Multidimensional and Dynamic Moderation in Drug Abuse and HIV Studies

ABSTRACT: Research Project II ? Multidimensional and Dynamic Moderation in Drug Abuse and HIV Studies To accelerate the fight against drug abuse and HIV, it is critical to gain a deeper understanding of heterogeneity in intervention effects. The overarching goal of this project is to develop a powerful new framework that integrates ideas from latent class analysis and time-varying effect models to enable researchers to discover (a) clinically useful, multidimensional moderators and (b) moderators that play a dynamic (i.e., age- and time-varying) role in the effects of interventions on univariate and multidimensional outcomes. The specific aims for this project are as follows. (1) Latent class moderation: To develop an approach to estimate and test the effects of complex, multidimensional moderators. We will develop a new approach that will enable researchers to identify moderators that represent multidimensional profiles of characteristics rather than a single variable. (2) Dynamic moderation: To develop an approach to identify moderators with age- and time-varying effects. We will develop a new type of moderation analysis to enable identification of baseline moderators with dynamic effects on interventions for observed and latent class moderators. (3) Dynamic moderation with a latent class outcome: To develop methods for identifying dynamic effects of moderators on complex, multidimensional outcomes. The approach in Aim 2 will be extended to enable researchers to identify dynamic effects of baseline moderators on multidimensional outcomes across age or time. (4) Dissemination and software development. Methodological and empirical findings will be presented in drug abuse and HIV prevention, intervention, health services, and quantitative outlets. The research team will work with the Dissemination, Software, and Technology Core to develop and disseminate software in SAS and Stata so that drug abuse and HIV researchers can readily apply these new methods. The techniques will be evaluated using computer simulation studies and demonstrated using data from a randomized drug abuse and HIV prevention program for Hispanic youth, a longitudinal study of alcohol and drug addiction service utilization, and ecological momentary assessments from a six-arm smoking cessation trial. Complex questions to be addressed include: ?What configurations of risk factors correspond to greater effects of an HIV risk behavior intervention program?? ?How does gender function as a moderator of the effect of outpatient drug abuse services, differentially across ages 16 to 50?? and ?At what times during a quit attempt does baseline dependence moderate the effect of a smoking cessation intervention on withdrawal symptoms profiles?? Ultimately, these methods will enable researchers to gain critical knowledge that will inform development of more effective, targeted drug abuse and HIV interventions.