摘要：Objectives. Multiple baseline designs (MBDs) have been suggested as alternatives to group-randomized trials (GRT). We reviewed structural features of MBDs and considered their potential effectiveness in public health research. We also reviewed the effect of staggered starts on statistical power. Methods. We reviewed the MBD literature to identify key structural features, recent suggestions that MBDs be adopted in public health research, and the literature on power in GRTs with staggered starts. We also computed power for MBDs and GRTs. Results. The features that have contributed to the success of small MBDs in some fields are not likely to translate well to public health research. MBDs can be more powerful than GRTs under some conditions, but those conditions involve assumptions that require careful evaluation in practice. Conclusions. MBDs will often serve better as a complement of rather than as an alternative to GRTs. GRTs may employ staggered starts for logistical or ethical reasons, but this will always increase their duration and will often increase their cost. The randomized controlled trial (RCT) is the gold standard for causal inference for individual-level interventions. 1 When interventions are applied at the group level and outcomes are measured at the individual level, the cluster- or group-randomized trial (GRT) is the gold standard for causal inference. 2 , 3 However, GRTs are often costly and time-consuming, prompting researchers to look for alternatives. One alternative that has been suggested is the multiple baseline design (MBD), which has a venerable history in education and applied behavior research with individual-level interventions but is relatively new in public health research with group-level interventions. 4 – 6 The MBD makes repeated measurements over a period of time and introduces a sustained intervention on a staggered schedule; intervention effects synchronized with the staggered start times provide evidence for causal inference. Hawkins et al. described the MBD as “a viable alternative to the RCT” and suggested that it will be lower cost, use smaller sample sizes, and still be statistically rigorous. 5 Biglan et al. suggested complementary roles, with MBDs used to “develop and sort through potentially effective intervention methods, followed by evaluation in RCTs both to test efficacy and to determine the extent of generalizability.” 4 We review the structural features that have made MBDs useful in other fields and consider whether similar success is likely in public health. We also compare the statistical power of MBDs and GRTs.
