pennsylvania with the PM2.5 levels for your control scenario, the baseline scenario, and
the change between them. Note that positive values indicate a reduction in PM2.5 in the
control scenario.
Let's look at a specific county in Pennsylvania. Scroll down and find 'Montgomery
County' in the table, or filter the county level by clicking on the arrow next to 'County'
and selecting 'Montgomery' from the drop-down list. You will see that in Montgomery
County, the estimated ambient PM2.5 concentration in the control scenario is 10.624
[j,g/m3, compared to the estimated baseline concentration of 10.709 (j,g/m3. The difference
between the two estimated concentrations (Delta PM2.5) is 0.085 (J,g/m3, which is the
estimated change in air quality due to the 25% reduction in emissions from fuel
combustion electricity generating plants in the whole state (the change in concentration is
due to decreases in emissions from plants within the county plus plants in other counties).
Note that positive changes indicate a lower concentration in the control scenario. If the
Delta PM2.5 were negative, it would indicate an increase in concentration.
21
July 2013
-------
Ch. 2. Quick-Start Tutorial
File View Help
COBRA
Screening
Model
Analysis Year: 2017
Air Quality
Table Options
Current table:
Scenario Name:
Penn Utility
Reduction-3%
View:
Pennsylvania
View new table by;
| Pennsylvania ~^1
View
View Scenario Definition
Export Scenario Definition
Overview Emissions (ility Reducti
Air Quality: Tables | Health Effects: Tables | Results: Mfips |
Export current data view
FIPS
County
State
Control PM 2.5
Base PM 2.5
Delta PM 2.5 (
42091
Montgomery
PA
10.624
10.709
.085
42093
Montour
PA
9.481
9.561
.0798
42095
Northampton
PA
9033
9.152
.12
42097
Northumberlanc
PA
8.714
8.817
.103
42099
Perry
PA
8.123
8.214
.0908
42101
Philadelphia
PA
14.772
14.849
.0773
42103
Pike
PA
7.628
7.702
.0739
42106
Potter
PA
5.846
5.921
,0753|
42107
Schuylkill
PA
8.682
8.802
12
42109
Snyder
PA
8.354
8.449
.0945
42111
Somerset
PA
8.138
8.221
.0828
42113
Sullivan
PA
6.538
6.615
.0765
42115
Susquehanna
PA
6.372
6.445
.0727
42117
Tioga
PA
5.918
5.992
.0743
E
Data estimates for 2017. All values are in ug/m3. To sort by column, click on the column title. To filter
the data view, use the arrows on the state/county columns. A positive value indicates a decrease from
the base scenario.
Now click on the 'Health Effects: Tables' tab. Select 'Pennsylvania' in the list and click
View. This screen shows, for each county in Pennsylvania, the estimated change in health
effects caused by the estimated change in ambient PM2.5 levels, as reported on the Air
Quality tab. Each change in health effect also has an associated dollar value.
Scroll down in the list of counties (click on 'County' to sort alphabetically by county) or
use the filter to display Montgomery County, which had a 0.085 jjg/m3 reduction in
ambient PM2.5, as shown on the Air Quality tab. This table shows the reductions in health
effects associated with that 0.085 (ig/m3 decrease. Note that positive values indicate a
decrease in impacts (that is, fewer cases of illness/premature mortality or avoided
economic loss). In the case of Montgomery County, the change in PM2.5 was associated
with total avoided health effects that range in value from approximately $30 million to
$67 million. The numbers at the bottom show the totals for the current view; in this case,
the totals for all of Pennsylvania.
22
July 2013
-------
Ch. 2. Quick-Start Tutorial
File View Help
COBRA
Screening
Model
Analysis Year: 2017
Health Effects
Table Options
Current table:
Scenario Name:
Penn Utility
Reduction - 3%
View:
Pennsylvania
View new table by;
| Pennsylvania ~^1
View
View Scenario Qefinition
Overview Emissions lility Reducti
Export Scenario Definition
Air Quality: Tables Health Effects: Tables Results: Mfips
Export current data view
Id
State
County
FIPS |
$ Total Health Effects (low) $ Total Health Effects (hie
PA
Montgomery
42091
29.625.285.54
66.792.4H
PA
Montour
42093
898.749 77
2.028.0
PA
Northampton
42095
16.389.971.49
36,949.3
PA
Northumberland 4209?
5.482.801.16
12.409.2
PA
Perry
42099
2.022,983.92
4,573.
PA
Philadelphia
42101
49.885.312.48
113.591.0
PA
Pike
42103
1.764.515.97
3,987,0] I
PA
Potter
42105
734,614.81
1,658.11I
PA
Schuylkill
42107
10.981.758.96
24.846.
$551,183,311.81
$1.246.516.0jpi
hi
~1
T o sort by column, click on the column title. T o filter the data view, use the arrows on the state/county columns.
This table piesents cases of health effects avoided (in columns with blue text) and the monetary values of those
benefits (in columns with black text). Any negative values indicate costs. Please refer to the User Manual lor
further details.
- COBRA provides two estimates of total health effects (low and high) which reflect two sets of assumptions about
the sensitivity of both adult mortality and adult myochardial infarction to changes in ambient PM2.5 levels. Please
refer to the User Manual for further details.
Scrolling over to the right, you can view the estimated reductions in specific health
effects (shown in blue text) and their associated dollar values (shown in black text). The
screenshot below shows this color-coding for respiratory-related hospital admissions.
23
July 2013
-------
Ch. 2. Quick-Start Tutorial
gWCOBRA
File View Help
COBRA
Screening
Model
Overview
Emissions tility Reducti
Air Quality: Tables Health Effects: Tables Results: Mfips |
Analysis Year: 2017
Health Effects
Table Options
Current table:
Scenario Name:
Penn Utility
Reduction - 3%
View:
Pennsylvania
View new table by;
| Pennsylvania ~^1
View
View Scenario Qefinition
Export current data view
Id
State
County
FIPS
Resp. Hosp. Adm.
$ Resp. Hosp. Adm.
CVDH
PA
Montgomery
42091
6665
17.746.47
PA
Montour
42093
.0134
363.85
PA
Northampton
42095
.5056
13.966.33
PA
Northumberland 4209?
.1099
3.056.21
PA
Perry
42099
.0392]
1,070.59
PA
Philadelphia
42101
2.5405
54.634 84
PA
Pike
42103
.0232
605.79
PA
Potter
42105
.0171
475.36
PA
Schuylkill
42107
.2466
6.706.32
14.97711
$389,676.94
F1
ra 1 1 iti
Export Scenario Definition
T o sort by column, click on the column title. T o filter the data view, use the arrows on the state/county columns.
This table presents cases of health effects avoided (in columns with blue text) and the monetary values of those
benefits (in columns with black text). Any negative values indicate costs. Please refer to the User Manual for
further details.
- COBRA provides two estimates of total health effects (low and high) which reflect two sets of assumptions about
the sensitivity of both adult mortality and adult myochardial infarction to changes in ambient PM2.5 levels. Please
refer to the User Manual for further details.
Note that the health effects table includes low and high estimates for the changes in the
number of cases and the corresponding economic values for adult mortality, non-fatal
heart attacks, and total health effects. The low and high estimates are derived using two
sets of assumptions about the sensitivity of adult mortality and non-fatal heart attacks to
changes in ambient PM2.5 levels. Specifically, the high estimates are based on studies that
estimated a larger effect of changes in ambient PM2 5 levels on the incidence of these
health effects. For further details on the calculation of low and high estimates, see the
description of the health effects table in Chapter 5 and the detailed assumptions in
Appendix C.
We will use three health effects to demonstrate the interpretation of the change in health
effects and their economic values for Montgomery County, Pennsylvania: adult mortality,
non-fatal heart attacks, and respiratory hospital admissions.
Respiratory Hospital Admissions. In COBRA, most health effects and their
economic values are expected to occur in the year of analysis. For instance, our
scenario results in approximately one avoided case of respiratory hospital
admissions in Montgomeiy County. This avoided case and its economic value
(approximately $18,000) would occur in 2017.
24
July 2013
-------
Adult Mortality. In contrast to respiratory hospital admissions, all avoided cases
of adult mortality are not expected to occur in the year of analysis. Therefore,
COBRA uses the 3% discount rate you selected in Step 4 to calculate the value of
all avoided cases of adult mortality in present terms (in Montgomery County, a
low estimate of approximately three avoided cases of adult mortality are valued at
a total of approximately $29 million).
Non-fatal Heart Attacks. Another special case is non-fatal heart attacks. All
avoided cases of non-fatal heart attacks are expected to occur in the year of
analysis, but the costs associated with this health effect would occur over multiple
years. Thus, while our scenario results in a range of less than one to more than
three cases of non-fatal heart attacks in 2017, all economic benefits associated
with this change ($41,000 to $376,000) would not accrue in that same year.
In addition, remember that although we changed emissions only for Pennsylvania,
COBRA calculates changes in PM2.5 for the whole country. (However, the detectable
changes are probably only in states bordering on Pennsylvania.) If you would like to
examine the results for any of states bordering Pennsylvania, simply change your
selection in the box in the left-hand panel and click View.
It is also important to remember that emissions in some states and regions are "capped"
and that firms may trade emission allowances. As a result, if we assume an emission
reduction among power plants in Pennsylvania, then it is likely that emissions will
increase from other entities, unless the emission allowances are retired as part of the
assumed emission reduction. COBRA does not automatically capture this potential effect
so care should be exercised when interpreting the impacts of an emissions change in a
given location.
The 'Results: Maps' tab shows the results from the previous two tables on a map. When
you click on the tab you will see a map of the United States. The default quantity
displayed is Delta PM2.5, the change in the particulate matter concentration between the
baseline and control scenarios. The darker the shade of blue, the greater the change in
concentration. As in the other results tables, a positive number indicates a decrease from
the baseline scenario.
25
July 2013
-------
Ch. 2. Quick-Start Tutorial
File View Help
Change numeric ranges:
Change |
View Scenario Definition
Export Scenario Definition
COBRA
Screening
Model
Analysis Yeai: 2017
Results Map Options
Current map view:
Scenario Name:
Penn Utility
Reduction - 3%
Quantity:
Delta PM 2.5 (ug/m3)
Change map quantity:
| Delta PM 2.5 (ugAri3)
Overview Emissions lility Reducti
Air Quality: Tables | Health Effects: Tables Results: Maps |
Zoom tools: zoom in | Qx zoom out | Cm full exjent |
Export Map
You can also view any of the other results on the map by selected from the drop-down
list under 'Change map quantity' and clicking View. You can also change the divisions in
the numeric range by clicking the Change button under 'Change numeric ranges' (see
Chapter 5 for details on changing the divisions).
Step 6. Export and save your results.
You may want to look at and manipulate the results data outside of COBRA. You can
export small data sets into formats for use with spreadsheet programs (larger data sets can
be exported into database formats). To export the health effects table for Pennsylvania
into an Excel format, click on the 'Health Effects: Tables' tab. If the table is filtered,
click the arrow next to 'County' and select 'All'. Click on the Export current data view
button. In the following window, browse to the file location where you want to save your
data, and select 'EXCEL files' in the 'Save as type' drop-down box. You can then enter a
name for the file in the box above; type 'Penn Utility Reduction 2017 - 3%_health'
(COBRA will add the Excel file extension for you). Click Save and COBRA will save
the file.
26
July 2013
-------
You can also save your entire results set to use again in a future COBRA session. You
might want to compare it to a different scenario involving different emissions changes. At
the top of the screen, click File... Save, and then select 'Penn Utility Reduction 2017 -
3%.' In the following window, browse to the file location where you would like to save
the results file and type 'Penn Utility Reduction 2017 - 3% results.' COBRA will add a
.erf (COBRA results file) extension for you. Click Save. To load these results in a future
session, go to File... Load and browse to the results file you just saved.
You could then compare the results from this session to a scenario where power plants
switched only ten percent of their generation to renewable sources, instead of 25%. These
kinds of comparisons allow you to estimate the magnitude of the changes in health effects
and benefits that result from emissions scenarios. See Appendix G for additional
examples.
27
July 2013
-------
CHAPTER 3.
Exploring Baseline Emissions Data
The COBRA model contains detailed 2017 baseline emissions data for every county in
the U.S., by state, county, tier category, and pollutant type (see Appendix A for details on
the data). You can explore the emissions data for the baseline air quality scenario by
clicking on the Emissions button in the top menu bar. You will see a screen with two
tabs: 'Base Emissions: Tables' and 'Base Emissions: Maps.' Each tab is described below.
Baseline Emissions: Tables
When you click on the Emissions button, the information window will show a table
summarizing the baseline emissions data for the U.S for the baseline air quality scenario
you selected. Your year of analysis (2017) is shown at the bottom of the screen.
The default view shows emissions (in tons) by tier category and county for the entire U.S.
Using the control panel at the left of the screen, you can change the table view by
selecting an option in the drop-down box under 'View new table by:.' In the first box you
can choose from the following options:
All States. This view is the default, and shows the pollutants for all counties in
the U.S. by tier category (in alphabetical order by state, county).
Individual State. Select an individual state, click View, and the right-hand
window will display pollutant emissions for that state, alphabetically by county.
The second selection box will become active. The default selection in the second
box is 'All Counties,' which lists pollutant values for each county in the state by
tier category. Alternatively, you may select any individual county from the drop-
down list to see the breakdown by tier category for that county only.
Each time you change your selection, click the View button to update the table display on
the right. The left-hand panel displays a description of the table currently on display
under 'Current table.' You can navigate through the table data in several ways:
Scroll through the data using the scroll bars.
Change the column order by clicking on the column name and dragging it to a
new position. When you see two green arrows, you can drop the column there.
Note that the sort order of the table will not change. The State and County
columns will remain frozen in the view when you scroll horizontally in the table.
Those two columns cannot be moved out of the frozen view, and no other
columns can be added into the frozen view.
28
July 2013
-------
Ch. 4. Creating a New Emissions Scenario
Change the width of a column by moving your mouse to the column header and
pointing to the dividing line between two columns. The mouse cursor will change
to two arrows, indicating that you can drag the column line to the left (to shorten)
or right (to expand).
File View Help
COBRA
Screening
Model
Analysis Year: 2017
Base Emissions
Table Options
Current table:
Data for: United States
View new table by:
| All Slates ~^ |
|state level options
Overview Emissions lility Reducti
Base Emissions: Iables | Base Emissions: Maps |
Export current data view
~
State County
Tierl
|Tier2
AL
Autauga
FUEL COMB. INDUSTRIAL
COAL
AL
Autauga
FUEL COMB. INDUSTRIAL
OIL
AL
Autauga
FUEL COMB. INDUSTRIAL
OIL
AL
Autauga
FUEL COMB INDUSTRIAL
GAS
AL
Autauga
FUEL COMB INDUSTRIAL
| OTHER
AL
Autauga
FUEL COMB. OTHER
COMMERCIAL/INSTITUTIONAL CC
AL
Autauga
FUEL COMB. OTHER
| COMMERCIAL/INSTITUTIONAL Oil
AL
Autauga
FUEL COMB. OTHER
COMMERCIAL/INSTITUTIONAL
AL
Autauga
FUEL COMB. OTHER
MISC. FUEL COMB. (EXCEPT RESI
AL
Autauga
FUEL COMB, OTHER
RESIDENTIAL WOOD
AL
Autauga
FUEL COMB . OTHER
RESIDENTIAL WOOD
AL
Autauga
FUEL COMB . OTHER
RESIDENTIAL WOOD
AL
Autauga
FUEL COMB. OTHER
RESIDENTIAL OTHER
AL
Autauga
FUEL COMB. OTHER
RESIDENTIAL OTHER p^l
hi
1 !~[
You can export the table into formats that can be read by spreadsheet and database
programs by clicking on the Export Current Data View button. This will export only
the data from the current view; to export the full baseline emissions database, make sure
the current view is set to 'All States' (note that the full dataset is very large and may take
a substantial amount of time to export, depending on your computer). When you click the
Export Current Data View button, COBRA will export the table and then open the
exported file using whichever program your computer has set as the default program for
the file type you selected.
Exporting the emissions data is especially useful as a reference when creating a new
scenario; simply open the data in a spreadsheet or database program and toggle back and
forth between it and COBRA. The emissions data are very large; if you see a 'File not
loaded completely' message when the file opens in your default program, then the data
are too large to be completely read by that program. In this case, simply open the data in
29
July 2013
-------
Ch. 4. Creating a New Emissions Scenario
a different program. Database programs such as Microsoft Access, Dbase, or Paradox can
open larger files.
Baseline Emissions: Maps
By clicking on the Base Emissions tab, you can see the geographic distribution of
pollutant concentrations in the baseline scenario. The values are mapped at the county
level and show the total pollutant concentration in each county from all tier categories.
You can change the pollutant being mapped by selecting a pollutant in the 'Display new
map:' drop-down box.
File View Help
COBRA
Screening
Model
Analysis Yeai: 2017
Base Emissions
Map Options
Current map view:
Pollutant:
PM 2.5 (tons)
Overview Emissions lility Reduct
Base Emissions: Tables Base Emissions: Maps ]
Zoom tools: S^joomin zoom out CSfullgxtent
Export Map
Display new map:
PM 2 5 (tons)
View
E
Change numeric ranges:
Change |
62.75 - 6178.31
6178.31-12293.88
12293.88 -18409.45
18409.45 ฆ 24525.02
24525.02 - 30640.58
The zoom tools across the top of the map allow
you to zoom in and out, as well as return to the
full extent of the map, which shows the
continental U.S. Using the mouse, you can pan
by clicking and dragging the map to the portion
you wish to view. To zoom in on a specific area, use the right mouse button to draw a
box around the area of interest.
Keyboard alternative:
Pari by holding the ALT key down
while using the arrow keys to move in
a specific direction.
30
July 2013
-------
The default map displays concentrations broken into five divisions based on ranges. By
clicking on the Change button at the bottom of the control panel, you can change the
number and type of numeric ranges. First enter the number of ranges that you wish to see
displayed in the box at the top, and then select the type of break. The following options
for the type of break are available:
Range. This method takes the range of the values, and, using the number of
breaks you selected, splits the range into equal intervals.
User Defined. This method allows you to manually enter the upper range of each
break. First select the 'User Defined' option, and then enter the numbers into the
boxes at the right. Note that the number of breaks selected at the top determines
how many boxes into which you can manually enter maximum values. The first
range starts at zero and ends with the maximum value you enter in the first box
(the range includes the value you enter). The next range starts with but does not
include the previous maximum and ends with the maximum you enter. The
maximum value must increase with each break level; the model will not accept a
maximum that is lower than the previous one. The displayed values will end with
the last maximum you enter, even if the range of values in the set exceeds that
maximum.
Click Apply once you have selected the number and type of break desired, and the map
display will be updated with your selections. The legend in the upper left corner describes
the numeric breaks and the color used to display each division on the map. To see a range
highlighted on the map, simply click on the desired row in the legend; to deselect the
range, right-click on the same row. If you change the pollutant displayed, the map will
use the same number and type of breaks.
You can export the currently displayed map by clicking on the Export Map button at the
top right corner of the window. COBRA will export the map as a bitmap file, which can
be viewed in many different programs including most word processors. Note that the
bitmap will match the zoom level and extent currently displayed. To export a map of the
entire country, click the Full Extent button before you export.
31
July 2013
-------
Ch. 4. Creating a New Emissions Scenario
CHAPTER 4.
Creating a New Emissions Scenario
COBRA allows users to define new emissions scenarios and investigate the related
changes in air quality and health effects. You can create up to five scenarios per session,
making it easy to compare outcomes between scenarios. Scenarios can also be saved and
loaded back into COBRA at a later date (see the 'Saving results' section of Chapter 5).
Additionally, the results tables and maps can be exported into various formats for future
use or archiving purposes.
The steps to creating a new scenario are simple:
Step 1. Select the Geography. You can specify emissions changes at the national,
state, or county level.
Step 2. Enter Emissions Changes. Changes can be made for each state or county, or
the entire nation. Within each state, you can group counties and make changes to
them together, or make different changes to each county.
Step 3. Run the Scenario. COBRA will calculate the changes in ambient PM2.5
between your selected baseline scenario and the new control scenario, and calculate
the associated changes in health effects and monetary impacts.
Step 4. Examine the results. See Chapter 5.
Selecting Scenario Geography
First, select the geographic area(s) for
which you would like to change
emissions estimates. Click on the
appropriate choice in the control panel on
the left side of the Overview screen under
'Scenario Options.' Then click Start. The
sections below describe the options for
each type of run.
Nationwide. If you select 'nationwide,' the model will display the emissions
entry screen for your selection, where you can specify the changes that you wish
to make to the baseline scenario (see 'Defining scenario emissions' below). The
changes will be made to the emissions for the entire country.
Individual State. If you wish to make emissions changes to just one state, click
'for individual states,' then select your state in the drop-down box and click Start.
The geographic areas you select determine
where your emissions changes are made.
Your selection does not affect the geographic
area for which you can see results.
Regardless of the geographic area you select
for emissions changes, you can see results
for the entire country.
32
July 2013
-------
The model will then ask you if you want to make changes statewide or to
individual counties. If you choose 'statewide,' the model will take you to the
emissions entry screen for your selection, where you can specify your desired
changes to the baseline scenario (see 'Defining scenario emissions' below). The
changes will be made to the emissions for the entire state. If you click 'for
individual counties' and select only one county, you will be taken to the emissions
entry screen. If you select two or more counties, you will be given the option of
creating county groups before proceeding. Note that COBRA only allows you to
enter individual emission changes for up to 10 counties per state. If you select
more than 10 counties in a state, you must apply the same emission changes to all
counties or place the counties into groups. See 'Grouping counties,' below.
Multiple States. If you wish to make emissions changes to more than one state,
click 'for individual states', then select your states in the drop-down box and click
Start. The model will then display a screen with one tab for each state you
selected. Each tab works separately. On each of these tabs, you must select
whether you want your changes to be applied statewide or to individual counties.
If you select 'statewide' or if you select 'for individual counties' and only pick
one county, the model will take you to the emissions entry screen for your
selection, where you can specify the changes to the baseline scenario that you
wish to make (see 'Defining scenario emissions' below). If you select two or
more counties, you will be given the option of creating county groups before
proceeding. Again, note that COBRA allows you to enter individual emission
changes for up to 10 counties per state; for more than 10 counties, you must apply
the same changes to all selected counties or place the counties into groups. See
'Grouping counties,' below. You must go through the selection process with
each state tab.
Grouping Counties
If you select three or more individual counties on the tab for any of your selected states,
the model will ask if you wish to: (1) apply different changes to each individual county,
(2) apply the same changes across all selected counties, or (3) group the counties into two
or more (up to four) groups and define changes for each group. Selecting the first option
will bring up the emissions entry screen with a tab for each selected county. You may
only select the first option for up to 10 counties per state. If you select more than 10
counties in a state, you must use the second or third options. The second option will also
bring up the emissions entry screen, with a tab that says 'Group:' and lists all of your
selected counties in a single group.
33
July 2013
-------
Ch. 4. Creating a New Emissions Scenario
AL AZ AR
CA
Create county groups:
Instructions: Drag and drop counties into group. Hold Ctrl while clicking to select multiple
counties. Alternatively highlight the county and press the number key matching
the desired target group.
Selected Counties:
Group 1:
Group 2:
Group 3:
Group 4:
Autauga
Baldwin
Barbour
Bibb
Blount
Bullock
Butler
Calhoun
Chambers
Cherokee
Chilton
Choctaw
< Back
Continue -->
If you select the third option, select the number of groups you wish to create in the drop-
down box (no more than four), and click Continue. This will bring up a screen where
you can click and drag your selected counties into boxes representing the defined groups.
If you drop a county name into the wrong box, simply click and drag it into the correct
one.
Each county must be assigned to only
one group, and all counties must be
assigned to groups before you can
continue. If you decide you do not
want to group your selected counties,
simply hit the 'Back' button and
change your choice. When you click Continue at the bottom right of the screen, the
emissions entry screen will be displayed, with one tab for each group you have created.
Keyboard alternative:
Highlight a county name and then type the number
of the group to which it should be assigned (e.g.,
for "Group 1", type 1). The county name will
automatically move into that group's box.
34
July 2013
-------
Ch. 4. Creating a New Emissions Scenario
Defining Scenario Emissions
Once you have selected your geographic areas and grouped them if appropriate, you can
create a new emissions scenario by defining changes to the baseline emissions scenario.
The emissions entry screen contains one tab for each state you have selected. On each
state tab, there is another set of tabs for the individual or grouped counties in that state.
You must enter emissions changes for each tab on each state separately.
Define scenario
AR
CA
Group 4: Cherokee. Chilton, Choctaw, Clarke, Clay, Cleburne, Coffee, Colbert, Conecuh, Coosa, Covington, Crenshaw, Cullman
Group 1: Autauga. Baldwin, Barbour | Group 2: Bibb, Blount, Bullock ] Group 3: Butler, Calhoun. Chambers
To change emissions estimates, click on a source category and enter your changes in the panel below. You MUST click the Apply
Edits button after editing each source category for your changes to be recorded.
Currently active category:
|(No selected category)
ฎ CHEMICAL & ALLIED PRODUCT MFG
a FUEL COMB. ELEC. UTIL
B FUEL COMB. INDUSTRIAL
a FUEL COMB. OTHER
i HIGHWAY VEHICLES
B METALS PROCESSING
E) MISCELLANEOUS
S NATURALSOURCES
S OFF-HIGHWAY
S OTHER INDUSTRIAL PROCESSES
E PETROLEUM & RELATED INDUSTRIES
IS SOLVENT UTILIZATION
0 STORAGE & TRANSPORT
S WASTE DISPOSAL & RECYCLING
< reduce by r,; 77 percent
. ' (enter amount) r,
i increase by 1 < tons
< reduce by
C increase by
(~ percent
C tons
S02:
| (enter amount)
(* reduce by
C increase by
( percent
C tons
NOx
| (enter amount)
( reduce by
C increase by
percent
C tons
NH3:
| (enter amount)
( reduce by
r increase by
< percent
C tons
VOC:
| (enter amount)
Apply Edits
< Back
Summarize Edits
Run Scenario ~>
The box on the left contains a directory tree with three levels for tier 1, 2, and 3 emissions
categories (for more information about emissions categories, see Appendix A). If you
hold your mouse over any of the tiers in the directory tree, a yellow box will pop up,
displaying a summary of the baseline data (total tons) for each pollutant emitted by
sources in that tier category for the current geographic area. If the box says
'calculating..,', the program is still in the process of summarizing the information; the
actual values will appear in a few seconds. Note that each tier category always includes
all of the categories indented underneath it.
35
July 2013
-------
Ch. 4. Creating a New Emissions Scenario
You can select a category at any tier level, then use the boxes at the right to enter changes
to the baseline emissions for each of the five pollutants potentially emitted by sources in
that category. You can enter emissions increases or decreases by percent or by tons in
whole numbers or decimals. A change entered for a category applies to all of the
branches under it, but you must enter changes individually for categories on separate
branches. Click Apply Edits to save your changes. Once saved, the name of the edited
category and any branches under it will turn blue; this will help you visually track which
categories you have already modified.
Once your changes are entered for a category, you can go back to it and change it by
clicking on it again in the directory tree. Your previously saved changes will be displayed
in the boxes on the right when you click on the category name. At any time, you can click
on the Summarize Edits button to see the changes you have made so far. If you have
decided to apply different changes to your county groups, repeat the above steps for each
tab that is visible.
If you click the Back button from the 'Define scenario' screen, you will be sent back to the
previous screen, and any changes that you have entered (even if you have clicked the Apply
Edits button) will be lost.
When you have made all of your desired changes, click the Run Scenario button at the
bottom right. The model will prompt you choose a discount rate (3% or 7%) for the
COBRA session. The model will then prompt you to enter the name of the scenario.
Enter a name (with no extension) and click OK. The time to generate your results will
vary, depending on the speed of your computer and on how many counties and tier
categories are affected by your scenario. Once the results are generated, you will see a
new button at the top with your scenario name. Click the button to view the results of
your run (see Chapter 5).
36
July 2013
-------
CHAPTER 5.
Viewing Results
Once you have defined your new scenario (called the control scenario) and run the
comparison between the baseline and your scenario, you can view the results. In this
section, you can review your scenario definition and see the changes in air quality and
health effects between the baseline and control scenarios. When the button with your new
scenario's name is depressed, three tabs will be visible to navigate between the screens.
Viewing Scenario Definition
If you need to remind yourself of the edits you made to the baseline emissions data in
order to create this scenario, click the View Scenario Definition button at the bottom of
the 'Air Quality: Tables' tab. This will display all of your changes by tier category. If the
scenario is acceptable and you wish to save it for future reference, click the Export
Scenario Definition button. This will export a comma-delimited file that contains the
same information as shown when you click the View Scenario Definition button.
Exporting this file is especially useful if you export any other tables from the results tabs;
at a later date you will have a handy reference for what the results tables and maps are
based on.
Air Quality: Tables
This tab describes the changes in particulate matter concentration between the baseline
emissions scenario and your scenario (the control scenario). For each county, the table
lists the annual average PM2.5 concentration for the control scenario and the baseline
scenario, as well as the change between the two scenarios (Delta PM2.5). The default table
view lists all counties, but you can choose any single state in the drop-down box on the
left panel to limit the view to the counties in that state. You can navigate through the
table data in several ways:
Scroll through the data using the scroll bar on the right.
Change the column order by clicking on the column name and dragging it to a
new position. When you see two green arrows, you can drop the column there.
Note that the sort order of the table will not change. The State and County
columns will remain frozen in the view when you scroll horizontally in the table.
Those two columns cannot be moved out of the frozen view, and no other
columns can be added into the frozen view.
37
July 2013
-------
Ch. 5. Viewing Results
Change the width of a column by moving your mouse to the column header and
pointing to the dividing line between two columns. The mouse cursor will change
to two arrows, indicating that you can drag the column line to the left (to shorten)
or right (to expand).
Filter column variables by clicking on the arrows on the FIPS (for 'Federal
Information Processing Standards' codes), County and State columns. You can
filter down to a specific state and county. The dark bar at the bottom of the table
displays any filters you have defined. The filter can be turned on or off by
checking the box next to it in the dark bar. You can delete a filter by clicking the
'X'.
Change the sort order by clicking on the heading of any column. Click once to
sort in descending order then click again for ascending order.
File View Help
COBRA
Screening
Model
Analysis Year 2017
Air Quality
Table Options
Current table:
Scenario Name:
Test Run
View:
Alabama
View new table by:
Alabama
View Scenario Definition
Export Scenario Definition
Overview Emissions Test Run
Air Quality: Tables Health Effects: Tables | Results: Maps I
Export current data view
FIPS
County
State
Control PM 2.5
Base PM 2.5
Delta PM 2.5
~
01001
Autauga
AL
9.903
9.903
.0002
01019
Cherokee
AL
10.391
10.391
.0003
01021
Chilton
AL
10.04
10.04
.0004
01023
Choctaw
AL
9.425
9.425
.0001
01025
Clarke
AL
8.781
8.781
0001
01027
Clay
AL
10.599
10.6
.0004
01029
Cleburne
AL
10.571
10.571
0004
01031
Coffee
AL
9.057
9.058
0003
01033
Colbert
AL
12.395
12.396
.0003
01035
Conecuh
AL
9.538
9.539
.0002
01037
Coosa
AL
10.113
10113
0003
01003
Baldwin
AL
9.289
9.289
0001
01039
Covington
AL
8.76
876
.0002
01041
Crenshaw
AL
9.612
9 612
.0003
....
Fl
Data estimates for 2017. All values are in ug/m3. To sort by column, click on the column title. T o filter
the data view, use the arrows on the state/county columns. A positive value indicates a decrease from
the base scenario.
Note that if you have run a state-specific scenario, changes in air quality for other states
will typically decrease as the distance from the state increases, since the emissions
changes were only made there. You can export the table by clicking on the Export
Current Data View button. This will export only the data from the current table; to
export the full air quality data set, make sure the current table shows 'All States.'
38
July 2013
-------
Ch. 5. Viewing Results
Health Effects: Tables
This tab displays the change in the number of cases for each health effect between the
baseline emissions scenario and your scenario. These changes are derived using the
health impact functions described in Appendix C. For each health effect, the table also
displays an estimate of the economic value of the change in the number of cases. For
more information, see Appendix F. Exhibit 2 describes the health endpoints and
valuations that are included in the health effects tables in COBRA.
Exhibit 2. Description of Health Effects and their Economic Values
Health Effect Description
Total Health EffcctsS (Low) Economic value of all health effects combined in Low Case, using a
discount rate of 3% or 7%
Total Health EffcctsS (High) Economic value of all health effects combined in High Case, using a
discount rate of 3% or 7%
Low estimate of the number of deaths, based on Krewski et al.
(2009)
Low estimate of the economic value of the number of deaths, using
Krewski et al. (2009) and a discount rate of 3% or 7%
High estimate of the number of deaths, based on Lepeule et al.
(2012)
High estimate of the economic value of the number of deaths, using
Lepeule et al. (2012) and a discount rate of 3% or 7%
Number of infant deaths
Economic value of the number of infant deaths
Low estimate of the number of non-fatal heart attacks, based on four
acute myocardial infarction (AMI) studies
Low estimate of the economic value of non-fatal heart attacks, based
on four AMI studies and a discount rate of 3% or 7%
High estimate of the number of non-fatal heart attacks, based on
Peter et al. (2001)
High estimate of the economic value of non-fatal heart attacks, using
Peter et al. (2001) and a discount rate of 3% or 7%
Number of respiratory-related hospitalizations (e.g., all respiratory,
asthma and COPD)
Economic value of respiratory-related hospitalizations
Number of cardiovascular-related hospitalizations (ICD codes 390-
409, 411-429). ICD code 410 (nonfatal heart attacks) is counted only
in 'Non-fatal Heart Attacks'
Adult Mortality (Low)
Adult Mortality $ (Low)
Adult Mortality (High)
Adult Mortality $ (High)
Infant Mortality
Infant Mortality $
Non-fatal Heart Attacks
(Low)
Non-fatal Heart Attacks $
(Low)
Non-fatal Heart Attacks
(High)
Non-fatal Heart Attacks $
(High)
Resp. Hosp. Adm.
Resp. Hosp. Adm. $
CVD Hosp. Adm.
39
July 2013
-------
Ch. 5. Viewing Results
Health Effect
Description
CVD Hosp. Adm. $
Acute Bronchitis
Acute Bronchitis $
Upper Resp. Symptoms
Upper Resp. Symptoms:
Lower Res. Symptoms
Lower Res. Symptoms $
Asthma ER Visits
Asthma ER Visits $
MRAD
MRAD $
Work Loss Days
Work Loss Days $
Asthma Exacerbations
Asthma Exacerbations $
Economic value of cardiovascular-related hospitalizations
Cases of acute bronchitis
Economic value of acute bronchitis cases
Episodes of upper respiratory symptoms (runny or stuffy nose; wet
cough; and burning, aching, or red eyes)
Economic value of episodes of upper respiratory symptoms
Episodes of lower respiratory symptoms: cough, chest pain, phlegm,
or wheeze
Economic value of episodes of lower respiratory symptoms
Number of asthma-related emergency room visits
Economic value of asthma-related emergency room visits
Number of minor restricted activity days (days on which activity is
reduced, but not severely restricted- e.g. missing work or being
confined to bed is too severe to be MRAD).
Economic value of minor restricted activity days
Number of work days lost due to illness
Economic value of work days lost due to illness
Shortness of breath, wheeze, and cough (in asthmatic individuals)
Economic value of episodes of asthma exacerbations
Notes: * For adult mortality and non-fatal heart attacks, COBRA contains multiple health impact functions
that relate PM2 5 and each health effect. Therefore, there are high and low estimates of the cases avoided
and their economic values for each of these health effects. The high and low estimates of the economic
value of total health affects avoided are based on the corresponding high and low estimates for adult
mortality and non-fatal heart attacks, along with the single estimates for all other health effects. More
details on the underlying health impact functions are available in Appendix C of the user manual. In
addition, future costs are calculated using a discount rate (3% or 7%) that you selected before running the
scenario.
The health effects table includes low and high estimates for the changes in the number of
cases and the corresponding economic values for adult mortality, non-fatal heart attacks,
and total health effects. The low and high estimates are derived using two sets of
assumptions about the sensitivity of adult mortality and non-fatal heart attacks to changes
in ambient PM2.5 levels. Specifically, the high estimates are based on studies that
estimated a larger effect of changes in ambient PM2.5 levels on the incidence of these
health effects. The low and high estimates for each of these values are derived as follows:
Adult Mortality. EPA (2009) recently used two studies when analyzing proposed
NO2 national ambient air quality standards; EPA presented the results separately
40
July 2013
-------
Ch. 5. Viewing Results
for each study. Following EPA, COBRA reports the results of two health impact
functions that relate PM2.5 and mortality: Krewski et al. (2009) and Lepeule et al.
(2012). In the health effects table, Adult Mortality (Low) and Adult Mortality $
(Low) represent estimates of deaths avoided and their economic value,
respectively, based on Krewski et al. (2009). Adult Mortality (High) and Adult
Mortality $ (High) represent estimates of deaths avoided and their economic
value, respectively, based on Lepeule et al. (2012). More details on the two
studies are available in Appendix C of the user manual.
Non-fatal Heart Attacks. COBRA calculates two estimates of the non-fatal heart
attack cases avoided (Non-fatal Heart Attacks) and their economic value (Non-
fatal Heart Attacks $). The low estimate is based on Peter et al. (2001), while the
high estimate is based on pooling of the effect estimates of the following four
studies: Sullivan et al. (2005), Pope et al. (2006), Zanobetti et al. (2009), and
Zanobetti & Schwartz (2006). More details on the studies are available in
Appendix C of the user manual.
Total Health Effects. The Total Health Effects $ (Low) includes estimates based on
Krewski et al. (2009) for adult mortality and Peter et al. (2001) for non-fatal heart attacks,
along with the single estimates for all other health effects. Similarly, Total Health
Effects $ (High) includes estimates based on Lepeule et al. (2012) for adult mortality and
pooling of the effect estimates of the four studies listed above for non-fatal heart attacks,
as well as the single estimates for all other health effects.
The value in each health effects column represents the total change in the number of
cases of each health endpoint in a county. A value of 3.00 in the Adult Mortality (low)
column, for instance, indicates that in your scenario there would be an estimated 3 fewer
cases of premature mortality compared to the baseline emissions scenario. Note, however,
that a negative number signifies an increase in cases. Therefore, - 3.00 in the Adult
Mortality (low) column indicates that in your scenario there would be 3 additional cases
of premature mortality compared to the baseline emissions scenario.
Interpreting positive and negative results:
In the health effects table, positive numbers indicate reductions in the number of cases of
adverse health effects and the associated monetary benefits of your scenario. Negative numbers
signify increases in the number of cases of health effects and the resulting costs.
41
July 2013
-------
Ch. 5. Viewing Results
COBRA
Screening
Model
Health Effects
Table Options
Current table:
Scenario Name:
Test Run
View:
All States
View new table by:
| choose state
Viewl
"31
View Scenario Definition
Export Scenario Definition
Overview Emissions Test Run
Air Quality: Tables Health Effects: Tables | Results: Maps |
Export current data view |
State a
County
FIRS
$ Total Health Effects (low)
$ Total Health Effects (hie
AL
Autauga
01001
6.335.85
16.
AL
Baldwin
01003
5.226.35
13.3
AL
Barbour
01005
2.79137
7.1
AL
Bibb
01007
9.385.64
2
AL
Blount
01009
51,368.8
131.3
AL
Bullock
01011
6,697.04
17.1
AL
Butler
01013
3.702.65
9.4
AL
Calhoun
01015
51.328.01
131.
AL
Chambers
01017
5,605.24
14.3
AL
Cherokee
01019
4.393.53
11.2
$2,063,295.36
$5,283.6
Fl
lซl 1 l>
T o sort by column, click on (he column title. T o filter the data view, use the arrows on the state/county columns.
ฆ This table presents cases of health effects avoided (in columns with blue text) and the monetary values of those
benefits (in columns with black text). Any negative values indicate costs. Please refer to the User Manual for
further details.
COBRA provides two estimates of total health effects (low and high) which reflect two sets of assumptions about
the sensitivity of both adult mortality and adult myochardial infarction to changes in ambient PM2.5 levels. Please
refer to the User Manual for further details.
The values in grey at the bottom of each column are the subtotals for the view that is
currently displayed in the table; therefore, to see the total national effect, be sure to set
your table view to 'All States.' Each row in the table represents the overall impact on the
entire population living in each county, so the number of adverse health effects can be
fairly large, particularly for the less severe health effects, such as work loss days (WLDs)
and minor restricted activity days (MRAJDs).
All health effects are monetized. However, to prevent double-counting, the calculation of
asthma exacerbations only includes asthma effects occurring in children aged 6-18 years.
This approach follows the recommendations of EPA's Science Advisory Board Health
Effects Subcommittee (SAB-HES) for valuing asthma exacerbations, as described in the
benefits analysis for the 2006 Regulatory Impact Analysis for the revised PM2.5 National
Ambient Air Quality Standard (U.S. EPA, 2006). Studies of the general population
include asthmatics, so estimates based solely on the adult asthmatic population cannot be
directly added to the general population numbers without double-counting. Instead,
asthma exacerbations occurring in adults were assumed to be accounted for in health
effects for the general population, such as WLDs and MRADs (U.S. EPA, 2006). Since
the health effects for the general population do not include asthma effects in children, the
42
July 2013
-------
analysis of asthma exacerbations for children does not lead to double-counting (see
Appendix C for details).
As in the 'Air Quality: Tables' screen, you can select different state views for your table
using the options in the lower left panel and filter the view down to the county level using
the arrows in the table columns. By clicking on the heading of any column you can sort
the table by that column. Click once to sort in descending order then click again for
ascending order. To go back to the default sort order, click on the first column. You can
also export the table to a comma-delimited file using the button at the top right of your
screen. See 'Air Quality: Tables' above for more detail.
Results: Maps
This screen displays the results of your scenario geographically. The left side of the
screen allows you to change the values shown on the map. You can display the change in
PM2.5 between your scenario and the baseline emissions (the same values shown as
'Delta PM2.5' in the 'Air Quality: Tables' screen), or any of the health endpoints included
in the model. The values displayed for each health endpoint are the change in the number
of cases (or deaths for 'Adult Mortality (Low)', 'Adult Mortality (High)', and 'Infant
Mortality') and the economic valuation of these cases from the scenario, as displayed in
the 'Health Effects: Tables' screen. Note that if your scenario represents an increase in
cases, the values are negative and are therefore displayed in parentheses on the map
legend.
43
July 2013
-------
Ch. 5. Viewing Results
Quantity:
Adult Mortality (low)
Change map quantity:
13
Delta PM 2.5 (ug/m3]
View |
Change numeric ranges:
Change |
View Scenario Definition
Export Scenario Definition
COBRA
File View Help
COBRA
Screening
Model
Analysis Yeai: 2017
Results Map Options
Overview Emissions Test Run
Air Quality: Tables | Health Effects: Tables Results: Maps |
Zoom tools: ฉN zoom in | zoom out | full extent |
Export Map
Current map view:
Scenario Name:
Test Run
0.0000 ฆ 0.0004
0.0004 ฆ 0.0007
0 0007 - 0.0011
0.0011 - 0.0014
0.0014 0 0018
You can select one of two options for your map's numeric breaks:
Range. This method takes the range of the values, and, using the number of
breaks you selected, splits the range into equal intervals.
User Defined. This method allows you to manually enter the upper range of each
break. First select the 'User Defined' option, then enter the numbers into the
boxes at the right, using a minus sign (rather than parentheses) if you are dealing
with negative numbers. Note that the number of breaks selected at the top
determines how many boxes into which you can manually enter maximum values.
For positive numbers, the first range starts at zero and ends with the maximum
value you enter in the first box (the range includes the value you enter). For
negative values, the first break starts at the smallest value for the health effect or
PM2.5 that you are mapping. The next range starts with, but does not include, the
previous break's maximum and ends with the maximum you enter. The maximum
value must increase with each break level; the model will not accept a maximum
that is lower than the previous one. The displayed values will end with the last
maximum you enter, even if the range of values in the set exceeds that maximum.
44
July 2013
-------
Ch. 5. Viewing Results
The zoom tools at the top left of the screen allow you to zoom in and out in the map. You
can also pan by clicking and dragging the map to the portion you wish to view (or from
the keyboard, use the ALT key plus arrow keys to indicate the direction to pan). You can
zoom in on a specific area by drawing a box with right mouse button depressed. Click on
Full Extent to zoom to full U.S. view. To highlight a range of values on the map, simply
click on the desired row in the legend; to deselect it, right click on the same row. The
highlighted data will be yellow
File View Help
COBRA
Screening
Model
Analysis Year 201?
Results Map Options
Current map view:
Scenario Name:
Test Run
Quantity:
$ Total Health Effect (Lov
Change map quantity:
| $ T otal Health E ffect (Lov * |
View |
Change numeric ranges:
Change |
View Scenario Definition
Overview ^missions Test Run
Air Quality: Tables | Health Effects: Tables Results:Maps
zoom out
Zoom tools:
full extent
Export Map
Export Scenario Definition
0.00 33147.49
33147.49 -66294.99
66294.99 ฆ 99442.48
99442.48 -132589 98
132589.98 -165737 47
You can export the map as currently displayed to a bitmap file by clicking on the Export
Map button at the top right. The bitmap file will show the map and the legend only; what
is being mapped and the scenario name will not be visible, so it is advisable that you
incorporate that information into the file name.
Saving Results
There are several options for saving your results. You can save the entire scenario for
future use within COBRA by clicking on File... Save and then selecting the scenario
from the list. This will store the scenario definition and results in a COBRA Results File
(.erf) format. If you want to use the scenario results in a future COBRA session, use the
45
July 2013
-------
File... Load option. You will see the loaded scenario appear as a button at the top of the
screen.
To save your results for use outside of the COBRA environment, you can export the air
quality and health effects tables (see 'Viewing Scenario Results' above). The maps can
also be customized within COBRA and then exported into bitmap format for use in
documents and presentations.
46
July 2013
-------
Glossary
Baseline scenario: The - emissions estimates for 2017 in absence of a policy, ambient
pollution levels and health impacts for 2017. The baseline scenario is compared to the control
scenario when running COBRA.
Control scenario: A hypothetical scenario that factors in user-specified emissions changes
(to 'control' emissions). In COBRA, the control scenario is compared to the baseline
scenario.
Delta PM2.5: The difference in ambient concentrations of particulate matter that is less than
or equal to 2.5 microns in diameter.
Health impact function: An equation that calculates the change in adverse health effects
associated with a change in exposure to air population. A typical health impact function has
inputs specifying the change in the air pollutant, an effect coefficient (specifying the percent
change in an adverse health effect per unit change of a pollutant), the age of the population
affected, and the incidence rate of the adverse health effect.
Mercury and Air Toxics Standards (MATS) Final Rule: An EPA regulation issued in
December 2011 to limit mercury and other toxic air pollution from coal and oil-fired power
plants. For more information on the rule, see
http://www.epa.gov/airqualitv/powerplanttoxics/actions.html.
Scenario definition: A table of all edits made to the baseline emissions when defining a
control scenario. The table can be viewed within COBRA or can be exported for future
reference.
Sensitivity analyses: Comparison of analyses performed with varied assumptions or
decisions to determine whether the assumptions/decisions have a major effect on the results
of the analysis.
Source-receptor matrix: An air quality model built into COBRA that calculates the change
in PM2.5 levels for any given change in emissions. Appendix A discusses this model in more
detail.
Tier category: Classification used by EPA for emission inventories. Additional information
on tier categories is available at: http://www.epa.gov/ttn/chief/codes/
47
July 2013
-------
APPENDICES
48
-------
Appendix A: Dispersion Modeling in COBRA
COBRA estimates particulate matter levels using the Phase II Source-Receptor (S-R) Matrix.
The S-R Matrix consists of fixed transfer coefficients that reflect the relationship between annual
average PM2.5 concentration values at a single receptor in each county (a hypothetical monitor
located at the county centroid) and the contribution by PM2.5 species to this concentration from
each emission source (E.H. Pechan & Associates Inc., 1994).
Levy et al. (2003) found that an earlier version of the S-R Matrix predicted public health benefits
that were similar to those predicted by CALPUFF, a comparatively more sophisticated model
often used in risk assessments. Using the emission impacts from seven power plants in northern
Georgia, Levy et al reported that the two models yielded generally similar results for sulfates or
primary PM2.5, with somewhat greater differences for nitrates. However, they carefully noted
that this result may differ depending on the location of the emissions, as temperature and
humidity are important considerations in the formation of ambient particles.
Because of the limited validation studies of the S-R Matrix, it should be treated as a screening
tool that provides a crude estimate of the likely impact of a change in emissions on ambient
PM2.5 levels. More sophisticated atmospheric dispersion models should be used to obtain
detailed estimates of ambient air quality changes.
The sections below summarize the development of the S-R matrix and the steps taken to apply
the matrix in COBRA in order to derive the changes in air quality resulting from changes in
emissions.
Development of the S-R Matrix
The S-R matrix is based on the Climatological Regional Dispersion Model (CRDM), which uses
assumptions similar to the Industrial Source Complex Short Term model (ISCST3), an EPA-
recommended short range Gaussian dispersion model (U.S. EPA, 1995). The CRDM
incorporates terms for wet and dry deposition of primary and secondary species that constitute
PM2.5 and uses meteorological summaries (annual average mixing heights and joint frequency
distributions of wind speed and direction) from 100 upper air meteorological sites throughout
North America. This analysis employs meteorological data collected in 1990.
Relative to more sophisticated and resource-intensive three-dimensional modeling approaches,
the CRDM does not fully account for all the complex chemical interactions that take place in the
atmosphere in the secondary formation of PM2.5. Instead it relies on more simplistic species
dispersion-transport mechanisms supplemented with chemical conversion at the receptor
location.
The CRDM uses Turner's sector-average approach (Turner, 1970), a probabilistic method in
which relative frequencies of occurrence of combinations of wind and stability conditions at the
emissions source are used to calculate the relative frequencies of transport in various sectors.
A -1
July 2013
-------
This method is recommended for the estimation of long-term average pollutant concentrations
(E.H. Pechan & Associates Inc., 1997).
The pollutant concentration in a destination sector is estimated as follows:
C,(r) = -^ฃp
yyj2x ijc up
- 1/
f \2
' H A
z,k J
(1)
joint frequency of wind speed class z, wind direction j, and stability
category k
where:
Cj (r) = atmospheric concentration in destination sector j at distance r
Q(r) = pollutant mass flux at distance r
y = sector width at distance r
fi,j,k
az,k = vertical diffusion coefficient for stability category k
= wind speed for wind class z
H = effective stack height of emissions source (= 0 for ground-level sources)
The sector width is calculated as:
r=(f) (2)
Primary emissions from a county are assumed to always impact the county source county itself
and are evenly distributed over a square with the same area as the county. A simple box model is
used for each combination of wind speed and stability category. The vertical diffusion
coefficient, crz, is then calculated at a downwind distance corresponding to the side of the
square.1 These assumptions are necessary since the spatial variation of emissions within a
county cannot be provided for a national scale model.2
Additional adjustments are made to ensure a consistent distribution of pollutant species among
areas in close proximity to the emissions source. Receptors at a distance less than the square root
of the source area are assumed to receive the same concentration of pollutants as the source area.
1 The vertical diffusion coefficient r>- was calculated using a subroutine from EPA's ISC3 model. Atmospheric stabilities
were assumed to be C class (slightly unstable) during the day and E class (slightly stable) at night. However, for wind
speeds in excess of 6 m/s, stability was assumed to be neutral (class D).
2 Actual measured concentrations would be expected to be higher than those modeled with these assumptions for a
monitor located in, or generally downwind from, a portion of the county with emission densities much higher than the
county average. On the other hand, concentrations would be expected to be lower if a monitor is located at the prevailing
upwind edge of the county or in an area of relatively low emission density.
A-2
July 2013
-------
In addition, the destination sector width is constrained to be at least equal to the square root of
the source area.
Equation (1) is applicable to both point and area sources, either ground-level or elevated, and
results in a Gaussian distribution of pollutant mass in the vertical dimension. However, for long-
range transport, emissions are distributed uniformly in the vertical between the top of the mixed
layer and the ground. This occurs when the vertical diffusion parameter, crz, is equal to the height
of the mixed layer, hm. For such long-range situations, the sector-average limited mixing model
of Turner (1970) estimates pollutant concentrations at a downward distance r from the source as:
C,(r)=Sซฃ (3)
Ky i* Ui
The mass flux of a directly emitted primary species at distance r from the source is a function of
the material initially emitted, the amount chemically converted to a secondary pollutant, and the
amount deposited by wet and dry processes during the period of transport (time t) from the
emission point to the receptor. This is calculated by solving the relevant differential equation
(Latimer, 1993):
QP(t) = Q0e-{kc+kp)t (4)
where:
<2,(0 = primary pollutant mass flux at transport time t
Q0 = initial emission rate
^ _ pseudo-first-order rate constant for chemical conversion of the primary
c species to the secondary species
^ = pseudo-first-order rate constant for deposition of primary species, equal
p to the sum of the dry and wet deposition rate constants (kpti + kpw)
t = transport time
The mass flux of secondary pollutants is dependent upon the fraction of the primary species that
is chemically converted in the atmosphere to the secondary species and the amount of the
secondary species that is deposited by wet and dry deposition processes during the transport time
t from the stack to the downwind receptor point at distance r. This is also calculated by solving
the relevant differential equation (Latimer, 1993):
Q,(ป=<5)
where:
0.(0 = mass flux of the secondary species at transport time t
A - 3
July 2013
-------
Qo
initial emission rate
kc
kP
ks
t
pseudo-first-order rate constant for chemical conversion of the primary
species to the secondary species
pseudo-first-order rate constant for deposition of primary species, equal
to the sum of the dry and wet deposition rate constants (kpd + kpK)
pseudo-first-order rate constant for deposition of secondary species,
equal to the sum of the dry and wet deposition rate constants {ksd + kSK)
transport time
The model parameters used to estimate mass flux are detailed in Exhibit A-l. Note that the
pseudo-first-order rate constant for deposition, kp, is estimated from the dry and wet deposition
velocities by dividing them by the mixing height (/?,).
Exhibit A-l. Pollutant-specific Model Parameters
PM2 5, SOA
SO,
NO,
NH3
Chemical Conversion Rate, kc (%/lir)
[RH = relative humidity (%)]
Dry Deposition Velocity (cm/s)
Wet Deposition Velocity (cm/s)***
[P = annual precipitation rate (in.)]
0.1
0.01 P
0.5 if RH < 40
1.5 if RH > 70
((RH - 40)/30) + 0.5
Otherwise
0.5
0.003 P
0.0003 P 0.0003 P
* Secondary organic aerosols.
** The chemical conversion rate for S02 was parameterized as a function of relative humidity to account for greater
atmospheric conversion rates in areas of the country with higher humidity.
*** Wet deposition velocities are from (Yamartino, 1985).
A-4
July 2013
-------
Details of S-R Matrix Implementation in COBRA
In subsections below we provide the following implementation details on: (i) processing of the
EPA's emissions data to create COBRA emissions baseline; (ii) meteorological data sources and
processing; (iii) generating S-R transfer coefficients; (iv) approach taken to model secondary
PM2.5 formation (atmospheric chemistry); and (v) calibration of dispersion model outputs to the
monitored PM2.5.
Emissions Data
We use emissions data from the control case of the EPA's Mercury and Air Toxics Standards
Final Rule3 (hereafter, the MATS rule) to forecast ambient 2017 PM2.5 levels in COBRA.4 The
assumptions underlying the emission inventories are detailed in the Emissions Modeling for the
Final Mercury and Air Toxics Standards Technical Support Document (U.S. EPA, 2011). The
2017 control case developed by EPA for the MATS rule includes:
electrical generating unit emissions (reflecting the implementation of both MATS
and the Cross-State Air Pollution Rule),5
mobile emissions (reflecting the impacts of implementation of the Energy
Independence and Security Act of 2007 and the Energy Policy Act of 2005 on
mobile source fuels), and
average year fire data.
In addition to the 2017 control case emission inventories for the MATS rule, we used a 2005
base case emissions inventory to help develop calibration factors (discussed in more detail in a
later section). Exhibit A-2a and Exhibit A-2b summarize the 2005 and 2017 emissions data for
the continental U.S. that we used.
3 77 FR 9304-9513
4 Note that 2005 county-level natural emissions (from plants and soil) and were estimated using the BEIS 3.12
model (U.S. EPA, 2003a).
5 On August 21, 2012, the U.S. Court of Appeals for the D.C. Circuit issued an opinion that would vacate the Cross
State Air Pollution Rule. On October 5, 2012 the United States filed a petition seeking rehearing of that decision.
Further information about CSAPR (77 FR 34830) is available at: http://www.epa.gov/airtransport/.
A-S
July 2013
-------
Exhibit A-2a. 2005 Emissions Inventory Summary, by Tier 1 (tons/year)
Tier 1
no2
so2
pm25
voc
nh3
Fuel Combustion Electric Utilities
3,606,858
10,247,563
494,069
37,133
17,849
Fuel Combustion Industrial
487,117
689,195
85,972
18,493
12,609
Fuel Combustion Other
623,336
477,948
409,684
573,465
16,294
Chemical & Allied Product Manuf.
63
9
24
114,430
61
Metals Processing
85
45
85
464
5
Petroleum & Related Industries
449,228
10,187
4,853
911,027
Other Industrial Processes
12,106
6,207
214,009
51,510
59,857
Solvent Utilization
111
23
1,683
3,912,489
59
Storage & Transport
7,297
172
448
1,384,849
22
Waste Disposal & Recycling
66,712
10,795
253,681
366,801
22,673
Highway Vehicles
8,233,725
168,452
301,030
3,267,313
144,381
Off-Highway Vehicles
3,953,869
349,313
258,049
2,873,916
2,743
Natural Sources
1,060,915
31,695,823
Miscellaneous
255,686
70,645
1,823,717
2,184,552
3,312,043
Total
18,757,108
12,030,555
3,847,303
47,392,264
3,588,597
Exhibit A-2b. 2017
Emissions Inventory Summary,
by Tier 1 (tons/year)
Tier 1
no2
so2
nh3
pm25
VOC
Fuel Combustion Electric Utilities
1,788,581
1,874,027
34,165
223,168
41,353
Fuel Combustion Industrial
455,286
669,180
11,754
84,980
15,527
Fuel Combustion Other
588,335
404,129
12,482
203,589
265,831
Chemical & Allied Product Manuf.
64,125
4,982
62
24
119,767
Metals Processing
85
45
5
85
464
Petroleum & Related Industries
445,567
492
3,241
1,162,367
Other Industrial Processes
11,796
2,944
59,797
217,227
107,299
Solvent Utilization
111
23
59
1,683
3,863,540
Storage & Transport
7,297
172
22
447
1,055,120
Waste Disposal & Recycling
66,711
10,795
22,673
253,677
346,884
Highway Vehicles
3,204,285
29,282
85,362
129,392
1,397,412
Off-Highway Vehicles
2,443,196
9,878
3,359
148,006
1,556,261
Natural Sources
1,060,915
31,695,823
Miscellaneous
255,686
70,645
3,565,464
1,830,393
2,188,495
Total
10,391,977
3,076,594
3,795,204
3,095,911
43,816,143
A - 6
July 2013
-------
We estimate the formation of SOA using a fixed relationship between SOA and VOC for each
Tier 3 emission category.6 The inventory for the MATS rule estimated VOC but did not estimate
SOA, so we developed a simple approach to estimate the conversion of VOC to SOA, though
this conversion actually depends upon a number of factors including climate and the type of
VOC. We used the 2010 base case inventory of SOA and VOC emissions generated for the Clear
Skies Act (CSA) of 2003 (U.S. EPA, 2003b). For each Tier 3 emission category in this
inventory, we calculated the ratio of SOA to VOC. We then used these Tier 3 category-specific
ratios to estimate SOA in the MATS emissions inventory:
/SOAcsa Tier 3^\
SOAMATS, Tier 3 = VOCMATS, Tier 3 " I 777^ 1 )
\vuksA,Tier3/
When modeling emission sources, we categorized them into elevated point sources and
area/mobile sources. For each, we calculate an "effective stack" height, which takes into account
the actual stack height, gas temperature and velocity, stack diameter, and other factors. The
effective stack height is important as it is one of the greatest determinants7 of how far emissions
will disperse - generally the taller the effective stack the further the emissions might travel from
the source. In calculating effective stack height, we assume an average wind speed of 5 meters
per second using the plume rise algorithm from ISCST3 (U.S. EPA, 1995).
We group stationary point source emissions for each county into three groups based on effective
stack height: (1) less than 250 meters, (2) 250 to 500 meters, and (3) greater than 500 meters. We
assume that emissions from the two groups less than 500 meters originate from the center of the
county in which they are located. For point sources with effective stack heights greater than 500
meters, we use their true latitude and longitude coordinates when modeling the dispersion of
emissions.8
Emissions from both ground-level mobile and area sources in the contiguous U.S. are combined
at the county-level and modeled as emissions from stacks with an effective stack height of zero
located at the source county centroid. Exhibit A-3 summarizes these emission categories.
6 The emissions inventory in COBRA has fourteen broad Tier 1 categories (e.g., on-road motor vehicles), and within
each of these larger categories there are Tier 2 (e.g., diesels), and Tier 3 (e.g., heavy duty diesels) categories.
7 The other determinants include wind speed and direction as well as atmospheric chemistry.
8 For some counties, the emissions inventory contained more than one emission source with stack height greater than
500m. These emission sources normally have different locations and stack heights. To create a composite county-
level emissions source with stack height greater than 500m, we used the latitude and longitude of the source with the
tallest stack, whereas the composite stack height was an emissions-weighted average.
A - 7
July 2013
-------
Exhibit A-3. Emissions Categories for the S-R Matrix
Emissions Category
Effective Stack Height
Modeled Location
U.S. area and mobile emissions
0 m
County center
U.S. elevated point emissions
0-250 m
County center
U.S. elevated point emissions
250-500 m
County center
U.S. elevated point emissions
>500 m
True location
Meteorological Data
Meteorological variables were calculated from rawinsonde data on the NAMER-WINDTEMP
tapes9 obtained from the National Climatic Data Center. Winds for each of 100 sites throughout
North America were averaged for the following layers: the surface to 250 meters above ground
level (m AGL), 250-500 m AGL, 500-1000 m AGL, 1000-2000 m AGL, and 2000-4000 m
AGL. For each of these levels and for each of the 100 meteorological sites, a joint frequency
distribution of wind direction (16 cardinal directions) and wind speeds (11 speeds in 1 m/s
increments) was calculated for 1990.
These distributions were calculated separately for the twice-daily soundings. The early morning
soundings were assumed to be associated with the E stability category, and the late afternoon
soundings were assumed to be associated with the C stability category. Mixing heights were
determined from each sounding by calculating the virtual potential temperature. The annual
average afternoon mixing heights were calculated for each of the 100 meteorological sites and
were used to calculate the upper limit of vertical diffusion (/?,). The appropriate wind layer for
concentration calculations was determined using the centroid of the diffusing plume: oz for a
ground-based plume that has not yet mixed uniformly in the vertical, H for an elevated source,
and hJ2 for a uniformly mixed plume (E.H. Pechan & Associates Inc., 1994).
S-R Transfer Coefficients
The S-R matrix used in COBRA estimated the transport of the following emissions species: (1)
directly emitted PM2.5 and secondary organic aerosols (SOA), (2) sulfur dioxide (SO2), (3)
nitrogen dioxide (NO2), and (4) ammonia (NH3). These species were then used in the calculation
of ambient concentrations of PM2.5.
A matrix of source-receptor coefficients (in units of s/m3) spanning the entire contiguous U.S.
was developed for each of the four pollutants using the CRDM. For a unique combination of
source and receptor sites, a S-R transfer coefficient represents the incremental ambient air quality
impact in (J,g/m3 at the receptor resulting from a 1 |ig/s unit emission from the source. The S-R
matrix therefore provides a link between emission reductions and resulting air quality
concentrations. Concentration reductions that occur in proportion to a decrease in emissions at a
source are determined by the S-R coefficients for a given source and all receptors.
9 Refers to North America wind and temperature. These are standard data tapes for upper-air (rawinsonde) data
collected twice daily throughout North America. Rawinsondes are radar-tracked wind balloons.
A - 8
July 2013
-------
The pollutant concentration at a destination county is given by:
D]=TLKnrF^Fm, (6)
i c
where:
Concentration of pollutant s at destination county j ([j,g/m3)
Emission of pollutant 5 from emissions category c in source county i
(tons/year)
Transfer coefficient for pollutant 5 from source county i to destination
county j for emissions category c (sec/m3)
Ionic conversion factor for pollutant .s
Unit conversion factor (28,778 (j,g-year/ton-sec)
The ionic conversion factors are molecular weight ratios used to adjust the transfer coefficients
to reflect the concentration of precursors to secondarily-formed particulate species. Standard
molecular weights along with the ionic conversion factors used in this analysis are given in
Exhibit A-4 and Exhibit A-5.
Exhibit A-4. Standard Molecular Weights
Specie
Symbol
Standard molecular weight10
Nitrate ion
N03"
62.0049
Sulfate ion
S042"
96.0626
Bisulfate
HSO4
97.07054
Sulfur Dioxide
so2
64.0638
Nitrogen Dioxide
N02
46.0055
Ammonia
nh3
17.03052
Ammonium ion
nh4+
18.03846
Ammonium Nitrate
NH4N03
80.04336
Ammonium Bisulfate
NH4HS04
115.109
Ammonium Sulfate
(NH4)2S04
132.13952
DJ
K
rj-rS
Uj ~
Fs
111 Standard atomic weights from Coursey, et al. (2011).
A - 9
July 2013
-------
Exhibit A-5. Ionic Conversion Factors
Species
Ionic conversion factor, Fs
pm25, soa
1
S02 -~ SO42"
96.0626/64.0638
N02^N03"
62.0049 / 46.0055
nh3 -> nh4+
18.03846/ 17.03052
Atmospheric Chemistry
This section describes how secondary reactions are modeled in COBRA, including formation of
ammonium bisulfate (NH4HSO4), ammonium sulfate ((NH4)2S04), and ammonium nitrate
(NH4NO3). Note that the COBRA treats atmospheric chemistry involved in the formation of
these pollutants in a more simplified fashion than state-of-the-art air quality models11 (e.g.,
CALPUFF, AERMOD, CMAQ). We try to address this problem by calibrating COBRA
modeling results to measured PM2.5 concentrations as described in later in this Appendix.
Nevertheless uncertainty remains.
For the atmospheric chemistry in COBRA, in the presence of sulfate (SO42") and nitrate (MV),
ammonium (NH4+) reacts preferentially with S042" to form NH4HSO4 and (NH4)2S04. NH4NO3
is only formed under conditions of excess NH4+ and low temperatures. In each destination
county, the relative amounts of each secondary particle are subject to the following assumptions:
S042" is always assumed to be a particle;
NO3" is assumed to be a gas, unless is combines with NH4+;
NH4+ reacts first with SO42". The nature of the reaction depends on the relative amounts
of NH4+ and S042":
o If there is a little NH4+, then S042" will be converted to NH4HSO4 with potentially
some leftover SO42";
o If there is an intermediate amount of NH4+, then a combination of NH4HSO4 and
(NH4)2S04 will be obtained;
o If there is a lot of NH4+, then S042" will be completely converted to (NH4)2S04;
After all reactions between NH4+ and S042" occur, any remaining NH4+ reacts with N03"
to form NH4NO3.
Below we lay out the specifics of our approach:
Step 1: Calculate the mole ratio of NH4+ to SO42".
R = (NH|/18.03846)/(SO|-/96.0626):
11 See U.S. EPA (2012).
A -10
July 2013
-------
a) If R < 1 then we assume that a portion of SO42" converts to NH4HSO4
(S042" + NH4+ ~ NH4HSO4), while the rest remains as S042"
- Resulting concentration of NH4HSO4 is
15.109 ฆ min{(NH4 /18.03846), (SC>4~/96.0626)}
- Resulting concentration of remaining SO42" is
96.0626 ฆ ((SC>4~/96.0626) - (NH^/18.0 3 846))
b) If 1 < R < 2 then we assume that all SO42" converts to NH4HSO4
(SO42" + NH4+ > NH4HSO4) and a portion of NH4HSO4 converts to (NH4)2S04
(NH4HSO4 + NH4+ > (NH4)2S04). The second reaction will occur if there is
enough NH4+ remaining after the first reaction.
- Resulting concentration of NH4HSO4 is
115.109 ฆ (2(SC>4~/96.0626) - (NH^/18.03846))
- Resulting concentration of (NH4)2S04 is
132.13952 ฆ ((NH^/18.03846) - (SC>4~/96.0626))
c) If R > 2 then we assume that all SO42" converts to (NH4)2S04
(S042" + 2NH4+ -> (NH4)2S04).
- Resulting concentration of (NH4)2S04 is
132.13952 ฆ (S0|-/96.0626)
- Resulting concentration of NH4+ (remaining) is
18.03846 ฆ ((NH4 /18.03846) - 2(SO|_/96.0626))
Step 2: If NH4+remains after Step 1 (c), then NH4NO3 formation can take place. The
number moles of NO3" neutralized in this reaction will be:
moles ofNC>3 (neutralized) = min{(NH4 (remaining)/18.03846), (NO3 /62.0049)}.
Step 3: Particulate NH4NO3 is stable at relatively low temperatures. Following prior
usage of the S-R Matrix (e.g., NOx SIP Call), we assume that nitrate converts to
ammonium nitrate only a quarter of the time (i.e., the winter months). The annual average
concentration of NH4NO3 formed by the neutralization process is therefore:
80.04336 ฆ 0.25 ฆ moles ofNO3 (neutralized).
Step 4: The concentration of PM2 5 at the destination county is estimated as the sum of
concentrations of primary PM2 5, SOA, remaining SO42" (if any) and secondary
NH4HSO4, (NH4)2S04, and NH4N03:
PM2i5 (total) = PM2i5 (primary) + SOA
A-11
July 2013
-------
+SO4 (remaining)
+NH4HSO4 + (NH4)2S04 + NH4NO3
Calibration of S-R Matrix Outputs to Monitoring Data
We calibrated the S-R Matrix model estimates to actual monitoring data obtained from EPA. The
county-level calibration factors were estimated using a 2005 emissions inventory developed for
the MATS rule and 2005 data from EPA Federal Reference Method (FRM) monitor sites and
EPA/National Park Service Visibility Interagency Monitoring of Protected Visual Environments
(IMPROVE) program monitor sites.12
First, we used the S-R Matrix with the 2005 emissions inventory to estimate PM2.5 levels at the
center of each county. Second, we spatially interpolated the PM2.5 monitor data to generate a
monitor-based estimate for each county center as follows:
1. We pre-processed EPA motoring data to ensure that it did not contain any values flagged
as invalid and that minimum number of daily measurements per quarter was ll;13
2. We calculated quarterly average PM2.5 concentrations for all monitoring sites with
sufficient data;
3. For each quarter, we used an automatic kriging routine from R project package 'automap'
(Hiemstra, 2012) to interpolate quarterly average PM2.5 values to county centroids;
4. At each county centroid we then average over the interpolated quarterly average PM2.5
values to generate an annual average PM2.5 value.
We calculated a "calibration factor" for each county by dividing our monitor estimate by the
model estimate. These county-level calibration factors ranged from 0.35 to 3.02 with a mean
value of 1.25. For each state, Exhibit A-6 gives the average of the county-level monitor and
model values as well as the ratio of the two (the ratio being the average of the calibration
factors).
When calculating future year PM2.5 levels in COBRA, we use the calibration factors to adjust our
model estimate for each county in the following way:
PM2.5(calibrated model, 2017)
/PM7 q(interpolated monitor, 2005)\
= PM2 5(model, 2017) ฆ ^-^-7 rr
\ PM2.5(model, 2005) J
To sum up, the steps involved in the calculation of 2017 ambient PM2.5 levels in COBRA are the
following. We start the process by running the CRDM model, which generates the S-R Matrix
transfer coefficients. Emissions data for 2005 are run through the S-R Matrix and atmospheric
12 These data are available with the distribution of EPA's Modeled Attainment Test Software (U.S. EPA, 2010b).
13 The choice of 11 as the minimum number of site-days per valid quarter corresponds to > 75% completeness for
monitors on a 1 in 6 day schedule. This is a minimum number of samples that is routinely used in calculations of
quarterly average concentrations by EPA.
A-12
July 2013
-------
chemistry calculations applied to generate un-calibrated 2005 model estimates. Monitoring data
for 2005 were interpolated to the county-level, and were then compared with the 2005 model
estimates to generate calibration factors. Estimates of 2017 ambient PM2.5 levels can then be
generated by running the 2017 emissions data through the S-R Matrix. The resulting 2017 model
PM2.5 levels are then multiplied with the previously generated calibration factors to calculate a
best estimate of 2017 calibrated ambient PM2.5 levels.
Exhibit A-6. Monitor and Model Average PM2.S Levels (ug/m3) in 2005 and Average of Monitor to Model
Ratios by State
State
Monitor
Model
Mean
State
Monitor
Model
Ratio
AL
13.91
12.70
1.10
MT
5.13
3.33
1.55
AR
13.68
10.59
1.30
NC
13.63
12.85
1.08
AZ
6.13
5.16
1.26
ND
5.02
3.83
1.31
CA
7.86
7.79
1.15
NE
8.30
5.77
1.45
CO
5.25
5.62
0.97
NH
9.25
9.31
1.02
CT
11.50
12.35
0.93
NJ
13.27
15.95
0.85
DC
14.57
22.79
0.64
NM
5.36
5.44
1.01
DE
13.89
14.96
0.94
NV
6.60
4.57
1.58
FL
10.45
10.86
1.00
NY
11.89
11.35
1.13
GA
13.18
14.41
0.94
OH
15.98
14.03
1.15
IA
11.90
7.24
1.66
OK
10.29
8.68
1.21
ID
6.27
5.58
1.20
OR
5.93
4.72
1.45
IL
15.11
10.90
1.40
PA
14.41
13.53
1.10
IN
16.02
14.09
1.15
RI
10.77
10.35
1.05
KS
9.27
7.48
1.29
SC
13.06
12.87
1.02
KY
15.94
12.99
1.24
SD
7.09
4.49
1.58
LA
12.98
9.94
1.36
TN
15.21
12.48
1.22
MA
10.39
12.21
0.90
TX
9.87
6.99
1.50
MD
14.40
16.69
0.89
UT
5.40
5.52
1.01
ME
7.46
6.46
1.22
VA
14.50
14.15
1.04
MI
11.83
8.90
1.36
VT
9.04
7.98
1.13
MN
8.18
7.49
1.16
WA
5.74
3.81
1.65
MO
13.58
9.83
1.41
WI
10.92
8.08
1.37
MS
13.96
11.58
1.21
WV
15.73
12.83
1.24
WY
5.36
4.33
1.26
A-13
July 2013
-------
Appendix B: Derivation of Health Impact Functions
This appendix reviews the steps we performed in taking models from the epidemiological study
and converting them into health impact functions, which we then use to quantify the change in
adverse health effects due to a change in air pollution exposure. The most common functional
forms the log-linear and logistic, with a linear model used in some cases. All three are discussed
below.
Note that the log-linear and logistic generally produce comparable results, so the fact that some
health impacts are estimated with a logistic function and others with a log-linear function is not a
cause for concern. Indeed, in some circumstances, such as for small changes in air pollution, the
logistic and log-linear produce essentially the same result.
The Linear Model
A linear model between the adverse health effect, y, and the pollutant concentration, x, is of the
form
y = a + P ฆ x
A linear model includes the factors that are believed to affect the incidence of the health effect,
of which the pollutant would be one. So, the variable "a" in the linear function consists of all the
other independent variables in the regression, typically evaluated at their mean values, times
their respective coefficients.
The function describing the relationship between a change in x and the corresponding change in
incidence (rate) of the health effect from the baseline level (y&) to the post-control level (yc) is
then:
Ay = yb-yc =p-(xb-xc) = 0-Ax
Ify denotes an incidence rate, then Ay denotes the change in the incidence rate. If denotes an
incidence count, then the fi is first divided the baseline study population to generate an incidence
rate. Ax is the difference between the baseline level of the pollutant concentration and the
control level of the pollutant concentration: Xb - xc. (Note that typically a control strategy is
intended to decrease the pollutant levels, so we expect Ax to be positive.) The expected number
of cases avoided would then be calculated by multiplying Ay by the relevant population:
CasesAvoided = /3 ฆ Ax population
The coefficient, fi, and standard error of fi {op) are reported directly in studies presenting results
from linear regression models.
B-l
July 2013
-------
The Log-linear Model
The most commonly used functional form for criteria air pollutant concentration-response
functions is the log-linear model. It defines the relationship between x and >' to be of the form:
y = B- exp(/? x)
or, equivalently,
In (y) = a + /3-x,
where the parameter B is the incidence (rate) corresponding to the zero pollutant concentration (x
= 0); the coefficient /> is the effect of pollutant x on the natural logarithm of the incidence (rate) >'
- In(y); and a = ln^g).1
Estimating Avoided Cases
The relationship between Ax and Ay is:
ky = yb -yc = 5(exp(/^)- exp(A"J)
This may be rewritten as:
Ay=yb-
l
p(/?-Ax)y
ex
where yb is the baseline incidence (rate) of the health effect - i.e., the incidence (rate) before the
change in x. Ify is incidence rate rather than incidence count, then the change in incidence rate,
Ay, must be multiplied by the relevant population to get the expected number of cases avoided.
For example, if y denotes the annual number of cases of the adverse health effect per 100,000
population then the expected number of cases avoided is calculated as:
CasesAvoided = population
100,000
i- 1
p(/?-Ax)y
ex
Estimating the Coefficient (/?)
Epidemiological studies that estimate log-linear concentration-response functions often report a
relative risk for a specific Ax, rather than the coefficient, /?, in the function itself. The relative
risk (RR) is simply the ratio of two risks corresponding to two levels of pollutant concentration -
the "high" riskyhigh (corresponding to the higher pollutant level, x = x/.,and the lower risk y/ov
(corresponding to the lower pollutant level, x = x/ow):
1 Other covariates besides pollution clearly affect mortality. The parameter B might be thought of as containing these
other covariates, for example, evaluated at their means. That is, B = B0exp{fiiXi + ... + where B0 is the incidence ofy
when all covariates in the model are zero, and xb ... , xn are the other covariates evaluated at their mean values. The
parameter B drops out of the model, however, when changes in y are calculated, and is therefore not important.
B-2
July 2013
-------
_ y high
y tow
Using the original log-linear function above, it can be shown that the relative risk associated with
a specific change in pollutant concentration of Ax* = Xhigh - xiow can be written as
RR\^=^L = exp(j3-Ax*)
yiow
Taking the natural log of both sides, the coefficient in the function underlying the relative risk
can be derived as:
q In (RR)
Ax*
Once the pollutant coefficient, fi, has been calculated, the change in incidence (rate), Ay,
corresponding to any change in pollutant concentration, Ax, can be calculated, using the
relationship between Ax and Ay given above, the baseline incidence (rate) and assessment
population.
There are instances when epidemiological studies report percent increase in the relative risk,
rather than relative risk itself (see for example, Moolgavkar (2003)). Given a reported x percent
increase in the relative risk, we calculate the relative risk as RR = l+exp(x/100). Then we
proceed to calculating /? as described above.
Estimating the Standard Error of /? (op)
The standard error of fi {op) is not often directly reported in studies presenting results from log-
linear regression models. Results are most commonly presented as a relative risk and 95%
confidence interval. The 95% confidence interval is defined as follows:
CI95% = exp (j3 ฆ Ax ฑ 1.96 g p Ax)
Based on this equation, the standard error of fi {op) can be estimated from the relative risk (RR),
upper limit of the 95% confidence interval (UL), and lower limit of the 95% confidence interval
(LL), as follows:
_ Pugh ~P _ (ln((IL)/Ax-\n(RIi)/Ax) _ P~ Plow _ (ln(i?i?)/Ax-ln(ZZ)/Ax)
^fi,high ~ 1% " 1% an V" 196 ~ 196
^ 6, high ^ ji. low Phigh Plow
ฐP= 1 0r ฐP ='
3.92
Some studies report only a central effect estimate and ^-statistic. The ^-statistic describes the
strength of the observed pollutant-health effect association. It is defined as the ratio of the
B-3
July 2013
-------
coefficient, /?, to the standard error of ft (op). The standard error of ft (op) can, therefore, be
estimated from the ^-statistic as follows:
CT'=7
B-4
July 2013
-------
The Logistic Model
In some epidemiological studies, a logistic model is used to estimate the probability of an
occurrence of an adverse health effect. Given a pollutant level, x, and a vector of other
explanatory variables, Z, the logistic model assumes the probability of an occurrence is:
D( I a exp(/?-x)exp(a-Z)
v = P\occurrence\B -x,a- Z) = ^^,
V ' ' 1 + exp(/? x)exp(a z)
where (J> is the coefficient of the pollutant concentration, x, and a is a vector of coefficients of the
variables in the vector Z.2
Estimating Avoided Cases
The change in the probability of an occurrence (Ay) corresponding to a change in the level of the
pollutant from Xb to xc (= Ax), all other covariates held constant, may be derived from the original
C-R function above:
Ay = yb-yc = yb
rt ! A
(\-yb)-exp(/J-Ax)+yb
Once again, to calculate the expected number of avoided cases of the adverse effect, it is
necessary to multiply by the population:3
( 1 ^
Cases Avoided = yb 1 population
(l->fc)'exp(/?-Axj + j^ J
Estimating the Coefficient (/?)
The estimated pollutant coefficient, fi, in the original function is typically not reported in studies
that use the logistic model. Instead, the odds ratio (OR) corresponding to a specific change in x
is reported.
The odds of an occurrence is defined as:
Odds = ฆ ^
i-y
It can be shown that:
Odds = - = exp(/? x)exp(a ฆ Z)
i-y
2 Greene (1997, Chapter 19) presents models with discrete dependent variables; in particular, page 874 presents the logit
model. See also Judge et al. (1985, p. 763).
3 Note that because Ay here is a change in probability of occurrence (rather than a change in the rate per 100,000
population), it is necessary to multiply by the population rather than by the population/100,000.
B-5
July 2013
-------
The odds ratio is just the ratio of the odds when the pollutant is at a specified higher level, Xhigh,
to the odds when the pollutant is at a specified lower level, x/ow:
QR = exp(/^ )exp(g.z) = exp^J =
exPl/?' xiow )exp(cir Z) exp(/?-xtoMJ
Often the odds ratio corresponding to a specified change in x, call it Ax*, is the only measure of
the effect of x reported from a study using a logistic model (just as the relative risk
corresponding to a specified change in x is often the only measure of the effect of x reported
from a study using a log-linear model). However, it is easy to calculate the underlying pollutant
coefficient, /?, from the odds ratio as follows:
OR\^* = exp(/?-Ax*) -> ln((9i?) = /? Ax* -> /? =
Given the pollutant coefficient, fi, and the baseline probability of occurrence, y/,, the change in
the probability, Ay, associated with any change in pollutant concentration, Ax, can be derived
using the equation for Ay above. The expected number of avoided cases of the adverse effect is
then obtained by multiplying by the population.
Estimating the Standard Error of /? (op)
The standard error of fi {op) is not often directly reported in studies presenting results from
logistic regression models. Results are most commonly presented as an odds ratio and 95%
confidence interval. The 95% confidence interval is defined as follows:
CI95% = exp (j3 ฆ Ax ฑ 1.96 g p Ax)
Based on this equation, the standard error of fi {op) can be estimated from the odds ratio (OR),
upper limit of the 95% confidence interval (UL), and lower limit of the 95% confidence interval
(LL), as follows:
Push ~ P (in ((//,)/Ax - ln((9i?)/Ax) _ A _ _ /?-Plow _ (ln(Oi?)/Ax - In(/./.)/Ax)
PMgh 1% 1% ft Jaw jgg jgg
^ ft, high ^ ft,low Phigh Plow
ฐ ft = 1 0r ฐ ft = '
3.92
Some studies report only a central effect estimate and ^-statistic. The ^-statistic describes the
strength of the observed pollutant-health effect association. It is defined as the ratio of the
coefficient, fi, to the standard error of (J> (erg). The standard error of (J> (erg) can, therefore, be
estimated from the ^-statistic as follows:
CT'=7
B-6
July 2013
-------
Appendix C: Health Impact Functions
A reduction in ambient PM2.5 levels is associated with reductions in a number of adverse health
effects, or "endpoints." This appendix discusses the calculation of avoided adverse health effects.
The health impact functions in the COBRA model were prepared by Abt Associates in close
consultation with EPA and rely on an up-to-date assessment of the published scientific literature
to ascertain the relationship between particulate matter and adverse human health effects. We
evaluated studies using a variety of selection criteria, including: study location and design, the
characteristics of the study population, and whether the study was peer-reviewed (Exhibit C-l).
Exhibit C-l. Summary of Considerations Used in Selecting Studies
Consideration Comments
Peer reviewed research Peer reviewed research is preferred to research that has not undergone the peer review
process.
Among studies that consider chronic exposure (e.g., over a year or longer) prospective
cohort studies are preferred over cross-sectional studies because they control for important
individual-level confounding variables that cannot be controlled for in cross-sectional
studies.
Studies examining a relatively longer period of time (and therefore having more data) are
preferred, because they have greater statistical power to detect effects. More recent
studies are also preferred because of possible changes in pollution mixes, medical care,
and life style over time.
Studies examining a relatively large sample are preferred because they generally have
more statistical power to detect small magnitude effects. A large sample can be obtained
in several ways, either through a large population, or through repeated observations on a
smaller population, e.g. through a symptom diary recorded for a panel of asthmatic
children.
U.S. studies are more desirable than non-U.S. studies because of potential differences in
pollution characteristics, exposure patterns, medical care system, population behavior and
life style.
For this analysis, C-R functions based on PM2 5 are preferred to those based on PMn,
(particulate matter less than 10 microns in aerodynamic diameter) because reductions in
emissions from diesel engines are expected to reduce fine particles and not have much
impact on coarse particles.
Economically valuable Some health effects, such as changes in forced expiratory volume and other technical
health effects measurements of lung function, are difficult to value in monetary terms. These health
effects are therefore not quantified in this analysis.
Non-overlapping Although the benefits associated with each individual health endpoint may be analyzed
endpoints separately, care must be exercised in selecting health endpoints to include in the overall
benefits analysis because of the possibility of double counting of benefits. Including
emergency room visits in a benefits analysis that already considers hospital admissions,
for example, will result in double counting of some benefits if the category "hospital
admissions" includes emergency room visits.
Model Selection
In many epidemiological studies of air pollution and health, researchers estimate and present
numerous single pollutant and multi-pollutant models for the same pollutant and health endpoint.
These models may differ from each other in a number of characteristics, including: the functional
Study type
Study period
Study size
Study location
Measure of PM
C-l
July 2013
-------
form of the model, the covariates included in the model, the pollutant exposure metric, the lag
structure, and the study population.
For the purposes of estimating health benefits associated with pollutant changes, it is neither
realistic nor advantageous to include every model presented in each study. However, it is
important that a relatively objective process be used to select from among models. Described
below are the criteria that were used as guidance in the selection of a particular model from
among several models presented in a study. It is not possible in all cases to select a model using a
completely objective and mechanical process. In many cases, professional judgment and an
understanding of the study context are necessary as well to select the most appropriate models.
Exhibit C-2 summarizes the selection criteria that we used.
Exhibit C-2. Description of Selection Criteria
Selection Criteria
Description
Goodness-of-fit statistics
If an appropriate measure of goodness of fit (i.e., how well the model fit the data)
is reported for each of several models in a study, then this measure may be used
as the basis on which to select a model.
Best captures distributed lag
Select the model that appears to best capture a distributed lag effect, as described
below. If multiple single-lag models and/or moving average models are specified,
select the model with the largest effect estimate, all else equal.
Best set of control variables
Select the model which includes temporal variables (i.e. season, weather patterns,
day of the week) and other known non-pollutant confounders, all else equal.
Select the model which uses the most sophisticated methods of capturing the
relationship between these variables and the dependent variable (e.g., affords the
most flexibility in fitting possible nonlinear trends).
Useful for health effects
modeling
The model must be in a form that is useful for health effects modeling (e.g., the
pollutant variable should be a continuous variable rather than a categorical
variable).
Sample size
Select the model estimated with the larger sample size, all else equal.
Distributed Lag Effect
The question of lags and the problems of correctly specifying the lag structure in a model has
been discussed extensively (U.S. EPA, 2002, Section 8.4.4). In many time-series studies, after
the basic model is fit (before considering the pollutant of interest), several different lags are
typically fit in separate single-lag models and the most significant lag is chosen. The 2002 draft
PM2.5 CD notes that "while this practice may bias the chance of finding a significant association,
without a firm biological reason to establish a fixed pre-determined lag, it appears reasonable"
(U.S. EPA, 2002, p. 8-237).
There is recent evidence (Schwartz, 2000) that the relationship between PM25 and health effects
may best be described by a distributed lag (i.e., the incidence of the health effect on day n is
influenced by PM25 concentrations on day n, day n-1, day n-2 and so on). If this is the case, a
model that includes only a single lag (e.g., a 0-day lag or a 1-day lag) is likely to understate the
total impact of PM25. The 2002 draft PM2 5 CD makes this point, noting that "if one chooses the
C-2
July 2013
-------
most significant single lag day only, and if more than one lag day shows positive (significant or
otherwise) associations with mortality, then reporting a RR [relative risk] for only one lag would
also underestimate the pollution effects" (U.S. EPA, 2002, p. 8-241). The same may hold true for
other pollutants that have been associated with various health effects.
Several studies report similar models with different lag structures. For example, Moolgavkar
(2000a) studied the relationship between air pollution and respiratory hospital admissions in
three U.S. metropolitan areas. The author reports models with PM2.5 lagged from zero to five
days. Since the lagging of PM2 5 was the only difference in the models and the relationship is
probably best described using a distributed lag model, any of single-lag effect estimates are
likely to underestimate the full effect. Therefore, we selected the model with the largest effect
estimate.
Pooling
There is often more than one study that has estimated a health impact function for a given
pollutant-health endpoint combination. Each study provides an estimate of the pollutant
coefficient, fi, along with a measure of the uncertainty of the estimate. Because uncertainty
decreases as sample size increases, combining data sets is expected to yield more reliable
estimates of fi, and therefore more reliable estimates of the incidence change predicted using /?.
Combining data from several comparable studies in order to analyze them together is often
referred to as meta-analysis.
For a number of reasons, including data confidentiality, it is often impractical or impossible to
combine the original data sets. Combining the results of studies in order to produce better
estimates of fi provides a second-best but still valuable way to synthesize information. This is
referred to as pooling. Pooling /f s requires that all of the studies contributing estimates of fi use
the same functional form for the health impact function. That is, the /f s must be measuring the
same thing.
To be consistent with the recent EPA benefits analyses, COBRA uses a random-/ fixed- effects
pooling procedure (see U.S. EPA, 2009, p. 5-18), which is a method for weighting estimates
involving using their variances. Variance takes into account both the consistency of data and the
sample size used to obtain the estimate, two key factors that influence the reliability of results.
The method is based on DerSimonian and Laird (1986).
Fixed Effect Weights
The fixed effects model assumes that there is a single true concentration-response relationship
and therefore a single true value for the parameter fi that applies everywhere. Differences among
/f s reported by different studies are therefore simply the result of sampling error. That is, each
reported fi is an estimate of the same underlying parameter. The certainty of an estimate is
reflected in its variance (the larger the variance, the less certain the estimate). Fixed effects
pooling therefore weights each estimate under consideration in proportion to the inverse of its
variance:
C-3
July 2013
-------
n=1 vn / n=\ v
Where
N- number of studies;
fin - estimate provided by study //;
vn - variance of the estimate provided by study //;
Pfe - pooled fixed effects estimate.
Random- / Fixed- Effect Weights
An alternative to the fixed effects model is the random effects model, which allows the
possibility that the estimates [J> from the different studies may in fact be estimates of different
parameters, rather than just different estimates of a single underlying parameter. In studies of the
effects of PM2.5 on hospitalizations for COPD, for example, if the composition of PM2.5 varies
among study locations the underlying relationship between the frequency of hospitalizations for
COPD and PM2.5 may be different from one study location to another. This would violate the
assumption of the fixed effects model.
It is possible to test whether it is appropriate to base the pooling on the random effects model (vs.
the fixed effects model). A test statistic, Qw, the weighted sum of squared differences of the
separate study estimates from the pooled estimate based on the fixed effects model /?/e, is
calculated as:
Under the null hypothesis that there is a single underlying parameter, /?, of which all the /?' s are
estimates, Qw has a chi-squared distribution with TV-1 degrees of freedom. (Recall that TV is the
number of studies in the meta-analysis.) If Qw is greater than the critical value corresponding to
the desired confidence level, the null hypothesis is rejected. That is, in this case the evidence
does not support the fixed effects model, and the random effects model is assumed, allowing the
possibility that each study is estimating a different /?. We use a five percent one-tailed test.
The random effect model-based pooling must take into account not only the within-study
variances (used in a meta-analysis based on the fixed effects model) but the between-study
variance as well. The between-study variance, if, is given by:
v
n=1
n
C - 4
July 2013
-------
(i.e., if Qw < N-1), then if is a negative number, and it is not possible to calculate a random
effects estimate. In this case, however, the small value of Qw would presumably have led to
accepting the null hypothesis described above, and the meta-analysis would be based on the
fixed effects model. The remaining discussion therefore assumes that if is positive.
Given a value for if, the random effects estimate is calculated in almost the same way as the
fixed effects estimate. However, the pooled estimate now incorporates both the within-study
variance (v) and the between-study variance (if):
N B /N 1
+vn/ +v
Where
N- number of studies;
[-> - estimate provided by study n,
vn - variance of the estimate provided by study n,
if - within-study variance;
Pre - pooled random effects estimate.
The weighting scheme used in a pooling based on the random effects model is basically the same
as that used if a fixed effects model is assumed, but the variances used in the calculations are
different. This is because a fixed effects model assumes that the variability among the estimates
from different studies is due only to sampling error (i.e., each study is thought of as representing
just another sample from the same underlying population), while the random effects model
assumes that there is not only sampling error associated with each study, but that there is also
between-study variability - each study is estimating a different underlying p. Therefore, the sum
of the within-study variance and the between-study variance yields an overall variance estimate.
Thresholds
Health impact functions have been developed with and without explicit thresholds. A threshold
means that air pollution levels below the specified threshold have no adverse health effects. In
some prior regulatory impact assessments (e.g., U.S. EPA, 2006) assumed a threshold of 10
|ig/m3 for PM25. However, EPA's most current understanding of the scientific literature is that
there is no threshold in the relationship between PM2.5 and adverse health impacts. In its recent
analysis of proposed NO2 national ambient air quality standards, U.S. EPA (2009) used a no-
threshold model to calculate PM2.5 co-benefits down to the lowest modeled PM2.5 air quality
levels.
Following EPA's updated methodology, we also assume there is no threshold for modeling
PM2.5-related health effects. This is supported by the National Research Council (2002) in its
review of methods for estimating the public health benefits of air pollution regulations. They
concluded that there is no evidence for any departure from linearity in the observed range of
C - 5
July 2013
-------
exposure to PMio or PM2.5, nor is there any indication of a threshold. They cite the weight of
evidence available from both short- and long-term exposure models and the similar effects found
in cities with low and high ambient concentrations of PM2.5. More recently, Schwartz et al
(2008) reached the same conclusion, finding a linear relationship between PM2.5 and premature
mortality with no evidence of a threshold.
In addition, U.S. EPA completed an "expert elicitation" analysis in which it elicited opinions
from 12 experts (in epidemiology, toxicology, and medicine) on the nature of this relationship
(see: Industrial Economics Incorporated (IEc), 2006). The experts were asked how likely they
thought it is that the relationship between PM2.5 and mortality is causal, and if it is causal, what is
the functional form of the C-R relationship, including whether there is a threshold. Eleven of the
twelve experts thought that, although each individual may have a threshold, there is insufficient
empirical evidence for a threshold for the population, which is the entity of interest in a C-R
function. Only one expert did include the possibility of a population threshold, assigning a
probability of 50 percent to there being a threshold and, if there is a threshold, an 80 percent
chance that it is less than or equal to 5 |ig/m3 (which is below the level of PM2.5 observed in
epidemiological studies), and a 20 percent chance that it is between 5 and 10 |ig/m3.
C-6
July 2013
-------
Summary of Health Impact Functions Used in COBRA
In this Appendix, we present the health impact functions used to estimate PM2.5-related adverse
health effects. Exhibit C-3 summarizes the epidemiological studies in COBRA used to estimate
adverse health impacts of PM2.5. Each sub-section has an exhibit with a brief description of the
health impact function and the underlying parameters. Following each exhibit, we present a brief
summary of each study and any information that is unique to that study.
Exhibit C-3. Epidemiological Studies Used to Estimate Adverse Health Impacts of PM2 5
Endpoint
Author
Age
Mortality, All Cause
Krewski et al. (2009)
30-99
Mortality, All Cause
Lepeule et al. (2012)
25-99
Mortality, All Cause
Woodruff et al. (1997)
Infant
Acute Myocardial Infarction, Nonfatal
Peters et al. (2001)
18-99
Acute Myocardial Infarction, Nonfatal
Pope et al. (2006)
18-99
Acute Myocardial Infarction, Nonfatal
Sullivan et al. (2005)
18-99
Acute Myocardial Infarction, Nonfatal
Zanobetti and Schwartz (2006)
18-99
Acute Myocardial Infarction, Nonfatal
Zanobetti et al. (2009)
18-99
HA, All Cardiovascular (less Myocardial Infarctions)
Bell et al. (2008)
65-99
HA, All Cardiovascular (less Myocardial Infarctions)
Moolgavkar (2000b)
18-64
HA, All Cardiovascular (less Myocardial Infarctions)
Peng et al. (2008)
65-99
HA, All Cardiovascular (less Myocardial Infarctions)
Peng et al. (2009)
65-99
HA, All Cardiovascular (less Myocardial Infarctions)
Zanobetti et al. (2009)
65-99
HA, All Respiratory
Zanobetti et al. (2009)
65-99
HA, All Respiratory
Kloog et al. (2012)
65-99
HA, Asthma
Babin et al. (2007)
0-17
HA, Astluna
Sheppard (2003)
0-17
HA, Chronic Lung Disease
Moolgavkar (2000a)
18-64
Emergency Room Visits, Astluna
Mar et al. (2010)
0-99
Emergency Room Visits, Astluna
Slaughter et al. (2005)
0-99
Emergency Room Visits, Astluna
Glad et al. (2012)
0-99
Acute Bronchitis
Dockery et al. (1996)
8-12
Astluna Exacerbation. Cough
Mar et al. (2004)
6-18
Astluna Exacerbation, Cough
Ostro et al. (2001)
6-18
Astluna Exacerbation, Shortness of Breath
Mar et al. (2004)
6-18
Astluna Exacerbation, Shortness of Breath
Ostro et al. (2001)
6-18
Astluna Exacerbation, Wheeze
Ostro et al. (2001)
6-18
Minor Restricted Activity Days
Ostro and Rothschild (1989)
18-64
Lower Respiratory Symptoms
Schwartz and Neas (2000)
7-14
Upper Respiratory Symptoms
Pope et al. (1991)
9-11
Work Loss Days
Ostro (1987)
18-64
Note that Appendix B mathematically derives the standard types of health impact functions that
we encountered in the epidemiological literature, such as, log-linear, logistic and linear, so we
simply note here the type of functional form. Appendix D presents a description of the sources
for the incidence and prevalence data used in these health impact functions.
C-7
July 2013
-------
Mortality
Health researchers have consistently linked air pollution, especially PM2.5, with excess mortality.
Although a number of uncertainties remain to be addressed, a substantial body of published
scientific literature recognizes a correlation between elevated PM2.5 concentrations and increased
mortality rates. Based on the scientific evidence, EPA's Integrated Science Assessment
determined a causal relationship between PM2.5 and premature mortality
(http://www.epa.gov/ncea/isa/).
Both long- and short-term exposures to ambient levels of particulate matter air pollution have
been associated with increased risk of premature mortality. It is clearly an important health
endpoint because of the size of the mortality risk estimates, the serious nature of the effect itself,
and the high monetary value ascribed to avoiding mortality risk. Because of the importance of
this endpoint and the considerable uncertainty among economists and policymakers as to the
appropriate way to estimate PM-related mortality risks, this section discusses some of the issues
surrounding the estimation of premature mortality associated with PM2.5.
Particulate matter has been linked with premature mortality in adults in multiple studies
throughout the world (Jerrett et al., 2005; Katsouyanni et al., 2001; Laden et al., 2006; Pope et
al., 2002; Samet, Dominici, Curriero, Coursac, & Zeger, 2000) as well as infants (Bobak &
Leon, 1999; Conceicao, Miraglia, Kishi, Saldiva, & Singer, 2001; Loomis, Castillejos, Gold,
McDonnell, & Borja-Aburto, 1999; Woodruff, Darrow, & Parker, 2008; Woodruff et al., 1997).
To estimate premature mortality in adults, we use an epidemiological analysis of the American
Cancer Society cohort by Krewski et al. (2009) and analysis of the Six-City cohort by Lepeule et
al. (2012). To estimate premature mortality in infants, we used a study by Woodruff et al. (1997).
Exhibit C-4. Health Impact Functions for Particulate Matter and All-Cause Mortality
Author
Year
Location
Age
Metric
Beta
Std Err
Functional
Form
Krewski et al.
2009
116 U.S.
30-99
Annual
0.005827
0.000963
Log-linear
cities
Lepeule et al.
2012
6 Eastern
25-99
Annual
0.013103
0.003347
Log-linear
cities
Woodruff et al.
1997
86 cities
0-0
Annual
0.003922
0.001221
Logistic
Note that COBRA does not pool Krewski et al. (2009) and Lepeule et al. (2012) to estimate
premature mortality in adults. In recent analysis of proposed NO2 national ambient air quality
standards, U.S. EPA (2009) used Pope et al. (2002)1 and Laden et al. (2006) to estimate the
PM2.5 mortality-related co-benefits and presented the results separately for each study:
"These are logical choices for anchor points in our presentation because, while both
studies are well designed and peer reviewed, there are strengths and weaknesses inherent
1 Krewski et al. (2009) is an extended and updated analysis of Pope et al. (2002).
C-8
July 2013
-------
in each, which we believe argues for using both studies to generate benefits estimates.
Previously, EPA had calculated benefits based on these two empirical studies, but derived
the range of benefits, including the minimum and maximum results, from an expert
elicitation of the relationship between exposure to PM2.5 and premature mortality (Roman
et al., 2008). Within this assessment, we include the benefits estimates derived from the
concentration-response function provided by each of the twelve experts to better
characterize the uncertainty in the concentration-response function for mortality and the
degree of variability in the expert responses. Because the experts used these cohort
studies to inform their concentration-response functions, benefits estimates using these
functions generally fall between results using these epidemiology studies (see Figure
5.9). In general, the expert elicitation results support the conclusion that the benefits of
PM2.5 control are very likely to be substantial." p. 5-25.
Mortality, All Cause (Krewski et al., 2009)
This cohort study consists of approximately 360,000 participants residing in areas of the country
that have adequate monitoring information on levels of PM2.5 for 1980 and about 500,000
participants in areas with adequate information for 2000. The causes of death that were analyzed
included all causes, cardiopulmonary disease (CPD), ischemic heart disease (IHD), lung cancer,
and all remaining causes. Data for 44 personal, individual-level covariates, based on participants'
answers to a 1982 enrollment questionnaire, were also used for the analyses. The authors also
collected data for seven ecologic (neighborhood-level) covariates, each of which represents local
factors known or suspected to influence mortality, such as poverty level, level of education, and
unemployment (at both zip code and city levels). Long-term average exposure variables were
constructed for PM2.5 from monitoring data for two periods: 1979-1983 and 1999-2000. Similar
variables were constructed for long-term exposure to other pollutants of interest from single-year
(1980) averages, including total suspended particles, ozone, nitrogen dioxide, and sulfur dioxide.
Exposure was averaged for all monitors within a metropolitan statistical area (MSA) and
assigned to participants according to their Zip Code area (ZCA) of residence.
The authors chose the standard Cox proportional-hazards model (and a variation to allow for
random effects) to calculate hazard ratios for various cause-of-death categories associated with
the levels of air pollution exposure in the cohort. They extended the random effects Cox model
to accommodate two levels of information for clustering and for ecologic covariates. Three main
analyses were conducted: a Nationwide Analysis, Intra-Urban Analyses in the New York City
(NYC) and Los Angeles (LA) regions, and an analysis designed to investigate whether critical
time windows of exposure to pollutants might have affected mortality in the cohort. Using a
multi-pollutant model (03, S04, S02, TSP, and PM2.5), the authors reported a relative risk (1.06)
for all-cause mortality and the corresponding 95% confidence interval (95% CI: 1.04-1.08) for a
10 |ig/m3 increase in the average of PM2.5 exposure level for 1999-2000 (Krewski et al., 2009,
Commentary Table 4). The results were adjusted for the 44 individual-level covariates and the 7
ecologic covariates at the MSA & DIFF levels.
Functional Form: Log-linear
C-9
July 2013
-------
Coefficient: 0.005827
Standard Error: 0.000963
Incidence Rate: county-specific annual all-cause mortality rate per person ages 30 and older
Population: population of ages 30 and older
Mortality, All Cause (Lepeule et al., 2012)
Lepeule et al. (2012) is an extended and updated analysis of Laden et al. (2006). The authors
performed an extended mortality follow-up from 1979-2009 using data from the Harvard Six
Cities adult cohort study. They used annual city-specific PM2.5 concentrations and assigned for
each participant until death or censoring. The authors replicated the previously applied Cox
regression (as used in Laden et al., 2006), and examined different time lags, the shape of the
concentration-response relationship using penalized splines, and changes in the slope of the
relation over time. Then they conducted Poisson survival analysis with time-varying effects for
smoking, sex, and education. The authors found a significant increase in the overall mean
mortality associated with a 10-[j,g/m3 increase in PM2.5
The coefficient and standard error are estimated from the relative risk (1.14) and 95% confidence
interval (1.07-1.22) associated with a 10-[^g/m3 increase in PM2.5 (Lepeule et al., 2012, Table 2).
Functional Form: Log-linear
Coefficient: 0.013103
Standard Error: 0.003347
Incidence Rate: county-specific annual all-cause mortality rate per person ages 25 and older
Population: population of ages 25 and older
Infant Mortality (Woodruff et al., 1997)
In a study of four million infants in 86 U.S. metropolitan areas conducted from 1989 to 1991,
Woodruff et al. (1997) found a significant link between PMi0 exposure in the first two months of
an infant's life with the probability of dying between the ages of 28 days and 364 days. PMi0
exposure was significant for all-cause mortality. PMio was also significant for respiratory
mortality in average birth-weight infants, but not low birth-weight infants.
The coefficient and standard error are based on the odds ratio (1.04) and 95% confidence interval
(1.02-1.07) associated with a 10 (j,g/m3 change in PMio (Woodruff et al., 1997, Table 3).
Functional Form: Logistic
Coefficient: 0.003922
Standard Error: 0.001221
Incidence Rate: county-specific annual post-neonatal2 infant deaths per infant under the age of
one
Population: population of infants under one year old
2 Post-neonatal refers to infants that are 28 days to 364 days old.
C-10
July 2013
-------
Non-Fatal Heart Attack
Non-fatal heart attacks have been linked with short-term exposures to PM2.5 in the U.S. (Peters et
al., 2001) and other countries (Poloniecki, Atkinson, de Leon, & Anderson, 1997).3 We used the
C-R functions reported in five studies as shown in Exhibit C-5.
The finding of a specific impact on heart attacks is consistent with hospital admission and other
studies showing relationships between fine particles and cardiovascular effects both within and
outside the U.S. These studies provide a weight of evidence for this type of effect. Several
epidemiological studies (Gold et al., 2000; Liao et al., 1999; Magari et al., 2001) have shown that
heart rate variability (an indicator of how much the heart is able to speed up or slow down in
response to momentary stresses) is negatively related to PM2.5 levels. Lack of heart rate
variability is a risk factor for heart attacks and other coronary heart diseases (Dekker et al., 2000;
Liao et al., 1997; Tsuji et al., 1996). As such, the reduction in heart rate variability due to PM2.5
is consistent with an increased risk of heart attacks.
Exhibit C-5. Health Impact Functions for Particulate Matter and Non-fatal Heart Attack
Author Year Location Age Metric
Beta Std Error Fuปctional
Form
Peters et al.
Pope et al.
Sullivan et al.
Zanobetti and
Schwartz
Zanobetti et al.
2001
2006
2005
2006
2009
Boston, MA
Greater Salt Lake
City, Utah
King County,
Washington
Greater Boston
area (Middlesex,
Norfolk, Suffolk
Counties)
26 U.S.
Communities
18-99
24-hr avg
0.024121
0.009285
Logistic
All
24-hr avg
0.00481
0.001992
Logistic
All
24-hr avg
0.001980
0.002241
Logistic
All 24-hr avg 0.005300 0.002213 Logistic
All 24-hr avg 0.00225 0.000592 Log-linear
COBRA reports two sets of incidence results: (1) incidence results based on C-R function from
Peters et al. (2001); (2) pooled incidence based on other four studies using random/fixed effects
pooling method.
Acute Myocardial Infarction (Heart Attacks), Nonfatal (Peters et al., 2001)
Peters et al. (2001) studied the relationship between increased particulate air pollution and onset
of heart attacks in the Boston area from 1995 to 1996. The authors used air quality data for PM10,
PM10-2.5, PM2.5, "black carbon", O3, CO, NO2, and SO2 in a case-crossover analysis. For each
subject, the case period was matched to three control periods, each 24 hours apart. In univariate
analyses, the authors observed a positive association between heart attack occurrence and PM2.5
levels hours before and days before onset. The authors estimated multivariate conditional logistic
3 Non-fatal heart attacks are considered chronic illness although they are related to short-term exposure because the
impact is long-lasting and this is reflected in its valuation (discussed in Appendix F).
C -11
July 2013
-------
models including two-hour and twenty-four hour pollutant concentrations for each pollutant.
They found significant and independent associations between heart attack occurrence and both
two-hour and twenty-four hour PM2.5 concentrations before onset. Significant associations were
observed for PMio as well. None of the other particle measures or gaseous pollutants was
significantly associated with acute myocardial infarction for the two hour or twenty-four hour
period before onset.
The patient population for this study was selected from health centers across the United States.
The mean age of participants was 62 years old, with 21% of the study population under the age
of 50. In order to capture the full magnitude of heart attack occurrence potentially associated
with air pollution and because age was not listed as an inclusion criteria for sample selection, we
apply an age range of 18 and over in the C-R function. According to the National Hospital
Discharge Survey, there were no hospitalizations for heart attacks among children <15 years of
age in 1999 and only 5.5% of all hospitalizations occurred in 15-44 year olds (Popovic, 2001,
Table 10).
The coefficient and standard error are calculated from an odds ratio of 1.62 (95% CI 1.13-2.34)
for a 20 (j,g/m3 increase in twenty-four hour average PM2.5 (Peters et al., 2001, Table 4, p. 2813).
Functional Form: Logistic
Coefficient: 0.024121
Standard Error: 0.009285
Incidence Rate: We use the county-specific daily AMI hospitalization rate (ICD-9 code 410) for
the population of individuals aged 18 years and older as the estimate for the incidence rate of
nonfatal heart attack, assuming all heart attacks that are not instantly fatal will result in a
hospitalization. We did not adjust for fatal AMIs in the incidence rate estimation, due to the way
that the epidemiological studies are designed. Those studies consider total admissions for AMIs,
which includes individuals living at the time the studies were conducted. Therefore, we use the
definition of AMI that matches the definition in the epidemiological studies.
Population: population of ages 18 and older
Adjustment: As some fraction of the admitted individuals die in the hospital, we apply a
survival rate of 93% in calculating the avoided cases of AMI in order to avoid double counting
(once in the calculation of AMI cases and once in the calculation of PM-related mortality).
Acute Myocardial Infarction (Heart Attacks), Nonfatal (Pope et al., 2006)
Pope et al. (2006) evaluated the association between short-term exposure to PM2.5 and acute
ischemic heart disease events, including acute nonfatal myocardial infarction, all acute coronary
events, and subsequent myocardial infarctions in individuals living in greater Salt Lake City,
Utah. In a case-crossover study, these ischemic events were assessed in relation to a 10 |ig/m3
increase in PM2.5
Using a single-pollutant model the coefficient and standard error were estimated from the
percent increase (4.81%) and 95% confidence interval (95% CI: 0.98-8.79) for a 10 |ig/m3
increase in daily 24-hour mean PM2.5 (Pope et al., 2006, Table 3).
Functional Form: Logistic
C -12
July 2013
-------
Coefficient: 0.00481
Standard Error: 0.001992
Incidence Rate: We use the county-specific daily AMI hospitalization rate (ICD-9 code 410) for
the population of individuals aged 18 years and older as the estimate for the incidence rate of
nonfatal heart attack, assuming all heart attacks that are not instantly fatal will result in a
hospitalization. We did not adjust for fatal AMIs in the incidence rate estimation, due to the way
that the epidemiological studies are designed. Those studies consider total admissions for AMIs,
which includes individuals living at the time the studies were conducted. Therefore, we use the
definition of AMI that matches the definition in the epidemiological studies.
Population: The study examined population of all ages. We apply the results to people ages 18
and older. We apply the results to people of ages 18 and older. Since the vast majority of AMI
occur among population 65-99, over-counting may not be an issue when applying the risk
coefficient to 18+.
Adjustment: As some fraction of the admitted individuals die in the hospital, we apply a
survival rate of 93% in calculating the avoided cases of AMI in order to avoid double counting
(once in the calculation of AMI cases and once in the calculation of PM-related mortality).
Acute Myocardial Infarction (Heart Attacks), Nonfatal (Sullivan et al., 2005)
Sullivan et al. (2005) studied the relationship between onset time of acute myocardial infarction
and the preceding hourly PM2.5 concentrations in 5,793 confirmed cased of myocardial infarction
through King County, Washington. In this case-crossover study from 1988-1994, air pollution
exposure levels averaged 1 hour, 2 hours, 4 hours, and 24 hours before onset of myocardial
infarction were compared to a set of time-stratified referent exposures from the same day of the
week in the month of the case event. The authors estimated that an associated risk of 1.01 (95%
CI: 0.98-1.05) for myocardial infarction onset could be attributed to a 10 |ig/m3 increase in
PM2 5 the hour before MI onset. No increased risk was found in all cases with preexisting cardiac
diseases with an odds ratio of 1.05 (95% CI: 0.95-1.16). Furthermore, stratification for
hypertension, diabetes, and smoking status did not modify the association between PM25 and
onset of myocardial infarction.
Using a single-pollutant model, the coefficient and standard error were estimated from the odds
ratio (1.02) and 95% confidence interval (95% CI: 0.98-1.07) for a 10 |ig/m3 increase in daily
24-hour mean PM2.5 lagged 1 day (Sullivan et al., 2005, Table 3).
Functional Form: Logistic
Coefficient: 0.001980
Standard Error: 0.002241
Incidence Rate: We use the county-specific daily AMI hospitalization rate (ICD-9 code 410) for
the population of individuals aged 18 years and older as the estimate for the incidence rate of
nonfatal heart attack, assuming all heart attacks that are not instantly fatal will result in a
hospitalization. We did not adjust for fatal AMIs in the incidence rate estimation, due to the way
that the epidemiological studies are designed. Those studies consider total admissions for AMIs,
which includes individuals living at the time the studies were conducted. Therefore, we use the
definition of AMI that matches the definition in the epidemiological studies.
Population: The study examined population of all ages. We apply the results to people ages 18
and older. We apply the results to people of ages 18 and older. Since the vast majority of AMI
C-13
July 2013
-------
occur among population 65-99, over-counting may not be an issue when applying the risk
coefficient to 18+.
Adjustment: As some fraction of the admitted individuals die in the hospital, we apply a
survival rate of 93% in calculating the avoided cases of AMI in order to avoid double counting
(once in the calculation of AMI cases and once in the calculation of PM-related mortality).
Acute Myocardial Infarction (Heart Attacks), Nonfatal (Zanobetti & Schwartz,
2006)4
Zanobetti and Schwartz (2006) analyzed hospital admissions through emergency department for
myocardial infarction (ICD-9 code 410) and pneumonia (ICD-9 codes 480-487) for associations
with fine particulate air pollution, ozone, black carbon, nitrogen dioxide, PM2.5 not from traffic,
and CO in the greater Boston area from 1995-1999. The authors used a case-crossover analysis
with control days matched on temperature. Significant associations were detected for NO2 with a
12.7% increase 95% CI: 5.8-18.0), PM2.5 with an 8.6% increase (95% CI: 1.2-15.4), and black
carbon with an 8.3% increase (95% CI: 0.2-15.8) in emergency myocardial infarction
hospitalizations. Similarly, significant associations were identified for PM2.5 with a 6.5%
increase (95% CI: 1.1-11.4) and CO with a 5.5% increase (95% CI: 1.1-9.5) in pneumonia
hospitalizations.
Using a single-pollutant model, the coefficient and standard error are estimated from the percent
change in risk (8.65%) and 95% confidence interval (95% CI: 1.22-15.38%) for a 16.32 ug/m3
increase in daily 24-hour mean PM2.5 for an average of the 0- and 1-day lag (Zanobetti &
Schwartz, 2006, Table 4).
Functional Form: Logistic
Coefficient: 0.005300
Standard Error: 0.002213
Incidence Rate: We use the county-specific daily AMI hospitalization rate (ICD-9 code 410) for
the population of individuals aged 18 years and older as the estimate for the incidence rate of
nonfatal heart attack, assuming all heart attacks that are not instantly fatal will result in a
hospitalization. We did not adjust for fatal AMIs in the incidence rate estimation, due to the way
that the epidemiological studies are designed. Those studies consider total admissions for AMIs,
which includes individuals living at the time the studies were conducted. Therefore, we use the
definition of AMI that matches the definition in the epidemiological studies.
Population: The study examined population of ages 65 and older. We apply the results to people
of ages 18 and older. Since the vast majority of AMI occur among population 65-99, over-
counting may not be an issue when applying the risk coefficient to 18+.
Adjustment: As some fraction of the admitted individuals die in the hospital, we apply a
survival rate of 93% in calculating the avoided cases of AMI in order to avoid double counting
(once in the calculation of AMI cases and once in the calculation of PM-related mortality).
4 The study looked at hospital admissions of AMI through ER. Under the assumption that all heart attacks will end
in hospitalization, we consider the endpoint as heart attack events to be consistent with other studies.
C-14
July 2013
-------
Acute Myocardial Infarction (Heart Attacks), Nonfatal (Zanobetti et al., 2009)
Zanobetti et al. (2009) examined the relationship between daily PM2.5 levels and emergency
hospital admissions for cardiovascular causes, myocardial infarction, congestive heart failure,
respiratory disease and diabetes among 26 U.S. communities from 2000-2003. The authors used
meta-regression to examine how this association was modified by season- and community-
specific PM2.5 composition while controlling for seasonal temperature as a substitute for
ventilation. Overall, the authors found that PM2.5 mass higher in Ni, As, and Cr as well as Br and
organic carbon significantly increased its effects on hospital admissions. For a 10 ug/m3 increase
in 2-day averaged PM2.5, a 1.89% (95% CI: 1.34-2.45) increase in cardiovascular disease
admissions, a 2.25% (95% CI: 1.10-3.42) increase in myocardial infarction admissions, a 1.85%
(95%) CI: 1.19-2.51) increase in congestive heart failure admissions, a 2.74% (95% CI: 1.30-
4.20) increase in diabetes admissions, and a 2.01% (95% CI: 1.20-2.95) increase in respiratory
admissions were observed. The relationship between PM2.5 and cardiovascular admissions was
significantly modified when the mass of PM2 5 was high in Br, Cr, Ni, and sodium ions, while
mass high in As, Cr, Mn, organic carbon, Ni and sodium ions modified the myocardial infarction
relationship and mass high in As, orgarnic carbon, and sulfate ions modified the diabetes
admission rates.
Using a single-pollutant model, the coefficient and standard error are estimated from the percent
change in risk (2.25%) and 95% confidence interval (95% CI: 1.10-3.42) for a 10 ug/m3 increase
in 2-day averaged PM2.5 (Zanobetti et al., 2009, Table 3).
Functional Form: Log-linear
Coefficient: 0.00225
Standard Error: 0.000592
Incidence Rate: We use the county-specific daily AMI hospitalization rate (ICD-9 code 410) for
the population of individuals aged 18 years and older as the estimate for the incidence rate of
nonfatal heart attack, assuming all heart attacks that are not instantly fatal will result in a
hospitalization. We did not adjust for fatal AMIs in the incidence rate estimation, due to the way
that the epidemiological studies are designed. Those studies consider total admissions for AMIs,
which includes individuals living at the time the studies were conducted. Therefore, we use the
definition of AMI that matches the definition in the epidemiological studies.
Population: The study examined population of ages 65 and older. We apply the results to people
of ages 18 and older. Since the vast majority of AMI occur among population 65-99, over-
counting may not be an issue when applying the risk coefficient to 18+.
Adjustment: As some fraction of the admitted individuals die in the hospital, we apply a
survival rate of 93% in calculating the avoided cases of AMI in order to avoid double counting
(once in the calculation of AMI cases and once in the calculation of PM-related mortality).
C-15
July 2013
-------
Hospitalizations
We include two main types of hospital admissions - respiratory (all respiratory, COPD, and
asthma) and cardiovascular (all cardiovascular less myocardial infarctions).
Respiratory and cardiovascular hospital admissions are the two broad categories of hospital
admissions that have been related to PM2.5 exposure. Although the benefits associated with
respiratory and cardiovascular hospital admissions are estimated separately in the analysis, the
methods used to estimate changes in incidence and to value those changes are the same for both
broad categories of hospital admissions.
Due to the availability of detailed hospital admission and discharge records, there is an extensive
body of literature examining the relationship between hospital admissions and air pollution.
Because of this, we pooled some of the hospital admission endpoints, using the results from a
number of studies. Specifically, we used the following pooling procedure.
> For respiratory hospital admissions (HA): Babin et al. (2007) and Sheppard (2003) were
used to estimate C-R functions for asthma hospitalizations (ICD-9 code: 493) for ages 0-
18 in Washington D.C and Seattle, WA, respectively. We pooled the C-R functions from
these two studies using the random/fixed effects method. We then pooled results from
Zanobetti et al. (2009) and Kloog et al. (2012) using subjective weights pooling method
(i.e., 0.5 for each study) to estimate incidence for all-respiratory admissions for the
elderly (age 65 and up). We then aggregated incidence estimates from the following three
non-overlapping categories: (1) pooled asthma hospitalization (ages 0-18) from above,
(2) pooled all-respiratory admissions for the elderly (age 65 and up) from above, and (3)
COPD less asthma admissions for ages 18-64 from Moolgavkar (2000a).
> For HA for cardiovascular diseases less myocardial infarctions (ICD-9 codes: 390-409,
411-429): Peng et al. (2008) and Peng et al. (2009) reported C-R functions for people age
65 years and older in 108 U.S. counties and 119 U.S. urban counties, respectively. We
assigned equal weights to the estimates from these two studies (i.e., 0.5 for each study)
and used the weighted average. We then assigned a weight of 0.33 to the results from
each of two other studies that look at population of 65 years and older - Zanobetti et al.
(2009) and Bell et al. (2008) - and pooled these results with the pooled results from Peng
et al. (2008) and Peng et al. (2009).
C -16
July 2013
-------
Exhibit C-6. Health Impact Functions for Particulate Matter and Hospital Admissions
Endpoint
Author
Year
Location
Age
Metric
Beta
Std
Error
Functional
Form
All
Zanobetti et
2009
26 U.S.
65+
24-hr avg
0.00189
0.00028
Log-linear
Cardiovascular
al.
communities
(less AMI)3
All
Peng et al.
2008
108 U.S. counties
65+
24-hr avg
0.00071
0.00013
Log-linear
Cardiovascular
(less AMI)3
All
Peng et al.
2009
119 U.S. urban
65+
24-hr avg
0.00068
0.00021
Log-linear
Cardiovascular
counties
(less AMI)3
All
Bell et al.
2008
202 US Counties
65+
24-hr avg
0.0008
0.00011
Log-linear
Cardiovascular
(less AMI)3
All
Moolgavkar
2000b
Los Angeles, CA
18-64
24-hr avg
0.0014
0.00034
Log-linear
Cardiovascular
(less AMI)3
HA, All
Zanobetti et
2009
26 U.S.
65+
24-hr avg
0.00207
0.00045
Log-linear
Respiratory13
al.
communities
HA, All
Kloog et al.
2012
New England area
65+
24-hr avg
0.0007
0.00096
Log-linear
Respiratory13
(6 states)
HA, Asthmab
Babin et al.
2007
Washington, D.C.
0-17
24-hr avg
0.002
0.00434
Log-linear
HA, Asthmab
Sheppard
2003
Seattle, WA
0-17
24-hr avg
0.00332
0.00104
Log-linear
HA, COPDb
Moolgavkar
2000a
Los Angeles, CA
18-64
24-hr avg
0.0022
0.00073
Log-linear
a These studies were pooled to generate pooled incidence estimates for cardiovascular hospital admissions.
b These studies were pooled to generate pooled incidence estimates for respiratory hospital admissions.
Hospital Admissions for All Cardiovascular (Bell et al., 2008)
Bell et al. (2008) evaluated the association between short-term exposure to PM2.5 and the risk of
cardiovascular (ICD-9 codes 410-414, 26-427, 428, 429, 430-438, and 440-449) hospital
admissions among Medicare enrollees >65 years old varied by season and geographic region in
202 U.S. counties with populations greater than 200,000 from 1999-2005. Three time-series
models were used to provide three key variables: consistent PM2.5 effects across the year,
different PM2.5 effects by season, and smoothly varying PM2.5 effects throughout the year. A
two-stage Bayesian hierarchical model was used to estimate the association between PM2.5 and
hospitalization rates, with the first stage estimating the association within a single county and the
second stage combining county-specific estimates to obtain national estimates. The authors
found statistically significant evidence of seasonal and regional variation. The strongest
association was for the northeast.
We use the national estimate for the all-year reported in Table 2 of Bell et al. (2008). The single
pollutant coefficient and standard error are calculated from the estimated 0.8 percent increase in
risk and 95% confidence interval (0.59-1.01 percent) for a 10 |ig/m3 increase in same-day (lag 0)
daily 24-hour mean PM2.5 (Bell et al., 2008, Table 2).
Note that Bell et al. (2008) considered a broader range of ICD-9 codes and estimated the risk of
both cardiovascular events and cerebro- and peripheral vascular disease. For comparability to
C -17
July 2013
-------
other studies, EPA decided to apply a baseline hospitalization rate for ICD-9 codes 390-409 and
411-429 when using this C-R function in quantifying impacts.
Functional Form: Log-linear
Coefficient: 0.0008
Standard Error: 0.00011
Incidence Rate: county-specific daily hospital admission rate for all cardiovascular admissions
less AMI per person ages 65+ (ICD-9 codes 390-409, 411-429)
Population: population of ages 65+
Hospital Admissions for All Cardiovascular (Peng et al., 2008)
Peng et al. (2008) examined the risk of hospital admissions for cardiovascular diseases (ICD-9
codes 426-427, 428, 430-438, 410-414, 429, 440-448) in relation to particulate matter (PM10-2.5
and PM2.5). To accomplish this, the authors utilized a database of 108 U.S. counties with daily
emergency hospital admission rates for cardiovascular diseases among Medicare enrollees living
9 miles from air, temperature, and dew-point temperature monitors. PM10-2.5 and PM2.5
concentrations were calculated by using monitoring data from January 1, 1999 through
December 31, 2005. Overall, there were 3.7 million cardiovascular disease-related hospital
admissions for the time period assessed. The authors found significant associations of PM2.5 and
PM10-2.5 with cardiovascular disease admissions.
In a single-pollutant model, the coefficient and standard error are calculated from the estimated
percent change in daily admission (0.44%) and 95% posterior interval (95% PI: 0.06-0.82%) for
a 10 |ig/m3 increase in daily 24-hour mean PM2.5 concentrations for the same day (Peng et al.,
2008, page 2175).
Note that Peng et al. (2008) considered a broader range of ICD-9 codes and estimated the risk of
both cardiovascular events and cerebro- and peripheral vascular disease. For comparability to
other studies, EPA decided to apply a baseline hospitalization rate for ICD-9 codes 390-409 and
411-429 when using this C-R function in quantifying impacts.
Functional Form: Log-linear
Coefficient: 0.00071
Standard Error: 0.00013
Incidence Rate: county-specific daily hospital admission rate for all cardiovascular admissions
less AMI per person ages 65+ (ICD-9 codes 390-409, 411-429)
Population: population of ages 65+
Hospital Admissions for All Cardiovascular (Peng et al., 2009)
Peng et al. (2009) investigated the relationship between hospital admissions for cardiovascular
and the chemical components of PM2.5 across 119 U.S. urban communities for 12 million
Medicare enrollees using log-linear Poisson regression models. This was achieved using a
national database with daily data from 2000-2006 on emergency hospital admissions of
cardiovascular outcomes, ambient levels of PM2.5 components and weather variables. Bayesian
hierarchical statistical models were used to estimate the associations. Three scenarios for PM2.5
C-18
July 2013
-------
exposure were assessed which were as follows: 1) for the period 2000-2006 and including only
days with available measurements for all 7 PM25 components from the Speciation Trends
network (STN); 2) PM2 5 measured by the STN for the period 2000-2006 and including only
days with available measurements for all 7 PM25 components from the STN and 3) PM25
estimated as the sum of the 7 largest components of PM2 5 mass for the period 2000-2006.
Results of percent increases in emergency admissions associated with PM2 5 at lag 0 under these
scenarios were showed in Figure 2 and the results for the components of PM2 5 from both single
and multi-pollutant models were showed in Figure 3.
In a single-pollutant model, the coefficient and standard error are calculated from the estimated
percent change in daily admission (0.68%) and 95% posterior interval (95% PI: 0.26-1.10%) for
a 10 |ig/m3 increase in daily 24-hour mean PM2 5 concentrations for the same day (Peng et al.,
2009, page 960).
Note that Peng et al. (2008) considered a broader range of ICD-9 codes and estimated the risk of
both cardiovascular events and cerebro- and peripheral vascular disease. For comparability to
other studies, EPA decided to apply a baseline hospitalization rate for ICD-9 codes 390-409 and
411-429 when using this C-R function in quantifying impacts.
Functional Form: Log-linear
Coefficient: 0.00068
Standard Error: 0.00021
Incidence Rate: county-specific daily hospital admission rate for all cardiovascular admissions
less AMI per person ages 65+ (ICD-9 codes 390-409, 411-429)
Population: population of ages 65+
Hospital Admissions for All Cardiovascular (Zanobetti et al., 2009)
Zanobetti et al. (2009) examined the relationship between daily PM2 5 levels and emergency
hospital admissions for cardiovascular causes, myocardial infarction, congestive heart failure,
respiratory disease and diabetes among 26 U.S. communities from 2000-2003. The authors used
meta-regression to examine how this association was modified by season- and community-
specific PM2.5 composition while controlling for seasonal temperature as a substitute for
ventilation. Overall, the authors found that PM2 5 mass higher in Ni, As, and Cr as well as Br and
organic carbon significantly increased its effects on hospital admissions. The relationship
between PM2 5 and cardiovascular admissions was significantly modified when the mass of PM2 5
was high in Br, Cr, Ni, and sodium ions, while mass high in As, Cr, Mn, organic carbon, Ni and
sodium ions modified the myocardial infarction relationship and mass high in As, orgarnic
carbon, and sulfate ions modified the diabetes admission rates.
The single-pollutant coefficient and standard error are calculated from the estimated percent
change in risk (1.89 percent) and 95% confidence interval (1.34-2.45) for a 10 |ig/m3 increase in
2-day averaged PM2 5 (Zanobetti et al., 2009, Table 3).
Note that Zanobetti et al. (2009) report results for ICD-9 codes 390-429. In the benefit analysis,
avoided nonfatal heart attacks are estimated separately. In order to avoid double counting heart
C -19
July 2013
-------
attack hospitalizations, we have excluded ICD-9 code 410 from the baseline incidence rate used
in this function.
Functional Form: Log-linear
Coefficient: 0.00189
Standard Error: 0.00028
Incidence Rate: county-specific daily hospital admission rate for all cardiovascular admissions
less AMI per person ages 65+ (ICD-9 codes 390-409, 411-429)
Population: population of ages 65+
Hospital Admissions for All Cardiovascular (Moolgavkar, 2000b)
Moolgavkar (2000b) examined the association between air pollution and cardiovascular hospital
admissions (ICD-9 390-448) in the Chicago, Los Angeles, and Phoenix metropolitan areas. He
collected daily air pollution data for ozone, S02, N02, CO, and PMi0 in all three areas. PM2.5
data was available only in Los Angeles. The data were analyzed using a Poisson regression
model with generalized additive models to adjust for temporal trends. Separate models were run
for 0 to 5 day lags in each location. In a single pollutant model, PM25 was statistically significant
for lag 0 and lag 1. In co-pollutant models with CO, the PM2 5 effect dropped out and CO
remained significant. For ages 20-64, S02 and CO exhibited the strongest effect and any PM2 5
effect dropped out in co-pollutant models with CO. Among the 65+ age group, the gaseous
pollutants generally exhibited stronger effects than PMio or PM2 5. The strongest overall effects
were observed for S02 and CO.
The single pollutant coefficient and standard error are calculated from an estimated percent
change of 1.4 and t-statistic of 4.1 for a 10 (j,g/m3 increase in PM2 5 in the zero lag model for ages
18-64 (Moolgavkar, 2000b, Table 4).
Note that Moolgavkar (2000b) reported results that include ICD-9 code 410 (heart attack). In the
benefits analysis, avoided nonfatal heart attacks are estimated separately. In order to avoid
double counting heart attack hospitalizations, we have excluded ICD-9 code 410 from the
baseline incidence rate used in this function.
Functional Form: Log-linear
Coefficient: 0.0014
Standard Error: 0.000341
Incidence Rate: county-specific daily hospital admission rate for all cardiovascular admissions
per person ages 18 to 64 (ICD-9 codes 390-409, 411-429)
Population: population of ages 18 to 64
Hospital Admissions for All Respiratory (Zanobetti et al., 2009)
Zanobetti et al. (2009) examined the relationship between daily PM2 5 levels and emergency
hospital admissions for cardiovascular causes, myocardial infarction, congestive heart failure,
respiratory disease and diabetes among 26 U.S. communities from 2000-2003. The authors used
meta-regression to examine how this association was modified by season- and community-
specific PM2.5 composition while controlling for seasonal temperature as a substitute for
C-20
July 2013
-------
ventilation. Overall, the authors found that PM2.5 mass higher in Ni, As, and Cr as well as Br and
organic carbon significantly increased its effects on hospital admissions. The relationship
between PM2.5 and cardiovascular admissions was significantly modified when the mass of PM2.5
was high in Br, Cr, Ni, and sodium ions, while mass high in As, Cr, Mn, organic carbon, Ni and
sodium ions modified the myocardial infarction relationship and mass high in As, orgarnic
carbon, and sulfate ions modified the diabetes admission rates.
In a single-pollutant model, the coefficient and standard error are estimated from the percent
change in risk (2.07%) and 95% confidence interval (1.2% - 2.95%) for a 10 [^g/m3 increase in 2-
day averaged PM2.5 (Zanobetti et al., 2009, Table 3).
Functional Form: Log-linear
Coefficient: 0.00207
Standard Error: 0.00045
Incidence Rate: county-specific daily hospital admission rate for all respiratory admissions per
person ages 65+ (ICD-9 codes 460 - 519)
Population: population of ages 65+
Hospital Admissions for All Respiratory (Kloog et al., 2012)
Kloog et al. (2012) investigated both the long and short term effects of PM2.5 exposure on
hospital admissions across New England for all residents aged 65 and older. The authors
performed separate Poisson regression analysis for each admission type: all respiratory,
cardiovascular disease (CVD), stroke and diabetes. Daily admission counts in each zip code were
regressed against long and short-term PM2.5 exposure, temperature, socio-economic data and a
spline of time to control for seasonal trends in baseline risk. They observed associations between
both short-term and long-term exposure to PM2.5 and hospitalization for all of the outcomes
examined.
In a single-pollutant model, the coefficient and standard error are estimated from the percent
change in risk (0.10%) and 95% confidence interval (0.35% - 0.52%) for a 10 [^g/m3 increase in
short-term (same day) PM2.5 exposure (Kloog et al., 2012, Table 3).
Functional Form: Log-linear
Coefficient: 0.0007
Standard Error: 0.00096
Incidence Rate: county-specific daily hospital admission rate for all respiratory admissions per
person ages 65+ (ICD-9 codes 460 - 519)
Population: population of ages 65+
Hospital Admissions for Asthma (Sheppard, 2003)
Sheppard et al. (1999) studied the relation between air pollution in Seattle and nonelderly (<65)
hospital admissions for asthma from 1987 to 1994. They used air quality data for PM10, PM2.5,
coarse PM10-2.5, SO2, ozone, and CO in a Poisson regression model with control for time trends,
C - 21
July 2013
-------
seasonal variations, and temperature-related weather effects.5 They found asthma hospital
admissions associated with PMio, PM2.5, PM10-2.5, CO, and ozone. They did not observe an
association for S02. They found PM2.5 and CO to be jointly associated with asthma admissions.
The best fitting co-pollutant models were found using ozone. However, ozone data was only
available April through October, so they did not consider ozone further. For the remaining
pollutants, the best fitting models included PM2.5 and CO. Results for other co-pollutant models
were not reported.
In response to concerns that the work by Sheppard et al. (1999) may be biased because of
concerns about the (S-plus) software used in the original analysis, Sheppard (2003) reanalyzed
some of this work; in particular Sheppard reanalyzed the original study's PM2.5 single pollutant
model.
The coefficient and standard error are based on the relative risk (1.04) and 95% confidence
interval (1.01-1.06) for a 11.8 (J,g/m3 increase in PM2.5 in the 1-day lag GAM stringent model
(Sheppard, 2003, pp 228-299).
Functional Form: Log-linear
Coefficient: 0.003324
Standard Error: 0.001045
Incidence Rate: county-specific daily hospital admission rate for asthma admissions per person
(ICD-9 code 493)
Population: population of ages 0 -176
Hospital Admissions for Asthma (Babin et al., 2007)
Babin et al. (2007) examined pediatric asthma-related emergency room (ER) visits and hospital
admissions (ICD-9 code 493) in Washington, D.C. from 2001-2004 and their short-term
associations with ozone, particulate matter, socioeconomic status, and age group. Applying
Poisson regression analyses, the authors found significant associations between asthma ER visits
and outdoor ozone concentrations for the 5-12 year old age group. The association between
PM2.5 and asthma hospitalization was found statistically insignificant.
The single pollutant coefficient and standard error are calculated from the estimated percent
increase in risk (0.2 percent) and 95% confidence interval (-0.6 - 0.1 percent) for a 1 (J,g/m3
increase in same-day (lag 0) daily 24-hour mean PM2.5 based on single-pollutant models (Babin
et al., 2007, Table 2).
Functional Form: Log-linear
Coefficient: 0.002
Standard Error: 0.00434
5 PM2 5 levels were estimated from light scattering data.
6 Although Sheppard (2003) reports results for the <65 year old age range, for comparability to other studies, we
apply the results to the population of ages 0 to 18.
C-22
July 2013
-------
Incidence Rate: county-specific daily hospital admission rate for asthma admissions per person
(ICD-9 code 493)
Population: population of ages 0-17
Hospital Admissions for Chronic Lung Disease (Moolgavkar, 2000a)
Moolgavkar (2000a) examined the association between air pollution and COPD hospital
admissions (ICD-9 490-496) in the Chicago, Los Angeles, and Phoenix metropolitan areas. He
collected daily air pollution data for ozone, SO2, NO2, CO, and PM10 in all three areas. PM2.5
data was available only in Los Angeles. The data were analyzed using a Poisson regression
model with generalized additive models to adjust for temporal trends. Separate models were run
for 0 to 5 day lags in each location. Among the 65+ age group in Chicago and Phoenix, weak
associations were observed between the gaseous pollutants and admissions. No consistent
associations were observed for PMi0. In Los Angeles, marginally significant associations were
observed for PM2.5, which were generally lower than for the gases. In co-pollutant models with
CO, the PM2.5 effect was reduced. Similar results were observed in the 0-19 and 20-64 year old
age groups.
The PM2.5 C-R functions for the 20-64 age group are based on the single-pollutant model. Since
the true PM2.5 effect is most likely best represented by a distributed lag model, any single lag
model should underestimate the total PM2.5 effect. As a result, we selected the lag models with
the greatest effect estimates for use in the C-R functions.
The single pollutant coefficient and standard error are calculated from an estimated percent
change of 2.2 and t-statistic of 3.0 for a 10 (j,g/m3 increase in PM2.5 in the two-day lag model
(Moolgavkar, 2000a, Table 4).
Functional Form: Log-linear
Coefficient: 0.0022
Standard Error: 0.000733
Incidence Rate: county-specific daily hospital admission rate for chronic lung disease
admissions per person 18-64 (ICD-9 codes 490-496)
Population: population of ages 18 to 647
7 Although Moolgavkar (2000a) reports results for the 20-64 year old age range, for comparability to other studies,
we apply the results to the population of ages 18 to 64.
C-23
July 2013
-------
Emergency Room Visits
To estimate the effects of PM2.5 air pollution reductions on asthma-related ER visits, we use the
C-R functions based on Mar et al. (2010), Slaughter et al. (2005) and Glad et al. (2012). COBRA
estimates the incidence results for ER visits by pooling these three studies using random/fixed
pooling method. Exhibit C-7 below summarizes the attributes of the C-R functions used in
COBRA.
Exhibit C-7. Health Impact Functions for Particulate Matter and Emergency Room Visits
Author
Year
Location
Age
Metric
Beta
Std Error
Functional
Form
Mar et al.
Slaughter et
al.
Glad et al.
2010
2005
Greater Tacoma,
Washington
Spokane, Washington
0-99
0-99
24-hr avg
24-hr avg
0.0056
0.0029
0.0021
0.0027
Log-linear
Log-linear
2012
Pittsburgh, PA
0-99
24-hr avg
0.0039
0.0028
Logistic
Emergency Room Visits for Asthma (Mar et al., 2010)
Mar et al. (2010) assessed the effect of particulate matter air pollution, including emissions from
diesel generators, on emergency room visits for asthma in the greater Tacoma, Washington area
from January 3, 1998 to May 30, 2002 using Poisson regression models. Health data were
collected for individuals of all ages from 6 Tacoma hospitals. The authors also assessed the
impacts of diesel generator use on emergency room visits for asthma from January 24, 2001 to
June 2, 2001. Overall, the researchers found an association between daily PM2.5 levels and
emergency room visits for asthma at lag days 2 and 3, with a relative risk for lag day 2 of 1.04
(95% CI: 1.01-1.07) and a relative risk for lag day 3 of 1.03 (95% CI: 1.0-1.06). No significant
association between emergency room visits for asthma and increased use of the diesel generators
was observed.
In a single-pollutant model, the PM2.5 coefficient and standard error are estimated from the
relative risk (1.04) and 95% confidence interval (95% CI: 1.01-1.07) for a 7 |ig/m3 increase in
daily 24-hour mean PM2.5 at lag day 2 (Mar et al., 2010, Table 4).
Functional Form: Log-linear
Coefficient: 0.0056
Standard Error: 0.0021
Incidence Rate: county-specific daily asthma emergency room rate per person (The study didn't
report ICD-9 code but we assume ICD-9 code 493)
Population: population of all ages
Emergency Room Visits for Asthma (Slaughter et al., 2005)
Slaughter et al. (2005) examined the short-term association of particulate matter (PMi, PM2.5,
PM10, and PM10-2.5) and carbon monoxide with hospital admissions and emergency room visits
for respiratory and cardiac outcomes and mortality in Spokane, Washington from January 1995
C-24
July 2013
-------
to June 2001 using a log-linear generalized linear model. The authors found no association
between respiratory emergency room visits and any size fraction of PM2.5, but there was a
suggestive relationship between fine PM2.5 and respiratory effects when compared to coarse
PM2.5. No association between cardiac hospital admissions or mortality and any size fraction of
PM2.5 or CO was observed at the 0- to 3-day lag. CO, on the other hand, was found to be
associated with all respiratory emergency room visits and visits for asthma at the 3-day lag.
In a single-pollutant model, the coefficient and standard error are estimated from the relative risk
(1.03) and 95% confidence interval (95% CI: 0.98-1.09) for a 10 |ig/m3 increase in daily 24-hour
mean PM2.5 at 1-day lag (Slaughter et al., 2005, Table 4).
Functional Form: Log-linear
Coefficient: 0.0029
Standard Error: 0.0027
Incidence Rate: county-specific daily asthma emergency room rate per person (ICD-9 code 493)
Population: population of all ages
Emergency Room Visits for Asthma (Glad et al., 2012)
Glad et al. (2012) investigated the relationship between air pollution and emergency department
(ED) visits for asthma in the Pittsburgh, Pennsylvania area between 2002 and 2005 using a case-
crossover methodology with a logistical model. The authors found a 2.5% increase in asthma ED
visits for each 10 ppb increase in the 1-hour maximum ozone level on day 2 (odds ratio [OR] =
1.025, p < .05). Particulate matter with an aerodynamic diameter <2.5 [j,m (PM2.5) had an effect
both on the total population on day 1 after exposure (1.036, p < .05), and on African Americans
on days 1, 2, and 3. PM2.5 had no significant effect on Caucasian Americans alone. The disparity
in risk estimates by race may reflect differences in residential characteristics, exposure to
ambient air pollution, or a differential effect of pollution by race.
In a single-pollutant model, the coefficient and standard error are estimated from the relative risk
(1.040) and 95% confidence interval (95% CI: 0.984-1.100) for a 10 |ig/m3 increase 6-day
average of daily PM2.5 (Glad et al., 2012, Table 3).
Functional Form: Logistic
Coefficient: 0.0039
Standard Error: 0.0028
Incidence Rate: county-specific daily asthma emergency room rate per person (ICD-9 code 493)
Population: population of all ages
C-25
July 2013
-------
Minor Effects
We include functions to estimate acute bronchitis, lower respiratory symptoms, minor restricted
days, and work loss days.
Exhibit C-8. Health Impact Functions for Particulate Matter and Acute Effects
Endpoint Name Author Year Location Age Metric Beta Functional
' B Error horm
Minor Restricted Ostro & 1989 Nationwide 18-64 24-hr avg 0.007410 0.000700 Log-linear
Activity Days Rothschild
Acute Bronchitis Dockery et 1996 24 communities 8-12 Annual 0.027212 0.017096 Logistic
al.
Work Loss Days Ostro 1987 Nationwide 18-64 24-hr avg 0.004600 0.000360 Log-linear
Lower Respiratory Schwartz 2000 6 U.S. cities 7-14 24-hr avg 0.019012 0.006005 Logistic
Symptoms and Neas
Acute Bronchitis (Dockery et al., 1996)
Dockery et al (1996) examined the relationship between PM2.5 and other pollutants on the
reported rates of asthma, persistent wheeze, chronic cough, and bronchitis, in a study of 13,369
children ages 8-12 living in 24 communities in U.S. and Canada. Health data were collected in
1988-1991, and single-pollutant models were used in the analysis to test a number of measures of
particulate air pollution. Dockery et al. found that annual level of sulfates and particle acidity
were significantly related to bronchitis, and PM2.i and PMi0 were marginally significantly related
to bronchitis.8 They also found nitrates were linked to asthma, and sulfates linked to chronic
phlegm. It is important to note that the study examined annual pollution exposures, and the
authors did not rule out that acute (daily) exposures could be related to asthma attacks and other
acute episodes. Earlier work, by Dockery et al. (1989), based on six U.S. cities, found acute
bronchitis and chronic cough significantly related to PM15. Because it is based on a larger
sample, the Dockery et al (1996) study is the better study to develop a C-R function linking
PM2.5 with bronchitis.
Bronchitis was counted in the study only if there were "reports of symptoms in the past 12
months" (Dockery et al., 1996, p. 501). It is unclear, however, if the cases of bronchitis are acute
and temporary, or if the bronchitis is a chronic condition. Dockery et al. found no relationship
between PM2.5 and chronic cough and chronic phlegm, which are important indicators of chronic
bronchitis. For this analysis, we assumed that the health impact function based on Dockery et al.
is measuring acute bronchitis. The health impact function is based on results of the single
pollutant model reported in Table 1.
The original study measured PM21, however when using the study's results we use PM2 5. This makes only a negligible
difference, assuming that the adverse effects of PM21 and PM2 5 are comparable.
C - 26 July 2013
-------
The estimated logistic coefficient and standard error are based on the odds ratio (1.50) and 95%
confidence interval (0.91-2.47) associated with being in the most polluted city (PM2.i = 20.7
(j,g/m3) versus the least polluted city (PM2.i = 5.8 (J,g/m3) (Dockery et al., 1996, Tables 1 and 4).
The original study used PM2.i, however, we use the PM2.i coefficient and apply it to PM2 5 data.
Functional Form: Logistic
Coefficient: 0.027212
Standard Error: 0.017096
Incidence Rate: annual bronchitis incidence rate per person = 0.043 (American Lung
Association, 2002, Table 11)
Population: population of ages 8-12
Lower Respiratory Symptoms (Schwartz & Neas, 2000)
Schwartz and Neas (2000) used logistic regression to link lower respiratory symptoms and cough
in children with coarse PMio, PM2 5, sulfate and H+ (hydrogen ion). Children were selected for
the study if they were exposed to indoor sources of air pollution: gas stoves and parental
smoking. The study enrolled 1,844 children into a year-long study that was conducted in
different years (1984 to 1988) in six cities. The students were in grades two through five at the
time of enrollment in 1984. By the completion of the final study, the cohort would then be in the
eighth grade (ages 13-14); this suggests an age range of 7 to 14.
The coefficient and standard error are calculated from the reported odds ratio (1.33) and 95%
confidence interval (1.11-1.58) associated with a 15 (j,g/m3 change in PM2 5 (Schwartz & Neas,
2000, Table 2).
Functional Form: Logistic
Coefficient: 0.01901
Standard Error: 0.006005
Incidence Rate: daily lower respiratory symptom incidence rate per person = 0.0012 (Schwartz
et al., 1994, Table 2).
Population: population of ages 7 to 14
Minor Restricted Activity Days (Ostro and Rothschild, 1989)
Ostro and Rothschild (1989) estimated the impact of PM2 5 and ozone on the incidence of minor
restricted activity days (MRADs) and respiratory-related restricted activity days (RRADs) in a
national sample of the adult working population, ages 18 to 65, living in metropolitan areas.9 The
annual national survey results used in this analysis were conducted in 1976-1981. Controlling for
PM2 5, two-week average ozone has highly variable association with RRADs and MRADs.
9 The study population is based on the Health Interview Survey (HIS), conducted by the National Center for Health
Statistics. In publications from this ongoing survey, non-elderly adult populations are generally reported as ages 18-64.
From the study, it is not clear if the age range stops at 65 or includes 65 year olds. We apply the health impact function to
individuals ages 18-64 for consistency with other studies estimating impacts to non-elderly adult populations.
C - 27
July 2013
-------
Controlling for ozone, two-week average PM2.5 was significantly linked to both health endpoints
in most years.10 The health impact function for PM2.5 is based on this co-pollutant model.
The study is based on a "convenience" sample of non-elderly individuals. Applying the health
impact function to this age group is likely a slight underestimate, as it seems likely that elderly
are at least as susceptible to PM2.5 as individuals under 65.
Using the results of the two-pollutant model (O3 and PM2.5), we developed separate coefficients
for each year in the analysis, which were then combined for use in this analysis. The coefficient
is a weighted average of the coefficients in Ostro and Rothschild (1989, Table 4) using the
inverse of the variance as the weight:
The standard error of the coefficient is calculated as follows, assuming that the estimated year-
specific coefficients are independent:
Functional Form: Log-linear
Coefficient: 0.00741
Standard Error: 0.00070
Incidence Rate: daily incidence rate for minor restricted activity days (MRAD) = 0.02137
(Ostro & Rothschild, 1989, p. 243).
Population: adult population ages 18 to 64
Work Loss Days (Ostro, 1987)
Ostro (1987) estimated the impact of PM2.5 on the incidence of work-loss days (WLDs),
restricted activity days (RADs), and respiratory-related RADs (RRADs) in a national sample of
the adult working population, ages 18 to 65, living in metropolitan areas.11 The annual national
10 The study used a two-week average pollution concentration; the health impact function uses a daily average, which is
assumed to be a reasonable approximation.
11 The study population is based on the Health Interview Survey (HIS), conducted by the National Center for Health
Statistics. In publications from this ongoing survey, non-elderly adult populations are generally reported as ages 18-64.
From the study, it is not clear if the age range stops at 65 or includes 65 year olds. We apply the health impact function to
individuals ages 18-64 for consistency with other studies estimating impacts to non-elderly adult populations.
$ = "iiil j = 0 00741
z=1976 & p.
This reduces down to:
TV/ = 0.00070
C-28
July 2013
-------
survey results used in this analysis were conducted in 1976-1981. Ostro reported that two-week
average PM2.5 levels12 were significantly linked to work-loss days, RADs, and RRADs, however
there was some year-to-year variability in the results. Separate coefficients were developed for
each year in the analysis (1976-1981); these coefficients were pooled. The coefficient used in the
concentration-response function presented here is a weighted average of the coefficients in Ostro
(1987, Table 3) using the inverse of the variance as the weight.
The study is based on a "convenience" sample of non-elderly individuals. Applying the health
impact function to this age group is likely a slight underestimate, as it seems likely that elderly
are at least as susceptible to PM2.5 as individuals under 65. On the other hand, the number of
workers over the age of 65 is relatively small; it was approximately 3% of the total workforce in
2001 (U.S. Bureau of the Census, 2002).
The coefficient used in the health impact function is a weighted average of the coefficients in
Ostro (1987, Table 3) using the inverse of the variance as the weight:
1981 O
z A
p=-
1981 1
z
= 0.0046
z=1976 C^
The standard error of the coefficient is calculated as follows, assuming that the estimated year-
specific coefficients are independent:
f 1951 ^ >
^ - 2
Op - Var
z
1
f 1951 ฃ >
2i __ 2
= Var
7
= ^ Var
f Pr ^
z =1976
/ฆa
Pi J
i=\916<5p. J
This eventually reduces down to:
a p = VVr = 0.00036
Functional Form: Log-linear
Coefficient: 0.0046
Standard Error: 0.00036
Incidence Rate: daily work-loss-day incidence rate per person ages 18 to 64 = 0.00595 (Adams,
Hendershot, & Marano, 1999, Table 41; U.S. Bureau of the Census, 1997, No. 22)
Population: adult population ages 18 to 64
12 The study used a two-week average pollution concentration; the health impact function uses a daily average, which is
assumed to be a reasonable approximation.
C-29
July 2013
-------
Asthma-Related Effects
We pool the results of studies by Ostro et al. (2001) and Mar et al. (2004) to get an estimate of
asthma exacerbation in asthmatics. In addition to the lower respiratory estimate, we include an
upper respiratory estimate based on a study by Pope et al. (1991).
Exhibit C-9. Health Impact Functions for Particulate Matter and Asthma-Related Effects
Endpoint Name
Author
Year
Location
Age
Metric
Beta
Std
Error
Functiona
1 Form
Asthma Exacerbation,
Cough
Ostro
et al.
2001
Los
Angeles,
CA
6-18
24-hr avg
0.000985
0.000747
Logistic
Asthma Exacerbation,
Shortness of Breath
Ostro
et al.
2001
Los
Angeles,
CA
6-18
24-hr avg
0.002565
0.001335
Logistic
Asthma Exacerbation,
Wheeze
Ostro
et al.
2001
Los
Angeles,
CA
6-18
24-hr avg
0.001942
0.000803
Logistic
Asthma Exacerbation,
Cough
Mar et
al.
2004
Vancouver,
CAN
6-18
24-hr avg
0.01906
0.009828
Logistic
Asthma Exacerbation,
Shortness of Breath
Mar et
al.
2004
Vancouver,
CAN
6-18
24-hr avg
0.01222
0.013849
Logistic
Upper Respiratory
Symptoms
Pope et
al.
1991
Utah
Valley
9-11
24-hr avg
0.0036
0.0015
Logistic
Pooling Ostro et al. (2001) and Mar et al. (2004)
To characterize asthma exacerbations in children, we use two studies that followed panels of
asthmatic children. Ostro et al. (2001) followed a group of 138 African-American children in Los
Angeles for 13 weeks, recording daily occurrences of respiratory symptoms associated with
asthma exacerbations (e.g., shortness of breath, wheeze, and cough). This study found a
statistically significant association between PM2.5, measured as a 12-hour average, and the daily
prevalence of shortness of breath and wheeze endpoints. Although the association was not
statistically significant for cough, the results were still positive and close to significance;
consequently, we decided to include this endpoint, along with shortness of breath and wheeze, in
generating incidence estimates.
Mar et al. (2004) followed nine asthmatic children for over eight months in Spokane,
Washington. Data on respiratory symptoms and medication use were recorded daily by the
study's subjects, while air pollution data was collected by the local air agency and Washington
State University. The authors found a strong association between cough symptoms and several
metrics of particulate matter, including PM25.
We employed the following pooling approach in combining effect estimates from the two studies
to produce a single asthma exacerbation incidence estimate. First, we pooled (with a
C-30
July 2013
-------
fixed/random effects approach) the incidence estimates based on the two studies for "cough" and
"shortness of breath" separately. We then assigned an equal weight (i.e., 0.33) to the pooled
results for cough, the pooled results for shortness of breath, and the (un-pooled) results for
wheeze (from Ostro et al., 2001).
To prevent double-counting, we followed U.S. EPA (2005, p. 4-38) and focused the estimation
on asthma exacerbations occurring in children and excluded adults from the calculation. Asthma
exacerbations occurring in adults are assumed to be captured in the general population endpoints
such as work loss days and MRADs. Consequently, if we had included an adult-specific asthma
exacerbation estimate, this would likely have double-counted incidence for this endpoint.
However, because the general population endpoints do not cover children (with regard to
asthmatic effects), an analysis focused specifically on asthma exacerbations for children (6 to 18
years of age) could be conducted without concern for double-counting.
Asthma Exacerbation: Cough, Wheeze, and Shortness of Breath (Ostro et al.,
2001)
Ostro et al. (2001) studied the relation between air pollution in Los Angeles and asthma
exacerbation in African-American children (8 to 13 years old) from August to November 1993.
They used air quality data for PMi0, PM2.5, N02, and 03 in a logistic regression model with
control for age, income, time trends, and temperature-related weather effects.13 Asthma symptom
endpoints were defined in two ways: "probability of a day with symptoms" and "onset of
symptom episodes". New onset of a symptom episode was defined as a day with symptoms
followed by a symptom-free day. The authors found cough prevalence associated with PMi0 and
PM2.5 and cough incidence associated with PM25, PMio, and N02. Ozone was not significantly
associated with cough among asthmatics.
Note that the study focused on African-American children ages 8 to 13 years old. We apply the
function based on this study to the general population ages 6 to 18 years old.
Asthma Exacerbation, Cough
The coefficient and standard error are based on an odds ratio of 1.03 (95% CI 0.98-1.07) for a 30
[j,g/m3 increase in 12-hour average PM2 5 concentration (Ostro et al., 2001, Table 4, p. 204).
Functional Form: Logistic
Coefficient: 0.000985
Standard Error: 0.000747
Incidence Rate: daily cough rate per person (Ostro et al., 2001, p. 202) = 0.145
Population: Asthmatic population ages 6 to 18 = 10.70%.14
13 The authors note that there were 26 days in which PM2 5 concentrations were reported higher than PM10 concentrations.
The majority of results the authors reported were based on the full dataset. These results were used for the basis for the C-
R functions.
14 The American Lung Association (2010, Table 7) estimates asthma prevalence for children 5-17 at 10.70% (based on
data from the 2008 National Health Interview Survey).
C-31
July 2013
-------
Asthma Exacerbation, Shortness of Breath
The coefficient and standard error are based on an odds ratio of 1.08 (95% CI 1.00-1.17) for a 30
[j,g/m3 increase in 12-hour average PM2.5 concentration (Ostro et al., 2001, Table 4, p. 204).
Functional Form: Logistic
Coefficient: 0.002565
Standard Error: 0.001335
Incidence Rate: daily shortness of breath rate per person (Ostro et al., 2001, p. 202) = 0.074
Population: Asthmatic population ages 6 to 18 = 10.70%.
Asthma Exacerbation, Wheeze
The coefficient and standard error are based on an odds ratio of 1.06 (95% CI 1.01-1.11) for a 30
[j,g/m3 increase in 12-hour average PM25 concentration (Ostro et al., 2001, Table 4, p. 204).
Functional Form: Logistic
Coefficient: 0.001942
Standard Error: 0.000803
Incidence Rate: daily wheeze rate per person (Ostro et al., 2001, p. 202) = 0.173
Population: Asthmatic population ages 6 to 18 = 10.70%.
Asthma Exacerbation, Cough and Shortness of Breath (Mar et al., 2004)
Mar et al. (2004) studied the effects of various size fractions of particulate matter on respiratory
symptoms of adults and children with asthma, monitored over many months. The study was
conducted in Spokane, Washington, a semiarid city with diverse sources of particulate matter.
Data on respiratory symptoms and medication use were recorded daily by the study's subjects,
while air pollution data was collected by the local air agency and Washington State University.
Subjects in the study consisted of 16 adults - the majority of whom participated for over a year -
and nine children, all of whom were studied for over eight months. Among the children, the
authors found a strong association between cough symptoms and several metrics of particulate
matter, including PM25. However, the authors found no association between respiratory
symptoms and PM2.5 of any metric in adults. Mar et al. therefore concluded that the discrepancy
in results between children and adults was due either to the way in which air quality was
monitored, or a greater sensitivity of children than adults to increased levels of PM25 air
pollution.
Asthma Exacerbation, Cough
In a single-pollutant model, the coefficient and standard error are estimated from the odds ratio
(1.21) and 95% confidence interval (1.00-1.47) for a 10.0 [j,g/m3 increase in 1-day lagged
concentration of PM25 (Mar et al., 2004, Table 7).
C-32
July 2013
-------
Functional Form: Logistic
Coefficient: 0.019062
Standard Error: 0.009828
Incidence Rate: daily cough rate per person (Ostro et al., 2001) = 14.5%
Population: The study reported results for population ages 7-12. For comparability to other
studies, we apply the results to the population of ages 6 to 18. Asthmatic population ages 6 to 18
= 10.70%.15
Asthma Exacerbation, Shortness of Breath
In a single-pollutant model, the coefficient and standard error are estimated from the odds ratio
(1.13) and 95% confidence interval (0.86-1.48) for a 10.0 [j,g/m3 increase in current-day
concentration of PM2.5 (Mar et al., 2004, Table 7).
Functional Form: Logistic
Coefficient: 0.012222
Standard Error: 0.013849
Incidence Rate: daily shortness of breath rate per person (Ostro et al., 2001, p.202) = 7.4%
Population: The study reported results for population ages 7-12. For comparability to other
studies, we apply the results to the population of ages 6 to 18. Asthmatic population ages 6 to 18
= 10.70%.
Upper Respiratory Symptoms (Pope, 1991)
Using logistic regression, Pope et al. (1991) estimated the impact of PMio on the incidence of a
variety of minor symptoms in 55 subjects (34 "school-based" and 21 "patient-based") living in
the Utah Valley from December 1989 through March 1990. The children in the Pope et al. study
were asked to record respiratory symptoms in a daily diary. With this information, the daily
occurrences of upper respiratory symptoms (URS) and lower respiratory symptoms (LRS) were
related to daily PMi0 concentrations. Pope et al. describe URS as consisting of one or more of
the following symptoms: runny or stuffy nose; wet cough; and burning, aching, or red eyes.
Levels of ozone, NO2, and SO2 were reported low during this period, and were not included in
the analysis. The sample in this study is relatively small and is most representative of the
asthmatic population, rather than the general population. The school-based subjects (ranging in
age from 9 to 11) were chosen based on "a positive response to one or more of three questions:
ever wheezed without a cold, wheezed for 3 days or more out of the week for a month or longer,
and/or had a doctor say the 'child has asthma' (Pope, 1991, p. 669)." The patient-based subjects
(ranging in age from 8 to 72) were receiving treatment for asthma and were referred by local
physicians. Regression results for the school-based sample (Pope, 1991, Table 5) show PM10
significantly associated with both upper and lower respiratory symptoms. The patient-based
sample did not find a significant PMi0 effect. The results from the school-based sample are used
here.
15 The American Lung Association (2010, Table 7) estimates asthma prevalence for children 5-17 at 10.70% (based on
data from the 2008 National Health Interview Survey).
C-33
July 2013
-------
The coefficient and standard error for a one (J,g/m3 change in PMio is reported in Table 5.
Functional Form: Logistic
Coefficient: 0.0036
Standard Error: 0.0015
Incidence Rate: daily upper respiratory symptom incidence rate per person = 0.3419 (Pope,
1991, Table 2)
Population: Asthmatic population ages 6 to 18 = 10.70% 16 of population ages 9 to 11.
16 The American Lung Association (2010, Table 7) estimates asthma prevalence for children 5-17 at 10.70% (based on
data from the 2008 National Health Interview Survey).
C-34
July 2013
-------
Appendix D: Baseline Incidence Rates for Adverse Health
Effects
Health impact functions developed from log-linear or logistic models estimate the percent
change in an adverse health effect associated with a given pollutant change. In order to estimate
the absolute change in incidence using these functions, we need the baseline incidence rate of the
adverse health effect. For certain health effects, such as asthma exacerbation, we need a
prevalence rate, which estimates the percentage of the general population with a given ailment
like asthma. This appendix describes the data used to estimate baseline incidence rates and
prevalence rates for the health effects considered in this analysis.
Mortality
This section describes the development of county mortality rates for year 2020 for use in
COBRA.1 First, we describe the source of 2004-2006 individual-level mortality data and the
calculation of county-level mortality rates. Then we describe how we use national-level Census
mortality rate projections to develop county-level mortality rate projections for year 2020.
Mortality Data for 2004-2006
We obtained individual-level mortality data from 2004-2006 for the whole United States from
the Centers for Disease Control (CDC), National Center for Health Statistics (NCHS). The data
were compressed into a CD-ROM, which contains death information for each decedent,
including residence county FIPS, age at death, month of death, and underlying causes (ICD-10
codes).
Using the detailed mortality data combined with county-level inter-censal population estimates,2
we generated age-, cause-, and county-specific mortality rates using the following formula:
R A,m (2004) + A.m (2005) + A.m (2006)
PiJe (2004) + PiJe (2005) + PiJe (2006)
where is the mortality rate for age group z, cause j, and county k, D is the death count; and P
is the population.
Following CDC Wonder (http://wonder.cdc.gov), we treated mortality rates as "unreliable" when
the death count is less than 20.3 For each combination of age group and mortality cause, we used
the following procedure to deal with the problem of "unreliable" rates:
1 We use projected 2020 mortality rates for year 2017 in COBRA.
2 The detailed mortality data obtained from CDC do not include population. The county-level inter-censal
population estimates are based on US Census of Population and Housing 2010 and forecasts developed by Woods &
Poole (2011).
D-l
July 2013
-------
For a given state, we grouped the counties where the death count (i.e., the numerator on
the right-hand side of the above equation) was less than 20 and summed those death
counts across those counties. If the sum of deaths was greater than or equal to 20, we then
summed the populations in those counties, and calculated a single rate for the "state
collection of counties" by dividing the sum of deaths by the sum of populations in those
counties. This rate was then applied to each of those counties.4
If the sum of deaths calculated in the above step was still less than 20, the counties in the
"state collection of counties" were not assigned the single rate from the above step.
Instead, we proceeded to the regional level (see Exhibit D-l for region definition). In
each region, we identified all counties whose death counts were less than 20 (excluding
any such counties that were assigned a rate in the previous step). We summed the death
counts in those counties. If the sum of deaths was greater than or equal to 20, we then
summed the populations in those counties, and calculated a single rate for the "regional
collection of counties" by dividing the sum of deaths by the sum of populations in those
counties. This rate was then applied to each of those counties in the "regional collection
of counties."5
Exhibit D-l. Regional Definitions from U.S. Census
Region States Included
Maine, New Hampshire, Vermont, Massachusetts,
Northeast Rhode Island, Connecticut, New York, New Jersey,
Pennsylvania
Ohio, Indiana, Illinois, Michigan, Wisconsin, Minnesota,
Midwest Iowa, Missouri, North Dakota, South Dakota,
Nebraska, Kansas
Delaware, Maryland, District of Columbia, Virginia,
goutja West Virginia, North Carolina, South Carolina,
Georgia, Florida, Kentucky, Tennessee, Alabama,
Mississippi, Arkansas, Louisiana, Oklahoma, Texas
Montana, Idaho, Wyoming, Colorado, New Mexico,
West Arizona, Utah, Nevada, Washington, Oregon,
California, Alaska, Hawaii
If the sum of deaths calculated in the previous (regional) step was still less than 20, the
counties in the "regional collection of counties" were not assigned the single rate from
the above step. Instead, we proceeded to the national level, identifying all counties in the
nation whose death counts were less than 20 (excluding any such counties that were
assigned a rate in the previous steps). We summed the death counts in those counties and
Among all the calculated age-, cause-, and county-specific mortality rates, there were about 67% "unreliable"
3
rates.
4
After this adjustment, there were 17% unreliable rates left.
After this regional adjustment, there were 7% unreliable rates left.
D-2
July 2013
-------
divided by the sum of the populations in those counties to derive a single rate for the
"national collection of counties." This rate was then applied to each of those counties in
the "national collection of counties."6
Exhibit D-2 shows the resulting national average all-cause mortality rates.
Exhibit D-2. National All-Cause Mortality Rates (per 100 people per year) by Age Group
Mortality Category
Infant 1 1
A 1 1 /
18-
24
25-
34
35-
44
45-
54
55-
64
65-
74
75-
84
85+
Mortality, All Cause
0.241 0.028
0.089
0.106
0.194
0.430
0.902
2.126
5.234
14.654
* We estimate post-neonatal mortality (deaths after the first month) for infants because the health impact
function (see Appendix F) estimates post-neonatal mortality.
Mortality Rate Projections to 2020
To estimate age- and county-specific mortality rates in year 2020, we calculated adjustment
factors, based on a series of Census Bureau projected national mortality rates (for all-cause
mortality), to adjust the age- and county-specific mortality rates calculated using 2004-2006 data
as described above. We used the following procedure:
For each age group, we obtained the series of projected national mortality rates from
2005 to 2050 (see the 2005 rate in Exhibit D-3) based on Census Bureau projected life
tables.7
We then calculated, separately for each age group, the ratio of Census Bureau national
mortality rate in year 2020 to the 2005 rate. These ratios are shown in Exhibit D-4.
Finally, to estimate mortality rates in year 2020 that are both age group-specific and
county-specific, we multiplied the county- and age-group-specific mortality rates for
2004-2006 by the appropriate ratio calculated in the previous step. For example, to
estimate the projected mortality rate in 2020 among ages 18-24 in Wayne County, MI, we
multiplied the mortality rate for ages 18-24 in Wayne County in 2004-2006 by the ratio
of Census Bureau projected national mortality rate in 2020 for ages 18-24 to Census
Bureau national mortality rate in 2005 for ages 18-24.
6 Even after this national adjustment, there were about 1% unreliable rates left. In these cases, we simply calculated
a single rate for the "national collection of counties, even though it was "unreliable," and assigned it to those
counties in the "national collection of counties."
7 For a detailed description of the model, the assumptions, and the data used to create Census Bureau projections,
see the working paper, "Methodology and Assumptions for the Population Projections of the United States: 1999 to
2100, Working Paper #38.", which is available on
http://www.census.gov/population/www/documentation/twps0038/twps0038.html (Hollman, et al. 2000).
D-3
July 2013
-------
Exhibit D-3. All-Cause Mortality Rate (per 100 people per year), by Source, Year, and Age Group
Source &
Infant*
1-17
18-
25-
35-
45-
55-
65-
75-
85+
Year
24
34
44
54
64
74
84
Calculated
CDC 2004-
2006
0.684/0.241
0.028
0.089
0.106
0.194
0.430
0.902
2.126
5.234
14.654
Census
Bureau 2005
0.654
0.029
0.088
0.102
0.183
0.387
0.930
2.292
5.409
13.091
* The Census Bureau estimate is for all deaths in the first year of life. COBRA uses post-neonatal mortality (deaths
after the first month, i.e., 0.23 per 100 people) because the health impact function (see Appendix F) estimates
postneonatal mortality. For comparison purpose, we also calculated the rate for all deaths in the first year, which is
0.684 per 100 people).
Exhibit D-4. Ratio of 2020 All-Cause Mortality Rate to 2005 Estimated All-Cause Mortality Rate, by Age
Group
Year
Infant
1-17
18-24
25-34
35-44
45-54
55-64
65-74
75-84
85+
2020
0.85
0.81
0.86
0.90
0.83
0.85
0.87
0.85
0.83
0.91
Hospitalizations
Hospitalization rates were calculated using data from the Healthcare Cost and Utilization Project
(HCUP). HCUP is a family of health care databases developed through a Federal-State-Industry
partnership and sponsored by the Agency for Healthcare Research and Quality (AHRQ).8 HCUP
products include the State Inpatient Databases (SID), the State Emergency Department
Databases (SEDD), the Nationwide Inpatient Sample (NIS), and the Nationwide Emergency
Department Sample (NEDS). HCUP databases can be obtained from the following data services:
The HCUP Central Distributor: Many of the HCUP databases are available for purchase
through the HCUP Central Distributor. The databases include detailed information for
individual discharges, such as primary diagnosis (in ICD-9 codes), patient's age and
residence county.
HCUP State Partners: Some HCUP participating states do not release their data to the
Central Distributor; however, the data may be obtained through contacting the State
Partners. Some State Partners (e.g., CA, TX, and NY) provided discharge-level data;
others (e.g., OH) provided summarized data.
HCUPnet: This is a free, on-line query system based on data from HCUP. It provides
access to summary statistics at the state, regional and national levels.
Exhibit D-5 shows the level of hospitalization data (e.g, discharge-level or state-level) for each
state. Note that for some states neither discharge-level nor state-level data were available. In such
cases we used regional statistics from HCUPnet to estimate hospitalization rates for those states.
8 More information about HCUP can be found at http://www.hcup-us.alira.gov/.
D-4
July 2013
-------
Exhibit D-5. Hospitalization Data from HCUP
Legend
| Dtscftarge-tevel Data
~ County-lev* Data
^ State-level Data
~ RegcnaWev* Data
The procedures for calculating hospitalization rates are summarized as follows:9
For states with discharge-level data:
o We calculated age-, health endpoint-, and county-specific hospitalization counts.10
o The above calculation excluded hospitalizations with missing patient age or
county FIPS, which may lead to underestimation of rates. Therefore we scaled up
the previously calculated age-, endpoint-, and county-specific counts using an
adjustment factor obtained as follows:
ฆ We first counted the number of discharges for a specific endpoint in the
state including those discharges with missing age or county FIPS.
ฆ We then counted the number of discharges for the endpoint in the state
excluding those records with missing age or county FIPS.
ฆ The adjustment factor is the ratio of the two counts.
9 The data year for most states is 2007; the exception is MA, for which the data year is 2006. We assume
hospitalization rates are reasonably constant from 2006-2007 and consider all as 2007 rates.
10 Ohio was the only state that, while not providing discharge-level data, did provide county-level data for each age
group-endpoint combination.
D-5
July 2013
-------
o We calculated hospitalization rates for each county by dividing the adjusted
county-level hospitalization counts by the Census estimated county-level
population for the corresponding year (2006 or 2007). Following CDC Wonder,
we treated rates as "unreliable" when the hospitalization count was less than 20,
using the same procedure we used for mortality rates above.
For states with summarized state statistics (from HCUPnet) we calculated the state-, age-,
endpoint- specific hospitalization rates and applied them to each county in the state. We
used the previously described procedure to adjust the "unreliable" rates.
For states without discharge-level or state-level data:
o We obtained the endpoint-specific hospitalization counts in each region from
HCUPnet/NIS (we refer to this count for the z'th endpoint in the /th region as
"TOTALi").
o For those states in the /th region that do have discharge-level or state-level data,
we summed the hospital admissions by endpoint (we refer to this count for the z'th
endpoint in the/'th region as "SUB
o We then estimated the hospitalization count for states without discharge or state
data for the z'th endpoint in the /th region as TOTALri - SUB f/. Note that while this
count is endpoint- and region- specific, it is not age-specific. We obtained the
distribution of hospital admission counts across age groups based on the National
Hospital Discharge Survey (NHDS) and assumed the same distribution for the
HCUP hospitalizations. We then applied this distribution to the estimated hospital
counts (i.e., TOTALy - SUB to obtain endpoint-, region-, and age-specific
counts.
o Using the corresponding age- and region-specific populations, we calculated age-
specific hospitalization rates for the z'th endpoint in the /th region and applied
them to those counties in the region that didn't have discharge-level or state-level
data.
Exhibit D-6 shows the resulting average national hospitalization rates by health endpoint and
age group.
D-6
July 2013
-------
Exhibit D-6. National Hospitalization Rates, by Health Endpoint and Age Group
Hospitalization
Category
ICD-9
Codes
Hospitalization Rate by Age Group
(admissions per 100 people per year)
0-17
18-44
45-64
65-84
85+
Respiratory
all respiratory
460-519
0.700
0.288
0.995
3.73
8.352
asthma
493
0.173
0.068
0.145
0.216
0.325
chronic lung disease
490-496
0.178
0.089
0.381
1.21
1.598
Cardiovascular
all cardiovascular
(less AMI)
390-
409,
411-430
0.019
0.234
1.356
4.974
10.051
Emergency Room Visits for Asthma
The data source for emergency department/room (ED or ER) visits is also HCUP, i.e., SID,
SEDD, and NEDS. The types of data providers are also the same as those described above for
hospitalizations. Exhibit D-7 shows the emergency department data in each state.
Exhibit D- 7. Emergency Department Data from HCUP
ri\
-
=otn -
"
U/
Legend
| DisctargcMovol Data
^ County-level Data
SUrte-lcve* Data
Regcnal-sev* Data
The calculation of ER visit rates is also similar to the calculation of hospitalization rates, except
for the following differences:
D-7
July 2013
-------
The SEDD databases include only those ER visits that ended with discharge. To identify
the ER visits that ended in hospitalization, we used a variable called "admission source"
in the SID databases. Admission source identified as "emergency room" indicates that the
hospital admission came from the ER - i.e., the ER visit ended in hospitalization. For
each combination of age group, endpoint and county, we summed the ER visits that
ended with discharge and those that resulted in hospitalization.
The data year varies across the states from 2005 to 2007 (see Exhibit D-7); we assumed
that ER visit rates are reasonably constant across these three years and consider them as
2006 rates.
Instead of using HCUPnet/NIS and NHDS in the last step as described for
hospitalizations, we used HCUPnet/NEDS and the National Ambulatory Medical Care
Survey (NAMCS) to calculate ER visit rates for states without discharge level or state
level data.
Exhibit D-8 shows the resulting average national rates of asthma emergency room visits by age
group.
Exhibit D-8. National Emergency Room Visit Rates for Asthma, by Age Group
ER Category
ICD-9
Codes
0-17
ER Visit Rate
(visits per 100 people per year)
18-44 45-64 65-84
85+
asthma
493
0.860
0.573 0.393 0.248
0.308
Nonfatal Heart Attacks
The relationship between short-term particulate matter exposure and heart attacks was quantified
in case-crossover analyses by Peters et al (2001), Pope et al. (2006), and Sullivan et al. (2005).
The study population was selected from heart attack survivors in a medical clinic. Therefore, the
applicable population to apply to the C-R function is all individuals surviving a heart attack in a
given year. Several data sources are available to estimate the number of heart attacks per year.
For example, several cohort studies have reported estimates of heart attack incidence rates in the
specific populations under study. However, these rates depend on the specific characteristics of
the populations under study and may not be the best data to extrapolate nationally.
An alternative approach to the estimation of heart attack rates is to use data from the HCUP,
assuming that all heart attacks that are not instantly fatal will result in a hospitalization.
According to the HCUPnet, in 2009 there were approximately 633,356 hospitalizations due to
D-8
July 2013
-------
heart attacks (acute myocardial infarction: ICD-9 code of 410, primary diagnosis).11 We used
county-level hospitalization rates over estimates extrapolated from cohort studies because the
former is part of a nationally representative survey with a larger sample size, which is intended
to provide reliable national estimates. The hospitalization section above describes the detailed
procedure for developing the incidence rates for hospitalization of AMI. As additional
information is provided regarding the American Heart Association methodology, we will
evaluate the usefulness of this estimate of heart attack incidence.
Rosamond et al. (1999) reported that approximately six percent of male and eight percent of
female hospitalized heart attack patients die within 28 days (either in or outside of the hospital).
We, therefore, applied a factor of 0.93 to the count of hospitalizations to estimate the number of
nonfatal heart attacks per year. Note that we did not adjust for fatal AMIs in the incidence rate
estimation, due to the way that the epidemiological studies are designed. Those studies consider
total admissions for AMIs, which includes individuals living at the time the studies were
conducted. Therefore, we use the definition of AMI that matches the definition in the
epidemiological studies.
Exhibit D-9 presents the national nonfatal heart attack incidence rates around year 2007 by age
group (Note: county-level rates around year 2007 are used in COBRA).
Exhibit D-9. Nonfatal Heart Attack Rates by Age Group
Nonfatal Heart Rate by Age Group
Endpoint
(admissions per 100 people per year)*
0-17
18-44 45-64 65-84
85+
Nonfatal heart attack 0.000 0.033 0.259 0.767 1.78
* Rates are based on data from the 2007 HCUP/SID.
Other Acute Effects
For many of the minor effect studies, baseline rates from a single study are often the only source
of information, and we assume that these rates hold for locations in the U.S. The use of study-
specific estimates is likely to increase the uncertainty around the estimate because they are often
estimated from a single location using a relatively small sample. These endpoints include: acute
bronchitis, upper respiratory symptoms, and lower respiratory symptoms. Exhibit D-10 presents
a summary of these baseline rates.
11 Source: Online query onHCUPnet website (AHRQ 2012), accessed 1-13-2012
http://hcupnet.ahrq.gov/HCUPnet.app/HCUPnet.jsp?Id=53F290DC050F1296&Form=SelLAY&GoTo=MAINSEL
&JS=Y
D-9
July 2013
-------
Exhibit D-10. Selected Acute Effects Rates
Endpoint
Age
Parameter
Rate
Source
Acute Bronchitis
8-12
Incidence
4.300
(American Lung Association 2002,
Table 11)
Lower Respiratory
Symptoms (LRS)
7-14
Incidence
43.8
(Schwartz et al., 1994, Table 2)
Minor Restricted Activity
Days (MRAD)
18-64
Incidence
780.0
(Ostro & Rothschild, 1989, p. 243)
Work Loss Day (WLD)
18-64
Incidence
217.2
(Adams, Hendershot, & Marano,
1999, Table 41; U.S. Bureau of the
Census, 1997)
Acute Bronchitis
The annual rate of acute bronchitis for children ages 5 to 17 was obtained from the American
Lung Association (2002). The authors reported an annual incidence rate per person of 0.043,
derived from the 1996 National Health Interview Survey.
Lower Respiratory Symptoms
Lower respiratory symptoms (LRS) are defined as two or more of the following: cough, chest
pain, phlegm, and wheeze. The proposed yearly incidence rate for 100 people, 43.8, is based on
the percentiles in Schwartz et al (1994, Table 2). The authors did not report the mean incidence
rate, but rather reported various percentiles from the incidence rate distribution. The percentiles
and associated per person per day values are 10th = 0 percent, 25th = 0 percent, 50th = 0 percent,
75th = 0.29 percent, and 90th = 0.34 percent. The most conservative estimate consistent with the
data are to assume the incidence per person per day is zero up to the 75th percentile, a constant
0.29 percent between the 75th and 90th percentiles, and a constant 0.34 percent between the 90th
and 100th percentiles. Alternatively, assuming a linear slope between the 50th and 75th, 75th and
90th, and 90th to 100th percentiles, the estimated mean incidence rate per person per day is 0.12
percent.12 We used the latter approach in this analysis, and then multiplied by 100 and by 365 to
calculate the incidence rate per 100 people per year.
Minor Restricted Activity Days (MRAD)
Ostro and Rothschild (1989, p. 243) provide an estimate of the annual incidence rate of MRADs
(7.8). We multiplied this estimate by 100 to get an annual rate per 100 people.
Work Loss Days
The yearly work-loss-day incidence rate per 100 people is based on estimates from the 1996
National Health Interview Survey (Adams et al., 1999, Table 41). They reported a total annual
work loss days of 352 million for individuals ages 18 to 65. The total population of individuals
of this age group in 1996 (162 million) was obtained from (U.S. Bureau of the Census, 1997).
12 For example, the 62.5th percentile would have an estimated incidence rate per person per day of 0.145 percent.
D-10
July 2013
-------
The average annual rate of work loss days per individual (2.17) was multiplied by 100 to obtain
the average yearly work-loss-day rate of 217 per 100 people.
Asthma-Related Health Effects
Several studies have examined the impact of air pollution on asthma development or
exacerbation in the asthmatic population. Many of the baseline incidence rates used in the C-R
functions are based on study-specific estimates. The baseline rates for the various endpoints are
described below and summarized in Exhibit D-l 1.
Exhibit D-ll. Asthma-Related Health Effects Rates
Endpoint
Age
Parametera
Rate
Source
Astluna Exacerbation, Cough
6-18
Incidence
Prevalence
24.46
14.50%
Astluna Exacerbation, Shortness of
Breath
6-18
Incidence
Prevalence
13.51
7.40%
(Ostro, Lipsett, Mann, Braxton-Owens,
& White, 2001, p. 202) b
Astluna Exacerbation, Wheeze
6-18
Incidence
Prevalence
27.74
17.3%
Astluna
6-18
Prevalence
10.70%
(American Lung Association, 2010,,
Table 7)c
Upper Respiratory Symptoms (URS)
9-11
Incidence
124.79
(Pope, Dockery, Spengler, & Raizenne,
1991, Table 2)
a The incidence rate is the number of cases per person per year. Prevalence refers to the fraction of people that
have a particular illness during a particular time period.
b the rates in the study were for African American children of ages 8-13. We apply it to children aged 6-18 to
match what was used in the selected epidemiological studies.
0 The American Lung Association (2010, Table 7) estimates asthma prevalence for children 5-17 at 10.70%
(based on data from the 2008 National Health Interview Survey). We apply to ages 6-18 because what was used in
the selected epidemiological studies.
D-ll
July 2013
-------
Appendix E: Population Forecasts
To estimate the change in population exposure to air pollution, we use projections based on US
Census of Population and Housing 2010 and forecasting models developed by Woods & Poole
(2011). The Woods and Poole (WP) database contains county-level projections of population by
age, sex, ethnicity, and race out to 2040. Projections in each county are determined
simultaneously with every other county in the United States to take into account patterns of
economic growth and migration. The sum of growth in county-level populations is constrained to
equal a previously determined national population growth, based on Bureau of Census estimates.
The projection year used for COBRA is 2017.
According to WP, linking county-level growth projections together and constraining to a
national-level total growth avoids potential errors introduced by forecasting each county
independently. County projections are developed in a four-stage process. First, national-level
variables such as income, employment, and populations are forecasted. Second, employment
projections are made for 172 economic areas defined by the Bureau of Economic Analysis, using
an "export-base" approach, which relies on linking industrial sector production of non-locally
consumed production items, such as outputs from mining, agriculture, and manufacturing with
the national economy. The export-based approach requires estimation of demand equations or
calculation of historical growth rates for output and employment by sector. Third, population is
projected for each economic area based on net migration rates derived from employment
opportunities and following a cohort component method based on fertility and mortality in each
area. Fourth, employment and population projections are repeated for counties, using the
economic region totals as bounds. The age, sex, ethnicity, and race distributions for each region
or county are determined by aging the population by single year of age by sex and race for each
year through 2040 based on historical rates of mortality, fertility, and migration.
E-l
July 2013
-------
Appendix F: Economic Value of Health Effects
This appendix presents the mean estimate of the unit values used in this analysis. Exhibit F-l
lists these unit values.
Exhibit F-l. Unit Values for Economic Valuation of Health Endpoints (2010 $)
Unit Value (2017 Income Level)
Health Endpoint
Age Range
3%DR
7%DR
Mortality3
25
-99
$8,434,924
$7,512,853
Infant Mortality13
0
-0
$9,401,680
$9,401,680
Acute Myocardial Infarction, Nonfatal0
0-
ฆ24
$33,259
$31,446
Acute Myocardial Infarction, Nonfatal0
25
-44
$45,085
$42,033
Acute Myocardial Infarction, Nonfatal0
45
-54
$50,689
$47,050
Acute Myocardial Infarction, Nonfatal0
55
-64
$134,003
$121,641
Acute Myocardial Infarction, Nonfatal0
65
-99
$33,259
$31,446
Acute Myocardial Infarction, Nonfatal
0-
ฆ24
$163,051
$163,051
Acute Myocardial Infarction, Nonfatal
25
-44
$174,876
$173,638
Acute Myocardial Infarction, Nonfatal
45
-54
$180,480
$178,655
Acute Myocardial Infarction, Nonfatal
55
-64
$263,795
$253,247
Acute Myocardial Infarction, Nonfatal
65
-99
$163,051
$163,051
HA, All Cardiovascular (less AMI)
18
-64
$41,002
$41,002
HA, All Cardiovascular (less AMI)
65
-99
$38,618
$38,618
HA, All Respiratory
65
-99
$32,697
$32,697
HA, Asthma
0-
ฆ 17
$15,430
$15,430
HA, Chronic Lung Disease
18
-64
$20,349
$20,349
Astluna ER Visits (Smith et al. (1997)
0-
ฆ99
$464
$464
Astluna ER Visits (Stanford et al. (1999)
0-
ฆ99
$388
$388
Acute Bronchitis
8-
ฆ 12
$477
$477
Lower Resp. Symptoms
7-
ฆ 14
$21
$21
Upper Resp. Symptoms
9-
ฆ 11
$33
$33
MRAD
18
-64
$68
$68
Work Loss Days
18
-64
$151
$151
Astluna Exacerbation (Cough, Shortness of
Breath, or Wheeze)
6-
ฆ 18
$57
$57
NOTE:a Mortality value after adjustment for 20-year lag.
b Infant mortality value is not adjusted for 20-year lag.
0 Based on Russell (1998)
d Based on Wittels (1990)
Selecting Unit Values for Monetizing Health Endpoints
The appropriate economic value for a change in a health effect depends on whether the health
effect is viewed ex ante (before the effect has occurred) or ex post (after the effect has occurred).
Reductions in ambient concentrations of air pollution generally lower the risk of future adverse
health effects by a small amount for a large population. The appropriate economic measure is
therefore ex ante WTP for changes in risk. However, epidemiological studies generally provide
estimates of the relative risks of a particular health effect avoided due to a reduction in air
pollution. A convenient way to use this data in a consistent framework is to convert probabilities
F-l
July 2013
-------
to units of avoided statistical incidences. This measure is calculated by dividing individual WTP
for a risk reduction by the related observed change in risk.
For example, suppose a measure is able to reduce the risk of premature mortality from 2 in
10,000 to 1 in 10,000 (a reduction of 1 in 10,000). If individual WTP for this risk reduction is
$100, then the WTP for an avoided statistical premature mortality amounts to $1 million
($100/0.0001 change in risk). Using this approach, the size of the affected population is
automatically taken into account by the number of incidences predicted by epidemiological
studies applied to the relevant population. The same type of calculation can produce values for
statistical incidences of other health endpoints.
For some health effects, such as hospital admissions, WTP estimates are generally not available.
In these cases, we use the cost of treating or mitigating the effect. For example, for the valuation
of hospital admissions EPA used the avoided medical costs as an estimate of the value of
avoiding the health effects causing the admission. These COI estimates generally understate the
true value of reductions in risk of a health effect, because, while they reflect the direct
expenditures related to treatment, they omit the value of avoiding the pain and suffering from the
health effect itself.
Updating Values for Inflation
The studies based on which the unit values were developed report estimates for a range for years
prior to 2010. To allow for the effect of inflation, we have adjusted these values to reflect prices
in 2010$. Because some functions are based on willingness to pay to avoid illness, while others
are based on cost of illness and/or lost wages, three different inflation indices are used. These are
the All Goods Index, the Medical Cost Index, and the Wage Index, respectively. Exhibit F-2
summarizes the types of inflation indices and their sources used to adjust different types of unit
values in COBRA.
Exhibit F-2. Type of Inflation Index Used for Adjust Unit Values for Health Effects Endpoints
Index
Source
Health Effects Endpoints
All Goods Index
Bureau of Labor Statistics'
Acute Bronchitis
(BLS) Consumer Price Index
Asthma Exacerbation
(CPI)
Lower Respiratory Symptoms
Mortality
Minor Restricted Activity Days
Upper Respiratory Symptoms
Medical Cost Index
BLS/CPI
Acute Myocardial Infarction
Emergency Room Visits
Hospital Admissions
Wage Index
BLS Employment Cost Index for
Acute Myocardial Infarction
private industry workers, 2001-
Hospital Admissions
2010
Work Loss Days
Growth in Unit Values Reflecting Growth in National Income
The unit value estimates reflect expected growth in real income over time. This is consistent with
economic theory, which argues that WTP for most goods (such as health risk reductions) will
F-2
July 2013
-------
increase if real incomes increase. There is substantial empirical evidence that the income
elasticity of WTP for health risk reductions is positive, although there is uncertainty about its
exact value (and it may vary by health effect). Although one might assume that the income
elasticity of WTP is unit elastic (e.g., a 10 percent higher real income level implies a 10 percent
higher WTP to reduce health risks), empirical evidence suggests that income elasticity is
substantially less than one and thus relatively inelastic. As real income rises, the WTP value also
rises but at a slower rate than real income.
The effects of real income changes on WTP estimates can influence benefits estimates in two
ways: through real income growth between the year a WTP study was conducted and the year for
which benefits are estimated, and through differences in income between study populations and
the affected populations at a particular time. Following the analysis in the 2006 PM2.5 NAAQS
regulatory impact assessment (U.S. EPA, 2006), we have focused on the former.
The income adjustment in COBRA follows the approach used by EPA (2005, p. 4-17), who
adjusted the valuation of human health benefits upward to account for projected growth in real
U.S. income. Faced with a dearth of estimates of income elasticities derived from time-series
studies, EPA applied estimates derived from cross-sectional studies.1 The available income
elasticities suggest that the severity of a health effect is a primary determinant of the strength of
the relationship between changes in real income and changes in WTP. As a result, EPA (2005, p.
4-18) used different elasticity estimates to adjust the WTP for minor health effects, severe and
chronic health effects, and premature mortality (see Exhibit F-3).
Exhibit F-3. Elasticity Values Used to Account for National Income Growth
Benefit Category
Central Elasticity
Estimate
Minor Health Effect
0.14
Severe & Chronic Health Effects
0.45
Premature Mortality
0.40
In addition to elasticity estimates, projections of populations and real gross domestic product
(GDP) are needed to adjust benefits to reflect real per capita income growth. COBRA uses
population and GDP projections developed by EPA, which are described in EPA (2005, p. 4-17).
To estimate national population growth rates for the years between 1990 and 1999, EPA used
national population estimates U.S. Census Bureau (Hollman, Mulder, & Kalian, 2000). These
population estimates are based on an application of a cohort-component model to 1990 U.S.
Census data projections (U.S. Bureau of the Census, 2000). For the years between 2000 and
2010, EPA applied growth rates based on the U.S. Census Bureau projections to the U.S. Census
estimate of national population in 2000. EPA used projections of real GDP provided in Kleckner
and Neumann (1999) for the years 1990 to 2010, and projections of real GDP (in chained 1996
dollars) provided by Standard and Poor's (2000) for the years 2010 to 2020.
1 Details of the procedure can be found in Kleckner and Neumann 1999.
F-3
July 2013
-------
Using the method outlined in Kleckner and Neumann (1999) and the population and income data
described above, EPA (2005, p. 4-18) calculated WTP adjustment factors for each of the
elasticity estimates. Benefits for each of the categories (minor health effects, severe and chronic
health effects, premature mortality, and visibility) are adjusted by multiplying the unadjusted
benefits by the appropriate adjustment factor.
Note that because of a lack of data on the dependence of COI on income, and a lack of data on
projected growth in average wages, no adjustments are made to benefits estimates based on the
COI approach or to work loss days benefits estimates. This lack of adjustment would tend to
result in an under-prediction of benefits in future years, because it is likely that increases in real
U.S. income would also result in increased COI (due, for example, to increases in wages paid to
medical workers) and increased cost of work loss days and lost worker productivity (reflecting
that if worker incomes are higher, the losses resulting from reduced worker production would
also be higher).
Valuation Pooling
In some cases there are multiple valuations available for a health effect, with no one valuation
clearly superior to another. In such cases we pooled valuations in COBRA.
> Smith et al. (1997) and Stanford et al. (1999) both evaluate asthma ER visits using COI.
We assign equal weight to each study (i.e., 0.5) and COBRA will then use the weighted
average to value ER visit.
> To value Acute Myocardial Infarction, we pool Russell (1998) and Wittels (1990) by
assigning equal weight (i.e., 0.5) to each.
> To value respiratory hospitalization, we sum across non-overlapping respiratory
hospitalization effects, i.e., Asthma HA (age 0-17), Chronic Lung Disease HA (age 18-
64), All Respiratory HA (age 65-99).
> Similarly, we sum across non-overlapping cardiovascular hospitalization effects, i.e., we
sum the value for cardiovascular less AMI hospitalization for ages 18-64 and that for
ages 65+.
Valuing Premature Mortality
To estimate the monetary value of risk change in premature death, we used the "value of
statistical lives" saved (VSL) approach, which is a summary measure for the value of small
changes in mortality risk for a large number of people. The VSL approach applies information
from several published value-of-life studies to determine a reasonable monetary value of
F-4
July 2013
-------
preventing premature mortality. Based on 26 published studies,2 the mean value of avoiding one
statistical death is estimated to be roughly $9.4 million (2010$ at 2017 income level).
There are a number of uncertainties in this estimate. The health science literature on air pollution
indicates that several human characteristics affect the degree to which mortality risk affects an
individual. For example, some age groups appear to be more susceptible to air pollution than
others (e.g., the elderly and children). Health status prior to exposure also affects susceptibility.
An ideal benefits estimate of mortality risk reduction would reflect these human characteristics,
in addition to an individual's WTP to improve one's own chances of survival plus WTP to
improve other individuals' survival rates.
The ideal measure would also take into account the specific nature of the risk reduction
commodity that is provided to individuals, as well as the context in which risk is reduced. To
measure this value, it is important to assess how reductions in air pollution reduce the risk of
dying from the time that reductions take effect onward and how individuals value these changes.
Each individual's survival curve, or the probability of surviving beyond a given age, should shift
as a result of an environmental quality improvement. For example, changing the current
probability of survival for an individual also shifts future probabilities of that individual's
survival. This probability shift will differ across individuals because survival curves depend on
such characteristics as age, health state, and the current age to which the individual is likely to
survive.
There are other potentially important factors that go beyond the scope of this discussion. For
additional details, EPA (2005, p. 4-57) has an in-depth discussion of the uncertainties underlying
mortality valuation.
Present Discounted Value of Avoiding Future Mortality
The delay, or lag, between changes in PM25 exposures and changes in mortality rates is not
precisely known. The current scientific literature on adverse health effects, such as those
associated with PM25 (e.g., smoking-related disease), and the difference in the effect size
estimated in chronic exposure studies versus daily mortality studies, suggests that it is likely that
not all cases of avoided premature mortality associated with a given incremental reduction in
PM2.5 exposure would occur in the same year as the exposure reduction.
Current EPA benefits analyses (U.S. EPA, 2006, p. 5-21) assume a 20-year lag structure, with 30
percent of premature deaths occurring in the first year, 50 percent occurring evenly over years 2
to 5 after the reduction in PM25, and 20 percent occurring evenly over years 6 to 20 after the
reduction in PM2 5. It should be noted that the selection of a 20-year lag structure is not directly
supported by any PM2 5-specific literature. Rather, it is intended to be a best guess at the
appropriate time distribution of avoided cases of PM2 5-related mortality. As noted by EPA, the
2 These 26 studies have been identified in the Section 812 Reports to Congress as "applicable to policy analysis."
This represents an intermediate value from a variety of estimates, and it is a value EPA has frequently used in
Regulatory Impact Analyses (RIAs) as well as in the Section 812 Retrospective and Prospective Analyses of the
Clean Air Act.
F-5
July 2013
-------
distribution of deaths over the latency period is intended to reflect the contribution of short-term
exposures in the first year, cardiopulmonary deaths in the 2- to 5-year period, and long-term lung
disease and lung cancer in the 6- to 20-year period. Finally, it is important to keep in mind that
changes in the lag assumptions do not change the total number of estimated deaths but rather the
timing of those deaths.
Specifying the lag is important because people are generally willing to pay more for something
now than for the same thing later. They would, for example, be willing to pay more for a
reduction in the risk of premature death in the same year as exposure is reduced than for that
same risk reduction to be received the following year. This time preference for receiving benefits
now rather than later is expressed by discounting benefits received later. There is an ongoing
discussion within the federal government about the choice of a discount rate in this context: a 3%
discount rate is recommended by EPA, while a 7% is recommended by OMB. Therefore, the
users now have the ability to specify the discount rate-3% or 7%-for a COBRA session.
Following EPA's Guidelines for Preparing Economic Analyses (U.S. EPA, 2010a), COBRA
users are recommended to calculate monetized health benefits using both discount rates and to
evaluate whether (and to what extent) the overall outcome of their analysis is affected by the
choice of discount rate.
Following EPA (2006, p. 5-21), COBRA assumes that some of the incidences of premature
mortality related to PM2.5 exposures occur in a distributed fashion over the 20 years following
exposure. To take this into account in the valuation of reductions in premature mortality, we
applied an annual 3 percent discount rate to the value of premature mortality occurring in future
years. Note that this lag adjustment does not apply to infant mortality, because Woodruff et al.
(1997) estimate the number of infant deaths occurring in the same year as the emissions change.
Valuing Non-Fatal Myocardial Infarction
We are not able to identify a suitable WTP value for reductions in the risk of non-fatal heart
attacks. Instead, we have used a cost-of-illness unit value with two components: the direct
medical costs and the opportunity cost (lost earnings) associated with the illness event. Because
the costs associated with a heart attack extend beyond the initial event itself, we considered costs
incurred over several years. For opportunity costs, we used values derived from Cropper and
Krupnick (Cropper & Sussman, 1990), originally used in the 812 Retrospective Analysis of the
Clean Air Act (U.S. EPA, 1997). For the direct medical costs, we found three possible sources in
the literature.
Wittels et al. (1990) estimated expected total medical costs of myocardial infarction over five
years to be $51,211 (in 1986$) for people who were admitted to the hospital and survived
hospitalization. (There does not appear to be any discounting used.) Using the CPI-U for medical
care, the Wittels et al. estimate is $163,050 in year 2010$. This estimated cost is based on a
medical cost model, which incorporated therapeutic options, projected outcomes and prices
(using "knowledgeable cardiologists" as consultants).
F-6
July 2013
-------
The model used medical data and medical decision algorithms to estimate the probabilities of
certain events and/or medical procedures being used. The authors noted that the average length
of hospitalization for acute myocardial infarction has decreased over time (from an average of
12.9 days in 1980 to an average of 11 days in 1983). Wittels et al. used 10 days as the average in
their study. It is unclear how much further the length of stay may have decreased from 1983 to
the present. The average length of stay for ICD code 410 (myocardial infarction) in 2009 is 4.9
days (Agency for Healthcare Research and Quality, 2010). However, this may include patients
who died in the hospital (not included among our non-fatal cases), whose length of stay was
therefore substantially shorter than it would be if they hadn't died.
Eisenstein et al. (2001) estimated 10-year costs of $44,663, in 1997$, or $73,950 in 2010$ for
myocardial infarction patients, using statistical prediction (regression) models to estimate
inpatient costs. Only inpatient costs (physician fees and hospital costs) were included.
Russell et al. (1998) estimated first-year direct medical costs of treating nonfatal myocardial
infarction of $15,540 (in 1995$), and $1,051 annually thereafter. Converting to year 2010$, that
would be $33,260 (3% discount rate) and $31,446 (7% discount rate) for a 5-year period.
As seen in Exhibit F-4, the three different studies provided significantly different values. We
have not adequately resolved the sources of differences in the estimates. Because the wage-
related opportunity cost estimates from Cropper and Krupnick (1990) cover a 5-year period, we
used a simple average of the two estimates for medical costs that similarly cover a 5-year period
(i.e., assign a subjective weight of 0.5 to each estimate). We added this to the 5-year opportunity
cost estimate. Exhibit F-5 gives the resulting estimates.
Exhibit F-4. Summary of Studies Valuing Reduced Incidences of Myocardial Infarction
Study
Direct Medical Costs
(2010 $, 3% DR)
Direct Medical Costs
(2010 $, 7% DR)
Over an x-year period, for
x =
Wittels et al., 1990a
$163,050
$163,050
5
Russell et al., 1998
$33,260
$31,446
5
Eisenstein et al., 2001
$73,950
$73,950
10
a Wittels et al. did not appear to discount costs incurred in future years.
F - 7
July 2013
-------
Exhibit F-5. Estimated Costs Over a 5-Year Period of a Non-Fatal Myocardial Infarction
Opportunity
Opportunity
Medical Cost
Medical Cost
Total Cost
Total Cost
Age Group Cost (2010 $,
Cost (2010$,
(2010 $, 3%
(2010 $, 7%
(2010 $, 3%
(2010 $, 7%
3% DR) a
7% DR) a
DR) b
DR) b
DR)
DR)
0-24
$0
$0
$98,155
$97,248
$98,155
$97,248
25-44
$11,825
$10,587
$98,155
$97,248
$109,980
$107,835
45-54
$17,429
$15,605
$98,155
$97,248
$115,584
$112,853
55-65
$100,744
$90,196
$98,155
$97,248
$198,899
$187,444
>65
$0
$0
$98,155
$97,248
$98,155
$97,248
aFrom Cropper and Krupnick (1990). Present discounted value of 5 years of lost earnings, adjusted from 1977$ to 2010$
using CPI-U "all items".
b An average of the 5-year costs estimated by Wittels et al. (1990) and Russell et al.(1998). Note that Wittels et al. appears
not to have used discounting in deriving a 5-year cost; Russell et al. estimated first-year direct medical costs and annual
costs thereafter. Medical costs were inflated to 2010$ using CPI-U for medical care.
Valuing Hospital Admissions
Society's WTP to avoid a hospital admission includes medical expenses, lost work productivity,
the non-market costs of treating illness (i.e., air, water and solid waste pollution from hospitals
and the pharmaceutical industry), as well as WTP of the affected individual, as well as of that of
relatives, friends, and associated caregivers, to avoid the pain and suffering.3
Because medical expenditures are to a significant extent shared by society, via medical
insurance, Medicare, etc., the medical expenditures actually incurred by the individual are likely
to be less than the total medical cost to society. The total value to society of an individual's
avoidance of hospital admission, then, might be thought of as having two components: (1) the
cost of illness (COI) to society, including the total medical costs plus the value of the lost
productivity, as well as (2) the WTP of the individual, as well as that of others, to avoid the pain
and suffering resulting from the illness.
In the absence of estimates of social WTP to avoid hospital admissions for specific illnesses
(components 1 plus 2 above), estimates of total COI (component 1) are typically used as
conservative (lower bound) estimates. Because these estimates do not include the value of
avoiding the pain and suffering resulting from the illness (component 2), they are biased
3 Some people take action to avert the negative impacts of pollution. While the costs of successful averting behavior
should be added to the sum of the health-endpoint-specific costs when estimating the total costs of pollution, these
costs are not associated with any single health endpoint. It is possible that in some cases the averting action was not
successful, in which case it might be argued that the cost of the averting behavior should be added to the other costs
listed (for example, it might be the case that an individual incurs the costs of averting behavior and in addition incurs
the costs of the illness that the averting behavior was intended to avoid). Because averting behavior is generally not
taken to avoid a particular health problem (such as a hospital admission for respiratory illness), but instead is taken
to avoid the entire collection of adverse effects of pollution, it does not seem reasonable to ascribe the entire costs of
averting behavior to any single health endpoint. However, omission of these averting behavior costs will tend to bias
the estimates downward.
F-8
July 2013
-------
downward. Some analyses adjust COI estimates upward by multiplying by an estimate of the
ratio of WTP to COI, to better approximate total WTP. Other analyses have avoided making this
adjustment because of the possibility of over-adjusting - that is, possibly replacing a known
downward bias with an upward bias. The COI values used in this benefits analysis will not be
adjusted to better reflect the total WTP.
Following the method used in the ง812 analysis (U.S. EPA, 1999), ICD-code-specific COI
estimates used in our analysis consist of two components: estimated hospital charges and the
estimated opportunity cost of time spent in the hospital (based on the average length of a hospital
stay for the illness). The opportunity cost of a day spent in the hospital is estimated as the value
of the lost daily wage, regardless of whether or not the individual is in the workforce. The
median daily wage was calculated by dividing the median weekly wage ($695 in 2007 dollar or
$564.3 in 2000 dollar) by 5. The median weekly wages for 2007 was obtained from the U.S.
Census Bureau, Statistical Abstract of the United States: 2009, Section 12, Table 626: "Full-
Time Wage and Salary Workers - Numbers and Earnings: 2000 to 2007".
For all hospital admissions endpoints available in this analysis, estimates of hospital charges and
lengths of hospital stays were based on discharge statistics provided by the Agency for
Healthcare Research and Quality's Healthcare Utilization Project National Inpatient Sample
(NIS) database (2007). The NIS is the largest inpatient care database in the United States, and it
is the only national hospital database containing charge information on all patients. It contains
data from a very large nationally representative sample of about eight million hospital
discharges, and therefore provides the best estimates of mean hospital charges and mean lengths
of stay available, with negligible standard errors. The sampling frame for the 2007 NIS is a
sample of hospitals that comprises approximately 90 percent of all hospital discharges in the
United States. Since the NIS is based on discharge samples, the discharge-level weight was used
to weight discharges in order to produce national estimates. The principle diagnoses (based on
ICD-9 codes) were used to define the health endpoints.
Since most pollution-related hospital admissions are likely unscheduled, the unit values of
avoided hospital admissions used in COBRA are based solely on unscheduled hospitalizations.
The total COI for an ICD-code-specific hospital stay lasting n days is estimated as the mean
hospital charge plus n times the daily lost wage. The hospital admissions for which unit values
are available in COBRA are given in Exhibit F-l.
Because of distortions in the market for medical services, the hospital charge may exceed "the
cost of a hospital stay." We use the example of a hospital visit to illustrate the problem. Suppose
a patient is admitted to the hospital to be treated for an asthma episode. The patient's stay in the
hospital (including the treatments received) costs the hospital a certain amount. This is the
hospital cost - i.e., the short-term expenditures of the hospital to provide the medical services
that were provided to the patient during his hospital stay. The hospital then charges the payer a
certain amount - the hospital charge. If the hospital wants to make a profit, is trying to cover
costs that are not associated with any one particular patient admission (e.g., uninsured patient
services), and/or has capital expenses (building expansion or renovation) or other long term
F-9
July 2013
-------
costs, it may charge an amount that exceeds the patient-specific short term costs of providing
services. The payer (e.g., the health maintenance organization or other health insurer) pays the
hospital a certain amount - the payment - for the services provided to the patient. The less
incentive the payer has to keep costs down, the closer the payment will be to the charge. If,
however, the payer has an incentive to keep costs down, the payment may be substantially less
than the charge; it may still, however, exceed the short-term cost for services to the individual
patient.
Although the hospital charge may exceed the short-term cost to the hospital of providing the
medical services required during a patient's hospital stay, cost of illness estimates based on
hospital charges are still likely to understate the total social WTP to avoid the hospitalization in
the first place, because the omitted WTP to avoid the pain and suffering is likely to be quite
large.
Valuing Emergency Room Visits for Asthma
To value asthma emergency room (ER) visits, we used a simple average of two estimates from
the literature. The first estimate comes from Smith et al.(1997), who reported that there were
approximately 1.2 million asthma-related ER visits made in 1987, at a total cost of $186.5
million, in 1987$. The average cost per visit was therefore $155 in 1987$, or $464 in 2010 $
(using the CPI-U for medical care to adjust to 2010 $). The second is from Stanford et al. (1999),
who examined data from asthmatics from 1996-1997, and reported an average cost of $388
(2010 $). We use a simple average of the two estimates, which yields a unit value of about $426
(2010 $).
In comparing their study to Smith et al. (1997), Stanford et al. (1999) noted that the data used by
Smith et al., "may not reflect changes in treatment patterns during the 1990s." In addition, its
costs are the costs to the hospital (or ER) for treating asthma rather than charges or payments by
the patient and/or third party payer. Costs to the ER are probably a better measure of the value of
the medical resources used up on an asthma ER visit.
Valuing Acute Symptoms and Illness Not Requiring Hospitalization
Several acute symptoms and illnesses have been associated with air pollution, including acute
bronchitis in children, upper and lower respiratory symptoms, and exacerbation of asthma (as
indicated by one of several symptoms whose occurrence in an asthmatic generally suggests the
onset of an asthma episode). In addition, several more general health endpoints which are
associated with one or more of these acute symptoms and illnesses, such as minor restricted
activity days and work loss days, have also been associated with air pollution.
F-10
July 2013
-------
Valuing Acute Bronchitis in Children
Estimating WTP to avoid a case of acute bronchitis is difficult for several reasons. First, WTP to
avoid acute bronchitis itself has not been estimated. Estimation of WTP to avoid this health
endpoint therefore must be based on estimates of WTP to avoid symptoms that occur with this
illness. Second, a case of acute bronchitis may last more than one day, whereas it is a day of
avoided symptoms that is typically valued. Finally, the C-R function used in the benefit analysis
for acute bronchitis was estimated for children, whereas WTP estimates for those symptoms
associated with acute bronchitis were obtained from adults.
In previous benefits analyses, such as in the ง812 Prospective analysis (U.S. EPA, 1999), acute
bronchitis was valued at $59.31 (in 2000 $ and at 1990 income level). This is the midpoint
between a low estimate and a high estimate. The low estimate is the sum of the midrange values
recommended by IEc (1994) for two symptoms believed to be associated with acute bronchitis:
coughing and chest tightness. The high estimate was taken to be twice the value of a minor
respiratory restricted activity day. For a more complete description of the derivation of this
estimate, see Abt Associates (2000, p. 4-30).
A unit value of $59.31 assumes that an episode of acute bronchitis lasts only one day. However,
this is generally not the case. More typically, it can last for 6 or 7 days. We therefore made a
simple adjustment, multiplying the original unit value of $59.31 by 6. The unit value thus
derived and used was $356 in 2000 $ and at 1990 income level (=$59.31 x 6) or $477 in 2010 $
and at 2017 income level.
Valuing Upper Respiratory Symptoms (URS) in Children
Willingness to pay to avoid a day of upper respiratory symptoms is based on symptom-specific
WTPs to avoid those symptoms identified by Pope et al. (1991) as part of the complex of upper
respiratory symptoms. Three contingent valuation studies have estimated WTP to avoid various
morbidity symptoms that are either within the complex defined by Pope et al. (1991), or are
similar to those symptoms. In each CV study, participants were asked their WTP to avoid a day
of each of several symptoms. The WTP estimates corresponding to the morbidity symptoms
valued in each study are presented in Exhibit F-7.
The three individual symptoms listed in Exhibit F-7 that were identified as most closely
matching those listed by Pope, et al. (1991) for upper respiratory symptoms are cough,
head/sinus congestion, and eye irritation, corresponding to "wet cough," "runny or stuffy nose,"
and "burning, aching or red eyes," respectively. A day of upper respiratory symptoms could
consist of any one of the seven possible "symptom complexes" consisting of at least one of these
three symptoms. These seven possible symptom complexes are presented in Exhibit F-8. We
assumed that each of these seven complexes is equally likely.4 The point estimate of WTP is just
an average of the seven estimates of WTP for the different complexes.
4 With empirical evidence, we could presumably improve the accuracy of the probabilities of occurrence of each
type of URS. Lacking empirical evidence, however, a uniform distribution seems the most reasonable "default"
assumption.
F-ll
July 2013
-------
A unit value of $32.99 (at 2017 income level and in 2010 $) assumes that an episode of the
symptoms lasts only one day.
Exhibit F-7. Median WTP Estimates and Derived Midrange Estimates (2017 income level, 2010 $)a
Symptom b
Dickie et al.
(1987)
Tolley et al.
(1986)
Loehman et al.
(1979)
Mid-Range
Estimate
Throat congestion
6.66
28.85
-
17.65
Head/sinus congestion
7.77
31.07
14.46
17.65
Coughing
2.22
24.43
8.79
12.36
Eye irritation
-
27.72
-
27.72
Headache
2.22
44.40
-
17.65
Shortness of breath
0.00
-
18.64
8.81
Pain upon deep inhalation
(PDI)
7.79
-
-
7.79
Wheeze
4.45
-
-
4.45
Coughing up phlegm
4.86 c
-
-
4.86
Chest tightness
11.12
-
-
11.12
a Values were inflated to 2010 $ using CPI-U for "all items"and adjusted to 2017 income level.
b All estimates are WTP to avoid one day of symptom. Midrange estimates were derived by IEc (1993).
b 10% trimmed mean.
Exhibit F-8. Estimates of WTP to Avoid Upper Respiratory Symptoms (2017 income level, 2010 $)a
Symptom Combinations Identified as URS by Pope et al.
(1991)
WTP to Avoid
Symptom(s)
Coughing
$12.36
Head/Sinus Congestion
$17.65
Eye Irritation
$27.72
Coughing, Head/Sinus Congestion
$30.01
Coughing, Eye Irritation
$40.08
Head/Sinus Congestion, Eye Irritation
$45.37
Coughing, Head/Sinus Congestion, Eye Irritation
$57.73
Average: $32.99
a Values were inflated to 2010 $ and adjusted to 2017 income level.
Valuing Lower Respiratory Symptoms (LRS) in Children
Schwartz et al. (1994, p. 1235) defined lower respiratory symptoms as at least two of the
following symptoms: cough, chest pain, phlegm, and wheeze. To value this combination of
symptoms, we used the same method as we did for upper respiratory symptoms. We chose those
F -12
July 2013
-------
individual health effects that seem most consistent with lower respiratory symptoms, we derived
all of the possible combinations of these symptoms, and then we valued these combinations.
The symptoms for which WTP estimates are available that reasonably match lower respiratory
symptoms are: cough (C), chest tightness (CT), coughing up phlegm (CP), and wheeze (W). A
day of lower respiratory symptoms could consist of any one of the 11 combinations of at least
two of these four symptoms.5 We assumed that each of the eleven types of lower respiratory
symptoms is equally likely,6 and the mean WTP is the average of the WTPs over all
combinations. Exhibit F-9 presents resulting estimate.
Exhibit F-9. Estimates of WTP to Avoid Lower Respiratory Symptoms (2017 income level and 2010 $)
Symptom Combinations Identified as LRS by Schwartz et al.
(1994, p. 1235)
WTP to Avoid
Symptoms
Coughing, Chest Tightness
$23.46
Coughing, Coughing Up Phlegm
$17.20
Coughing, Wheeze
$16.79
Chest Tightness, Coughing Up Phlegm
$15.96
Chest Tightness, Wheeze
$15.56
Coughing Up Phlegm, Wheeze
$9.31
Coughing, Chest Tightness, Coughing Up Phlegm
$28.33
Coughing, Chest Tightness, Wheeze
$27.92
Coughing, Coughing Up Phlegm, Wheeze
$21.66
Chest Tightness, Coughing Up Phlegm, Wheeze
$20.42
Coughing, Chest Tightness, Coughing Up Phlegm, Wheeze
$32.77
Average: $20.85
Valuing Work Loss Days (WLDs)
Willingness to pay to avoid the loss of one day of work was estimated by dividing the median
weekly wage ($576 in 2000$) by 5. Using the Bureau of Labor Statistics' (BLS) Employment
Cost Index for Wages & Salaries in Private Industry Workers (ECIWAG), the 2000$ was
inflated to 2010$, which resulted a unit value of $151 (2010 $) (BLS 2012). The median weekly
5 Because cough is a symptom in some of the upper respiratory symptom clusters as well as some of the lower
respiratory symptom clusters, there is the possibility of a very small amount of double counting - if the same
individual were to have an occurrence of upper respiratory symptoms which included cough and an occurrence of
lower respiratory symptoms which included cough both on exactly the same day. Because this is probably a very
small probability occurrence, the degree of double counting is likely to be very minor. Moreover, because upper
respiratory symptoms is applied only to asthmatics ages 9-11 (a very small population), the amount of potential
double counting should be truly negligible.
6 As with URS, if we had empirical evidence we could improve the accuracy of the probabilities of occurrence of
each type of LRS. Lacking empirical evidence, however, a uniform distribution seems the most reasonable "default"
assumption.
F-13
July 2013
-------
wage was obtained from U.S. Census Bureau, Statistical Abstract of the United States: 2001,
Section 12, Table 621: "Full-Time Wage and Salary Workers - Numbers and Earnings: 1985 to
2000." Valuing the loss of a day's work at the wages lost is consistent with economic theory,
which assumes that an individual is paid exactly the value of his labor.
The use of the median rather than the mean, however, requires some comment. If all individuals
in society were equally likely to be affected by air pollution to the extent that they lose a day of
work because of it, then the appropriate measure of the value of a work loss day would be the
mean daily wage. It is highly likely, however, that the loss of work days due to pollution
exposure does not occur with equal probability among all individuals, but instead is more likely
to occur among lower income individuals than among high income individuals. It is probable, for
example, that individuals who are vulnerable enough to the negative effects of air pollution to
lose a day of work as a result of exposure tend to be those with generally poorer health care.
Individuals with poorer health care have, on average, lower incomes.
To estimate the average lost wages of individuals who lose a day of work because of exposure to
PM2.5 pollution, then, would require a weighted average of all daily wages, with higher weights
on the low end of the wage scale and lower weights on the high end of the wage scale. Because
the appropriate weights are not known, however, the median wage was used rather than the mean
wage. The median is more likely to approximate the correct value than the mean because means
are highly susceptible to the influence of large values in the tail of a distribution (in this case, the
small percentage of very large incomes in the United States), whereas the median is not
susceptible to these large values.
Valuing Minor Restricted Activity Days (MRADs)
No studies are reported to have estimated WTP to avoid a minor restricted activity day (MRAD).
However, IEc (1993) has derived an estimate of WTP to avoid a minor respiratory restricted
activity day (MRRAD), using WTP estimates from Tolley et al. (1986) for avoiding a three-
symptom combination of coughing, throat congestion, and sinusitis. This estimate of WTP to
avoid a MRRAD, so defined, is $38.37 (1990 $), or after adjusting for inflation and income
growth $68 (2017 income level and 2010 $). Although Ostro and Rothschild (1989) estimated
the relationship between PM2.5 and MRADs, rather than MRRADs (a component of MRADs), it
is likely that most of the MRADs associated with exposure to PM2.5 are in fact MRRADs. For
the purpose of valuing this health endpoint, then, we assumed that MRADs associated with PM2.5
exposure may be more specifically defined as MRRADs, and therefore used the estimate of
mean WTP to avoid a MRRAD.
Any estimate of mean WTP to avoid a MRRAD (or any other type of restricted activity day other
than WLD) will be somewhat arbitrary because the endpoint itself is not precisely defined. Many
different combinations of symptoms could presumably result in some minor or less minor
restriction in activity. Krupnick and Kopp (1988) argued that mild symptoms will not be
sufficient to result in a MRRAD, so that WTP to avoid a MRRAD should exceed WTP to avoid
any single mild symptom. A single severe symptom or a combination of symptoms could,
however, be sufficient to restrict activity. Therefore WTP to avoid a MRRAD should, these
F -14
July 2013
-------
authors argue, not necessarily exceed WTP to avoid a single severe symptom or a combination of
symptoms. The "severity" of a symptom, however, is similarly not precisely defined; moreover,
one level of severity of a symptom could induce restriction of activity for one individual while
not doing so for another. The same is true for any particular combination of symptoms.
Valuing Asthma Exacerbations
Rowe and Chestnut (1986) surveyed asthmatics to estimate WTP for avoidance of a "bad asthma
day," as defined by the subjects. For purposes of valuation, an asthma attack is assumed to be
equivalent to a day in which asthma is moderate or worse as reported in the Rowe and Chestnut
study. Using the mean of average WTP estimates for the four severity definitions of a "bad
asthma day," the asthma exacerbation could be valued at $57.5 (2017 income level and 2010 $)
per incidence.
F-15
July 2013
-------
Appendix G: Additional Quick Start Tutorials on Sample
COBRA Scenarios
You can use COBRA to estimate the impacts of a new policy or program that results in a
change in air pollution. This appendix provides two case studies to give you a quick
introduction on how to work through the steps of a simple analysis for a few types of
policies: renewable energy supply goals or standards and energy efficiency programs.
Estimating the Benefits of Clean
Energy Policies
State and Local
Climate and Energy Program
Quickstart Tutorial: How To Use The
Co-Benefits Risk Assessment (COBRA)
Screening Model
Analytical Steps and Case Studies
IIHChMHi
fll mftiylli TiflliiWHtMll Til
United States
Environmental Protection
ป mAgency
July, 2013
G-l
July 2013
-------
&er<\ Overview of Presentation
iVl
fifr
How to conduct an analysis with COBRA
- Summarizes four key analytical steps
Two case studies illustrate how to apply these
steps in two clean energy scenarios:
1. Renewable Portfolio Standard
2. Energy Efficiency Programs in Public Buildings
How to Conduct an Analysis with
COBRA
State and Local
Climate and Energy Program
Analytical Steps and Relevant Resources
glilSaaiB
MhAOI
United States
Environmental Protection
Agency
G-2
July 2013
-------
&epa Steps in COBRA Analysis
2.
3.
Estimate where (e.g., in one or more counties or
states, regionally, nationally) and what emission
reductions will take place
Enter the location, types, and quantity of emission
reductions expected from the policy or activity in
COBRA
This presentation will:
Select a discount rate in COBRA
to appropriately discount the
value of future benefits
Run the model and review the
results
Walk you through
these steps, and
Lead you to other tools
and resources that can
help you develop your
inputs.
COBRA uses your inputs to estimate the air quality,
Mil health, and related economic impacts of the scenario
Ml
AEPA
Step 1: Estimate where and what
emissions reductions will take place
Decide on the geographic area
where emissions are expected to
change
COBRA can assess actions that
affect emissions in:
- a single county or state,
- groups of counties and states
(contiguous or otherwise), or
- the entire nation
COBRA allows you to vary the types
and amounts of emissions changes
expected to occur in different
locations
Estimating what and where electricity will
be displaced and emissions reduced
presents challenges due to the:
- Complex way electricity is generated
and transmitted across the U.S.
- Uncertainty about future emissions in
places with market-based
environmental programs, such as cap
and trade
Simplifying assumptions can be made
when using COBRA but a highly
sophisticated energy analysis of the
impacts of a clean energy policy on a
location will generate more reliable
results
For more information about the
complexity of the energy system, see
Chapters 3 and 4 of Assessing the Multiple
Benefits of Clean Energy: A Resource for
States, available at
http://www.eoa.gov/statelocalciimate/resoiirces
/benefits.html
G - 3
July 2013
-------
SEPA
Step 1: Estimate where and what
emissions reductions will take place
In COBRA, you can enter the emission changes as a
percentage or in absolute terms
- A percentage can be used when a policy is expected to
reduce emissions or use of an energy source by a specific
proportion
For example, for an increase in the use of renewable
electricity generation by 20%, you could assume that the
use of existing fuels for electricity generation would be
reduced by 20%
- An absolute number can be used for policies that do not lend
themselves easily to percentage reductions or when you want
to enter more specific emission changes
For example, 5,000 tons of sulfur dioxide
AEPA
Resources for Calculating Emissions Reductions
from Electricity-related Policies
If you do not have absolute emission reduction estimates,
you can use:
- A basic approach or tool, such as:
Applying an emission factor obtained from EPA's Emissions &
Generation Resource Integrated Database (eGrid)
http://www.epa.gov/cleanenergv/energv-resources/egrid/iridex.html. or
EPA's Power Plant Emissions Calculator (P-PEC)
http://www.epa.gov/airqualitv/eere/quantify.html
- More sophisticated approaches, such as those described in EPA
guides:
Assessing the Multiple Benefits of Clean Energy: A Resource for
States, Chapter 4
http://www.epa.eov/statelocalclimate/documents/pdf/epa assessing benefits ch4.
odf
Roadmapfor Incorporating Energy Efficiency/Renewable Energy
Policies and Programs into State and Tribal Implementation Plans,
Appendix I http://www.ep3.E0v/airgualitv/eere/manual.html
G-4
July 2013
-------
SEPA
Step 2: Select and enter the types, location and
quantity of emission reductions expected
m
You will need to know what source categories of emissions
will be affected by the policy
The emissions inventory in COBRA includes the 14 major
emissions source categories (i.e., "tiers") of criteria
pollutants included in the National Emissions Inventory
(NEI):*
Chemical and Allied Product
Manufacturing
Fuel Combustion - Electric Utility
Fuel Combustion - Industry
Fuel Combustion - Other
Highway Vehicles
Metal Processing
Miscellaneous
Natural Sources (Biogenics)
- Off-Highway
- Other Industrial Processes
- Petroleum & Related Industries
- Solvent Utilization
- Storage & Transport
- Waste Disposal & Recycling
*For more on the 2008 NEI, see:
http://www.epa.gov/ttnchiel/net/2008inve
ntorv.html
AEPA
Step 2: Select and enter the types, location and
quantity of emission reductions expected
Often, clean energy investments, such as those that
increase the use of renewable energy or energy
efficiency, will affect the "fuel combustion from
electric utilities" category
Within each category, there are fuel choices, such as
coal, gas, and oil
If you know the specific fuel will be affected, you may
choose it
- If not, you can use the broader category
Enter the estimated emission reductions by the
appropriate types and locations, ensuring that you
save your inputs once you are finished
G - 5
July 2013
-------
vปEPA
Step 3: Select a discount rate
A discount rate is used to appropriately discount the
value of future benefits
Not all benefits occur in the year of analysis, and
people are generally willing to pay more for
something now than for the same thing later
COBRA accounts for this time preference by
discounting benefits received later
10
Step 3: Select a discount rate
EPA's Guidelines for Economic Analysis recommend
using both 3% and 7% discount rates to see how the
conclusions of your analysis change. Both rates are
available in COBRA
The discount rate will affect the value of the benefits
A higher discount rate favors investments with
immediate benefits and reduces the value of future
benefits
A lower discount rate places a greater value on future
benefits to society
You can run your scenario with both rates and then
evaluate the effect of the change in discount rate on
the results
ii
G-6
July 2013
-------
SEPA
Step 4: Run the model and review the
results
Once you have completed these four steps, you are
ready to run the model, which will take a few
minutes depending on the speed of your computer
You can view the results for the changes in air
quality, health effects, and related economic value in
table and map forms
You can export results as tables and copy/paste
screenshots into reports and presentations
12
AEPA
Key Considerations when Interpreting
Results
COBRA is intended as a screening tool
- COBRA does not predict the future but can be
used to obtain ballpark health benefits estimates
and to compare or rank options
When more detailed analyses are required,
consider using more sophisticated modeling
approaches
TO*
13
G-7
July 2013
-------
SEPA
Key Considerations when Interpreting
Results
There is uncertainty surrounding the values of
key assumptions embedded in COBRA (i.e.,
emissions inventory, health impact functions,
and economic values)
- You should review the limitations and assumptions
described in the COBRA User Manual
14
AEPA
Key Considerations when Interpreting
Results
Emissions in some states and regions are
"capped" and firms may trade emission
allowances
- If you assume an emission reduction among power
plants in a state, emissions from other power plants
may increase unless emission allowances are retired
as part of the assumed emission reduction
COBRA does not automatically capture this
potential effect; it would need to be calculated
in another model
15
G-8
July 2013
-------
Case Study 1:
Renewable Portfolio Standard
State and Local
Climate and Energy Program
This case study illustrates how to conduct an analysis of a clean energy policy
with COBRA using a renewable portfolio standard as an example.
SEPA
United States
Environmental Protection
Agency
AEPA
Analyzing the Health Benefits of a
Renewable Portfolio Standard with COBRA
A renewable portfolio standard (RPS) requires
electric utilities to switch a particular
percentage of electricity generation to
renewable sources
If electricity had previously been generated
with fossil fuels, the RPS will result in criteria air
pollutant reductions and health benefits
17
G-9
July 2013
-------
SEPA
Analyzing the Health Benefits of a
Renewable Portfolio Standard with COBRA
The next slides describe how to estimate
the health and related economic benefits of
a state or local RPS
- Specifically, we assume a state (Michigan) has
established an RPS requirement that 10% of
electricity generation must be from renewable
sources by 2015
We also could have looked at a county with a
renewable target or requirement
18
AEPA
Step 1: Estimate where and what emissions
reductions will take place
Select what geographic locations you expect
to be affected by the emissions change
- You can enter emissions changes at the national,
regional, state or county levels
- If you know that specific plants will be affected,
you can enter emissions changes only in those
counties
- Or you could use more sophisticated energy
modeling approaches or tools to identify any and
all plants that may be affected by a state or local
RPS and manually enter those changes for the
counties with affected plants
TO*
19
G-10
July 2013
-------
SEPA
Step 1: Estimate where and what emissions
reductions will take place
For the Michigan RPS, we assume that all
emission changes will occur statewide
In COBRA, we create a scenario for an
individual state and select Michigan
20
AEPA
Step 1: Estimate where and what emissions
reductions will take place
To determine the emissions reduced, you can:
- Assume that a switch of 10% of electricity
generation from fossil fuels to renewable sources
that do not generate air pollution will reduce 10%
of all pollutants, or
- Estimate absolute emission reductions using:
An emission factor approach as described earlier
A more sophisticated modeling approach, if
available
TO*
21
(i - U
July 2013
-------
SEPA
Step 1: Estimate where and what emissions
reductions will take place
For this example, we use emissions factors from EPA's
Emissions & Generation Resource Integrated Database
(eGrid)* to develop an absolute estimate
- Using "eGRID2012 year 2009 Summary Tables (PDF)," we
found:
Net electric generation in Michigan: 88 million MWh
Non-baseload output emissions rates for Michigan:
S02: 6.6348 lbs. per MWh
NOx: 1.9392 lbs. per MWh
Percentage of electric generation that already comes from
renewable sources in Michigan: 3.1%
! eGRID Is available at http://www.epa.gov/cleanenergv/engrgy-resQurces/egrid/indgx.html
22
AEPA
TO*
Step 1: Estimate where and what emissions
reductions will take place
Since 3.1% of electric generation already comes from
renewable sources, we assume our scenario will reduce
emissions by:
10% - 3.1% = 6.9%
We calculate the reduction in MWh:
6.9% x 88 million MWh = 6 million MWh
Assuming the renewable energy used does not emit any
air pollution, we calculate the emission reductions as:
S02: 6 million MWh x 6.6348 per MWh = 40 million lbs.
= 20,000 tons
NOx: 6 million MWh x 1.9392 per MWh = 12 million lbs.
= 6,000 tons
[Note that 1 ton = 2,000 lbs.]
23
G-12
July 2013
-------
SEPA
Step 2: Set up Scenario in COBRA
(a) Location of Emission Reductions Expected
File View Help
COBRA
3 Screening
Wil Model
Analysis Year 2017
Scenario Options
Run a new scenario:
."flr*11
for individual slates:
5>
Maryland
Mississippi
_ Missouri
Montana
Nebraska
Nevada
New Hampshire
New Jersey
Start I
Overview Emissions
Welcome to the Co-Benefits Risk Assessment
Screening Model (COBRA)
To begin using COBRA, you may
1) Explore the analysis year 2017 emissions data.
This data can be accessed in table and map form by clicking on the
"Emissions" button above Viewing the baseline data first can help
you decide what changes you want to make in your own scenario.
2) Create your own scenario
Vou can create a new scenario through the left panel of this page
or load in a previously saved scenario through 'File' -> 'Load'
24
AEPA
Step 2: Set up Scenario in COBRA
(b) Types of Emission Reductions Expected
A RPS affects the fuel combustion from electricity
generation category
- These categories include fuel choices (e.g., gas, coal)
You can select specific fuel choices that are expected to
be affected if known or assume all fuel choices are
affected
For the Michigan RPS example, we assume that all fuel
sources would be affected by the RPS (i.e., not just
natural gas or just coal) and select the "fuel combustion
from electricity generation" category
25
G-13
July 2013
-------
oEPA
Step 2: Set up Scenario in COBRA
(b) Types of Emission Reductions Expected
Ml
T o change emissions estimates, cfck on a source category and enter your chan
Edits button alter edihng each source category for your changes to be recorded
Currently active category:
|FUEL COM0 ELEC UTIL
t category and enter your changes in the panel below You MUST cfck the Apply
i uour ehanoes to be recorded
Ifl rtlELOOMU.il 1UUJII-
12 FUEL COMB. OTHER
ffl HIGHWAY VEHICLES
ffl METALS PROCESSING
IS MISCELLANEOUS
ฎ NATURALSOURCES
ฎ OFF-HIGHWAY
ffi OTHER INDUSTRIAL PROCESSES
B PETROLEUM & RELATED INDUSTRIES
m SOLVENT UTILIZATION
- STORAGE & TRANSPORT
WASTE DISPOSAL & RECYCLING
PM2.5:
S02:
NOx
NH3:
VOC.
< reduceby r
r increase by '
reduceby np
C increase by '
f? reduceby rp
C increase by '
6 reduceby np
r increase by '
C reduceby rp
C increase by '
ฆ <* percent
C tons
percent
C tons
1 < percent
f* tons
ฆ < percent
C tons
percerV
C tons
Apply Edits
26
Step 2: Set up Scenario in COBRA
(c) Quantity of emission reductions expected
Ml
Al Counties |
I o change emissions estimates, dick on a source category and enter your changes r> the panel below You MUST cfck the Apply
Edits button after editing each source category for you changes to be recorded
Currently active category
IFUELCOMB ELEC UTIL
ฎ CHEMICAL & ALUED PRODUCT MFG
ffi FUEL COMB ELEC UTIL
ฎ FUEL COMB INDUSTRIAL
ffl FUEL COMB OTHER
0 HIGHWAY VEHiaES
ffl METALS PROCESSING
ฉ MISCELLANEOUS
S NATURALSOURCES
ฎ OFF-HIGHWAY
OTHER INDUSTRIAL PROCESSES
ffi PETROLEUM & RELATED INDUSTRIES
SOLVENT UTIUZATION
STORAGE S TRANSPORT
WASTE DISPOSALS RECYCUNG
^c"" I
Sun Scenario ฆ>J
27
G- 14
July 2013
-------
vปEPA
Step 3: Select a discount rate
A discount rate is used to appropriately
discount the value of future benefits
In this case study, we use a 3% discount rate
This discount rate provides an upper bound
for the estimated benefits and places a
greater value on future benefits to society,
compared to higher discount rates
28
Step 3: Select a discount rate
COBRA estimates the economic value of current and future avoided deaths and
illnesses expected based on emissions reductions in the year 2017. Emission reductions
require investments and, like aB investments, there are trade-offs, or opportunity costs, of
picking one investment over another, each with their own set and schedule of expected
benefits. T o reflect the opportunity costs of the investments foregone by investing in
emission reductions and to figure out how much future benefits are worth today, COBRA
users must select a discount rate.
Rather than using just a single rate, EPA's Guidelines for Economic Analysis recommend that analysts use
a bounding approach to discounting, developing an upper and lower bound for their estimates. They
advise use of both.
a 3% rate, reflecting the interest rate consumers might earn on Government backed securities, and
a 7% fate, reflecting the opportunity cost of private capital, based on estimates from the Office of
Management arid Budget.
NOTE: A higher discount rate favors those investments with immediate benefits and reduces the value of
future benefits more than a lower discount rate, which places a greater value on future benefits to society.
For more information on discount rates and how EPA uses them in monetizing health benefits, see the
User Manual
In order to run the COBRA model, please select a discount rate to use in this COBRA session.
Continue
29
G-15
July 2013
-------
*>EPA
Step 4: Run the model and review the
results
We calculated absolute
emissions reductions of
Michigan's renewable
portfolio standard of 10%.
Annual Emission
Reductions (shorttons)
COBRA (1) converted emissions
reductions into air quality
improvements, and (2) estimated
annual adverse health impacts avoided.
COBRA monetized the value
or benefits of the avoided
adverse health effects.
Annual Adverse Health Impacts Avoided I Annual Benefits (20io,$i,ooos)
Pollutant
Sulfur Dioxide (S02)
Nitrogen Oxides
(NOJ
Amount
20,000
6,000
Outcome Number
Mortality 120 - 272
Asthma Exacerbations 3,267
Heart Attacks 15 -139
Hospital Admissions 79
Acute Bronchitis 169
Respiratory Symptoms 5,233
Asthma ER Visits 69
Minor Restricted Activity Days 88,325
Work Days lost 14,784
total
* Don't forget to consider the caveats from slides 13 through 15
Case Study 2:
Energy Efficiency Programs in
Public Buildings
This case study illustrates how to conduct an analysis of a clean energy program
with COBRA using an energy efficiency program as an example.
Dollar Value
$1,012,503 -$2,291,598
$187
$1,841 - $17,103
$2,647
$81
$147
$29
$5,981
$2,230
$1,025,644 - $2,320,002
30
State and Local
Climate and Energy Program
United States
Environmental Protection
Agency
G- 16
July 2013
-------
SEPA
Using COBRA to Evaluate Energy
Efficiency Programs in Public Buildings
Energy efficiency programs are intended to
reduce electricity use in homes or businesses
If the electricity had previously been generated
with fossil fuels, the programs can lead to
criteria air pollutant reductions and health
benefits
32
AEPA
Using COBRA to Evaluate Energy
Efficiency Programs in Public Buildings
The next slides describe how to estimate the
health and related economic benefits of a state
or local energy efficiency program in buildings
- Specifically, we assume a city (Richmond, Virginia)
has decided to explore the benefits associated with
increasing the efficiency of municipal facilities and
operations
33
G-17
July 2013
-------
SEPA
Step 1: Estimate where and what
emissions reductions will take place
Select what geographic locations you expect to
be affected by the emissions change
- You can enter emissions changes at the national,
regional, state or county levels
- If you know that specific plants will be affected, you
can enter emissions changes only in those counties
- Or you could use more sophisticated energy
modeling approaches or tools to identify any and all
plants that may be affected by a state or local energy
efficiency program for buildings and enter those
changes in manually
34
AEPA
Step 1: Estimate where and what
emissions reductions will take place
For this example, we assume that the energy
savings will take place in buildings within
Richmond
Due to the interconnectedness of the grid,
these savings will affect electricity providers
and emissions beyond Richmond
TO*
35
G-18
July 2013
-------
SEPA
Step 1: Estimate where and what
emissions reductions will take place
To estimate the electricity reductions expected from
the program, you can either:
Estimate how many kWhs you expect to save, or
Find a similar program and use their results as a proxy
In this hypothetical example, we assume that
Richmond bases their program upon and achieves
similar reductions to a program in Arlington County,
VA (Fresh AIRE - Arlington Initiative to Reduce
Emissions)
Fresh AIRE had an estimated a reduction in annual electricity
consumption of more than 3,000,000 kWh in county
buildings
For more details, see:
Fresh AIRE case study in EPA's "Energy Efficiency in Local Government Operations" report, available at
I Ml I http://www.epa.gov/statelocalclimate/documents/Ddf/ee municipal operations.pdf. and
iLZjjjjB, * Arlington County's Fresh AIRE website: at http://freshaireva.us/2011/10/countv-buildings/.
SSS
36
AEPA
Step 1: Estimate where and what
emissions reductions will take place
To estimate the emissions reduced from the
3,000,000 kWh of generation avoided each year,
you can use:
An emission factor approach as described earlier
- A more sophisticated modeling approach, if available
TO*
37
G- 19
July 2013
-------
ฐEPA ^t0P Est'mate where and what
emissions reductions will take place
For this example, we use EPA's Draft Power Plant
Emissions Calculator (P-PEC)* to:
- Determine that Richmond City is in the SERC Virginia/
Carolina eGRID subregion
- Apply a 3,000,000 KWh energy savings goal for the power
plants within this eGRID subregion which spans across GA,
NC, SC, VA, and WV
- Calculate the emission reductions for each state and
counties within the states in SERC Virginia/Carolina eGRID
subregion
P-PEC is available at http://www.epa.gov/airqualitv/eere/quantifv.html
38
Step 1: Estimate where and what emissions
reductions will take place
Emission reductions from 3,000,000 Kwh
energy savings using P-PEC:
- North Carolina: 0.69 tons of NOx, 0.28 tons of S02
- South Carolina: 0.48 tons of NOx, 0.19 tons of S02
- Virginia: 0.56 tons of NOx, 0.19 tons of S02
- Grant County, WV: 0.05 tons of NOx
ifa
Bp
i
is
5P
s
oEPA
G- 20
July 2013
-------
SEPA
Step 2: Set up Scenario in COBRA
(a) Location of Emission Reductions Expected
Emission reductions in North Carolina, South
Carolina, and Virginia are calculated at the state
level
<
~ CซimI
: i bmi
-------
SEPA
Step 2: Set up Scenario in COBRA
(b)Types of Emission Reductions Expected
Since energy efficiency programs affect
electricity use, the affected emissions
category is "fuel combustion from electricity
generation"
This categories includes fuel choices (e.g.,
gas, coal)
Since all fuel sources would be affected by
the energy efficiency program, select the
"fuel combustion from electricity generation"
category
42
AEPA
Step 2: Set up Scenario in COBRA
(b)Types of Emission Reductions Expected
Select emissions category for each affected state or
county
c |sc
A1 Court** |
Currently active category;
< ITFUEL COMB ELEC UTlQ
hUtLLUMB INUUb. I t-tiAL
ฆ_rhf M\rM s ai i iFnPftnni irrnrr.
< reduce by r
C rKic&ze by '
ffl FUEL COMB OTHER
S HIGHWAY VEHICLES
ffi METALS PROCESSING
2 MISCELLANEOUS
ซB NATURAL SOURCES
S OFF-HIGHWAY
83 OTHER INDUSTRIAL PROCESSES
ffi PETROLEUM & RELATEO INDUSTRIES
S SOLVENT UTTUZATION
a STORAGE & TRANSPORT
S3 WASTE DISPOSAL & RECYCLING
.I"5"
< reduecby rj
C irweatebp '
ItF
a leduceby |jj
[ftoVMVTl
43
G-22
July 2013
-------
SEPA
Step 2: Set up Scenario in COBRA
(c) Quantity of emission reductions expected
Enter emission reductions for each affected state or
county
- Do not forget to enter and click "Apply Edits" for each state/
county
t
AlCotrtm |
I wv
Currently acbva coloqory
|FUEL COMB ELEC UTH.
ฆ CHEMICAL & ALLIED PRODUCT MFG
: FUEL COMB ELEC UTIL
J FUEL COMB INDUSTRIAL
i FUEL COMB OTHER
? HIGHWAY VEHICLES
3 METALS PROCESSING
3 MISCELLANEOUS
NATURALSOURCES
i OFF-HIGHWAY
i OTHER INDUSTRIAL PROCESSES
; FETROLEUM & RELATED INDUSTRIES
3 SOLVENT UTILIZATION
i STORAGE & TRANSPORT
3 WASTE DISPOSAL & RECYCLING
NO*
NH3
VOC
44
ซปEFฅ\ step 3: Select a discount rate
A discount rate is used to appropriately
discount the value of future benefits
In this case study, we use a 3% discount rate
This discount rate provides an upper bound for
the estimated benefits and places a greater
value on future benefits to society, compared to
higher discount rates
G-23
July 2013
-------
x>ERi\ step 3: Select a discount rate
111
oEPA
Select a Discount Rate for the Scenario
COBRA estimates the economic value of current and future avoided deaths and
illnesses expected based on emissions reductions in the year 2017. Emission reductions
require investments and. like aH investments, there are trade-offs, or opportunity costs, of
picking one investment over another, each with their own set and schedule of expected
benefits. T o reflect the opportunity costs of the investments foregone by investing in
emission reductions and to figure out how much future benefits are worth today, COBRA
users must select a discount rate.
Rather than using just a single rate, EPA's Guidelines for Economic Analysis recommend that analysts use
a bounding approach to discounting, developing an upper and lower bound for their estimates. They
advise use of both:
a 2% rate, reflecting the interest rate consumers might earn on Government backed securities, and
a 1% rate, reflecting the opportunity cost of private capital, based on estimates from the Office of
Management and Budget
NOTE A higher discount rate favors those investments with immediate benefits and reduces the value of
future benefits more than a lower discount rate, which places a greater value on future benefits to society.
For more information on discount rates and how EPA uses them in monetizing health benefits, see the
User Manual.
In order to run the COBRA model, please select a discount rate to use in this COBRA session.
<*p!
rJx
Continue
_l
46
Step 4: Run the model and review the
results
We used eGrid to calculate
the emissions reductions
due to a 3,000,000 kWh
reduction in electricity use.
COBRA (1) converted emissions
reductions into air quality
improvements, and (2) estimated
annual adverse health impacts avoided.
COBRA monetized the value
or benefits of the avoided
adverse health effects.
Annual Emission Reductions
| Annual Adverse Health Impacts Avoided |
| Annual Benefits (20io, Si,ooos|
(short tons)
Outcome Number
Dollar Value
Pollutant Amount
Mortality 0.006 - 0.013 $48 - $109
Sulfur Dioxide (S02)
0.66
Nitrogen Oxides
(NO,)
1.78
^!8
Asthma Exacerbations
Heart Attacks
Hospital Admissions
Acute Bronchitis
Respiratory Symptoms
Asthma ER Visits
Minor Restricted Activity Days
Work Days Lost
* Don't forget to consider the caveats from slides 13 through 15
0.16
o.oo - 0.01
0.00
Note: These reductions are
aggregated across all affected
states.
0.01
0.26
0.00
4.36
0.73
total
$0.01
$0.08 - $0.78
$0.12
$0
$0.01
$0
$0.30
$0.11
$49 - $110
47
G - 24
July 2013
-------
vปEPA
How Can I Learn More?
Visit Our Website:
http://www.epa.Eov/statelocalclimate/resources/cobra.html
Contact Us:
Denise Mulholland
EPA State and Local Climate and Energy Programs
(202)343-9274
Mulholland.Denise@epa.gov
State and Local
Climate and Energy Program
G - 25
July 2013
-------
Appendix H: References
Abt Associates Inc. (2000). Final Heavy Duty Engine/Diesel Fuel Rule: Air Quality Estimation, Selected
Health and Welfare Benefits Methods, and Benefit Analysis Results. Bethesda, MD: Prepared for
U.S. EPA, Office of Air Quality Planning and Standards, Research Triangle Park, NC.
Adams, P. F., Hendershot, G. E., & Marano, M. A. (1999). Current Estimates from the National Health
Interview Survey, 1996. Vital Health Stat, 70(200), 1-212.
Agency for Healthcare Research and Quality. Healthcare Cost and Utilization Project (HCUP), from
http ://www .hcup-us. ahrq .gov/
Agency for Healthcare Research and Quality. (2007). Healthcare Cost and Utilization Project. National
Inpatient Sample (NIS). Rockville, Maryland.
Agency for Healthcare Research and Quality. (2012). HCUPnet, Healthcare Cost and Utilization Project,
from Agency for Healthcare Research and Quality, Rockville, MD http://hcupnet.ahrq.gov/
American Lung Association. (2002, September 2002). Trends in Morbidity and Mortality: Pneumonia,
Influenza, and Acute Respiratory Conditions, from http://www.lungusa.org/data/pi/PI_l.pdf
American Lung Association. (2010). Trends in Asthma Morbidity and Mortality: American Lung
Association Epidemiology and Statistics Unit Research and Program Services Division.
Arlington County Government. (2012). County Buildings Retrieved June 15, 2012, from
http://freshaireva.us/2011/10/county-buildings/
Babin, S. M., Burkom, H. S., Holtry, R. S., Tabernero, N. R., Stokes, L. D., Davies-Cole, J. O., . . . Lee,
D. H. (2007). Pediatric patient asthma-related emergency department visits and admissions in
Washington, DC, from 2001-2004, and associations with air quality, socio-economic status and
age group. Environ Health, 6, 9.
Bell, M. L., & Dominici, F. (2008). Effect Modification by Community Characteristics on the Short-term
Effects of Ozone Exposure and Mortality in 98 U.S. Communities. American Journal of
Epidemiology, 7(57(8), 986-997.
Bell, M. L., Ebisu, K., Peng, R. D., Walker, J., Samet, J., Zeger, S. L., & Dominici, F. (2008). Seasonal
and Regional Short-term Effects of Fine Particles on Hospital Admissions in 202 US Counties,
1999-2005. American Journal of Epidemiology, 7(55(11), 1301-1310.
Bobak, M., & Leon, D. A. (1999). The effect of air pollution on infant mortality appears specific for
respiratory causes in the postneonatal period [see comments]. Epidemiology, 10(6), 666-670.
Bureau of Labor Statistics. Consumer Price Index, from http://www.bls.gov/cpi/
Bureau of Labor Statistics. (2012, January 31, 2012). Employment Cost Index: Wages & Salaries: Private
Industry Workers, from http://research.stlouisfed.org/fred2/data/ECIWAG.txt
H-1
July 2013
-------
Conceicao, G. M., Miraglia, S. G., Kishi, H. S., Saldiva, P. H., & Singer, J. M. (2001). Air pollution and
child mortality: a time-series study in Sao Paulo, Brazil. Environ Health Perspect, 709(Suppl 3),
347-350.
Coursey, J. S., Schwab, D. J., Tsai, J. J., & Dragoset, R. A. (2011). Atomic Weights and Isotopic
Compositions with Relative Atomic Masses, from
http ://physics .nist.gov/PhysRefData/ Compositions
Cropper, M. L., & Krupnick, A. J. (1990). The Social Costs of Chronic Heart and Lung Disease.
Washington, DC: Resources for the Future.
Cropper, M. L., & Sussman, F. G. (1990). Valuing Future Risks to Life. Journal of Environmental
Economics and Management, 19, 160-174.
Dekker, J. M., Crow, R. S., Folsom, A. R., Hannan, P. J., Liao, D., Swenne, C. A., & Schouten, E. G.
(2000). Low heart rate variability in a 2-minute rhythm strip predicts risk of coronary heart
disease and mortality from several causes: the ARIC Study. Atherosclerosis Risk In
Communities. Circulation, 102(11), 1239-1244.
DerSimonian, R., & Laird, N. (1986). Meta-Analysis in Clinical Trials. Controlled Clinical Trials, 7, 177-
188.
Dickie, M., & Gerking, S. (1987). Reconciling Averting Behavior and Contingent Valuation Benefit
Estimates of Reducing Symptoms of Ozone Exposure (draft), as cited in Neumann, J.E., M.
Dickie, and R.E. Unsworth. 1994.: Prepared by Industrial Economics. Prepared for Jim
DeMocker, U.S. EPA, Office of Air and Radiation.
Dockery, D. W., Cunningham, J., Damokosh, A. I., Neas, L. M., Spengler, J. D., Koutrakis, P., . . .
Speizer, F. E. (1996). Health Effects of Acid Aerosols On North American Children - Respiratory
Symptoms. Environmental Health Perspectives, 104(5), 500-505.
Dockery, D. W., Speizer, F. E., Stram, D. O., Ware, J. H., Spengler, J. D., & Ferris, B. G., Jr. (1989).
Effects of Inhalable Particles on Respiratory Health of Children. Am Rev Respir Dis, 139, 587-
594.
E.H. Pechan & Associates Inc. (1994). Development of the OPPE Particulate Programs Implementation
Evaluation System. Springfield, VA: Prepared for Christopher Knopes, U.S. Environmental
Protection Agency, Office of Policy, Planning and Evaluation.
E.H. Pechan & Associates Inc. (1997). Integrated Ozone and Particulate Matter and Regional Haze Costs
Analysis: Methodology and Results. Prepared for Innovative Strategies and Economics Group,
Office of Air Quality Planning and Standards, US. Environmental Protection Agency. Research
Triangle Park, NC.
Eisenstein, E. L., Shaw, L. K., Anstrom, K. J., Nelson, C. L., Hakim, Z., Hasselblad, V., & Mark, D. B.
(2001). Assessing the clinical and economic burden of coronary artery disease: 1986-1998. Med
Care, 39(8), 824-835.
GeoLytics Inc. (2002). CensusCD 2000 Long Form Profile Release 1.0.
H - 2
July 2013
-------
Glad, J.A., L.L. Brink, E.O. Talbott, P.C. Lee, X. Xu, M. Saul, and J. Rager. 2012. The Relationship of
Ambient Ozone and PM2.5 Levels and Asthma Emergency Department Visits: Possible Influence
of Gender and Ethnicity. Archives of Environmental & Occupational Health. Vol 62 (2): 103-
108.
Gold, D. R., Litonjua, A., Schwartz, J., Lovett, E., Larson, A., Nearing, B., . . . Verrier, R. (2000).
Ambient pollution and heart rate variability. Circulation, 707(11), 1267-1273.
Grabow, M. L., Spak, S. N., Holloway, T., Stone, B., Mednick, A. C., & Patz, J. A. (2012). Air quality
and exercise-related health benefits from reduced car travel in the midwestern United States.
Environ Health Perspect, 720(1), 68-76. doi: 10.1289/ehp.l 103440
Greene, W. H. (1997). Econometric Analysis. Upper Saddle River, NJ: Prentice Hall.
Hiemstra, P. (2012). Package 'automap': Automatic interpolation package. Version 1.0-12. For R version
2.14.1 (2011-12-22) Copyright (C) 2011 The RFoundation for Statistical Computing ISBN 3-
900051-07-0.
Hollman, F. W., Mulder, T. J., & Kalian, J. E. (2000). Methodology and Assumptions for the Population
Projections of the United States: 1999 to 2100. Washington, D.C.: Population Projections Branch,
Population Division, U.S. Census Bureau, Department of Commerce.
Industrial Economics Incorporated (IEc). (1993). Memorandum to Jim DeMocker, U.S. Environmental
Protection Agency, Office of Air and Radiation, Office of Policy Analysis and Review.
Industrial Economics Incorporated (IEc). (1994). Linkage Between Health Effects Estimation and
Morbidity Valuation in the Section 812 Analysis ~ Draft Valuation Document: Memorandum to
Jim DeMocker, U.S. Environmental Protection Agency, Office of Air and Radiation, Office of
Policy Analysis and Review. Prepared by J.E. Neumann, M.T. Dickie, and R.E. Unsworth.
Industrial Economics Incorporated (IEc). (2006). Expanded Expert Judgment Assessment of the
Concentration-Response Relationship Between PM2.5 Exposure and Mortality, Final Report.
Cambridge, MA: Prepared for U.S. Environmental Protection Agency, Office of Air and
Radiation, Office of Policy Analysis and Review. Prepared by IEc.
Jerrett, M., Burnett, R. T., Ma, R., Pope, C. A., 3rd, Krewski, D., Newbold, K. B., . . . Thun, M. J. (2005).
Spatial analysis of air pollution and mortality in Los Angeles. Epidemiology, 16(6), 727-736.
Judge, G. G., Griffiths, W. E., Hill, R. C., Lutkepohl, H., & Lee, T.-C. (1985). The Theory and Practice
of Econometrics (2nd ed.). New York: John Wiley and Sons.
Katsouyanni, K., Touloumi, G., Samoli, E., Gryparis, A., Le Tertre, A., Monopolis, Y., . . . Schwartz, J.
(2001). Confounding and effect modification in the short-term effects of ambient particles on
total mortality: results from 29 European cities within the APHEA2 project. Epidemiology, 72(5),
521-531.
Kleckner, N., & Neumann, J. (1999). Recommended Approach to Adjusting WTP Estimates to Reflect
Changes in Real Income: Memorandum to Jim Democker, U.S. EPA/OPAR.
H-3
July 2013
-------
Kloog, I., B.A. Coull, A. Zanobetti, P. Koutrakis, J.D. Schwartz. 2012. Acute and Chronic Effects of
Particles on Hospital Admissions in New-England. PLoS ONE. Vol 7 (4): 1-8.
Krewski, D., Jerrett, M., Burnett, R. T., Ma, R., Hughes, E., Shi, Y., . . . Tempalski, B. (2009). Extended
follow-up and spatial analysis of the American Cancer Society study linking particulate air
pollution and mortality. Res Rep Health Eff Inst(\40), 5-114; discussion 115-136.
Krewski, D., Jerrett, M., Burnett, R. T., Ma, R., Hughes, E., Shi, Y., . . . Tempalski, B. (2009). Extended
follow-up and spatial analysis of the American Cancer Society study linking particulate air
pollution and mortality. Res Rep Health Eff Inst(\40), 5-114; discussion 115-136.
Krupnick, A. J., & Kopp, R. J. (1988). The Health and Agricultural Benefits of Reductions in Ambient
Ozone in the United States. Washington, DC: Resources for the Future.
Laden, F., Schwartz, J., Speizer, F. E., & Dockery, D. W. (2006). Reduction in Fine Particulate Air
Pollution and Mortality: Extended follow-up of the Harvard Six Cities Study. Am JRespir Crit
Care Med, 173(6), 667-672.
Latimer, D. A. (1993, June 13-18). Development of Regional Haze Screening Models. Paper presented at
the 86th Annual Meeting and Exhibition, Denver, CO.
Lepeule J, Laden F, Dockery D, Schwartz J. Chronic exposure to fine particles and mortality: an extended
follow-up of the Harvard Six Cities study from 1974 to 2009. Vol 120(7). 965-970.
Levy, J. I., Wilson, A. M., Evans, J. S., & Spengler, J. D. (2003). Estimation of Primary and Secondary
Particulate Matter Intake Fractions for Power Plants in Georgia. Environ Sci Technol, 37, 5528-
5536.
Liao, D., Cai, J., Rosamond, W. D., Barnes, R. W., Hutchinson, R. G., Whitsel, E. A., . . . Heiss, G.
(1997). Cardiac autonomic function and incident coronary heart disease: a population-based case-
cohort study. The ARIC Study. Atherosclerosis Risk in Communities Study. Am J Epidemiol,
745(8), 696-706.
Liao, D., Creason, J., Shy, C., Williams, R., Watts, R., & Zweidinger, R. (1999). Daily variation of
particulate air pollution and poor cardiac autonomic control in the elderly. Environ Health
Perspect, 107(7), 521-525.
Loehman, E. T., Berg, S. V., Arroyo, A. A., Hedinger, R. A., Schwartz, J. M., Shaw, M. E., . . . Green, A.
E. S. (1979). Distributional Analysis of Regional Benefits and Cost of Air Quality Control.
Journal of Environmental Economics and Management, 6, 222-243.
Loomis, D., Castillejos, M., Gold, D. R., McDonnell, W., & Boija-Aburto, V. H. (1999). Air pollution
and infant mortality in Mexico City. Epidemiology, 10(2), 118-123.
Magari, S. R., Hauser, R., Schwartz, J., Williams, P. L., Smith, T. J., & Christiani, D. C. (2001).
Association of heart rate variability with occupational and environmental exposure to particulate
air pollution. Circulation, 104(9), 986-991.
Mar, T. F., Koenig, J. Q., & Primomo, J. (2010). Associations between asthma emergency visits and
H-4
July 2013
-------
particulate matter sources, including diesel emissions from stationary generators in Tacoma,
Washington. Inhal Toxicol, 22(6), 445-448.
Mar, T. F., Larson, T. V., Stier, R. A., Claiborn, C., & Koenig, J. Q. (2004). An analysis of the
association between respiratory symptoms in subjects with asthma and daily air pollution in
Spokane, Washington. Inhal Toxicol, 7(5(13), 809-815.
Moolgavkar, S. H. (2000a). Air Pollution and Hospital Admissions for Chronic Obstructive Pulmonary
Disease in Three Metropolitan Areas in the United States. Inhalation Toxicology, 72(Supplement
4), 75-90.
Moolgavkar, S. H. (2000b). Air pollution and hospital admissions for diseases of the circulatory system in
three U.S. metropolitan areas. J Air Waste Manag Assoc, 50(1), 1199-1206.
Moolgavkar, S. H. (2003). Air Pollution and Daily Deaths and Hospital Admissions in Los Angeles and
Cook Counties. In: Revised Analyses of Time-Series Studies of Air Pollution and Health (pp.
183-198). Boston, MA: Health Effects Institute.
National Climatic Data Center. NAMER-WINDTEMP (North America wind and temperature) tapes.
Ashville, NC.
National Research Council. (2002). Estimating the Public Health Benefits of Proposed Air Pollution
Regulations. Washington, D.C.: The National Academies Press.
Ostro, B., Lipsett, M., Mann, J., Braxton-Owens, H., & White, M. (2001). Air pollution and exacerbation
of asthma in African-American children in Los Angeles. Epidemiology, 12(2), 200-208.
Ostro, B. D. (1987). Air Pollution and Morbidity Revisited: A Specification Test. Journal of
Environmental Economics and Management, 14, 87-98.
Ostro, B. D. (1989). Estimating the Risks of Smoking, Air Pollution, and Passive Smoke On Acute
Respiratory Conditions. Risk Anal, 9(2), 189-196.
Ostro, B. D., & Rothschild, S. (1989). Air Pollution and Acute Respiratory Morbidity - an Observational
Study of Multiple Pollutants. Environ Res, 50(2), 238-247.
Peng, R. D., Bell, M. L., Geyh, A. S., McDermott, A., Zeger, S. L., Samet, J. M., & Dominici, F. (2009).
Emergency admissions for cardiovascular and respiratory diseases and the chemical composition
of fine particle air pollution. Environ Health Perspect, 117(6), 957-963.
Peng, R. D., Chang, H. H., Bell, M. L., McDermott, A., Zeger, S. L., Samet, J. M., & Dominici, F.
(2008). Coarse particulate matter air pollution and hospital admissions for cardiovascular and
respiratory diseases among Medicare patients. Jama, 299(18), 2172-2179.
Peters, A., Dockery, D. W., Muller, J. E., & Mittleman, M. A. (2001). Increased particulate air pollution
and the triggering of myocardial infarction. Circulation, 103(23), 2810-2815.
Poloniecki, J. D., Atkinson, R. W., de Leon, A. P., & Anderson, H. R. (1997). Daily time series for
cardiovascular hospital admissions and previous day's air pollution in London, UK. Occup
H-5
July 2013
-------
Environ Med, 54(8), 535-540.
Pope, C. A. (1991). Respiratory hospital admissions associated with PM10 pollution in Utah, Salt Lake,
and Cache Valleys. Arch Environ Health, 46(2), 90-97.
Pope, C. A., 3rd, Burnett, R. T., Thun, M. J., Calle, E. E., Krewski, D., Ito, K., & Thurston, G. D. (2002).
Lung cancer, cardiopulmonary mortality, and long-term exposure to fine particulate air pollution.
Jama, 287(9), 1132-1141.
Pope, C. A., Dockery, D. W., Spengler, J. D., & Raizenne, M. E. (1991). Respiratory Health and PmlO
Pollution - a Daily Time Series Analysis. American Review of Respiratory Disease, 144(3), 668-
674.
Pope, C. A., 3rd, Muhlestein, J. B., May, H. T., Renlund, D. G., Anderson, J. L., & Home, B. D. (2006).
Ischemic heart disease events triggered by short-term exposure to fine particulate air pollution.
Circulation, 114(23), 2443-2448.
Popovic, J. R. (2001). 1999 National Hospital Discharge Survey: annual summary with detailed diagnosis
and procedure data. Vital Health Stat 73(151), i-v, 1-206.
Roman, H. A., Walker, K. D., Walsh, T. L., Conner, L., Richmond, H. M., Hubbell, B. J., & Kinney, P. L.
(2008). Expert judgment assessment of the mortality impact of changes in ambient fine
particulate matter in the U.S. Environ Sci Technol, 42(1), 2268-2274.
Rosamond, W., Broda, G., Kawalec, E., Rywik, S., Pajak, A., Cooper, L., & Chambless, L. (1999).
Comparison of medical care and survival of hospitalized patients with acute myocardial infarction
in Poland and the United States. Am J Cardiol, 53(8), 1180-1185.
Rowe, R. D., & Chestnust, L. G. (1986). Oxidants and Asthmatics in Los Angeles: A Benefits Analysis ~
Executive Summary. Washington, DC: Prepared for U.S. Environmental Protection Agency,
Office of Policy Analysis. Prepared by Energy and Resource Consultants, Inc.
Russell, M. W., Huse, D. M., Drowns, S., Hamel, E. C., & Hartz, S. C. (1998). Direct medical costs of
coronary artery disease in the United States. Am J Cardiol, 81(9), 1110-1115.
Samet, J. M., Dominici, F., Curriero, F. C., Coursac, I., & Zeger, S. L. (2000). Fine Particulate Air
Pollution and Mortality in 20 U.S. Cities, 1987-1994. N Engl J Med, 343(24), 1742-1749.
Schwartz, J. (2000). The distributed lag between air pollution and daily deaths. Epidemiology, 11(3), 320-
326.
Schwartz, J., Coull, B., Laden, F., & Ryan, L. (2008). The effect of dose and timing of dose on the
association between airborne particles and survival. Environ Health Perspect, 77(5(1), 64-69.
Schwartz, J., Dockery, D. W., Neas, L. M., Wypij, D., Ware, J. H., Spengler, J. D., . . . Ferris, B. G.
(1994). Acute Effects of Summer Air Pollution On Respiratory Symptom Reporting in Children.
Am JRespir Crit Care Med, 150(5), 1234-1242.
Schwartz, J., & Neas, L. M. (2000). Fine particles are more strongly associated than coarse particles with
H-6
July 2013
-------
acute respiratory health effects in schoolchildren. Epidemiology, 77(1), 6-10.
Sheppard, L. (2003). Ambient Air Pollution and Nonelderly Asthma Hospital Admissions in Seattle,
Washington, 1987-1994. In: Revised Analyses of Time-Series Studies of Air Pollution and
Health (pp. 227-230). Boston, MA: Health Effects Institute.
Sheppard, L., Levy, D., Norris, G., Larson, T. V., & Koenig, J. Q. (1999). Effects of ambient air pollution
on nonelderly asthma hospital admissions in Seattle, Washington, 1987-1994. Epidemiology,
70(1), 23-30.
Slaughter, J. C., Kim, E., Sheppard, L., Sullivan, J. H., Larson, T. V., & Claiborn, C. (2005). Association
between particulate matter and emergency room visits, hospital admissions and mortality in
Spokane, Washington. J Expo Anal Environ Epidemiol, 75(2), 153-159.
Smith, D. H., Malone, D. C., Lawson, K. A., Okamoto, L. J., Battista, C., & Saunders, W. B. (1997). A
national estimate of the economic costs of asthma. Am J Respir Crit Care Med, 156(3 Pt 1), 787-
793.
Standard and Poor's. (2000). The U.S. Economy: The 25 Year Focus.
Stanford, R., McLaughlin, T., & Okamoto, L. J. (1999). The cost of asthma in the emergency department
and hospital. Am J Respir Crit Care Med, 160(1), 211-215.
Sullivan, J., Sheppard, L., Schreuder, A., Ishikawa, N., Siscovick, D., & Kaufman, J. (2005). Relation
between short-term fine-particulate matter exposure and onset of myocardial infarction.
Epidemiology, 7(5(1), 41-48.
Tolley, G. S., & et al. (1986). Valuation of Reductions in Human Health Symptoms and Risks: Prepared
for U.S. Environmental Protection Agency.
TranSystems|E.H. Pechan. (2012). The Emissions & Generation Resource Integrated Database for 2012
(eGrid2012) Technical Support Document. Prepared for U.S. Environmental Protection Agency,
Office of Atmospheric Programs, Clean Air Markets Division. Retrieved from
http://www.epa.gov/cleanenergy/documents/egridzips/eGRID2012_year09_TechnicalSupportDoc
ument.pdf.
Tsuji, H., Larson, M. G., Venditti, F. J., Jr., Manders, E. S., Evans, J. C., Feldman, C. L., & Levy, D.
(1996). Impact of reduced heart rate variability on risk for cardiac events. The Framingham Heart
Study. Circulation, 94(11), 2850-2855.
Turner, D. B. (1970). Workbook of Atmospheric Dispersion Estimates, AP-26. Research Triangle Park,
NC: U.S. EPA, Office of Air Programs.
U.S. Bureau of the Census. (1997). Statistical Abstract of the United States: 1997 (117 ed.). Washington,
DC.
U.S. Bureau of the Census. (2000). Population Projections for States by Age, Sex, Race, and Hispanic
Origin: 1999 to 2100. Washington, DC: Population Projections Program, Population Division,
U.S. Census Bureau.
H- 7
July 2013
-------
U.S. Bureau of the Census. (2002). Statistical Abstract of the United States: 2001. Washington DC.
U.S. Bureau of the Census. (2008). Statistical Abstract of the United States: 2009. Washington, DC.
U.S. EPA. (1995). User's Guide for the Industrial Source Complex (ISC3) Dispersion Models. EPA-
454/B-95-003a. Research Triangle Park, NC.
U.S. EPA. (1997). The Benefits and Costs of the Clean Air Act: 1970 to 1990. Washington, DC: U.S.
EPA, Office of Air and Radiation, Office of Policy, Planning and Evaluation.
U.S. EPA. (1999). The Benefits and Costs of the Clean Air Act: 1990 to 2010: EPA Report to Congress.
Washington, DC: U.S. EPA, Office of Air and Radiation, Office of Policy.
U.S. EPA. (2002). Air Quality Criteria for Particulate Matter, Third External Review Draft. Research
Triangle Park, NC: National Center for Environmental Assessment, Office of Research and
Development.
U.S. EPA. (2003a). Biogenic Emissions Inventory System (BEIS) Modeling: BEIS Version 3.12.
U.S. EPA. (2003b). Technical Support Document for the Clear Skies Act 2003 Air Quality Modeling
Analyses. Research Triangle Park, NC: Office of Air Quality Planning and Standards, Emissions
Analysis and Monitoring Division.
U.S. EPA. (2005). Regulatory Impact Analysis for the Final Clean Air Interstate Rule. Research Triangle
Park, NC: Office of Air Quality Planning and Standards (OAQPS), Emission, Monitoring, and
Analysis Division and Clean Air Markets Division.
U.S. EPA. (2006). Final Regulatory Impact Analysis: PM2.5 NAAQS. Research Triangle Park, NC:
Office of Air and Radiation, Office of Air Quality Planning and Standards.
U.S. EPA. (2009). ProposedN02 NAAQS Regulatory Impact Analysis (RIA). Research Triangle Park,
NC.: Office of Air and Radiation, Office of Air Quality Planning and Standards Retrieved from
http ://www.epa.gov/ttn/ecas/ria.html.
U.S. EPA. (2010a). Guidelines for Preparing Economic Analyses. Washington, DC: National Center for
Environmental Economics, Office of Policy Economics and Innovation.
U.S. EPA. (2010b, August 18, 2010). Photochemical Modeling Tools, from
http://www.epa.gov/scram001/modelingapps_mats.htm
U.S. EPA. (201 la). Assessing the Multiple Benefits of Clean Energy: A Resource for States. (EPA-430-R-
11-014). Washington, DC: Retrieved from
http://www.epa.gov/statelocalclimate/documents/pdf/epa_assessing_benefits.pdf.
U.S. EPA. (201 lb). Emissions Modeling for the Final Mercury and Air Toxics Standards Technical
Support Document. Washington, DC.
U.S. EPA. (201 lc). Energy Efficiency in Local Government Operations. (EPA-430-R-09-033).
Washington, DC: Retrieved from
H-8
July 2013
-------
http://www.epa.gov/statelocalclimate/documents/pdf/ee_municipal_operations.pdf.
U.S. EPA. (2012a). 2008 National Emissions Inventory, version 2: Technical Support Document.
Research Triangle Park, NC: Retrieved from
http://www.epa.gov/ttn/chief/net/2008neiv2/2008_neiv2_tsd_draft.pdf.
U.S. EPA. (2012b, May 31, 2012). Calculations and References Retrieved June 15, 2012, from
http://www.epa.gov/cleanenergy/energy-resources/refs.html
U.S. EPA. (2012c, March 2, 2012). Preferred/Recommended Models, from
http://www.epa.gov/scram001/dispersion_prefrec.htm
U.S. EPA. (2012d, March 15, 2012). Renewable Portfolio Standards Fact Sheet Retrieved June 15, 2012,
from http: //www. epa.gov/chp/state -policy/renewable_fs .html
U.S. EPA. (2012e, July 2012). Roadmap for Incorporating Energy Efficiency/Renewable Energy Policies
and Programs into State and Tribal Implementation Plans. (EPA-456/D-12-001j). Research
Triangle Park, NC: Retrieved from http://epa.gov/airquality/eere/manual.html.
U.S. EPA. (2012f, December 2012). U.S. EPA-default options for PM2 5 impact assessments: BenMap
configuration (.cfg) file. Office of Air Quality Planning and Standards, Risk and Benefits Group,
Research Triangle Park, NC: Accessible at
http ://www.epa.gov/airquality/benmap/models/pm_ria_final .cfg.
U.S. EPA. (2012g, December 2012). U.S. EPA-default options for PM2 5 impact assessments: BenMap
aggregation, pooling and valuation (.apv) file. Office of Air Quality Planning and Standards, Risk
and Benefits Group, Research Triangle Park, NC: Accessible at
http://www.epa.gov/airquality/benmap/models/pm_ria_final.apv.
Wittels, E. H., Hay, J. W., & Gotto, A. M., Jr. (1990). Medical costs of coronary artery disease in the
United States. Am J Cardiol, 65(1), 432-440.
Woodruff, T. J., Darrow, L. A., & Parker, J. D. (2008). Air pollution and postneonatal infant mortality in
the United States, 1999-2002 .Environ Health Perspect, 116(1), 110-115. doi: 10.1289/ehp. 10370
[doi]
Woodruff, T. J., Grillo, J., & Schoendorf, K. C. (1997). The relationship between selected causes of
postneonatal infant mortality and particulate air pollution in the United States. Environmental
Health Perspectives, 105(6), 608-612.
Woods & Poole Economics Inc. (2011). 2012 Complete U.S. Demographic Database Files. Washington,
DC.
Yamartino, R. J. (1985). Atmospheric Pollutant Deposition Modeling. In D. D. Houghton (Ed.),
Handbook of Applied Meteorology. New York: John Wiley & Sons.
Zanobetti, A., Franklin, M., & Schwartz, J. (2009). Fine particulate air pollution and its components in
association with cause-specific emergency admissions. Environmental Health 8, 58-60.
H-9
July 2013
-------
Zanobetti, A., & Schwartz, J. (2006). Air pollution and emergency admissions in Boston, MA. J
Epidemiol Community Health, 60(10), 890-895. doi: 60/10/890 [pii] 10.1136/jech.2005.039834
[doi]
H- 10
July 2013
------- |