Cluster-Randomized Crossover Experiment

A Cluster-Randomized Crossover Experiment is a cluster randomized experiment that is a crossover experiment.

• AKA: Cluster-Crossover Study, Crossover CRT.
• Context:
  • It can range from being a Categorical Outcome Cluster Randomized Crossover Experiment to being a Continuous Outcome Cluster Randomized Crossover Experiment.
  • It can be analyzed by a Cluster-Randomized Crossover Experiment Evaluation Task.
  • It can be characterized by </math>(C \times P, N)</math>, where:
    • [math]\displaystyle{ C(c=1,…,C) }[/math] is the number of cluster randomized experiment clusters.
    • [math]\displaystyle{ P(p=1,…,P) }[/math] is the number of randomized crossover experiment randomizations (times each cluster is randomized during the trial).
    • [math]\displaystyle{ N(i=1,…,N) }[/math] is the total number of subjects enrolled in the study.
  • …
• Example(s):
  • a Continuous Outcome Cluster Randomized 2x2 Crossover Experiment, such as eEPC Lift Evaluation 2x2 Crossover Experiment.
  • a Crossover-based Randomized eEPC Predictor Lift Analysis Experiment.
  • the Parent Baby Interaction Programme (PBIP) (Johnson, Whitelaw et al., 2009) http://www.newton.ac.uk/programmes/DAE/seminars/2011081516351.pdf
  • compare 2 policies for ordering chest X-rays for ventilated intensive care patients (Lancet 2009; 374:1687-93)
  • evaluate real-time audio-visual feedback about cardio-pulmonary resuscitation performed outside hospital (BMJ 2011; 342: d512)
• Counter-Example(s):
  • a Crossover Subject-level Randomized Experiment.
  • a Non-Crossover Group Randomized Experiment.
  • a Quasi Experimental Study.
  • a Stepped Wedge Cluster Randomized Trial.
• See: Randomization Algorithm.

---

---

References

2007

• (Parienti & Kuss, 2007) ⇒ Jean-Jacques Parienti, and Oliver Kuss. (2007). “Cluster-Crossover Design: A method for limiting clusters level effect in community-intervention studies.” In: Contemporary Clinical Trials, 28(3). doi:10.1016/j.cct.2006.10.004
  • QUOTE: The cluster-crossover design can be used for clinical trials comparing two or more interventions in a naturally formed study population, i.e. a cluster. This design differs from that of a crossover study, in that the treatment sequence is allocated at the cluster level. A cluster-crossover study can thus be considered as a cluster-randomized controlled study with additional periodic cluster-randomization(s) or treatment permutation(s) during the study. The data must be analyzed with hierarchical models with random effects in order to allow for different outcome probabilities in each period, cluster and cluster-period. Original data from two published field studies of hospital infection control based on this design are used here to illustrate the impact of different statistical models on the interpretation of the results. …

    .... Three characteristics, designated (C × P, N), describe a cluster-crossover design: C ([math]\displaystyle{ c= 1,…,C }[/math]) is the number of clusters participating in the trial. … [math]\displaystyle{ P(p= 1,…,P) }[/math] is the number of times each cluster is randomized during the trial. … N (i = 1,…, N) is the total number of subjects enrolled in the study. …

    … Quasi-experimental studies are defined as observational studies in which the study factor is manipulated without randomization [7]. Before–after studies can be considered as (1 × 2) cluster-crossover studies, but the order is always predetermined (no intervention, followed by the intervention). However, more sophisticated designs with untreated control groups, pretests and replications can strongly resemble cluster-crossover studies, particularly when replications are repeated several times. The main difference between quasi-experimental studies and cluster-crossover studies is the presence of group-randomization in the latter.

2006

• (Hussey & Hughes, 2006) ⇒ Hussey MA, and Hughes JP. (2006). “Design and analysis of stepped wedge cluster randomized trials.” In: Contemp Clin Trials.
• (Turner et al., 2006) ⇒ Turner RM, White IR, and Croudace T. (2006). “Analysis of cluster randomized cross-over trial data: a comparison of methods.” In: Stat Med.

---