Table 1.

Variations of the Cluster-Randomized Trial (CRT) Design

Type	Description	Advantages	Disadvantages
Parallel CRT	Clusters are randomly assigned to intervention or control in parallel	Straight-forward design and analysis	Clusters in the control arm do not have an opportunity to receive the intervention during the study period
Matched CRT	Matched pairs of clusters are randomly assigned to intervention or control. Examples of matching variables include hospital size, rural/urban location.	Addresses confounders at the design stage Offers some statistical advantage	If one cluster withdraws, the other hospital/ICU must also be dropped from the study.
Stratified CRT	Clusters are randomly assigned to intervention or control within a strata. Example: unit of randomization = infectious disease clinic, stratification variable = country or state	Ensures that an equal number of clusters from each group are in the intervention vs control groups Controls for confounding at the design stage	Requires a certain number of clusters in each group
CRT with stepped-wedge	All clusters eventually crossover but only from the control to the intervention at a time point determined at random.	Most appropriate when there is some evidence to support the intervention. All clusters eventually receive the intervention. Allows clusters to be enrolled gradually enrolled over time	Some argue that this is really a quasiexperimental design and is not truly randomized because no clusters get randomized to control.
CRT with crossover	All clusters receive both intervention and control in a sequence determined at random.	Reduces total number of cluster required	Requires a longer study duration Washout period is often not long enough resulting in bias.
CRT with crossover and multiple periods	All clusters receive both intervention and control multiple times in a sequence determined at random.	Reduces total number of cluster required Can assess carryover effect Allows determination of the true treatment effect vs a period effect	Adds another source of variance (between cluster, within cluster AND between cluster-periods) Requires longer study duration
Factorial CRT	Clusters are randomly allocated to 1 of 4 groups: receiving both interventions, intervention A, intervention B, or no intervention.	Can test whether intervention A is better than no intervention A or whether intervention B is better than no intervention B Because there is a combination of interventions A and B, it is also possible to test whether interventions A and B work together in synergy or are antagonistic. Used to assess the effects of 2 interventions in the same study and to explore interactions between interventions	Logistically and analytically complex
Fractional factorial CRT	Derived from a full factorial by dropping some conditions (combinations of factors) when resources are too limited to implement all possible combinations of factors or because some combinations cannot or should not be implemented	When conditions are removed, certain effects become completely confounded with each other and cannot be estimated separately.	Logistically and analytically complex