Nonparametric Inferences on a Novel Measure of Diagnostic Accuracy for Diagnosing Incipient Diseases

• Activity Number: 561
• Type: Contributed
• Date/Time: Wednesday, August 6, 2014 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Chengjie Xiong*+ and Jingqin Luo
• Companies: Washington University in St. Louis and Washington University in St. Louis
• Keywords: Estimating Equation ; diagnostic accuracy ; diagnostic threshold

Abstract

For practicing clinicians who examine subjects on a regular basis for their earliest possible signs of chronic diseases such as Alzheimer's (AD), it is crucial to have a biomarker with optimum diagnostic accuracy and a well established clinical threshold to diagnose incipient disease. An utmost consideration for making such diagnosis by clinicians is the intra-subject change over consecutive assessments, which makes the routine analysis using Receiver Operating Characteristic Curve inappropriate as the latter requires a cross sectional case-control design. We propose a novel measure of diagnostic accuracy for diagnosing incipient disease under a matched pair design when subjects make transition from normal to incipient disease. Unlike the summary measure such as the area under the ROC curve, our proposed measure of diagnostic accuracy is directly linked to an optimum threshold for diagnostic purpose, allowing us to jointly estimate these two most important summary statistics through an estimating equation. We present asymptotical distribution of the estimates and provide the confidence interval estimates, both with and without adjustments for the covariates. We then implement the