Cluster randomized trials for pharmacy practice research

Abstract

Introduction Cluster randomized trials (CRTs) are now the gold standard in health services research, including pharmacy-based interventions. Studies of behaviour, epidemiology, lifestyle modifications, educational programs, and health care models are utilizing the strengths of cluster randomized analyses. Methodology The key property of CRTs is the unit of randomization (clusters), which may be different from the unit of analysis (individual). Subject sample size and, ideally, the number of clusters is determined by the relationship of between-cluster and within-cluster variability. The correlation among participants recruited from the same cluster is known as the intraclass correlation coefficient (ICC). Generally, having more clusters with smaller ICC values will lead to smaller sample sizes. When selecting clusters, stratification before randomization may be useful in decreasing imbalances between study arms. Participant recruitment methods can differ from other types of randomized trials, as blinding a behavioural intervention cannot always be done. When to use CRTs can yield results that are relevant for making “real world” decisions. CRTs are often used in non-therapeutic intervention studies (e.g. change in practice guidelines). The advantages of CRT design in pharmacy research have been avoiding contamination and the generalizability of the results. A large CRT that studied physician–pharmacist collaborative management of hypertension is used in this manuscript as a CRT example. The trial, entitled Collaboration Among Pharmacists and physicians To Improve Outcomes Now (CAPTION), was implemented in primary care offices in the United States for hypertensive patients. Limitations CRT design limitations include the need for a large number of clusters, high costs, increased training, increased monitoring, and statistical complexity.