Evaluating the relationship between socioeconomic disparity and potential PFAS contamination in the United States with machine learning: Implications for public health and environmental justice

Abstract

PFAS (Per-and Polyfluoroalkyl Substances), or synthetic “forever chemicals,” pose a dangerous threat to public health, contaminating food, water, and air. PFAS testing regulations vary widely across states, remaining inadequate and inconsistent. This study sought to investigate the relationship between the presence of PFAS-handling industries and socioeconomic factors to identify areas with the highest vulnerability to PFAS contamination and prioritize testing at those locations. A Random Forest classifier was developed for predicting proximal PFAS industry presence based on six socioeconomic features and uncovered non-linear and non-monotonic relationships using partial dependence plots and Shapley Additive Explanations analysis. With the notable exception of direct income level, it was found that more disadvantaged socioeconomic conditions generally yielded a higher likelihood of PFAS contamination in communities. More specifically, for most features, it was determined that individuals of generally middle to lower socioeconomic conditions, not the lowest, may be at the greatest risk of PFAS exposure, contrary to traditional expectations. Our results therefore reveal a necessity for greater nuance in the identification of particularly vulnerable communities for more effective prioritization of PFAS testing areas. This study hopes to quantitatively inform the implementation of consistent and targeted PFAS testing to advance public health across the United States.

Keywords: PFAS, socioeconomic, public health risk, variations, contamination, exposure, testings, eco-friendly