Review of Recent Methodological Developments in Group-Randomized Trials: Part 2—Analysis

Abstract

In 2004, Murray et al. published a review of methodological developments in both the design and analysis of group-randomized trials (GRTs). Over the last 13 years, there have been many developments in both areas. The goal of the current paper is to review developments in analysis, with a companion paper to focus on developments in design. As a pair, these papers update the 2004 review. This analysis paper includes developments in topics included in the earlier review, such as methods for parallel-arm GRTs, inference for conditional and marginal effects, and new topics including methods to account for multiple levels of clustering and alternative estimation methods such as augmented GEE, targeted maximum likelihood and quadratic inference functions. We also examine developments in dealing with missing outcome data, including doubly robust approaches, software available for analysis, and analysis of alternative group designs (including stepped wedge GRTs, network-randomized trials, pseudo-cluster randomized trials and individually-randomized group treatment trials). These alternative designs, like the parallel-arm GRT, require clustering to be accounted for in both their design and analysis.

The mixed model for repeated measures for cluster randomized trials: a simulation study investigating bias and type I error with missing continuous data

“…There have been multiple reviews that have asserted that the analyses for CRTs are incorrect based on whether a mixed-model ANOVA is used, if there are more than two time points, unless a randomcoefficients model is used [25,26,39]. Our results contradict this, as the MMRM-CRT appears, as a whole, to have good statistical properties.…”

“…We extend the MMRM to cluster trials (MMRM-CRT) by simply adding a random effect for cluster. While this model is not necessarily new; for example, Littell discusses this model in the context of repeated measures with clustering due to schools [24], the choice of similar models has been criticized when outcomes are measured at more than two time points (as mentioned above) [23,25,26]. Furthermore, to our knowledge, this model has not been investigated for its statistical properties when outcome data are incomplete.…”

An evaluation of constrained randomization for the design and analysis of group‐randomized trials with binary outcomes

“…41 Although variations in group sizes should not affect the validity of permutation tests, 41 the appropriate choice of model-based tests may depend on the degree of group size variability. 35,47 Another possible limitation is that we only considered group-level binary covariates in generating the outcome data. The group-level binary covariates are used in the data generation for illustration purposes and can be easily generalized to group-level continuous covariates.…”

Reducing Cardiovascular Disparities Through Community-Engaged Implementation Research

“…88,99 For example, use of the steppedwedge design, in which all communities or sites receive the intervention in a phased approach, can be a helpful strategy to maintain community engagement when the alternative of potential randomization to a nonintervention comparison group is viewed unfavorably. [100][101][102] Researchers may use low-burden measures to assess the quality, quantity, speed, and extent of implementation 103 in the real-world systems and find ways that are practical, feasible, and align with the needs and mission of the service provider or community. Unobtrusive measures, 104 which involve data sources already being collected during service delivery, can be invaluable.…”