AN INDEPENDENT PUBLIC HEALTH READING ROOM
Places. People.
Possibilities.

A Cluster Trial Assigns Groups, So Count the Groups as Well as the People

Paper collage of a notebook, tablet, magnifying glass and source card.
AI-created conceptual illustration of a fictional setting; not a record of people or outcomes.

A trial involving hundreds of people may have randomised only a small number of clinics, schools or communities. The number of participants and the number of independently assigned groups describe different parts of its design.

The National Cancer Institute’s course on cluster randomised designs distinguishes group-level assignment from traditional patient-level randomisation. This design can be useful when an intervention is delivered to a whole setting.

Consider a fictional trial assigning twelve clinics to two reminder systems, with one hundred patients observed in each clinic. The study includes twelve hundred patients, but it did not make twelve hundred independent random assignments.

People within a clinic may share staff, procedures or local conditions. Their outcomes can therefore be related. An analysis needs to account appropriately for the grouping rather than treating every patient as an unrelated source of information.

Look for the number of clusters, their sizes, the allocation method and the timing of participant recruitment. If people entered the study after clinics knew their assignment, recruitment procedures deserve particular attention. The participant total alone cannot answer those methodological questions.

For a reading note, draw one box per clinic and place the participant counts inside. Mark the intervention at the box level. This simple diagram makes the unit of assignment visible before you read an effect estimate.

Cluster randomisation is not inherently a weaker or stronger label than individual randomisation; suitability depends on the question and implementation. A public summary should explain what was assigned and what was measured, while leaving personal treatment decisions to appropriate clinical advice. Counting both groups and people gives readers a more accurate picture of how the evidence was produced.

Visual note: the image illustrates the subject, not a documented event.