Cluster-randomized design for simulation-based evaluation of complex healthcare interventions

Abstract

In this paper we discuss the application of cluster-randomized intervention studies to simulation-based evaluation of surgical care policies, arguing that the methodological rigour of evaluative studies should be applied to the design and analysis of simulation experiments. We introduce a framework and study design to evaluate methods for improving the surgical care process with the use of patient flow models that simulate the steps in service delivery and response pathways for individual patients. Because patient-level outcomes in a given simulation run may be correlated, we suggest a cluster-randomized design of experiment for determining how many simulation runs are required and how input factors should vary across the runs. In such a design, simulation runs rather than simulated individuals are randomized to different study groups. For patient outcomes that vary more across simulation runs than within each run, we provide formulas to be adapted in sample size calculations to allow for clustering of responses. As an example, we report on the design and analysis of a simulation study comparing two methods of booking admission dates for patients who are to undergo elective surgery.