BACKGROUND AND OBJECTIVES: The SNDS (‘Système National Des Données De Santé’) is a claims database covering 99% of the French population. This rich data source has been highlighted as a powerful tool for pharmacoepidemiology. In this study we assessed the feasibility of deriving data from SNDS for cost-effectiveness models. METHODS: We explored the feasibility to derive target input data for a cost-effectiveness model using the SNDS database. The hospital costs, primary care costs, drug costs, drug doses, intervention costs and time-to-event data were explored for a target population. RESULTS: Hospital discharge summaries database provides both admission/discharge dates and the cost data for each hospitalization, including standard and home hospitalization. The mathematical relationship between length-of-stay and hospital costs could be analyzed with such data. Primary care costs could be obtained from outpatient claims database. The price and the quantity of drugs dispensed through outpatient and hospital pharmacies were available. However, those drugs used within hospitalization were not recorded, except for expensive or ‘Temporary Authorisation for Use (ATU)’ drugs. Detailed information on drugs could be found and thus it was possible to estimate the doses for expensive or ATU drugs. In hospital discharge summaries database, only the costs of certain interventions specified within diagnosis-related-group system were available, instead of costs for any intervention. Also, interventions were generally under-reported in the hospital database. Instead, intervention costs in primary care setting could be identified directly from outpatient claims database. Time-to-event data, such as from certain intervention or diagnosis to death, could be estimated. However, the date of diagnoses were not reported and the admission date had to be used as a proxy for index date. CONCLUSIONS: SNDS is a powerful database to provide the real-world inputs for cost-effectiveness models, which gives more credibility to the model results than using assumptions.