Abstract
This chapter discusses validation of simulation models used to inform health policy. Confidence in a model’s validity can be weaker or stronger depending on several factors. These factors include verifying whether model specifications were implemented correctly, evaluating the extent to which model-predicted results are consistent with empirical results, and examining whether model predictions are robust to alternative structural assumptions. Systematic evaluation of these factors can be used to gauge the extent to which a model is validated for a given application. It reviews types of validation, discusses the related concepts of calibration and nonidentifiability, takes a deeper dive into cancer model validation studies, and concludes with questions that consumers of models should ask (and modelers should answer) to inform judgment about a model’s fitness for purpose. Final judgments about when model results can be trusted ultimately rely on the evolving understanding of the disease and intervention effects, available data relevant to the application, and access to reporting of model validation exercises.