Table 3.

Summary of Key Design and Analysis Considerations When Developing a Cluster-Randomized Trial (CRT) in Infection Control and Hospital Epidemiology

Epidemiological Principle
1- Design
Select the appropriate type of CRT (see Table 1) Report the justification/rationale for using this design in the introduction section of the paper
2- Sample size estimates
Account for clustering by including a design effect such as ICC or CV in the sample size estimate Report the ICC or CV used to estimate sample size Report the effect size used to estimate sample size
3- Consent
Consider seeking consent at the cluster level Determine whether it is appropriate and feasible to seek consent from representatives of the cluster rather than from individuals within each cluster
4- Level of inference
Define whether the level of inference is at the individual or the cluster level
5- Matching and/or stratification
Consider using an appropriate matching factor to improve power However, use caution when employing this approach, especially if the number of clusters is large Consider using a stratified design if appropriate
6- Bias and/or contamination
Consider the following techniques to reduce the potential for bias and/or contamination: Use an appropriate wash-out period if using a crossover design Implement the study in areas where clusters are distinct and well separated Use control group clusters that are external to the experimental trial Use multiple crossover periods
7- Analysis
If the level of inference is the individual, account for clustering in the analysis Consider using statistical techniques such as mixed-effects models