Using EPA's Environmental Benefits Mapping and Analysis Program
(BenMAP) for Global Health Impact Analysis

Susan Casper
2008 NNEMS Fellow
University of North Carolina at Chapel Hill
Department of Environmental Science & Engineering
scasper(a), email, unc. edu


-------
Susan Casper

2008 NNEMS Fellow

Using EPA's Environmental Benefits Mapping and Analysis Program
(BenMAP) for Global Health Impact Analysis

1.	Introduction

The Environmental Benefits Mapping and Analysis Program (BenMAP) is a tool for
performing customized health benefits analysis. BenMAP is used by the Environmental
Protection Agency (EPA) to estimate domestic health impacts of U.S. air quality
regulations. Researchers in countries around the world also use BenMAP by customizing
the data inputs to their country or region. As understanding of the global nature of air
quality and climate change increases, new applications have arisen for health impact
analysis on a global scale. However, global scale analyses also present new
considerations for methods and underlying uncertainties. This document describes the
data inputs necessary to run BenMAP on a global scale, using methods consistent with
the traditional U.S. setup. It will then discuss considerations unique to the global scale,
such as aggregation levels and valuation. Finally, major uncertainties associated with
global health impact analyses are highlighted.

2.	Data and Methods

Global BenMAP analyses use health impact functions to relate changes in air pollution
concentrations with health outcomes. Health impact functions take into account exposed
population, baseline incidence rates, pollutant concentrations, and concentration-response
factors identified by the epidemiology literature (Figure 1). The method used for
calculating health impacts using health impact functions is consistent with the U.S.
BenMAP setup and, therefore, will not be discussed in detail here. However, the data
inputs are specific to global analyses and are described below.

1


-------
Susan Casper

2008 NNEMS Fellow

2.1.	Concentration-Response Factors

Concentration-Response Factors (CRFs) relate a unit change in air pollutant
concentrations to a change in health endpoint incidence. They are identified by short- and
long-term epidemiology studies. BenMAP can be run with the user's choice of CRF
(pre-programmed or customized), though some CRFs may be better suited for use on a
global scale than others. In its current version, BenMAP applies each chosen CRF to the
entire geographical range included in the analysis.

Many time-series studies have associated ozone with mortality and can be used in global
health analysis. It is recommended that CRFs from Bell et al., 2004, are used, because
they are based on 95 large US communities and may not be subject to publication bias
like the single-city studies that are incorporated into meta-analyses. Unlike ozone, the
effects of PM2.5 are most pronounced after chronic exposure. Long-term PM2 5 health
effects are demonstrated by long-term cohort studies. While three PM25 cohort studies
have been conducted in the U.S. (Dockery et al., 1993; Pope et al., 2002; Laden et al.,
2006) and one Europe (Hoek et al., 2002), to date no cohort study has yet been conducted
in a developing country.

However, several studies of short-term ozone and PM2.5 effects in developing countries
have shown similar results to North American and European short-term studies (HEI
International Scientific Oversight Committee, 2004). Until more complete data for the
relationship between PM25 and mortality in developing countries exists, it can be
assumed that mortality relationships found in the developed world are valid globally. Of
the four PM2 5 cohort studies, Pope et al., 2002, may be the most generalizeable because
it had the largest sample size (approximately 500,000). Therefore, it may be the most
appropriate cohort study to use for global analyses.

2.2.	Baseline Incidence Rates

Baseline incidence rates at the finest resolution available should be used to calculate
health impacts. At the global scale, the finest resolution is likely to be at the country-
specific level. However, many countries lack the infrastructure to collect data on
morbidity, and often lack medical professionals to diagnose illness in a systematic way.
Until morbidity data is improved around the world, global health impact assessments
should focus on mortality only.

Mortality rates are available for many countries and are made public by the World Health
Organization (WHO) (http://www.who.int/healthinfo/morttables/en/). However, mortality
rates are not available for all countries. Furthermore, in some cases, available country-
specific mortality rates are over a decade old and obsolete in the fast-changing social and
economic environments of developing countries. While country-specific rates have many
data quality problems, the WHO also publishes estimated regional mortality rates for 14
world regions (http://www.who.int/choice/demography/by_country/en/index.html). The
latest year available for cause-specific and age-specific mortality rates is 2002 (World
Health Report 2004, Annex 2). Since some epidemiology studies include only adults of a

2


-------
Susan Casper

2008 NNEMS Fellow

certain age, baseline mortality rates for the relevant population only must be used. Age-
specific baseline mortality rates for countries and regions can be found on the WHO
website.

Differences in total mortality rates experienced by countries and regions also present
some difficulty in estimating global health impacts. As explained in section 2.1, no cohort
studies to date relate PM2.5 concentrations and mortality in developing countries.
Therefore, cohort studies conducted in the U.S. or in Europe must be extrapolated across
the global population, including populations in developing countries. Since baseline
incidence rates and causes of total mortality vary widely between developed and
developing countries, it may be inappropriate to apply U.S. and European total mortality
CRFs to the developing world. Instead, it is more appropriate to assess cause-specific
mortality, such as cardiopulmonary and lung cancer mortality, because they are not
affected by differences in baseline mortality rates.

2.3.	Pollutant Concentrations

Pollutant concentrations may be provided by global atmospheric chemistry and transport
models or by global air quality monitor networks. Any global model, such as the Model
of Ozone and Related Tracers (MOZART) and the Goddard Earth Observing System
chemical model (GEOS-CHEM), can be used to supply pollutant concentrations. The
user must create a new grid definition for each global model used in the analysis.

Similarly, any global monitoring network may be used in BenMAP, provided it has
adequate sampling frequency and meets quality assurance criteria. Some monitoring
networks available are:

•	Clean Air Status and Trends Network (CASTNET)

•	European Monitoring and Evaluation Programme (EMEP)

•	Acid Deposition Monitoring Network in East Asia (EANET)

•	Climate Monitoring and Diagnostics Laboratory (CMDL)

•	Interagency Monitoring of Protected Visual Environments (IMPROVE)

•	Aerosol Robotic Network (AERONET)

2.4.	Population

Spatially-distributed population can be provided by the Landscan database operated by
Oak Ridge National Laboratory or by the Center for International Earth Science
Information Network (CIESIN), operated by the Earth Institute at Columbia University.
These programs apportion population to very fine grid cells (30" by 30") based on
likelihood coefficients, such as proximity to roads, slope, land cover, nighttime lights,
and other information. Fine resolution population data is very memory intensive and
slows down BenMAP considerably. Since global health impact analyses are necessarily
at a coarse resolution due to the coarseness in global modeling data and baseline
incidence rates, such fine scale resolution for population is an unnecessary burden on
computing memory and speed. Instead, fine resolution population data should be summed

3


-------
Susan Casper

2008 NNEMS Fellow

to the larger grid resolution of the global model used for analysis before loading the data
into BenMAP.

As previously mentioned, some epidemiology studies include adults of a certain age only.
The exposed population, therefore, should be limited to the age group included in the
epidemiology study selected to provide the CRFs for the analysis. Since Landscan and
CIESIN do not provide the age structure of population in each grid cell, the fraction of
the population within each age group for each of the 14 world regions should be
converted to grid cells at the air quality model resolution. These fractions can then
multiplied by population in each grid cell (outside of BenMAP) to obtain the correct
exposed population in each grid cell. Age-specific population is available from the WHO
for 2002 (World Health Organization, 2004).

3.	Aggregati on Level s

Global health impact analysis results may be aggregated to the country level or the
continent level. However, country level aggregation may be inappropriate given the
coarse grid resolution necessitated by the global air quality model. Results may also be
aggregated to the 14 WHO regions. Countries included in each region can be found on
the WHO website (http://www.who.int/choice/demography/bv country/en/index.htmD.
Each aggregation level must be added to BenMAP as a grid definition.

4.	Valuation

EPA often uses the Value of a Statistical Life (VSL) and other economic indicators to
quantify the economic benefits of U.S. air quality regulations. Monetary quantification
supports cost-benefit analysis techniques when deciding whether a regulation meets a
cost-benefit test. Similar analyses using global air quality models and country-specific
economic data could provide useful data on which the worldwide community can base
decisions. However, uncertainties and ethical concerns surround attempts to apply
valuation statistics to global health impact analyses. For example, inequity in salaries
between developed and developing countries may result in vastly different VSLs around
the world, causing mortalities in developing countries to be valued lower than the same
number of mortalities in developed countries. Users should consider such issues when
deciding whether to run the valuation step in BenMAP or whether to calculate health
impacts only.

5.	Sources of Uncertainty

Many uncertainties are associated with health impact analysis. Uncertainties are
associated with impact functions, pollutant concentrations, mortality risk, and baseline
incidence rates. These uncertainties are common for local, regional, and global health
impact assessments. Since most uncertainties have previously been described in detail
(PM RIA), only those that influence global health impact analysis differentially are
discussed here and summarized in Table 1.

4


-------
Susan Casper

2008 NNEMS Fellow

5.1. Scale

A major source of uncertainty in global health impact analysis is scale. Due to computing
power constraints, global air quality models are very coarse in grid size, making it
difficult to estimate actual exposure levels. Peaks in ozone and PM2.5 concentrations that
usually occur in urban areas are most likely subdued, due to averaging across the large
grid cell. If captured by the model, large urban populations would be exposed to these
concentration peaks, resulting in high mortalities. However, since the peak concentrations
are diluted within each grid cell in global air quality models, these high urban mortalities
may not be captured.

Table 1. Primary Sources of Uncertainty Affecting Global Benefits Analysis Differentially

Uncertainty

Description

Possible Effect on Health Outcome

Scale

Global air quality models
have coarse grid size,
causing urban peaks to be
diluted

Health outcome underestimated
because urban peak concentrations
and populations not captured

Extrapolation of
Health Impact
Function Across the
World

Cohort studies on PM
mortality have included
only U.S. and European
populations and must be
assumed to apply to
populations across the
world

Health outcome could be
overestimated because median age
in developing countries is much
lower than that in developed
countries, and the elderly are more
susceptible to air pollution
mortality (Schwartz, 2008). Health
outcome could also be
underestimated because populations
in developing countries exposed to
greater susceptibility factors, such
as indoor air pollution and disease.

Use of

Concentration
Response Factors for
Total PM

No PM cohort study to date
has examined the effects of
specific PM components
on mortality. Though some
PM components may have
more deleterious health
effects than others, relative
risks are for total PM and
must be applied to the total
PM concentration.

Direction of effect on health
outcome is unknown, but the
magnitude could be great due to
significant variability in PM
composition around the world and
between urban and rural regions.

5.2. Extrapolation of Health Impact Functions Across the World

Extrapolation of health impact functions across populations is another source of
uncertainty. CRFs found in the U.S. and Europe must be assumed to apply to all other
parts of the world. While some evidence exists to validate this assumption, no long-term

5


-------
Susan Casper

2008 NNEMS Fellow

cohort epidemiology study of PM2.5 and mortality has been conducted in developing
countries. Differences in health status and medical care between developing and
developed countries may have substantial consequences for air pollution-related
mortality. For instance, the differential exposure of populations in developing countries
to indoor air pollution, mainly due to wood-burning stoves and poor ventilation, may
influence the susceptibility of these populations to diseases caused by outdoor air
pollution. Similarly, many populations in developing countries lack access to adequate
medical care, which may result in a higher rate of mortality from any particular disease.
Differences in lifestyle and age structure are also likely to affect population-level health
response to air pollution. For these reasons, less uncertainty is associated with estimating
cause-specific mortality than with estimating total mortality. Therefore, cause-specific
mortality should be calculated when possible.

5.3. Use of Concentration Response Factors (CRFs) for Total PM

Lack of CRFs for PM2.5 components also influences uncertainty. To date, no
epidemiology study has assessed mortality related to specific PM2.5 components.
However, PM2.5 composition is drastically different throughout the world, depending
largely on emissions. While particle size is a critical factor for toxicity, composition is
also important. Though significant variability in PM2.5 composition exists between
countries and between rural and urban regions, all PM2.5 components must be assumed to
have the same CRFs until more data becomes available.

6. Future Directions

BenMAP is a useful tool for performing global health impact analyses with many
potential applications. However, in the future, steps should be taken to reduce
uncertainty, which may limit the strength of the results.

For example, the coarse resolution of global health impact analyses introduces significant
uncertainty in actual exposures. To diminish the impact of such coarse resolution, data
from global monitoring networks can be employed in areas where monitors are located.
BenMAP's Model and Monitor Relative function allows users to anchor model results to
actual observations and to calculate concentrations on a smaller grid sale through
statistical methods. Another method for improving scale is to use finer scale regional
models for regions with high quality data, such as the U.S., Europe, and Japan and coarse
resolution global models for the rest of the world.

Extrapolation of CRFs discovered in one country to the rest of the world also presents
uncertainty. As new epidemiological data for the developing world becomes available, a
method for applying unique CRFs to different areas around the world should be
developed. In this way, CRFs from developed countries would be applied only to
populations in developed countries, and the same would be true for the developing world.

Similarly, as epidemiological studies associating PM2.5 components with mortality
become available, these data should be applied to global health impact analyses. For

6


-------
Susan Casper	2008 NNEMS Fellow

example, to examine the global health impact of sulfates, a CRF specific to sulfates
should be applied, rather than the CRF for total PM2.5. Applying component-specific
CRFs would lessen the impact of uncertainty due to differential PM2.5 composition
around the world.

7


-------
Susan Casper

2008 NNEMS Fellow

References

Bell, M.L., A. McDermott, S.L. Zeger, J.M. Samet, and F. Dominici, 2004. Ozone and
short-term mortality in 95 US urban communities, 1987-2000. JAMA, 292(19), 2372-
2378.

Dockery, D.W., III, C.A. Pope, X. Xu, J.D. Spengler, J.H. Ware, M.E. Fay, B.G. Ferris,
Jr., F.E. Speizer, 1993. An association between air pollution and mortality in six U.S.
cities. New Engl. J. Med., 329, 1753-1759.

Health Effects Institute International Scientific Oversight Committee, 2004. Health
Effects of Outdoor Air Pollution in Developing Countries of Asia: A Literature Review.
Special Report 15. Health Effects Institute, Boston, MA.

Hoek, G., B. Brunekreef, S. Goldbohn, P. Fischer, P.A. van den Brandt, 2002.
Association between mortality and indicators of traffic-related air pollution in the
Netherlands: A cohort study. Lancet, 360, 1203-1209.

Laden, F., J. Schwartz, F.E. Speizer, D.W. Dockery, 2006. Reduction in fine particulate
air pollution and mortality. Am. J. Respir. Crit. Care Med., 173, 667-672.

Pope, C.A., R.T. Burnett, M.J. Thun, E.E. Calle, D. Krewski, K. Ito, G.D. Thurston,
2002. Lung cancer, cardiopulmonary mortality, and long-term exposure to fine
particulate air pollution. JAMA, 287(9) 1132-1141.

World Health Organization, 2004. The World Health Report 2004: Changing History.
World Health Organization, Geneva, Switzerland.

8


-------