Implications of different approaches for characterizing ambient air pollutant concentrations within the urban airshed for time-series studies and health benefits analyses

Background: In time-series studies of the health effects of urban air pollutants, decisions must be made about how to characterize pollutant levels within the airshed. Methods: Emergency department visits for pediatric asthma exacerbations were collected from Atlanta hospitals. Concentrations of carbon monoxide, nitrogen dioxide, ozone, sulfur dioxide, particulate matter less than 10 microns in diameter (PM10), particulate matter less than 2.5 microns in diameter (PM2.5), and the PM2.5 components elemental carbon, organic carbon, and sulfate were obtained from networks of ambient air quality monitors. For each pollutant we created three different daily metrics. For one metric we used the measurements from a centrally-located monitor; for the second we averaged measurements across the network of monitors; and for the third we estimated the population-weighted average concentration using an isotropic spatial model. Rate ratios for each of the metrics were estimated from time-series models. Results: For pollutants with relatively homogeneous spatial distributions we observed only small differences in the rate ratio across the three metrics. Conversely, for spatially heterogeneous pollutants we observed larger differences in the rate ratios. For a given pollutant, the strength of evidence for an association (i.e., chi-square statistics) tended to be similar across metrics. Conclusions: Given that the chi-square statistics were similar across the metrics, the differences in the rate ratios for the spatially heterogeneous pollutants may seem like a relatively small issue. However, these differences are important for health benefits analyses, where results from epidemiological studies on the health effects of pollutants (per unit change in concentration) are used to predict the health impacts of a reduction in pollutant concentrations. We discuss the relative merits of the different metrics as they pertain to time-series studies and health benefits analyses.