Cluster-Randomized Trials Explainer
A cluster-randomized trial assigns intact groups, rather than individual people, to study conditions. The clusters may be clinics, hospitals, schools, neighborhoods, families, or communities. Researchers use this design when an intervention is delivered to a whole group, when individual assignment is impractical, or when people in different study arms would influence one another and contaminate the comparison.
For example, a trial of a new handoff training program might randomize hospitals because the training changes team behavior throughout each site. Outcomes can still be measured in individual patients or staff, but the assignment occurred at the hospital level. The protocol must define how clusters and participants are recruited, when consent is obtained, and whether the same eligibility rules are applied before or after group assignment.
People within a cluster tend to resemble one another or share conditions, so their results are not statistically independent. The analysis and sample-size calculation must account for that correlation. A trial with many people but only a few clusters may have less reliable evidence than the head count suggests. Baseline differences between clusters, loss of whole sites, and recruitment after assignment can also introduce bias.
Cluster trials are powerful for studying care delivery, public-health programs, and policies that operate above the individual level. Their credibility depends on enough clusters, appropriate analysis, transparent reporting, and ethical attention to people affected by a group decision. Readers should check both the number of participants and the number of randomized groups, because the cluster—not each person—is the unit that received the random assignment.
A clinic, school, household, or community is allocated to the intervention or comparison condition as one unit.
Individual outcomes may still be measured within each group.
Members of the same group often share staff, environment, or behavior and therefore produce correlated outcomes.
Sample size and analysis must account for that similarity.
A large participant total can be misleading when only a small number of groups were randomized.
The balance, number, and retention of clusters strongly affect confidence in the result.
Explore more "Explainers"
Discover additional explainers across politics, science, business, technology, and other fields. Each explainer breaks down a complex idea into clear, everyday language—helping you better understand how major concepts, systems, and debates shape the world around us.
