Air Pollution Epidemiology to Support Environmental Regulations and Policies

Abstract

Research shows that air pollutants such as particulate matter (PM) are associated with heart and lung disease and other adverse health effects. A primary way the US Environmental Protection Agency (EPA) reduces air pollutant concentrations is by setting National Ambient Air Quality Standards (NAAQS), which regulate the maximum pollutant levels allowed in ambient air. The EPA also establishes source-specific emissions standards and pollution-reduction policies. Currently, the NAAQS regulates PM2.5 (<2.5 μm) and PM10 particles (<10 μm). Importantly, however, PM10 includes PM2.5 particles. As such, the EPA would rather replace the PM10 standard with a standard specifically for PM10-2.5 (≤ 10 μm and >2.5). Yet, the EPA has not done so, deciding that there was inadequate information on the causal impacts of PM10-2.5 on health. The limited amount of PM10-2.5 health research is largely due to insufficient availability of exposure data. However, data from National Aeronautics and Space Administration (NASA) satellites creates opportunities to estimate PM10-2.5 levels for assessing PM10-2.5 health impacts. The first two aims of this dissertation were designed to inform the science used by the EPA to set a PM10-2.5 NAAQS. In Aim 1, we estimated PM10-2.5 exposure using highly resolved satellite data and advanced spatiotemporal statistical modeling in six US metropolitan areas. In Aim 2, we paired these PM10-2.5 estimates with the detailed health data in the Multi-Ethnic Study of Atherosclerosis (MESA) to estimate cross-sectional associations between long-term PM10-2.5 exposure and levels of inflammation and coagulation, which are subclinical markers of cardiovascular disease. Aim 3 was designed to inform the EPA in its setting of a source-specific emissions reduction program. We focused on a diesel school bus program since they transport more than 25 million children – a highly sensitive subpopulation – in the US each year. The EPA developed the Diesel Emissions Reduction Act (DERA) School Bus Rebate Program, which randomly awards funding to replace older, more polluting diesel school buses with cleaner buses. In Aim 3 I quantified the attendance impacts of this EPA program. In Aim 1, our final predictions captured the long-term spatial patterns of PM10-2.5 very well in four of our study areas, and well to modestly in the others. Additionally, it outperformed two alternative exposure prediction methods spatially in all six areas. In Aim 2, we found no evidence that long-term PM10-2.5 exposure was associated with greater levels of inflammation and coagulation in the MESA cohort. In fact, contrary to our hypothesis, we found greater inflammation and coagulation with lower PM10-2.5, although results were sensitive to adjustment for chronic health conditions. In Aim 3, we found suggestive evidence that receiving rebate funds to replace or retrofit older school buses with cleaner buses was associated with increases in school district attendance. The attendance effects were strongest when larger proportions of students were impacted and when the oldest, most polluting buses were replaced. This dissertation provides evidence that using satellite data in spatiotemporal prediction modeling can successfully generate long-term average spatially resolved PM10-2.5 estimates. It also highlights the strengths and limitations of using satellite-based predictions to study the health impacts of PM10-2.5 exposure. Additionally, it provides evidence that a source-specific emissions reduction program, such as the EPA’s DERA School Bus Rebate Program, can have measurable impacts based on the suggestive evidence we saw of the program’s role in improving school district attendance.