ISEE/ISEA 2006 Conference Abstracts Supplement: Symposium Abstracts: Abstracts
Multisite time series studies of particulate matter (PM) and health have provided evidence that daily variation in air pollution levels is associated with daily variation in mortality and morbidity counts. Key areas of research concern: 1) assessing uncertainties in the adjustment for potential confounders and in the selection of the statistical model; 2) quantifying spatial and seasonal variation of PM10 health effects to understand toxicity of various PM components; and 3) developing a computational framework for reproducible research.
To address these questions, we: 1) explore statistical properties of semiparametric regressions to adjust for confounding bias; 2) extend the NMMAPS hierarchical semiparametric Poisson regression models to allow for seasonally and spatially varying PM effects; and 3) introduce a R software package (NMMAPS-R add web site), which implements the statistical methods described and serves as a platform for conducting systematic and reproducible research.
We apply methods to the National Morbidity, Mortality, and Air Pollution Study (NMMAPS), a national study covering 100 U.S. cities for the years 1987–2000 for PM10 and for the years 1999–2000 for PM2.5.
We found strong evidence that lag 1 exposure to PM10 continues to be associated with all-cause and cardiorespiratory mortality with the greatest effect in the eastern United States. Under the NMMAPS basic model, a 10-μg increase in PM10 at lag 1 is associated with a 0.19% and 0.24% increase in all-cause mortality and cardiorespiratory mortality (95% confidence interval [CI]: 0.10 to 0.28 and 0.13 to 0.36, respectively). In addition a 10-μg increase in PM2.5 at lag 1 is associated with 0.29% and 0.38% increases in all-cause mortality and cardiorespiratory mortality (95% CI: 0.01 to 0.57 and −0.01 to 0.82, respectively).
These findings provided key epidemiologic evidence for the EPA's review of the U.S. National Ambient Air Quality Standards for particulate matter.