A Sensitivity Analysis of Mortality/Pm-10 Associations in Los Angeles

Abstract

Several recent studies have reported statistically significant and quantitatively similar associations between daily mortality and PM-10 or total suspended particulates (TSP). These results have raised questions of biological plausibility, as well as questions regarding the statistical methods employed, which are relatively new and not widely understood. This article evaluates the sensitivity of mortality/PM-10 results to a range of statistical methods in a newly developed data set from Los Angeles county for the period 1985–1990. Data reported here include total daily deaths (excluding accidents and suicides), 24-h average PM-10, daily 1-h maximum 0₃ and carbon monoxide, maximum daily temperature, and mean daily relative humidity. Analyses were limited to the subset of days on which PM-10 data were available (every sixth day). Several alternative methods for addressing each of four issues were explored in this data set. These four issues were cyclic data variations, weather influences, other air pollutants, and the distribution of residuals. The associations between mortality and PM-10 concentrations, measured as relative risks associated with a 100-μml/m³ increase in PM-10, were only mildly sensitive to the alternative statistical methods. In particular, no difference was observed between the results of ordinary least squares and Poisson models. We observed a relative risk of about 1.05, which is similar to, but somewhat smaller than, the mortality/PM.-10 relative risks reported in recent studies. These new results add to the growing body of data suggesting that current levels of airborne particulate matter may contribute to excess deaths in the United States.