Sensitivity analysis for healthcare models fitted to data by statistical methods

After fitting complex models to data using statistical methods, a sensitivity analysis can be carried out. This determines which parts of a model are causing the bulk of the uncertainty in the model predictions (model ‘output’), and is a decision-support tool for the modeller who contemplates refining a model further or collecting additional data. A simple methodology for carrying out a sensitivity analysis is described. It is envisaged that such a relatively quick insight-generating step would precede the use of a more formal decision-theoretic approach that would address specific questions. Its use is illustrated using a model for breast cancer screening previously published in this journal. A simpler 3-parameter screening model is used in a simulation study of the error of the method as a function of sample size.