2016 Progress Report: Health Effects of Air Pollution and Mitigation ScenariosEPA Grant Number: R835873C005
Subproject: this is subproject number 005 , established and managed by the Center Director under grant R835873
(EPA does not fund or establish subprojects; EPA awards and manages the overall grant for this center).
Center: Center for Air, Climate, and Energy Solutions
Center Director: Robinson, Allen
Title: Health Effects of Air Pollution and Mitigation Scenarios
Investigators: Burnett, Richard T , Brauer, Michael , Ezzati, Majid , Pope, Clive Arden
Institution: Health Canada - Ottawa , Brigham Young University , Imperial College , University of British Columbia
EPA Project Officer: Chung, Serena
Project Period: May 1, 2016 through April 30, 2021
Project Period Covered by this Report: May 1, 2016 through April 30,2017
RFA: Air, Climate And Energy (ACE) Centers: Science Supporting Solutions (2014) RFA Text | Recipients Lists
Research Category: Air , Climate Change
Project 5 specific aims include (1) estimate multi-pollutant mortality risk surfaces using two large, unique, population-based US datasets (1986-2004 National Health Interview Surveys linked to mortality to 2011, and 1982-2011 county level mortality data) and (2) explore regional and temporal variability in those risk surfaces. Health – especially mortality reduction (increased life expectancy) – is a large motivator for air, climate, and energy policy; however, the specific relationships between mortality (life expectancy) and several ambient air pollutants, either singly or jointly, have not been sufficiently characterized to be useful in burden assessments, especially in nationally representative studies. Little is known about how the association between air pollution and health varies regionally in the United States.
Project 5 is composed of two major separate studies: National Health Interview Survey Cohort and the County Level Mortality Space-Time Study.
National Health Interview Survey Cohort: In preparation for applying for and analyzing the full National Health Interview Survey (NHIS) data, we have conducted a preliminary/preparatory study that evaluated associations between long-term PM2.5 exposure and mortality risk using cohorts of the U.S. adult population constructed from public-use National Health Interview Survey (NHIS) data. Two cohorts consisting of 392,107 and 162,262 individuals (with and without individual smoking data) were compiled from public-use NHIS survey data (1986-2001) with mortality linkage through 2011. Cohorts were restricted to persons who lived in a metropolitan statistical area (MSA), were 18-84 years of age, and had individual risk factor information. Modeled PM2.5 exposures were assigned as MSA-level mean ambient concentration for 1999 through 2008 using previously modeled pollution data obtained from related studies using the American Cancer Society Cancer Prevention Study II (ACS CPS-II) cohort.
Mortality hazard ratios (HRs) were estimated using Cox proportional hazards regression models, controlling for age, race, sex, income, marital status, education, body mass index, and smoking status. Estimated HRs for all-cause and cardiovascular mortality, associated with a 10 µg/m3 exposure increment of PM2.5, were 1.06 (1.01-1.11) and 1.34 (1.21-1.48) respectively in models that controlled for various individual risk factors including smoking. This preliminary study demonstrates that the NHIS survey data with mortality linkage can be effectively used to evaluate mortality associations with air pollution. It also provides evidence that elevated risks of mortality, especially cardiovascular disease mortality, are associated with long-term exposure to PM2.5 air pollution in U.S. nation-wide adult cohorts constructed from public-use NHIS data. A manuscript reporting on this analysis using public-use data has been prepared and it is currently in review.
County-Level Mortality Space-Time Study: We have established a time consistent set of 3,082 counties from the contiguous United States in order to remove the effects of boundary changes, mergers and splits of counties over the study period.
In order to identify the statistical model that will be used as the basis for modelling the health effect of pollutants, a series of models have been run on two different platforms. We have carried out test analyses for all models using age- and county-specific death rates based on national mortality and population data. All models were formulated to capture levels and trends of death rates in relation to age group and over time and space. A balance was required to be struck between model complexity, model fit and run-time. The effects of these choices on model performance and run time have been collated.
We have designed an algorithm to merge counties below a designated population threshold. The merging of small counties into larger multi-county units is desirable for three reasons; firstly, for modelling reasons it is desirable to avoid the zero population age groups which do occur in smaller counties; secondly, in order to lessen the difference in size between the largest and smallest counties and thirdly in order to reduce the computational burden. In conjunction with the model testing, we evaluated a number of population thresholds. Setting the threshold so that counties are merged if either male or female populations are less than 25,000 at any time during the study period, results in a reduction from 3,082 to 1,195 counties. In the final models describing the health effects of pollutants, sensitivity to the merging of counties will be assessed by carrying out age separate analyses. By analyzing age groups separately the ability to share information across age groups is lost but the reduced computational burden allows the full set of counties to be used.
A series of univariate pollutant models have been identified including age-specific and non-linear components. Variability in the effects of pollutants over space and time will also be tested. However, because we expect lower variability in health effects of pollutants due to space and time, making inference for these variables will be more difficult and differences between age groups will be the primary focus. Potential confounders and data sources have been identified.
We plan to exclude the Alaskan and Hawaiian counties from the analyses due to the lack of contiguity with the main bulk of counties. Contiguity of borders is a condition of the conditional auto-regressive model which will be used, and relevant for how air pollution disperses. As described above some counties will be merged and therefore inference will be made for merged counties.
We are currently preparing a Research Data Center (RDC) research proposal/application that will allow analysis that uses restricted NHIS data, that is linked with CACES generated air pollution data, and that substantially extends the analysis. The key extensions of our proposed analysis will include: 1. Use of full data including sampled individuals residing both inside and outside of MSAs. 2. Use of survey data from 1986 through 2014 with mortality follow-up through 2015. 3. CACES-generated pollution estimates will be at the census block of residence allowing for much greater spatial resolution. 4. The CACES estimated air pollution variables will expand to include NO2, CO, and O3 and PM2.5 (with subspecies EC, OC, nitrate, sulfate, ammonium and “source-resolved” species including gasoline vehicles, diesel vehicles, coal fired electrical generating units, biomass burning, anthropogenic SOA, biogenic SOA). 5. A coherent framework of multiple pollutant models of increasing complexity will be developed. As additional pollutants, subspecies and source specific species become available, these will be implemented. 6. Appropriate confounders will be added to the model. 7. A presentation of preliminary results will be made at the International Society for Environmental Epidemiology (ISEE) in September 2017.
Supplemental Keywords:mortality risk functions, multi-pollutant, population health burden
Main Center Abstract and Reports:R835873 Center for Air, Climate, and Energy Solutions
Subprojects under this Center: (EPA does not fund or establish subprojects; EPA awards and manages the overall grant for this center).
R835873C001 Mechanistic Air Quality Impact Models for Assessment of Multiple Pollutants at High Spatial Resolution
R835873C002 Air Quality Observatory
R835873C003 Next Generation LUR Models: Development of Nationwide Modeling Tools for Exposure Assessment and Epidemiology
R835873C004 Air Pollutant Control Strategies in a Changing World
R835873C005 Health Effects of Air Pollution and Mitigation Scenarios