Novel Methods to Inform mHealth Interventions for Substance Use - Project Summary Despite the availability of a variety of intervention approaches for substance use disorders (SUD), many indi- viduals do not benefit from existing interventions either because they are not offered at the time in which they can be most beneficial and/or they are not tailored to the changing needs of individuals. Modern technologies such as mobile and wearable devices provide unprecedented opportunities to deliver just-in-time adaptive in- terventions (JITAIs)–an intervention design that adapts intervention delivery to an individual’s rapidly changing internal state and context, in real-time, real-world settings. However, new research methodologies are needed to capitalize upon these opportunities and inform a new generation of interventions for SUD. Recent advances in Micro-Randomized Trials (MRTs) support the collection of data that can empirically inform the construction of JITAIs. Although the MRT is a major step forward in methodology for building empirically- based JITAIs, new data analytic methods are needed to fully realize the potential of these data. Specifically, there are several important gaps in existing data analytic approaches. First, these methods can be used to evaluate a set of pre-specified JITAIs and assess intervention effects on proximal outcomes (e.g., next-day self-reported substance use) but not to construct an optimal JITAI with respect to a distal outcome (e.g., substance use by a 6-month follow-up). This represents a major gap given that long-term benefits are the primary motivation of SUD treatments. Second, existing methods are limited to intention-to-treat (ITT) analyses, which may provide biased estimates of the effec- tiveness of an intervention when individuals are not fully engaged. The latter is a major challenge in developing effective JITAIs given that engagement in mobile health (mHealth) interventions is often suboptimal and declines over time. When the rate or the pattern of intervention engagement in the study vary from the rate or the pattern of engagement achieved when implementing the intervention in real-world, the results obtained by ITT analyses might not be reproducible. Third, there is no variable selection method in MRT settings to identify the important tailoring variables. These gaps can seriously limit the usefulness and generalizability of the results obtained from MRT data. We propose to close these gaps by developing new methods that can help SUD investigators leverage MRT data to construct a JITAI that is most beneficial in terms of a distal outcome (Aim 1). We further propose to develop new methods that accommodate intervention engagement when estimating the optimal JITAI (Aim 2). We will also develop a method that enables investigators to identify important tailoring variables for inclusion in the optimal JITAI (Aim 3). We will apply these methods to rich data from 3 funded studies, each employing an MRT to inform a JITAI for substance use. These data sets will be used to motivate the new methodologies, as well as to illustrate their utility as we disseminate them to SUD investigators via open-source, easy-to-use statis- tical software (Aim 4). The successful completion of this work will facilitate the development of empirically-based, effective, interpretable and generalizable JITAIs to prevent and treat SUD and potentially other chronic disorders.
