An Approach to Assessing Health Risks from Particulate Matter in Two Cities. L. Deck, E. S. Post, and M. Wiener, Abt Associates Inc., Hampden Square, Suite 600, 4800 Montgomery La., Bethesda, MD 20814; and E. Smith and H. Richmond, U.S. Environmental Protection Agency, Office of Air Quality Standards and Planning, Mail Drop 15, Research Triangle Park, NC 27711
An assessment of the current health risks due to particulate matter (PM) and the reduction in health risks associated with achieving alternative PM standards was carried out for two sample locations, Philadelphia and Los Angeles, as part of EPA’s periodic reevaluation of PM standards, mandated by the Clean Air Act. Premature mortality and morbidity (hospital admissions and symptoms) incidence associated with fine particles (PM 2.5) were estimated in the two cities for existing conditions and attainment of potential fine particle standards. Substantial uncertainty surrounds any estimates in such a risk assessment because of the incompleteness of information upon which estimates are necessarily based. The concentration-response (C-R) relationship between ambient concentrations of particulate matter and the population incidence of a given health endpoint is the source of considerable uncertainty. This relationship is likely to vary, moreover, from one location to another as the components of the relationship vary. The composition of particulate matter varies geographically, and the nature of the exposed population (e.g., its age distribution, behavioral patterns and the extent of susceptible subgroups) is likely to vary as well. Regional differences in the composition of PM may result in regional differences in health impacts, even if the exposed populations have identical characteristics. Similarly, PM of identical composition may have differential impacts on exposed populations with significantly different characteristics. While a C-R function estimated in a given location captures the association of the specific PM and the specific exposed population in that location, such location-specific functions are often not available. In the absence of C-R relationships estimated in a sample location, risk was assessed based on C-R relationships estimated in different locations. Assuming a model of geographic variability in concentration-response functions for each health endpoint, the distribution of these functions was estimated for each endpoint from C-R functions reported for that endpoint in the epidemiological literature. In the absence of more location-specific information, the point estimate of PM-related risk for a given health endpoint in a sample location was based on the mean of this estimated distribution. The fifth and ninety-fifth percentiles of the distribution constitute a 90 percent credible interval, an estimate of the interval within which the concentration-response function in a randomly selected location will fall with 90 percent probability. This credible interval was used to characterize the uncertainty surrounding an estimate of risk in a specific location due not only to sampling error but to geographic variability as well.
Work funded in part through EPA Contract No. 68-W4-0029.