Combining Information from Multiple Surveys To Improve on Analyses of Self-reported Data in Estimating Measures of Health Disparities

• Activity Number: 74
• Type: Contributed
• Date/Time: Sunday, August 7, 2005 : 8:00 PM to 9:50 PM
• Sponsor: Section on Government Statistics
• Author(s): Nathaniel Schenker*+ and Trivellore Raghunathan
• Companies: National Center for Health Statistics and University of Michigan
• Keywords: Bayesian ; Measurment Error ; Missing Data ; Public Health

Abstract

Despite advances that have improved the health of the United States population, disparities in health remain among various racial/ethnic and socioeconomic groups. Common data sources for measuring the prevalence of health conditions include large-scale surveys that often pose questions requiring a self-report, such as "Has a doctor or other health professional ever told you that you have diabetes?" Answers to such questions might not reflect the true prevalence of a condition (for example, if a respondent does not have access to a doctor or other health professional), and the resulting measurement error could affect inferences about measures of health disparities. In this paper, we first develop measurement error models using the National Health and Nutrition Examination Survey, which asks self-report questions during an interview and obtains clinical measures based on a physical examination. We then develop methods, using a Bayesian framework, for improving the analysis of self-reported data alone to draw inferences about measures of health disparities. The methods are applied to the National Health Interview Survey in examples involving diabetes and hypertension.