Estimating the standardized ROC curve adjusting for covariate effects with

 
   Lead Investigator:    Danping Liu
   Institution      :    University of Washington
   E-Mail           :    dpliu@uw.edu
   Proposal ID      :    165


   Publications:


1  Liu D, Zhou A
   ROC Analysis in Biomarker Combination with Covariate Adjustment
   Journal Article  Acad Radiol,20:874-882,2013,PMC3682803

   PUBLISHED





   Proposal Description:
   Background: In the study of a diagnostic test, the receiver operating characteristic (ROC) curve is widely used to evaluate the capability of the test in distinguishing diseased from healthy subjects. When the test accuracy or the decision threshold is associated with patient characteristics, it is necessary to account for the covariate effects. The commonly used covariate-specific ROC curve only evaluates the diagnostic accuracy for particular subgroup of subjects, but lacks the population-level interpretation. We propose the standardized ROC curve as a population level accuracy measure, which takes the covariate distribution into consideration. The idea is, the marginal distribution of the test result in the healthy population is replaced by a standardized version, which is obtained by reweighting each health subject to match the covariate composition of the diseased population. Specific Aims:  1. To propose the new measure of the overall diagnostic accuracy. 2. To examine the accuracy of MMSE score in predicting (a) the transition from normal cognition to mild cognitive impairment (b) the development of Alzheimer???s disease.  3. To explore possible combinations of biomarkers and screening tests that improve the early detection of dementia and AD.