Abstract:

Mobile health (mHealth) or digital health interventions are becoming increasingly common in tandem with advances in technology. In this talk, we will present an innovative trial design arising in mHealth, namely, the micro-randomized trial (MRT) that involves sequential, within-person randomization over many instances. MRTs often aim to estimate the proximal effects of “push”-type mHealth interventions, e.g., motivational text-messages to promote physical activity or other healthy behaviors. The basic MRT design can be further improved to make it adaptive, thereby enabling it to learn from accumulated data as the trial progresses. This is appealing from an ethical perspective since the adaptive learning tends to make better interventions available to the trial participants. Adaptive learning is often operationalized via contextual bandit algorithms. Specifically, we will discuss the role of a particular contextual bandit algorithm, namely, Thompson sampling, in designing adaptive MRTs with count-type outcomes. Simulation results will be shown to validate the proposed design approach. mHealth case studies will be discussed in detail.

About the Speaker:

Bibhas Chakraborty is an Associate Professor and Ex-Director of the Centre for Quantitative Medicine at the Duke-NUS Medical School, Singapore, an Associate Professor of Statistics and Data Science at the National University of Singapore, as well as an Adjunct Associate Professor of Biostatistics and Bioinformatics at Duke University. Previously (2009-13), he was an Assistant Professor of Biostatistics at Columbia University, after completing his Ph.D. in Statistics from the University of Michigan in 2009. He is a recipient of the Calderone Research Prize for Junior Faculty from Columbia University’s Mailman School of Public Health in 2011, and the Young Statistical Scientist Award from the International Indian Statistical Association in 2017. In 2022, he became an Elected Member of the International Statistical Institute. His core areas of research include dynamic treatment regimes, sequential decision making, reinforcement learning, adaptive clinical trial designs and mobile/digital health. He wrote the first textbook on dynamic treatment regimens.