Introduction
The Clean Air Act requires the US Environmental Protection Agency (EPA) to set National Ambient Air Quality Standards (NAAQS). In 1971, EPA set the first NAAQS to best protect public health and welfare. NAAQS are periodically reviewed/revised based on scientific evidence, resulting in a steady decrease in fine particle (PM2.5; particles with diameter ≤2.5 μm) concentrations. The association between long-term exposure to PM2.5 and mortality is well documented (1–4). Recent studies have examined effect estimates among people who are exposed to PM2.5 concentrations below the current NAAQS (3, 5). These studies found that exposure to PM2.5 below the US standard is associated with an increased mortality risk.
Some scientists, including the current chair of EPA’s Clean Air Scientific Advisory Committee (CASAC), have argued against including studies that use traditional statistical approaches to inform revisions of the NAAQS, and propose focusing on studies that apply causal inference approaches (6). Their main criticism is that traditional approaches that include potential confounders as covariates in the regression model do not inform causality. Recently, Goldman and Dominici argued that although “causal inference approaches tend to be more robust to violation of assumptions, […] air pollution regulations must be based on existing evidence and demonstrated inference methods that arise from review of existing literature” (7).
Our study bridges this divide. Using the largest air pollution study cohort to date (nationwide Medicare enrollees 2000–2016), we provide strong evidence of the causal link between long-term PM2.5 exposure and mortality under a set of assumptions necessary for causal inference (described in detail in the Materials and Methods section). We applied traditional and causal inference approaches to the same data, assessed sensitivity to modeling assumptions and study period, and made statistical software available.