OBJECTIVE: Disease models are of increasing interest and influence on decision support, outcomes research and health technology assessment in Europe. Validity and reliability of the incorporated medical evidence in arithmetical modeling software is crucial for the simulated outcomes. Currently the diabetes modeling software Accusim(r)(tm), first established in 1996, is being validated for the 2002 edition. METHODS: Input validity: The Australian MERGE checklist system is used for the literature assessment in order to assess the level of evidence and the likelihood of bias. Process validity: The model is based on Markov processes and has been published several times in peer-review journals. Outcome validity: The model was run against published population based diabetes data and is under constant review by independent medical experts. Currently an advisory board is being established. RESULTS: When calculating the expected life expectancy for representative type-2 diabetes patients aged 65 years against published American diabetes data the computed life expectancy of the model is 12.1 years versus 12.3 years published (2% deviation). The accuracy of simulated life expectancy decreases in younger patients with computed 17.4 years versus published 18.9 years in 55 year-old patients (8% deviation). From the literature screening, out of a total of 3512 literature references, 1007 were eligible for structured assessment, only 85 references (2.4% of all) were eligible for being used for the modeling software. CONCLUSION: Validation is a labour intensive and continuous process. Methods of evidence based medicine are needed for supporting the development and validation of disease models. Yet only few references meet the quality criteria for disease modeling. Disease models are able to compute reliable results in realistic scenarios. Disease modeling is important for outcomes research, monitoring and evaluation of disease management programmes, as well as for risk stratification and health technology assessment.