气候变化领域集成服务门户

Epidemiological analyses of air quality often estimate human exposure from ambient monitoring data, potentially leading to exposure misclassification and subsequent bias in estimated health risks. To investigate this, we conducted a case-crossover study of summertime ambient ozone and fine particulate matter (PM2.5) levels and daily respiratory hospitalizations in New York City during 2001-2005. Comparisons were made between associations estimated using two pollutant exposure metrics: observed concentrations and predicted exposures from the EPA's Stochastic Human Exposure and Dose Simulation (SHEDS) model. Small, positive associations between interquartile range mean ozone concentrations and hospitalizations were observed and were strongest for 0-day lags (hazard ratio (HR) = 1.013, 95% confidence interval (CI): 0.998, 1.029) and 3-day lags (HR = 1.006, 95% CI: 0.991, 1.021); applying mean predicted ozone exposures yielded similar results. PM2.5 was also associated with admissions, strongest at 2- and 4-day lags, with few differences between exposure metrics. Subgroup analyses support recognized sociodemographic differences in concentration-related hospitalization risk, whereas few inter-stratum variations were observed in relation to SHEDS exposures. Predicted exposures for these spatially homogenous pollutants were similar across sociodemographic strata, therefore SHEDS predictions coupled with the case-crossover design may have masked observable heterogeneity in risks. However, significant effect modification was found for subjects in the top exposure-to-concentration ratio tertiles, suggesting risks may increase as a consequence of infiltration or greater exposure to outdoor air.