A Hybrid Bayesian Hierarchical Model Combining Cohort and Case-Control Studies for Meta-Analysis of Diagnostic Test: Accounting for Disease Prevalence and Partial Verification Bias

• Activity Number: 236
• Type: Topic Contributed
• Date/Time: Monday, August 5, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Health Policy Statistics Section
• Author(s): Xiaoye Ma*+ and Haitao Chu and Yong Chen and Stephen R. Cole
• Companies: University of Minnesota and University of Minnesota School of Public Health and The University of Texas School of Public Health and The University of North Carolina at Chapel Hill
• Keywords: cohort and case-control study ; partial verification bias ; Bayesian method

Abstract

Bivariate random effects models have been recommended to jointly model sensitivities and specificities in meta-analysis of diagnostic tests accounting for between-study heterogeneity. Because the severity and definition of disease may differ across studies, the sensitivity and specificity of a diagnostic test may depend on disease prevalence. To account for the potential dependence, trivariate random effects models have been proposed. However, the proposed approach only includes cohort studies to estimate disease prevalence. In addition, some diagnostic accuracy studies select a subset of samples based on the test results to be verified by the reference test. It is known that ignoring unverified subjects can cause partial verification bias in the estimation of prevalence and test accuracy in a single study. However, the impact of this bias on the meta-analysis of diagnostic tests has not been investigated. As many diagnostic accuracy studies use case-control designs, we propose a hybrid Bayesian hierarchical model combining cohort and case-control studies to account for prevalence and to correct partial verification bias. We conduct simulation studies and present a case study.