Background: For economic models to be considered fit for purpose, it is vital that their outputs can be interpreted with confidence by clinicians, budget holders and other stakeholders. Consequently, thorough validation of models should be carried out to enhance confidence in their predictions. Here, we present results of external dependent and independent validations of the Core Obesity Model (COM), which was developed to assess the cost-effectiveness of weight management interventions.
Objective: The aim was to assess the external validity of the COM (version 6.1), in line with best practice guidance from the International Society for Pharmacoeconomics and Outcomes Research and the Society for Medical Decision Making.
Methods: For validation, suitable sources and outcomes were identified, and used to populate the COM with relevant inputs to allow prediction of study outcomes. Study characteristics were entered into the COM to replicate either the studies used to develop the model (dependent validation) or those not included in the model (independent validation). The concordance between predicted and observed outcomes was then assessed using established statistical methods and generation of mean error estimates.
Results: For most outcomes, the predictions of the COM showed good linear correlation with observed outcomes, as evidenced by the high coefficients of determination (R2 values). The independent validation revealed a degree of underestimation in predictions of cardiovascular (CV) disease and mortality, and type 2 diabetes.
Conclusion: The predictions generated by the risk equations used in the COM showed good concordance both with the studies used to develop the model and with studies not included in the model. In particular, the concordance observed in the external dependent validation suggests that the COM accurately predicts obesity-related event rates observed in the studies used to develop the model. However, the impact of existing CV risk, as well as mortality, is a key area for future refinement of the COM. Our results should increase confidence in the estimates derived from the COM and reduce uncertainty associated with analyses using this model.