VARIATIONS IN PM-10 CONCENTRATIONS WITHIN 2 METROPOLITAN-AREAS AND THEIR IMPLICATIONS FOR HEALTH-EFFECTS ANALYSES

Temporal variations of PM-10 levels at multiple sites between 1985 and 1990 in two major metropolitan areas, Cook County, IL, and Los Angeles County, CA, were characterized, and sensitivity of mortality-PM-10 associations to the choice of alternate sites was examined. In both cities, the correlation of PM-10 levels among multiple sites decreased as their distance increased. While averaging PM-10 concentrations over multiple sites generally improved the significance of PM-10-mortality associations, the highest PM-10-mortality association in Cook County was found for an individual site. In Cook County, the magnitude of the mortality association for the average of 6 PM-10 sites, as expressed as ''relative risk'' per 100 mu g/m(3) PM-10, was similar (RR = 1.06, 95% CI 1.01-1.10) to those reported in other PM-10 studies with similar model specifications. However, the significance of regression coefficients for individual PM-10 sites varied considerably it ratios range -0.62 to 3.30). Furthermore, every-6-days subsamples of the daily data at a site in Cook County showed a wide range in the significance of regression coefficients (t ratios range -0.17 to 3.44). This variability of significance among the six sites may be partly due to their small sample sizes (n approximate to 300), which raises concern regarding the potential for compromised statistical power of health effects analyses in ''short'' study periods (<6 yr) at the current every-6-days sampling frequency used for most PM-10 monitors in the United States. Also, the qualitative site information available, such as land use, location setting, and monitoring objective, did not show any coherent influence on the site's PM-10-mortality association's significance. Overall, it was found that the choice of PM-10 sites and sampling frequency can make a substantial difference in the calculated significance of such health effects time-series analysis.