Accelerating Smoking Relapse Research Using Longitudinal Models of EMA Data, D. McCarthy & S. Shiffman (Co-PIs)

Relapse is a central problem in smoking cessation and other areas of behavior change. Although our conceptual models of relapse and our methods of measuring behavior and its antecedents in real-time have grown in sophistication over the past 20 years, our analytical models have not followed suit. The gap between the richness of dynamic conceptual models of change, and the relatively simple, linear statistical models of change typically adopted has slowed progress in understanding and preventing relapse. Although research has identified individual differences that predict increased relapse risk, we know little about how (i.e., by what proximal mechanisms) such factors influence momentary smoking decisions. As a result, we do not know which proximal processes to monitor or target in smoking cessation interventions. In addition, we do not yet know how to identify smokers most vulnerable to unfavorable experiences when they quit smoking, in terms of subjective distress and demoralization. As such, we do not yet know how to improve the process of quitting while also effectively promoting abstinence. Reducing distress and demoralization during the process of quitting may have important implications for late relapse and recycling (or returning to abstinence following relapse). In the proposed project, the research team will bridge the gap between conceptual and analytic models of relapse and address these important, unanswered questions about the relapse process. To achieve these aims, the team will apply state-of-the-art statistical modeling paradigms to real-time data on smoking and its antecedents collected via ecological momentary assessment (EMA) from four samples of smokers engaged in assisted smoking cessation attempts. First, the team will conduct latent transition analyses to identify both distal and proximal predictors of key transitions in the smoking cessation process (i.e., a first lapse, relapse to regular smoking, and recycling). Second, the team will fit nonlinear dynamical systems models to the data to identify the combinations of distal, proximal, and contextual influences that predict non-linear increases in lapse and relapse risk. Third, the team will use latent growth mixture modeling to identify classes of trajectories in smoking and subjective distress or demoralization during the first 2-6 weeks of a quit attempt in an effort to identify predictors of unfavorable experiences that could be ameliorated with future treatments. Results of these analyses will extend knowledge of critical, distal determinants of important smoking and subjective outcomes, and will illuminate how these influences affect key transitions or trajectories in the smoking cessation process. Such information could suggest new treatment targets and new strategies for matching smokers to treatments or delivering just-in-time treatments during periods of elevated risk. Results from the proposed analyses may have implications for other addictive or health behavior changes, as well. In addition, the proposed application of state-of-the-art analytic modeling to behavior change data may serve as a model to other researchers, and thus may spur advances and innovations in diverse research areas.