Bayesian Hierarchical Models for Racial and Socioeconomic Predictors of Mortality in a Sample of the U.S. Medicare Population

• Activity Number: 199
• Type: Contributed
• Date/Time: Monday, August 7, 2006 : 2:00 PM to 3:50 PM
• Sponsor: Section on Statistics in Epidemiology
• Author(s): Yijie Zhou*+ and Francesca Dominici and Thomas A. Louis
• Companies: Johns Hopkins University and Johns Hopkins University and Johns Hopkins University
• Keywords: race ; socioeconomic status ; mortality ; multi-level model ; multiple imputation

Abstract

In this project, we investigate the association between individual race and risk of death, as well as whether this association is affected by both individual-level and zip code-level socioeconomic status (SES). We assemble a data set that links at individual level and zip code level three government databases: Medicare, US census and Medicare Current Beneficiary Survey (MCBS), for 2095 zip codes in the Northeast region of the United States. We develop and apply multi-level statistical models to this data set, and estimate population-level attributable risk, relative attributable risk and odds ratio of death comparing the black versus the white population. In addition, we develop multiple imputation methods for combining information across the Medicare, U.S. census and MCBS databases to impute missing individual-level SES data.