Department of Biostatistics
University of Michigan
Regression Analysis for Probabilistic Cause-of-Disease Assignment
Using Case-Control Diagnostic Tests
This talk concerns the estimation of cause-specific case fractions (CSCFs) for a disease of multiple causes. The unobserved causes of disease are, however, often imperfectly measured, because the diagnostic tests lack sensitivity or specificity. In this talk, I will first introduce recent methodological advances to address the scientific need of incorporating control data to estimate the CSCFs. I will then introduce a general hierarchical Bayesian framework for regression analysis using such etiologic data for estimating CSCFs as functions of covariates. Data from controls provide requisite information about measurement specificities and covariations, which is used to correctly assign cause-specific probabilities for each case given her measurements. A regression analysis of data from an etiology study of childhood pneumonia reveals the dependence of pneumonia etiology upon season, age, disease severity and HIV status. I will conclude the talk with the model’s connection to a promising family of restricted latent class models.
Friday, September 13, 2019, 11:30 a.m. BLOC 113