Selection and evaluation of air pollution exposure indicators based on geographic areas

Geographic exposure indicators (GEIs) use point estimates of ambient air pollutant concentrations to characterize the exposure of populations residing within a specified area. Both zone- and proximity-type GEIs have been widely employed in epidemiological studies and other applications to identify regions or populations at high risk. Their use requires a number of assumptions, for example, pollutant concentrations should be homogeneous within the area, and concentrations should differ between areas in a predictable manner. These assumptions have not been rigorously examined. This paper evaluates the most common types of GEIs as surrogate measures of ambient air pollutant exposures. Statistical measures proposed to evaluate GEIs include accuracy, homogeneity, misclassification and statistical power. GEIs and statistical measures are evaluated in two case studies that use different air pollution sources and an air quality dispersion model. The case studies show that pollutant levels may vary substantially within a small area, and significant errors and exposure misclassification may result if the GEI represents a large geographic area. GEIs based on residential proximity to a pollution source should not be used for elevated emission sources, and the use of proximity measures is discouraged for ground level sources. A systematic evaluation is suggested to evaluate and improve the accuracy of the GEIs used in epidemiological and other applications.