368 – Methods and Applications in High-Dimensional Data, Part 1
A Bayesian Joint Hierarchical Model for Long-Term Multiple Substance Use and Recovery from Substance Use
Li-Jung Liang
UCLA
Chi-hong Tseng
UCLA
Sitaram Vangala
UCLA
Yih-Ing Hser
UCLA
Research on how the patterns of addicts' early-period substance use can predict their recovery from long-term substance use, using data from natural history interview studies, is limited. We propose to use a Bayesian joint hierarchical model to investigate the association between patterns of addicts' early-period substance use (longitudinal) and time to recovery from substance use. This approach allows us to properly account for the correlations among multiple drugs within subjects and to provide efficient estimates for the association between time to recovery and long-term use of multiple drugs. A 33-year follow-up study was used to demonstrate our approach.