Stratified mortality risk estimation across the ozone distribution: interactions with PM 2.5 , NO 2 , and population characteristics

Studies examining mortality risk related to long-term ozone (O3) exposure, and the associated concentration response function, have yielded inconsistent results. In addition to different exposure metrics and averaging periods across studies, co-pollutant interactions and population characteristics varying with O3 exposure level increase the complexity of estimating independent O3 effects. We conducted a case-control study of Medicare beneficiaries, matching on O3 exposure decile, to examine all-cause mortality risk along the entire U.S. annual average O3 distribution during 2015–2016. We calculated odds ratios stratified by decile and examined demographic and co-pollutant characteristics within these deciles. Correlations between O3, fine particulate matter (PM2.5) and nitrogen dioxide (NO2) were highly variable within deciles and across O3 exposure metrics. Beneficiary- and zip code-level characteristics, including socioeconomic status (SES) and race/ethnicity, varied across O3 and co-pollutant exposure combinations. Risk estimates for O3 were generally null, but sensitive to stratification by demographic characteristics and co-pollutant concentrations including Medicaid-eligibility, age, sex, urbanicity, and PM2.5 and NO2 exposure levels. PM2.5 was consistently associated with mortality within most O3 exposure deciles. Whether our results reflect synergistic effects between O3 and PM2.5 and/or residual confounding of SES or co-pollutants, they may explain, at least in part, variability observed in previous studies.