Cluster Unit Randomized Trials

2. Introduction

More generally, contamination may be a serious concern in trials of lifestyle modification and counseling that are conducted using individual randomization if there is an opportunity for control group subjects to mingle with experimental group subjects, and therefore to compare or even share the interventions received.

Other reasons to adopt this design arise simply because of administrative convenience. For example, having set up a disease screening program in general practice, it could be logistically awkward and distracting to staff if a formal randomization scheme were implemented within a doctor's office. In this case the CRT design is attractive since it allows physicians and nurses to operate as they would normally on a day to day basis. That is, it allows the intervention to be given in a way that is more consistent with how the intervention would be given in practice. Randomization by practice also removes problems that could arise if some medical professionals have ethical qualms about offering an innovative health care program to only some of their patients.

Medical settings and communities tend to be the most common randomization units seen in the literature. However some CRTs have randomized more unusual clusters, as listed in Table 1.

Table 1

Examples of Unusual Clusters

Cluster Researcher
Religious institutions Lasater et al., 1997
Baseball teams Walsh et al., 1999
Sex establishments Fontanet et al., 1998
Student pubs Johnsson and Bergland, 2003
Boy scout troops Jago et al., 2006
Calendar weeks Mason et al., 2007
Grocery stores Hunt et al., 2003
Lasater T.M., Becker D.M., Hill M.N., Gans K.M. (1997). Synthesis of findings and issues from religious-based cardiovascular disease prevention trials. Annals of Epidemiology, 7:S46-S53.