Multivariate Left-Censored Bayesian Model for Predicting Exposure Using Multiple Chemical Predictors During the Deepwater Horizon Oil Spill Clean-Up

• Activity Number: 268
• Type: Contributed
• Date/Time: Monday, August 1, 2016 : 2:00 PM to 2:45 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Caroline Groth* and Sudipto Banerjee and Gurumurthy Ramachandran and Mark R. Stenzel and Dale P. Sandler and Aaron Blair and Lawrence S. Engel and Richard R. Kwok and Patricia P. Stewart
• Companies: University of Minnesota and University of California at Los Angeles and University of Minnesota and Exposure Assessment Applications and National Institute of Environmental Health Sciences and National Cancer Institute and The University of North Carolina at Chapel Hill and National Institute of Environmental Health Sciences and Stewart Exposure Assessments
• Keywords: Bayesian ; left-censoring ; Exposure assessment ; Deepwater Horizon oil spill ; linear models ; limits of detection

Abstract

On April 20, 2010 the Deepwater Horizon oil rig caught fire, exploded, and sank, sending approximately 5 million barrels of oil into the Gulf of Mexico over the ensuing 3 months. Thousands of workers were involved in the response and clean-up efforts. Many harmful chemicals were released into the air from crude oil including benzene, toluene ethylbenzene, xylene, and hexane. NIEHS's GuLF STUDY investigators are working to quantify the exposure the workers experienced related to the event and evaluate associations between the exposure and detrimental health outcomes. Approximately 150,000 personal exposure measurements were collected but a high percentage of the measurements were below the analytical methods' limit of detection and denoted as censored. In this presentation, we propose a model where one chemical is estimated by the other chemicals in a Bayesian linear regression setting, with each chemical having its own limits of detection. This multivariate extension of a simple linear Bayesian framework accounting for censoring in both X and Y, should allow for even stronger predictions and unbiased estimates of exposure for different groups of workers.