Association of particulate matter components with daily mortality and morbidity in urban populations

Indices of atmospheric particulate matter (PM) have been reported to be associated with daily mortality and morbidity in a large number of recent time-series studies. However, the question remains as to which components of PM are responsible for the reported associations. Multiple PM components rarely are measured simultaneously. To investigate PM effects on mortality and morbidity, we used the multiple PM components measured in Windsor, Ontario, at a site only a few miles from downtown Detroit, Michigan. This study focused primarily on two study periods in which multiple PM components were measured in Windsor: 1985 to 1990, when levels of total suspended particles (TSP), sulfate from TSP (TSP-SO4(2-)), PM less than 10 microns in diameter (PM10), and nonthoracic TSP (TSP-PM10) were measured throughout the year; and 1992 to 1994, when data on PM10, PM2.5 (PM less than 2.5 microns in diameter), PM10-2.5 (PM10 minus PM2.5), particle acidity (H+), and artifact-free sulfates (SO4(2-)) were available for mostly summer months. Mortality data were analyzed for the 1985 to 1990 study period, and data on both mortality and hospital admissions of elderly patients were analyzed for the 1992 through 1994 period. Poisson regressions were used to estimate the effects of these PM components and gaseous criteria pollutants on mortality (nonaccidental, circulatory, respiratory, and nonaccidental without circulatory and respiratory) and on hospital admissions of elderly patients (for pneumonia, chronic obstructive pulmonary disease [COPD], ischemic heart disease, dysrhythmias, heart failure, and stroke), adjusting for temperature and humidity, trends and seasonal cycles, and day of the week. Both PM10 and TSP were associated significantly with respiratory mortality for the 1985 to 1990 period, with similar relative risk (RR) estimates for PM10 (RR = 1.123; 95% confidence interval [CI] 1.0361-1.218) and TSP (RR = 1.109; 95% CI 1.028-1.197), per 5th to 95th percentile increment. The effect-size estimates for TSP-SO4(2-) and TSP-PM10 were smaller and less significant. In two-pollutant models, simultaneous inclusion of gaseous pollutants with PM10 or TSP reduced PM coefficients by 0 to 34%. The effect-size estimates for total mortality, circulatory mortality, and total minus circulatory and respiratory mortality were less than those for respiratory mortality. Ozone (O3) and nitrogen dioxide (NO2) also were associated significantly with total and circulatory mortality, but a simultaneous consideration of these pollutants with PM10 reduced PM10 coefficients only slightly, or even increased them. In these results, pollution coefficients often were positive at multiple lag days (0-day through 3-day lags were examined), but for PM indices, 1-day lag coefficients were most significant. However, when all combinations of multiple-day average exposures were examined, for cases in which multiple lag days were positive, the choice of single-day or multiple-day average exposure did not appreciably change the estimated effect sizes. An examination of temporal correlation showed that the order of spatial uniformity as expressed by the median site-to-site correlation was O3 (0.83), PM10 (0.78), TSP (0.71), NO2 (0.70), carbon monoxide (CO) (0.50), and sulfur dioxide (SO2) (0.49), which suggests less exposure error for O3 and PM10 than for the other measured pollutants. Thus, these results suggest that spatially homogeneous pollution indices show higher associations with measured health outcomes.