A Bayesian Framework for Estimating Disease Risk Due to Exposure to Uranium Mine and Mill Waste on the Navajo Nation

• Activity Number: 86
• Type: Contributed
• Date/Time: Sunday, August 3, 2014 : 4:00 PM to 5:50 PM
• Sponsor: Section on Statistics and the Environment
• Author(s): Lauren Hund*+ and Edward Bedrick and Curtis Miller and Gabriel Huerta and Johnnye Lewis
• Companies: University of New Mexico and University of Colorado and University of New Mexico and University of New Mexico and University of New Mexico
• Keywords: multivariate binary regression ; bayesian model averaging ; potential outcomes ; chronic disease ; environmental disease risk

Abstract

Abandoned sites from the uranium mining period are scattered across the Navajo Nation, resulting in exposure to environmental contaminants. Using the DiNEH survey of 1,304 Navajo residents, we examine the relationship between mining exposures and kidney disease, diabetes, and hypertension, adjusting for known contributors to disease. We use a Bayesian potential outcomes framework to estimate exposure effects, incorporating uncertainty in the confounding adjustment model using Bayesian model averaging (BMA). We fit a Bayesian multivariate-t model for correlated binary outcomes; this model results in marginal conditional log-odds ratio parameters and is computationally efficient. Conditioning on a relatively small known set of confounders, we select the functional form for regression adjustment using BMA, considering various interaction terms and smooth functions. We estimate conditional odds ratios; and multivariate sample average exposure effects using the counterfactual predictive distributions. In addition to the known risk factors, both historic mining era and current legacy waste exposures are associated with increased risk of the chronic diseases assessed in this study.