A1693
Title: Design issues for longitudinal, cluster randomized clinical trials with repeated measures
Authors: Florin Vaida - University of California San Diego (United States) [presenting]
Abstract: The aim is to investigate design issues for cluster-randomized, longitudinal clinical trials with repeated measures. Units grouped within clusters are randomized to two or more groups. Units are observed longitudinally, at baseline and follow-up visits. Repeated measures for the response of interest are obtained at each visit. We show that a specific imbalance allocation of measures between baseline and follow-up is optimal. The robustness of the design is also considered. Statistical analyses are based on analytical derivations. We show that the optimal analyses are those controlling for baseline as a covariate, or including baseline in a longitudinal analysis and assuming no baseline differences. We show analytically that the two approaches are equivalent for finite samples. This approach was applied to the Open \& Ask Study (NCT ID NCT03385512), a recently published US large-scale multi-center cluster-randomized controlled trial assessing the comparative effectiveness of three interventions to improve engagement of healthcare providers with their patients.
