DAILY MORTALITY ANALYSIS BY USING DIFFERENT REGRESSION-MODELS IN PHILADELPHIA COUNTY, 1973-1990

Earlier investigators found that in the city of Philadelphia there exi sts an association between daily levels of particulate matter and mort ality (Schwartz and Dockery, 1992a). To more fully explore this result , we analyzed the daily mortality data in Philadelphia using a host of independent variables including air quality indicators such as total suspended particulates (TSP), SO2, and O-3, and weather factors such a s temperature, humidity, barometric pressure, and precipitation for th e years 1973-1990. A wide variety of appropriate statistical models wa s applied to the data, including Poisson, multiple, and autoregression models. Our major findings on the association between mortality and d aily pollution levels are that (1) the estimates of this association a re related to the selection of the type of statistical model and the v ariables included in the model, and (2) this relationship differs acro ss age groups, causes of death, and seasons of the year. These results are not consistent with a unique point estimate of the association be tween air quality and mortality.