Advantages of Joint Modeling of Component HIV Risk Behaviors and Non-Response: Application to Randomized Trials in Cocaine-Dependent and Methamphetamine-Dependent Populations

Abstract

The HIV risk-taking behavior scale (HRBS) is an 11-item instrument designed to assess the risks of HIV infection due to self-reported injection-drug use and sexual behavior. A retrospective analysis was performed on HRBS data collected from approximately 1,000 participants pooled across seven clinical trials of pharmacotherapies for either the treatment of cocaine dependence or methamphetamine dependence. Analysis faced three important challenges. The sample contained a high proportion of missing assessments after randomization. Also, the HRBS scale consists of two distinct behavioral components which may or may not coincide in response patterns. In addition, distributions of responses on the subscales were highly concentrated at just a few values (e.g., 0, 6). To address these challenges, a single probit regression model was fit to three outcomes variables simultaneously – the two subscale totals plus an indicator variable for assessments not obtained (non-response). This joint-outcome regression model was able to identify that those who left assessment early had higher self-reported risk of injection-drug use and lower self-reported risky sexual behavior because the model was able to draw on information on associations among the three outcomes collectively. These findings were not identified in analyses performed on each outcome separately. No evidence for an effect of pharmacotherapies was observed, except to reduce missing assessments. Univariate-outcome modeling is not recommended for the HRBS.

Simultaneous Modeling of the Impact of Treatments on Alcohol Consumption and Quality of Life in the COMBINE Study: A Coupled Hidden Markov Analysis

“…This study constructed coupled HMMs to directly estimate the impact of treatment assignment on drinking and QOL while explicitly modeling the dynamic relationship between drinking and QOL over time. In addition to the statistical benefits afforded by latent‐variable models that were discussed above, the present joint analytic approach has the added benefits of potentially increased accuracy, and potentially improved statistical power, of model parameter estimates (Holmes et al., ).…”