Optimizing Serial Diagnostic Testing to Reduce Cost

• Activity Number: 126
• Type: Contributed
• Date/Time: Monday, July 30, 2012 : 8:30 AM to 10:20 AM
• Sponsor: Section on Statistical Computing
• Author(s): Dat Huynh*+ and Ron Brookmeyer
• Companies: University of California at Los Angeles and University of California at Los Angeles
• Keywords: logic regression ; serial testing ; diagnostic ; sensitivity ; specificity

Abstract

The diagnosis of disease can be a costly and time-consuming procedure. Often a combination of tests is used to make an accurate diagnosis; each test can improve the accuracy but also incurs an additional cost. Here we consider a serial testing scheme that improves the accuracy of diagnosis while managing the costs of administering tests and the discomfort of unnecessary procedures. We present a method to identify an optimal combination of tests that provides maximal accuracy under a defined loss function to account for cost. Our approach uses logic regression to identify a classification rule that is optimal under the loss function. This classification rule is then fit into a serial testing algorithm for diagnostic testing. This method is applied to data from multiple HIV cohort studies. We compare this approach to logistic regression and linear discriminant analysis. We find that we can maintain an acceptable level of accuracy while achieving a very significant reduction in cost.