Detecting and Assessing Multiple Correlated Diagnostic Tests with Receiver Operating Characteristic Curves

• Activity Number: 32
• Type: Contributed
• Date/Time: Sunday, August 7, 2005 : 2:00 PM to 3:50 PM
• Sponsor: Biometrics Section
• Author(s): Feng Gao*+ and Chengjie Xiong and Yan Yan and Kai Yu and Zhengjun Zhang
• Companies: Washington University in St. Louis and Washington University in St. Louis and Washington University in St. Louis and Washington University in St. Louis and Washington University in St. Louis
• Keywords: receiver operating characteristic (ROC) curve ; optimal linear combination test ; specificity ; sensitivity

Abstract

Receiver operating characteristic (ROC) methodology is widely used to evaluate diagnostic tests. It is not uncommon in medical practice that multiple diagnostic tests are applied to the same study samples. Usually, the potential correlations among these tests will be ignored and a single ROC curve is constructed and evaluated for each test separately. The area under curve (AUC) is one of the most frequently used measures to evaluate the performance of a ROC curve. For many clinical practitioners, however, a more relevant question of interest may be "what the sensitivity would be for a given specificity (say, 90%) or what the specificity would be for a given sensitivity." Under the framework of a ROC curve, we present a method to estimate the maximal sensitivity for a given specificity based on an optimal linear combination of the multiple diagnostic tests while assuming a multivariate normal distribution. We also propose an algorithm to identify an optimum subset of diagnostic tests by searching over all possible linear combinations.