Extending the Multiphase Optimization Strategy (MOST) to recursive and dynamic intervention optimization: A bayesian framework for developing effective, affordable, scalable and efficient multicomponent interventions

Abstract: Multicomponent interventions play a central role in behavioral health practice and are increasingly important in clinical medicine. Multicomponent interventions consist of a set of individual treatment components, each intended to affect a subset of behavioral or biological pathways relevant to one or more outcomes of interest. For example, a multicomponent smoking cessation intervention might contain components like pharmacotherapy, financial incentives, counseling, monitoring and feedback, and family support (see, e.g., Wen et al. 2019). Traditionally, multicomponent intervention development has followed a “treatment package” approach (Wyrick et al. 2014), where a set of individual components are assembled ad hoc according to investigator choice, tested in a pilot study, and then advanced to a traditional two-armed randomized controlled trial (RCT). Investigators and other stakeholders may have prior information about efficacy of some individual components from previous research, including potentially similar interventions used in treating other conditions, but often, little is known about how well the components work together to produce desired health outcomes. Recently, the Multiphase Optimization Strategy (MOST) (Collins 2018) has changed the landscape of intervention science by providing a framework for optimizing interventions prior to evaluation in an RCT. A common application of MOST involves screening out ineffective (or iatrogenic) individual components based on the magnitude of main and interaction effects estimated using classical analysis of variance in a factorial optimization trial. Our team is currently extending MOST into a framework for recursive and dynamic intervention optimization (RADIO). In this talk, we outline our vision for a process of continuous intervention development, implementation and adaptation of interventions to suit local constraints and preferences, based on sequential learning from Bayesian adaptive experiments and decision-making based on multi-criteria decision analysis (MCDA). The goal of MOST-RADIO is to develop effective, affordable, scalable and efficient multicomponent interventions while maximizing decision-relevant information per dollar spent in research.

Wen X, Eiden RD, Justicia-Linde FE, et al. A multicomponent behavioral intervention for smoking cessation during pregnancy: a nonconcurrent multiple-baseline design. Transl Behav Med. 2019;9(2):308-318. doi:10.1093/tbm/iby027.
Wyrick DL, Rulison KL, Fearnow-Kenney M, Milroy JJ, Collins LM. Moving beyond the treatment package approach to developing behavioral interventions: addressing questions that arose during an application of the Multiphase Optimization Strategy (MOST). Translational Behavioral Medicine. 2014;4(3):252-259. doi:10.1007/s13142-013-0247-7.
Collins LM. Optimization of Behavioral, Biobehavioral, and Biomedical Interventions: The Multiphase Optimization Strategy (MOST). Springer; 2018.

David J. Vanness, PhD, Jillian C. Strayhorn, MS – Pennsylvania State University, Linda M. Collins, PhD – New York University