A Doubly Robust Estimator in Comparing Predictive Values for Diagnostic Tests Under Missing at Random

• Activity Number: 344
• Type: Contributed
• Date/Time: Tuesday, August 3, 2010 : 10:30 AM to 12:20 PM
• Sponsor: Biopharmaceutical Section
• Author(s): Yoonjin Cho*+
• Companies: GlaxoSmithKline
• Keywords: Diagnostic Test ; Covariates ; Doubly Robust ; Missing at Random

Abstract

Predictive Value which is the conditional probability of population with positive disease given the population of diagnostic test positive is used in this talk as the accuracy measure of diagnistic test. When more than one diagnostic test is available, comparing diagnostic tests give the confidence of the diagnostic test. It is common that Disease status is missing and some additional information related to patients are available. This information can give us more confidence on accuracy of diagnostics. In this talk, when additional information is available, the methodology accounting for the covariates in comparing predictive values is discussed under the assumption of Missing at Random of disease status. This estimator has the property of Doubly Robust which means that this estimator is robust under misspecification of one of two models: model of missing disease or covariate model.