Walter Dempsey
University of Michigan
Twin revolutions in wearable technologies and smartphone-delivered digital health interventions have significantly expanded the accessibility and uptake of mobile health (mHealth) interventions in multiple domains of health sciences. Sequentially randomized experiments called micro-randomized trials (MRTs) have grown in popularity as a means to empirically evaluate the effectiveness of mHealth intervention components. MRTs have motivated a new class of causal estimands, termed “causal excursion effects”, that allow health scientists to answer important scientific questions about how intervention effectiveness may change over time or be moderated by individual characteristics, time-varying context, or past responses. In this talk, we revisit the estimation of causal excursion effects and present two new tools for improving efficiency. First, we will present a simple and intuitive method to improve the efficiency of moderated causal excursion effects by including auxiliary variables. This method extends the covariate-adjustment RCT literature to the time-varying setting. Second, we will consider a meta-learner perspective, where any supervised learning algorithm can be used to assist in the estimation of the causal excursion effect. Theoretical comparisons accompanied by extensive simulation experiments demonstrate the relative efficiency gains. Practical utility of the proposed methods is demonstrated by analyzing data from a multi-institution cohort of first year medical residents in the United States.
