Abstract
Background: Computer simulation models are increasingly used in complex chronic disease to prioritize treatments, study the effects of treatments, and guide policy and management. The accuracy of these models is doubtful, particularly in real-world use.
Methods: We designed EBMI to ‘think’ as an evidence-based clinician thinks, by combining the best user-specified trial evidence about treatment effectiveness with the best possible individualized estimates of patient risk. Local validity is much more likely because EBMI’s risk functions can be estimated from local data. The use of local data also allows the model to be used for real-time clinical prioritization in local settings.
Results: EBMI reproduces Kaiser Permanente Northwest data very closely and, in its only validation to date against a clinical trial, closely replicated the Heart Protection Study. More validations and studies using the model are underway.
Conclusions: It is possible to design a simulator that is both more evidence-based and more locally valid than existing models. EBMI and its source code are now available to other researchers, clinicians and managers.
- Received September 11, 2008.
