You are here

Designing optimal policies for reducing air pollution-related health inequities

Research Team: 
,

 Harvard University

Dr. Nethery’s study seeks to develop statistical methods for characterizing spatial and racial and ethnic variation in health effects associated with exposure to fine particulate matter (PM2.5) across the United States, and to design potential policies for reducing PM2.5-attributable health inequities.

Funded under
Status: 
Ongoing
Abstract

While the US EPA has committed to tailoring future regulatory policies to mitigate air pollution-related health inequities, existing statistical methods offer a limited toolkit for informing the design of national environmental justice-centered policies. Certain marginalized groups experience heightened exposure to PM2.5 and may also be more susceptible to its adverse health impacts as a result of social structural forces, such as high rates of outdoor work. Moreover, due to differing PM2.5 sources and climate patterns across the US, the health effects of PM2.5 may also vary over space. All of these factors must be simultaneously taken into account, alongside realistic policy constraints, in order to design optimal policies for mitigating pollution-related health inequities. Current methods are ill-equipped to rigorously characterize the heterogeneity in health impacts of pollutant exposures across space and groups, and are less capable still of leveraging information about effect heterogeneity and exposure distributions to inform optimal PM2.5 reduction policies to mitigate health inequities under resource constraints.


In the proposed work, we directly address these methodological and epidemiologic gaps by developing novel statistical methods to characterize spatial and racial/ethnic variation in PM2.5 health effects across the US and to design optimal policies for reducing PM2.5-attributable health inequities under resource constraints. Our methodological developments will integrate concepts from Bayesian causal inference, simulation-based approaches, and the constrained optimal treatment regimes and spatial optimization literatures. Taken together, they will establish a new sub-field of statistical methods for optimal environmental health policy design. To maximize the epidemiologic impact of our work, we will apply The methods to 2009-2018 nationwide Medicare claims and PM2.5 data. Our specific aims are:


Aim 1: Develop Bayesian causal inference methods to estimate area- and racial/ethnic group-specific causal exposure-response curves (ERC), building on techniques from the spatially-varying coefficient literature for information-sharing and Gaussian processes for confounding adjustment. We will apply them to Medicare data to estimate PM2.5 ERCs for each of the three largest racial/ethnic groups in the US (non-Hispanic white, Black, and Hispanic) for several different health outcomes.


Aim 2: Design and implement a Monte Carlo simulation approach to identify hypothetical PM2.5 reduction policies in the US that minimize racial/ethnic group-specific PM2.5-attributable health risks. This transparent and interpretable method will evaluate the health impacts of many carefully-designed, realistic hypothetical “policies”, i.e., geographic distributions of PM2.5 reductions, accounting for PM2.5
effect heterogeneity via the Aim 1 estimated ERCs, to identify the ones that minimize risks.


Aim 3: Develop and apply constrained optimization algorithms to identify hypothetical PM2.5 reduction policies that minimize group-specific PM2.5-attributable health risks. Emphasizing rigor, these approaches will utilize principled strategies to intelligently explore the space of hypothetical PM2.5 reduction policies under constraints with the aim of identifying globally optimal policies for minimizing risks.
The successful completion of these aims will transform the development of environmental justiceoriented air pollution policy. In particular, it will provide tools to communities and policymakers to design cost-effective interventions that maximize health benefits for marginalized groups. Moreover, the statistical methods developed here can be applied and expanded upon to inform optimal and equitable environmental health policy design more broadly. The proposed work directly addresses the accountability and environmental justice focus areas highlighted in the HEI’s 2020-2025 strategic plan.