Bayesian Multistate Models for Recurrent Episode Data of Illicit Drug Use

• Activity Number: 608
• Type: Contributed
• Date/Time: Wednesday, August 7, 2013 : 2:00 PM to 3:50 PM
• Sponsor: Section on Bayesian Statistical Science
• Author(s): Adam King*+ and Robert E Weiss
• Companies: University of California, Los Angeles and University of California, Los Angeles
• Keywords: Survival Analysis ; Competing Risks ; Multistate Models ; Transition Models ; Nonparametrics ; Discrete Time

Abstract

We propose a Bayesian Generalized Additive Mixed Model (GAMM) approach for discrete time episodic data of illicit drug use. We simultaneously model transition rates between three states: not-using the drug, using the drug, and incarcerated. Each transition rate may depend on multiple time variables (such as duration at risk for a transition, age, and calendar time), each of which is incorporated semiparametrically. In addition, transition rates may depend on the history of the drug use process up to the current time point. Methods are applied to retrospectively elicited lifetime histories of drug usage and related characteristics of 500 illegal drug users in Los Angeles.