TIME-SERIES ANALYSIS OF OZONE AND MORTALITY IN 95 U.S. CITIES USING A BAYESIAN HIERARCHICAL MODEL

Abstract

ISEE-122 Introduction: Ozone has been associated with various adverse health endpoints, including hospital admissions and exacerbation of respiratory illness. Although numerous time-series studies have examined the relationship between short-term changes in ozone and mortality, many found an association while others were inconclusive or found no association. Resolution of ozone time-series studies is critical to the regulatory process, which relies heavily on epidemiological evidence. Methods: Using methodology based on approaches developed for the National Morbidity, Mortality and Air Pollution Study (NMMAPS), we investigated the short-term effects of changes in ozone levels on mortality for 95 large United States cities from 1987 to 2000. We used a Bayesian hierarchical model to generate the national estimate, based on independent city-specific estimates under various model structures, including constrained and unconstrained distributed lag models. Our multi-city approach has several strengths: avoidance of selection bias, uniform analytic approaches, large statistical power, and information gained from heterogeneity of effects across cities. We explored the sensitivity of findings to potential confounding by particulate matter; the control for seasonality, long-term trend, and temperature; and model specification. Results: A 10 ppb increase in the previous week’s ozone was associated a 0.52% (95% confidence interval 0.27 to 0.77%) increase in daily mortality. Key results were robust to adjustment for particulate matter and vice versa, indicating an absence of confounding. The association between ozone and mortality was not confounded by temperature or the degree of control for seasonality and long-term trends. Risk for those age > 75 years or for cardiovascular and respiratory mortality were slightly higher than for total mortality for all age groups. The estimated effect was relatively consistent across various model structures including constrained and unconstrained distributed lag models. Results indicate a statistically significant association between short-term changes in ozone concentrations and mortality. Discussion: This research is one of the most extensive analyses to date of ozone and mortality. The strengths of this study, such as the lack of selection bias, provide evidence of the relationship between short-term changes in ozone concentrations and mortality, based on a large number of cities broadly representing the United States. This work has implications for policy-makers by providing evidence of the association between O3 and mortality at the national level over a long timeframe and strengthens the rationale for regulatory control of O3 through limits on emissions of precursors.