A National Cohort Study of Effects of PM2.5 and Ozone at Low Concentrations on Mortality in the Continental United States

Qian Di,Francesca Dominici,Antonella Zanobetti,Joel Schwartz
DOI: https://doi.org/10.1289/isee.2016.4040
2016-01-01
Abstract:INTRODUCTION The effect of ozone on mortality in cohort studies is still relatively unexplored, and the effects of both pollutants at levels below the National Ambient Air Quality Standards (NAAQS) need further exploration. METHOD We constructed an open cohort of all Medicare beneficiaries (n=67,682,479) in the continental United States from 2000 to 2012. We considered all-cause mortality as outcome. Medicare records include individual covariates like age at entry, race, sex and whether eligible for Medicaid, a proxy for socioeconomic status (SES). To control for confounders, we collected data on covariates at zip code and county level. We used a hybrid model to predict daily concentrations of PM2.5 and ozone for the study domain with a spatial resolution of 1 km and computed annual averages for each zip code. RESULTS We fitted a cox model through categorized propensity score to estimate the effects. For 1 µg/m3 increase in PM2.5 and 1 ppb increase in ozone, we found evidences for adverse effects of both PM2.5 (HR = 1.0168 (1.0167, 1.0170)) and ozone (HR = 1.0018 (1.0018, 1.0019)). If constraining PM2.5 below 12 µg/m3 and ozone below 75th percentile, the adverse effects are still significant at low concentrations (HR = 1.0235 (1.0231, 1.0239) for PM2.5; HR= 1.0013 (1.0011, 1.0014) for ozone). Race, sex and SES are effect modifiers for both pollutants. Both dose-response curves are linear, without any signal of cut-off points at low concentrations. Conclusions This study provides solid evidence for adverse effects of both PM2.5 and ozone, especially at low concentrations, supporting further adjusting NAAQS to even lower levels.
What problem does this paper attempt to address?