IMPACT OF SMOOTHING METHODS ON ESTIMATED EFFECTS OF AIR POLLUTION ON HUMAN DAILY MORTALITY

Abstract

ISEE-458 Introduction/Aim: Using human mortality and atmospheric data from the Atlanta, GA region, we compare estimated effects of air pollution via Generalized Linear Models (GLM) and Generalized Additive Models (GAM). Methods: We compare the estimated effects using various spans (GAM) and knot placement and frequency (GLM), lag structures, and extent of missing data for all decedents, decedents at least 65 years of age and women/men decedents separately. We will discuss the apparent effect of nearly 20 air quality indicators, as well as metals and other constituents of PM 2.5, on human mortality, by categories of causes of death. We will present our findings of individual and groups of AQI, metals and other constituents of PM 2.5 in terms of their own inherent toxicity, their relationships with other pollutants and weather variables, and as markers for contributions from specific types of sources. We use weather and metal and other constituents of PM 2.5 data captured continuously and summarized on a daily basis from the Aerosol Research and Inhalation Study (ARIES), and daily mortality from the two major counties of Atlanta, GA, for the years August 1998 – July 2000. Having collected the data on the causes of death directly from the death certificates, we also compare the estimated effects based on the underlying cause(s) of death and the acute cause of death. Results: Results will be summarized and presented in tabular and graphical formats, using colour representation, to easily ascertain differences/similarities in the estimated effects. Discussion: New aspects of these findings will include differences by gender of the decedents, underlying versus acute causes of death, and the incorporation of metals into the mix of AQI. This work follows from a 2004 article in Inhalation Toxicology by the same four authors.