Causal inference for Evaluating National or Regional Air Quality Interventions

Abstract

Introduction: Recent years have brought renewed scrutiny surrounding the effectiveness of specific regulatory programs for continuing to improve air quality and public health outcomes. Methods: We outline a modern perspective on causal inference to characterize the extent to which a designation of "nonattainment" with the 1997 National Ambient Air Quality Standards for particulate matter causally affected ambient pollution and health outcomes among over 10 million Medicare beneficiaries in the Eastern United States. Propensity scores were used to diagnose and mitigate the consequences of observed confounding, and principal stratification was used to evaluate whether causal health effects were heterogeneous across locations according to whether the intervention actually reduced ambient pollution. Analysis was anchored to publicly-available data on ambient pollution concentrations from 1999-2012, Medicare mortality and hospitalization outcomes, population demographics, weather, and regulatory indicators. Results: The results provide limited evidence that nonattainment designations causally reduced ambient pollution and health outcomes when averaged across all nonattainment locations. However, the principal stratification results suggest significant causal reductions in health burden among areas for which the nonattainment designation were estimated to have causally reduced ambient PM2.5. Conclusions: These findings provide targeted evidence of the effectiveness of nonattainment designations for the 1997 PM2.5 national ambient air quality standard that can be used in further refinement of air quality standards and enforcement procedures. The illustrated methods have relevance to the evaluation of a variety of national or regional air quality regulations.