Copyright 2016 American Medical Association. All Rights Reserved. Applicable FARS/DFARS Restrictions Apply to Government Use.
In Reply Dr Nunez and colleagues make important points regarding susceptibility of cluster randomized trials to selection bias. They also state that results from other studies support the effectiveness of checklists.1
We agree that design features of cluster randomized trials make them especially vulnerable to selection bias.2 Prospective identification and recruitment of individuals in cluster randomized trials often takes place after randomization, as opposed to all possible patients in the cluster being considered in the analysis. This characteristic may lead to selection bias. When identification of patients after randomization is necessary, methodologists recommend that recruitment be conducted by a professional who is blinded to treatment allocation or by a trained person who is not involved with patient care.2,3 In addition, bias is unlikely if eligibility criteria are simple, objective, and clear, so that recruiters cannot have a substantial effect on who is recruited.4 Finally, need of patients’ informed consent after randomization may also introduce bias.2
Cavalcanti AB, Machado FR, Berwanger O. Quality Improvement Intervention and Mortality of Critically Ill Patients—Reply. JAMA. 2016;316(8):879-880. doi:10.1001/jama.2016.9276