Learning Logic Rules Using an Iterative Algorithm, with an Application to Developing Criteria Sets for the Diagnostic and Statistical Manual of Mental Disorders (DSM)

• Activity Number: 351
• Type: Contributed
• Date/Time: Tuesday, August 5, 2014 : 10:30 AM to 12:20 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Christine Mauro*+ and Donglin Zeng and Katherine Shear and Yuanjia Wang
• Companies: Columbia University and University of North Carolina at Chapel Hill and Columbia University and Columbia University
• Keywords: Classification rules ; Diagnostic and Statistical Manual of Mental Disorders ; Statistical learning ; Support vector machine

Abstract

For psychiatric disorders, a clinically significant anatomical or physiological deviation cannot be used to determine disease status. Instead, clinicians rely on criteria sets from the Diagnostic and Statistical Manual of Mental Disorders (DSM) to make diagnoses. Each criteria set has several symptom domains, determined by expert opinion or psychometrics. In order to be diagnosed, an individual must meet the minimum number of symptoms required for each domain. We propose a novel approach to determine these minimum values. Given disease status and the number of symptoms present in each domain for a sample of individuals, we fit a linear discriminate function within each domain. Since one must meet the criteria for all domains, a positive diagnosis is only issued if the prediction in each domain is positive. The overall decision rule is the intersection of all the domain specific rules. We fit this model using the support vector machine and account for the unique logic structure of DSM criteria sets. This results in a non-convex minimization problem, which we approximate by an iterative algorithm. This algorithm shows good performance in simulations and the real data application.