Causally-Sufficient Dimensionality Reduction Methods for Assessing Joint Effects of Air Pollution Mixtures on Health Outcomes

ABSTRACT. Air pollution constitutes a multifaceted environmental challenge with far-reaching implications for public health and policy formulation. This research aims to deepen our understanding of the combined effects of air pollution mixtures on health outcomes. By quantifying the health impacts of pollution mixtures, we can improve risk as-

sessments, design more effective regulatory policies, and develop comprehensive strategies for monitoring and mitigating multiple pollutants. Air pollution presents distinctive analytical challenges due to its high-dimensional nature, as well as the presence of between-pollutant correlations due to common sources, transport, and atmo-

spheric chemistry. Conventional methodologies, which either isolate individual pollutant constituents or naively reduce dimensionality, often yield imprecise assessments of combined effects. These shortcomings are often due to oversights of confounding factors when performing dimension reduction, potentially leading to less ef-

fective policy decisions to protect public health. To address these challenges, this research proposes innovative dimension reduction techniques based on causal structural models. These techniques aim to preserve the causal relationships between air pollution mixtures and health outcomes while making minimal assumptions about data

distribution. Successful completion of these aims promises more accurate assessments of air pollution’s impact on health. We further intend to employ a diverse array of established techniques, alongside our own novel ap- proaches, on a comprehensive dataset of emergency department visits in Atlanta, spanning over two decades.

Specifically, building upon previous single-pollutant analyses, we will examine short-term health effects of differ- ent air pollution mixtures, including criteria pollutants, fine particulate species, and volatile organic compounds. This endeavor will serve to quantify the tangible health ramifications of air pollution within a real-world context.

Insights from this application will inform public health interventions, enhance air quality standards, and support evidence-based policy decisions. We also anticipate that the methodology can be applied to other environmental mixtures of concern beyond air pollution. In summary, this research bridges critical gaps in understanding the

health effects of air pollution by pioneering innovative dimension reduction techniques. By preserving causal relationships, it enhances the accuracy of assessments and offers more informed policy recommendations. With a team of leading experts in causal inference and air pollution epidemiology, this research is uniquely positioned

to improve community well-being through health protection.