The stepped-wedge cluster-randomized design has received substantial attention in pragmatic clinical trials and implementation science research. The key feature of the design is the unidirectional crossover of clusters from the control to intervention conditions on a staggered schedule. The stepped-wedge design (SWD) first appeared in a study of infectious disease in the 1980s. However, the statistical model used to justify the design and analysis was not formally introduced until 2007 in an article by Hussey and Hughes (2007). Statistical approaches to stepped-wedge trials were initially simple extensions of random-intercept models for parallel cluster randomized trials (CRTs), but attention to the longitudinal nature of the design has raised awareness of potential heterogeneity of time trends and treatment effects across clusters. Furthermore, causal inference methods are a foundation for the development of estimation methods and use the concept of potential outcomes that represent the outcomes a patient would have under each treatment option. However, in many settings, the ultimate outcome for a patient is determined both by the choice of specific treatment and by the context in which this treatment is received, such as a specific provider who may systematically influence results among their patient panel. In addition, research is needed to determine optimal and robust analysis methods for the SWD. Software for the computation of power and sample size is limited, and there is a need for tools that can approach general designs and discrete outcomes. Finally, the design of CRTs with a baseline observation period would enable consideration of whether randomization that constrains on the baseline outcomes is beneficial for improved precision, and the monitoring of accruing information over time in stepped-wedge trials should be detailed to inform interim analysis of safety or efficacy.
