The Effects of Air Pollution on Health: A Longitudinal Study of Los Angeles County Accounting for Measurement Error

TL;DR

This study employs a Bayesian hierarchical model to accurately assess the impact of various air pollutants on health outcomes in Los Angeles, accounting for measurement errors and variable selection.

Contribution

It introduces a flexible Bayesian framework that accounts for measurement error and selects relevant pollutants, improving the estimation of pollution-related health risks.

Findings

Significant associations between PM2.5, NO2, and health outcomes.

Measurement error correction enhances estimate precision.

Key pollutants identified without overfitting.

Abstract

This study develops a Bayesian hierarchical model to explore the effects of air pollution on respiratory and cardiovascular mortality in Los Angeles County. The model takes into account various pollutants such as PM2.5, PM10, CO, SO2, NO2 and O3, as well as a related meteorological factor: temperature. The objective is to identify the significant factors affecting selected health outcomes without including all variables in each model specification. This flexibility enables the model to capture key drivers of health risk without redundancy. To account for potential measurement error in pollution data due to imperfect monitoring or averaging, certain observed pollutant levels are treated as noise proxies for true exposure. By specifying priors for regression coefficients and measurement error parameters and estimating posterior distributions via Markov Chain Monte Carlo (MCMC) sampling, it leads to more precise and reliable estimates of the health risks associated with air pollution exposure in Los Angeles County by incorporating both the count nature of the health data and the uncertainties in pollution measurements.