Abstract
Stepped wedge cluster randomized trials (SW-CRTs) are a type of uni-directional crossover designs and are increasingly common in prevention and implementation research. Although sample size formulas have been developed to support the planning of SW-CRTs, almost no prior methods incorporated the baseline measurement of the outcome-a common feature in many randomized trials and, increasingly, in cross-sectional SW-CRTs. In this article, we systematically investigate the possibility of addressing a baseline outcome measurement in designing cross-sectional SW-CRTs. We provide three linear mixed modeling approaches to adjust for the baseline outcome and derive the corresponding variance formula of the treatment effect estimator under each. The derived formulas reveal the efficiency implications of including a baseline outcome measurement, and provide a natural vehicle for the efficiency comparisons across adjustment approaches to generate practical recommendations. We validate the power and sample size methods under each baseline adjustment approach using simulations and provide an illustrative sample size calculation with a baseline outcome using the context of a real SW-CRT.