Variable Selection for Disease Screening with and Without a Gold Standard Outcome

• Activity Number: 6
• Type: Invited
• Date/Time: Sunday, July 29, 2012 : 2:00 PM to 3:50 PM
• Sponsor: Social Statistics Section
• Author(s): Yuanjia Wang*+
• Companies: Columbia University
• Keywords: disease screening ; variable selection ; gold standard outcome ; support vector machine

Abstract

In many clinical settings, a commonly encountered problem is to assess accuracy of a screener for early detection of a disease or for identification of a patient population. In these applications, predictive performance of the test is of interest. Variable selection may be useful in assisting design of a clinical screener. An example is a research study conducted to design a new screening test of psychosis by selecting variables from an existing screener with a hierarchical structure among variables. We propose methods for variable selection with structured variables when predictive accuracy of a screener is the main focus. We maximize a direct summary measure of the predictive performance of the test, the area under a receiver operating characteristic curve, while controlling for overfitting. We discuss methods for both the observed gold-standard outcome variable setting and the unobserved outcome setting. We apply the methods to two studies: (1) designing a structured screener to be used in primary care clinics to refer potentially psychotic patients for further specialty diagnostics and treatment; (2) designing a criteria set to screen subjects affected with complicated grief.