• Sarah J Arnup, Joanne E McKenzie, David Pilcher, Rinaldo Bellomo, and Andrew B Forbes.
• School of Public Health and Preventive Medicine, Monash University, Melbourne, VIC, Australia. andrew.forbes@monash.edu.
• Crit Care Resusc. 2018 Jun 1; 20 (2): 117-123.

ObjectiveThe cluster randomised crossover (CRXO) design provides an opportunity to conduct randomised controlled trials to evaluate low risk interventions in the intensive care setting. Our aim is to provide a tutorial on how to perform a sample size calculation for a CRXO trial, focusing on the meaning of the elements required for the calculations, with application to intensive care trials.Data SourcesWe use all-cause in-hospital mortality from the Australian and New Zealand Intensive Care Society Adult Patient Database clinical registry to illustrate the sample size calculations.MethodsWe show sample size calculations for a two-intervention, two 12-month period, cross-sectional CRXO trial. We provide the formulae, and examples of their use, to determine the number of intensive care units required to detect a risk ratio (RR) with a designated level of power between two interventions for trials in which the elements required for sample size calculations remain constant across all ICUs (unstratified design); and in which there are distinct groups (strata) of ICUs that differ importantly in the elements required for sample size calculations (stratified design).ResultsThe CRXO design markedly reduces the sample size requirement compared with the parallel-group, cluster randomised design for the example cases. The stratified design further reduces the sample size requirement compared with the unstratified design.ConclusionsThe CRXO design enables the evaluation of routinely used interventions that can bring about small, but important, improvements in patient care in the intensive care setting.

11 keywords...

hide…