The proposed research project is an R01 project in response to PA-10-018 (Accelerating the pace of drug abuse research using existing epidemiology, prevention, and treatment research data). The goal of this project is to establish a framework for causal inference accounting for heterogeneity in longitudinal substance use trajectories in the context of prevention intervention trials. To accomplish this goal, we propose to integrate two powerful modeling frameworks, growth mixture modeling and causal modeling. Growth mixture modeling is a flexible tool for identifying heterogeneous trajectory strata and strata-specific intervention effects based on empirical model fitting. However, little is known about how we can use the growth mixture modeling results as strong evidence of causal intervention effects when their identification heavily relies on empirical fitting and parametric assumptions. Causal modeling refers to an inferential framework that focuses on clarification of assumptions that make causal interpretation possible. The strength of this approach is that the quality of causal inference can be evaluated by the scientific plausibility of the assumptions and the quality of sensitivity analysis based on these assumptions. Despite the significant potential benefit of integrating the two frameworks, little research has been conducted so far to examine such possibility. The proposed project is intended to guide this integration and to provide a practical framework for causal inference accounting for longitudinal heterogeneity in substance use development. Our investigations will be guided by existing data from two intervention studies: Adolescent Substance Abuse Prevention Study (AS- APS: Sloboda et al., 2009) and Johns Hopkins University Preventive Intervention Research Center Study (JHU PIRC: Ialongo et al., 1999). This project will provide extensive secondary analyses of these data using cutting-edge growth mixture modeling methods, in particular focusing on estimating intervention effects among groups with heterogeneous substance use trajectories. We expect that our study will promote high quality secondary analysis and improve the design of future substance use intervention trials by improving the evaluation of differential intervention effects as well as the identification of subpopulations who would benefit most from the intervention.
This project intends to improve evaluation of prevention intervention impacts by considering heterogeneity in longitudinal substance use trajectories and to improve identification of subpopulations that would benefit most from early interventions. The results of this project will improve the design and quality of future intervention trials, and therefore will have positive impact on public health.
| Jo, Booil; Findling, Robert L; Wang, Chen-Pin et al. (2017) Targeted use of growth mixture modeling: a learning perspective. Stat Med 36:671-686 |
| Wang, Chen-Pin; Lorenzo, Carlos; Habib, Samy L et al. (2017) Differential effects of metformin on age related comorbidities in older men with type 2 diabetes. J Diabetes Complications 31:679-686 |
| Jo, Booil; Findling, Robert L; Hastie, Trevor J et al. (2016) Construction of longitudinal prediction targets using semisupervised learning. Stat Methods Med Res :962280216684163 |
| Wang, Chen-Pin; Jo, Booil; Brown, C Hendricks (2014) Causal inference in longitudinal comparative effectiveness studies with repeated measures of a continuous intermediate variable. Stat Med 33:3509-27 |
| Wang, Chen-Pin; Jo, Booil (2013) Applications of a Kullback-Leibler Divergence for Comparing Non-nested Models. Stat Modelling 13:409-429 |
