*Geek Box: Cluster Randomisation

*Geek Box: Cluster Randomisation

The present study had a design that is slightly different to the traditional method of randomisation, using a cluster-randomisation method. What does “cluster” mean in this context? It means that the unit of randomisation is not an individual.

Rather, a cluster of individuals in a pre-specified group are randomised together. This method of randomisation is often used in education and public health policy research. For example, suppose you want to do an intervention in a school comparing the effects of a reading technique in children. If you randomised individuals in a classroom, you could have a situation where a child receiving the intervention is sitting next to a child in the control group; this would lead to what is called “contamination”, i.e., the potential for the treatment and control groups to mix and thus compromise the intervention.

To prevent contamination, and also for ease of implementation, it would be more useful to randomise the entire class; therefore, randomisation assigns the whole of Class A to the intervention and Class B is assigned to the control. Thus, the class is the “cluster”. In the present study, the residential care home was the unit of randomisation.