Causal Propositions on Air Pollution and Human Health

Abstract

Fine particulate matter (PM2.5), ozone (O3), and nitrogen dioxide (NO2) are considered leading threats to human health, largely based on historical associations calculated from traditional statistical analyses. These studies did not assess causality, have primarily investigated long- and short-term exposures separately, and have used multiplicative models. In addition, less is known about the potential health disparities in the impact of air pollution.

In this dissertation, we attempted to address these gaps. We began by simultaneously assessing causal associations of long- and short-term exposures to PM2.5, O3, and NO2 with all-cause mortality on additive scale among Medicare beneficiaries in Massachusetts using three generalized propensity score (GPS) adjustment approaches. We found that long- and short-term exposures to PM2.5, O3, and NO2 were positively associated with increased risk of death, even at levels well below the current national air quality standards. The general consistency between the parametric and nonparametric methods suggests that there were not many interaction effects and high-order nonlinearities among confounders. Under the assumption that both GPS model and outcome regression model were correctly specified, the analysis addressed a critically important question of how many early deaths were caused by air pollution.

In Chapter II, we evaluated causal dose-response (D-R) relations between chronic exposures to PM2.5, O3, NO2 and all-cause mortality using national Medicare cohort during 2000–2016. We proposed a decile binning approach to relax the contentious assumptions of conventional inverse probability weighting. Assuming the causal framework was valid, we found that higher levels of annual-averaged PM2.5, O3, and NO2 were causally associated with greater risk of mortality and that PM2.5 posed the greatest risk. This study provides significant implications for reviewing the national air quality standards, as the causal D-R curves essentially infer the number of potential early deaths prevent if air pollutants were reduced to specific levels.

In the last chapter, we evaluated the associations between short-term exposures to ambient PM2.5, O3, and NO2 and the risk of asthma hospitalization among the Medicaid population for the continental US over the years 2000–2012. Specifically, we estimated the percent increase in the risk of hospital admission for asthma associated with a unit increase in PM2.5, O3, and NO2 over lag 0–6 days using a time-stratified case-crossover design with an extensive confounding control. We found that the adverse effect of PM2.5 on asthma is more acute with the response mostly seen at the same day, while the effects of O3 and NO2 were less acute with the response mostly seen at one to three days after the exposures. Further, our findings suggest that air pollution exposures had larger effects on asthma onset than exacerbations, and community-level factors that contributed to disproportionate burdens of asthma due to air pollution include access to health services and healthy food.

This dissertation provided more rigorous causal evidence on adverse effects of air pollution on human health. The proposed methodologies will benefit future scientific work.