Predicting the Relapse of Psychotic Depression: A Bayesian Joint Modeling Approach

• Activity Number: 66 - Novel Bayesian Methodology with Health Applications
• Type: Contributed
• Date/Time: Monday, August 3, 2020 : 10:00 AM to 2:00 PM
• Sponsor: Mental Health Statistics Section
• Author(s): Yiyuan Wu* and Wenna Xi and Samprit Banerjee
• Companies: Weill Cornell Medical College and Weill Medical College, Cornell University and Weill Medical College, Cornell University
• Keywords: Joint modeling; survival analysis; longitudinal data; Bayesian Modeling; psychotic depression

Abstract

Patients with psychotic depression often experience relapses, even after they have fully recovered from a major episode. Medication treatments such as a combination of antidepressant and antipsychotic can help achieve remission and are usually continued afterward to prevent relapse. However, little is known about how the continuation of treatment after remission affects patients’ chance of relapse and how patients’ clinical biomarkers, such as total cholesterol, glucose, and other metabolic measures, can help identify patients who are at high risk of relapse. To predict the prognosis of patients’ relapse, we develop a Bayesian joint model that accounts for the dependency between two types of outcomes:; time to relapse and longitudinal metabolic measures. We also propose a new time-dependent metric to evaluate the discriminative ability of the joint model. We evaluate our proposed method via simulation studies and analyzing the data from the STOPPD study.