OBJECTIVES: For assessing the long-term health economic impact of obesity programs, modelling techniques are frequently applied to project obesity-associated events over time. To obtain information on the predictive quality of the applied event simulations, which ensures the acceptance of results from economic evaluations, published external event validation approaches were identified, reviewed and analyzed.

METHODS: A systematic review was performed in Pubmed and NHSEED to identify decision models for full health economic assessments (HEA) in obesity. For each included study we extracted information on the external event validation approach, using the best practice recommendations of the report on “Model Transparency and Validation” issued by the ISPOR-SMDM Modeling Good Research Practices Task Force.

RESULTS: We identified 87 papers and 83% (72 of 87) simulated obesity-associated events. Only ten models (≈11%) performed an external validation and only for one the predictiveness of the event simulation was investigated in a cohort of obese subjects. Considering other ISPOR best practice criteria we have found that for none of these external validation cases a systematic identification of suitable data sources was performed, and that a justification of the data source selection, due to predefined criteria, was identified only in three cases. However an adequate result presentation (simulation results provided for each source, presentation of discrepancies, and a qualitative measure on fit) was provided for most external validation cases (≥ eight cases).

CONCLUSIONS: We have found that only a limited number of published decision models for full HEAs in obesity have applied an external event validation. In addition, those who conducted external validation suffers from major limitations including the data source selection process, as only in one case, obesity cohorts were used as basis for the validation procedure. Thus in conclusion published obesity models lack information on the predictive quality of the applied event simulation approaches.