Simulation Trials for Evaluating Clinical Decision Support Systems

The evaluation of clinical decision support systems (CDSS) during randomized clinical trials in very long and complex to set up. Simulation trials consist of evaluating the CDSS on fictitious patients. While they cannot entirely replace clinical trials, they are simpler to perform and offer additional possibilities. In this paper, we present recommendations for the design of simulation trials evaluating CDSS, based on the lessons learned during two recent studies. We also propose a model of protocol that maximizes statistical power and avoids several common biases, and an R code template for statistical analysis using linear mixed models.