STRIDE: EXPERIMENTAL AND STATISTICAL DESIGN CONSIDERATIONS

摘要

The cluster-randomized STRIDE trial features several design complexities warranting special attention to preserve validity, and to insure adequate statistical power to detect a hypothesized 20% reduction in serious fall injuries. The relatively small number (n=86) and diversity of eligible practices necessitated a constrained randomization scheme preserving balance within and across healthcare systems. This scheme considered practice-level characteristics (e.g. practice size and geographic distribution) and participant-level characteristics (e.g. race and ethnicity). The resulting design and corresponding analytic plan take into account the estimated incidence of serious fall injuries in the target population; the prevalence of specific fall risk factors; the anticipated between-practice variation in number of participants eligible for enrollment; the intra-practice correlation in outcomes; and the potential for deaths and other competing risks to preclude event capture. Drawing upon the STRIDE framework, we will discuss implications for cluster randomized trials of practice-level interventions in older, at risk populations.