Joint longitudinal and survival models for intensive longitudinal data from mobile health studies of smoking cessation

点击翻译按钮获取中文摘要

Project Summary/Abstract Increasing collection of intensive longitudinal data (ILD) through mobile health (mHealth)-based approaches, such as ecological momentary assessment (EMA), present a rich source of information for understanding temporal variations in psychological states key to smoking cessation. However, advances in statistical methods are needed to fully leverage these rich data to assess interventions and inform the design of future interventions. In Aims 1 and 2, this proposal seeks to develop a novel statistical model (joint longitudinal recurrent-event model) and estimation method that will allow for analysis of EMA data from a smoking cessation study using low-dimensional interpretable states to describe the behavioral phenomenon and processes related to smoking cessation. By incorporating just-in-time adaptive interventions (JITAIs) into the model in Aim 3, this project will facilitate assessment of the impact of time-varying adaptive interventions on a subject’s risk of a future lapse in smoking cessation using data from the Mobile Assistance for Regulating Smoking (MARS) micro-randomized trial (U01CA229437; PIs: Nahum-Shani, Wetter). Training goals, which were developed with the mentorship team, include: (i) advancing technical training in statistical theory and computing, (ii) improving written and oral communication skills, (iii) building collaborative relationships, and (iv) attending conferences, workshops, and professional development activities. The proposed research and training will be conducted at the University of Michigan (UM) in the Department of Biostatistics, which has close ties to the UM Institute for Social Research and a reputation for excellence in research and training. Altogether, the statistical methodology proposed in this project will contribute to the science in two key ways: (i) it will allow for the integration of many different items (e.g. emotions, urge, motivation) in a way that facilitates interpretation when measuring vulnerability (e.g. risk of lapse) and (ii) its interpretability will subsequently help inform the design of evidence-based adaptive interventions (e.g. JITAIs) through increased understanding of the conditions that represent vulnerability. These scientific contributions directly promote the National Institute on Drug Abuse’s strategic goal of developing “new and improved treatments to help people with substance use disorders achieve and maintain a meaningful and sustained recovery”. Although presented in the context of a smoking cessation study, this methodological framework is highly flexible with broad applicability to mHealth studies of substance use disorders and in other health domains. This novel analytic method will be freely available in a user-friendly R package, thus facilitating the potential impact on drug-use research.