MS3 Reproduction of Published Health Economic Obesity Models: Assessment of Facilitators, Hurdles and Barriers

Abstract

This research aims to: (i) reprogram published health economics models in obesity; (ii) compare the reproduction results to the original; (iii) determine reproduction facilitators and hurdles; (IV) suggest model replication reporting standards to enhance model reproducibility. Using a published systematic review, high quality economic models were selected by applying published structural quality criteria. Of those all models simulating an adult UK population were selected for reproduction in TreeAge Pro 2020. The replication results were then compared to the original results by presenting the related minimum, maximum and mean percentage deviations relative to the original, focusing on cost-effectiveness outcomes. Facilitators and hurdles were determined and documented during each pre-defined step of model reprogramming and transferred to replication reporting standards. Four obesity models were rebuilt; all were state transition models, applied a UK NHS perspective and a quality adjusted life years (QALYs) approach. Comparing original and replication outcomes, the following range of deviations was observed; costs: -3.9% to 16.1% (mean over all model simulations: 3.78%), QALYs: -3.7% to 2.1% (mean: -0.11%), and average cost-effectiveness ratio: -3.0 to 17.9% (mean 4.28%). The incremental cost-effectiveness ratio showed stronger variations, although the same cost-effectiveness conclusions as for the original were reached. Key reproduction facilitators were input data tables, model diagrams including state transitions, descriptions of underlying equations and outcome presentations including clinical event frequencies. Key reproduction hurdles were missing input population parameters, missing standard deviations to estimate distribution parameters for probabilistic analysis and missing formulas for equations based on own calculations. It was possible to rebuild all identified health economic obesity models, but depending on the model reporting, minor to major assumptions were needed. Model replications can help to identify how transparently a health economic model was documented and can be a first step in to validate current model reporting practices.