You are here

Robust statistical approaches to understanding the causal effect of air pollution mixtures

Principal Investigator: 

University of Florida

This project seeks to develop statistical methodology that allows for complex relationships between air pollution and health outcomes to be used to estimate causal effects of multivariate exposures. Additionally, the proposed methodology will allow for evaluation of separate subgroups in the population to identify the most vulnerable subgroups.

Funded under

The majority of existing epidemiological evidence of the health effects of air pollution is focused on the analysis of single pollutants at a time. Recently, emphasis has been placed on understanding the joint impact of a large number of exposures simultaneously as humans are routinely exposed to a number of distinct environmental exposures. Despite this, significant statistical and epidemiological gaps remain in the analysis of the air pollution mixture. There has been little to no research utilizing causal inference methods to assess the health effects of simultaneous exposure to multiple pollutants. Additionally, sensitivity analyses for the analysis of multiple exposures have not yet been developed, and can provide crucial information about the robustness of findings to unmeasured confounding bias. We address these, and other concerns, in the following specific aims:

Aim 1: Define policy-relevant causal estimands for air pollution mixtures and develop novel statistical approaches for estimation. By formulating the problem in a causal inference context, we will be explicit about the assumptions required for identification of the effects of interest. Additionally, we will develop new methods that improve the power of existing statistical approaches to identify important components of the air pollution mixture.

Aim 2: Characterize population heterogeneity in policy-relevant causal effects of the air pollution mixture. Understanding which communities are most impacted by changes in the air pollution mixture will lead to an improved understanding of the nature of the relationship between air pollution and health, as well as identify vulnerable subpopulations that can lead to more targeted regulatory policy.

Aim 3: Develop new methods to evaluate causal mixture effects in the presence of interference. New causal estimands will be developed to inform policy decisions with evidence for how certain mixture components may spill across space to affect health at nearby (or distant) locations. We will present simplifying assumptions for identification of interference effects, and develop estimation strategies for direct and spillover effects of air pollution.

Aim 4: Develop approaches to assess the sensitivity of the causal analysis of mixtures to untestable assumptions about unmeasured confounding variables. Identifying causal effects from observational data necessarily relies on untestable assumptions, and it is crucially important to develop sensitivity analyses to understand the degree of robustness of our findings to unmeasured confounding bias. Developing novel statistical approaches rooted in causal inference methodology addresses a number of key HEI initiatives such as improving the statistical methodology for the air pollution mixture and performing accountability research. Additionally, by applying our methods to nationally representative data sets, such as the Medicare cohort, we will improve the existing epidemiological evidence on the causal effects of air pollution, particularly at low levels.