Predicting Recovery from Substance Abuse Using Histories of Early-Period Substance Use

• Activity Number: 621
• Type: Contributed
• Date/Time: Thursday, August 2, 2012 : 8:30 AM to 10:20 AM
• Sponsor: WNAR
• Author(s): Li-Jung Liang*+ and Richard Kwock and Chi-hong Tseng
• Companies: University of California at Los Angeles and University of California at Los Angeles and University of California at Los Angeles
• Keywords: Recovery of substance abuse ; Joint modeling ; Shared latent random effects model

Abstract

Understanding whether the patterns of addicts' early-period substance use (AEPSU) can predict their recovery from substance abuse, using data from natural history interview (NHI) studies, is of interest. The NHI studies collect long-term substance abuse history data and behaviors along with other potentially related information. To understand the patterns of AEPSU and how the patterns impact recovery from substance abuse, we propose to develop an analytical approach that uses a Bayesian framework to (1) construct various history functions based on AEPSU data, and (2) develop a joint hierarchical model for time to recovery and history function(s) of AEPSU using a shared latent random effects model to investigate factors that predict time to recovery from substance abuse. Our joint modeling approach considers pattern(s) of multiple substances simultaneously to properly account for the correlations among different substances within subjects. Also, this approach provides efficient estimates for the associations between time to recovery and history function(s) of AEPSU, leading to more accurate predictions. We demonstrate the proposed method using a long-term follow-up NHI study.