Selection bias in cluster randomised controlled trials

Professor David Torgerson, Director of the York Trials Unit, gives a talk for the Evidence Based Healthcare podcast series. He has published widely with over 250 peer reviewed papers many of them on the design of randomised trials including a student text book 'Designing Randomised Trials in Health Education and the Social Sciences' (2008, Palgrave MacMillan). He has a particular interest in the design and conduct of cluster randomised trials. Randomisation, if conducted properly, will abolish selection bias. Poor randomisation practice for individually randomised trials allows the allocation schedule to be predicted and can lead to subversion of the randomisation, which introduces selection bias. In cluster randomised controlled trials there is a particular problem with some trials in that cluster randomisation occurs before individual recruitment. When this occurs the allocation can become known to the potential participant and the recruiting clinician and research. This, then, allows selective recruitment to occur which means that selection bias is introduced at the level of the individual participant. In this talk the problem is illustrated in case studies and I discuss approaches to dealing with this potential source of selection bias.