B1405
Title: Analyzing Event-triggered Adaptive Interventions using Data from Sequentially Randomized Trials
Authors: Mason Ferlic - University of Michigan (United States) [presenting]
Abstract: A dynamic treatment regime consists of a protocolized sequence of decision rules used to guide an intervention across multiple stages of treatments contingent on the evolving status of the individual. Technological advances in mobile and digital health have made it possible to monitor dynamic treatment response in near real-time and adapt future treatment to individual needs. Technology-assisted adaptive interventions with a digital tailoring variable are becoming more commonplace. In such mobile monitoring environments, the set of decision rules is also allowed to vary with time, enabling researchers to answer more complex questions regarding the time-varying effect of treatment and the dynamic trajectory of response status. We introduce a new approach to analyzing technology-assisted adaptive interventions embedded in a sequential, multiple-assignment, randomized trial (SMART) on a continuous, longitudinal outcome. We propose a simple two-stage regression algorithm that adjusts for time-varying transitions to second-stage treatment. Through simulation studies, we illustrate the validity of estimated treatment effects and examine operating characteristics under different levels of model misspecification. We show that unadjusted standard errors are anti-conservative. Using data from a SMART, we illustrate our methodology in a case study involving digitally monitored weight loss treatment.
