A Simulation-Based Assessment of Alternative Explanations for Apparent Confounding in “PM Decomposition” Studies

28 May 2022
By Dr. Garrett Glasgow, Bharat Ramkrishnan, Dr. Anne E. Smith, et al.

In a recent article in Environmental Modeling & Assessment, NERA Associate Director Dr. Garrett Glasgow, Consultant Bharat Ramkrishnan, and Affiliated Consultant Dr. Anne E. Smith explore alternative explanations for apparent confounding in PM decomposition studies. “Confounding” is the presence of a third variable that affects both the dependent and independent variables in an analysis.

Generally, PM decomposition studies have found that, while particulate matter (PM2.5) and mortality are trending downward at the national level, areas with steeper declines in PM2.5 do not have correspondingly steep declines in mortality. This suggests recent declines in mortality are not directly caused by declines in PM2.5, and a confounding variable exists.

Alternative explanations have still proposed a causal link between PM2.5 and mortality. This study conducts simulation-based tests for four of the proposed alternative explanations: the omission of spatial variation in PM2.5 and mortality, confounding at the local level, measurement error, and an association between PM2.5 and mortality at a different time scale than those tested in previous PM decomposition studies. The authors find that none of the alternative explanations can reproduce the results in the PM decomposition studies while simultaneously allowing for a causal link between PM2.5 and mortality.

Glasgow, G., Ramkrishnan, B., Smith A.E., et al. A Simulation-Based Assessment of Alternative Explanations for Apparent Confounding in “PM Decomposition” Studies. Environmental Modeling & Assessment (2022).