Statistical Consulting: Exploring Bayesian Latent Class Models as a Potential Statistical Tool to Estimate Sensitivity and Specificity in Presence of an Imperfect or No Gold Standard (303253)

Keywords: Bayesian Latent Class Model, Imperfect Gold Standard

As a statistical collaborator at a medical center, you often encounter interesting projects whose analyses do not use main stream statistical methods. The focus of this presentation is to provide a brief overview of such a project from clinical microbiology. Assessment of a new assay or diagnostic test is generally performed using statistical measures such as sensitivity, specificity, negative predictive value, positive predictive value and area under the curve when an established gold standard exists. However, in some cases, the gold standard may be imperfect or may not exist. In such situations, Bayesian latent class models (BLCM) is proposed as a possible alternative. LCM does not assume any gold standard and a true disease state (present/absent) for each individual is also unknown. Bayesian methodology to LCM will be illustrated using a simple example of a real life data set from Clinical Microbiology research study. This approach is increasingly used to validate diagnostic tests with no established gold standard in the areas of infectious diseases and clinical microbiology research.