Background: Cancer registries track the number of new cases diagnosed among residents of a specific geographic area but cover only 13% of the French population. Estimating the national incidence based on the ratio of incidence to observed mortality in areas covered by a registry, and national mortality rates, is imprecise for thyroid cancer. The utility of a national database other than mortality data, such as that of the Medical Information Systems Program, should be considered but requires an assessment of its predictive value and sensitivity. Methods: Based on a selection of hospital stays involving a diagnosis of thyroid cancer and a surgical procedure on the thyroid, the predictive value and sensitivity of the 1999 and 2000 Medical Information Systems Program for thyroid cancers were estimated by comparison with the database of 10 cancer registries. The process of cross-referencing the two databases was carried out using nominative data transmitted by the relevant institutions. Using the registry database, individuals not recorded in the selection of the medicalization program of the information systems were compared to those recorded by the program using logistic regression. Results: Among the 973 individuals identified by the medical information system program, 866 corresponded to cases of thyroid cancer diagnosed in 1999 or 2000—true positives—and 107 to false positives. The positive predictive value was thus 89% (95% confidence interval: 87–91%) and did not differ significantly across departments (p = 0.80). According to the registry database, 322 cases diagnosed in 1999 or 2000 were not selected by the medicalization program of the information systems. Sensitivity was thus 73% (70–76%) and varied by the type of facility where the surgical procedure was performed (p < 0.01) and by department (p < 0.001). Conclusion: The positive predictive value of the 1999 and 2000 medical information system program for identifying incident thyroid cancers is high and consistent across departments. However, its sensitivity is lower and varies by facility category and by department, which limits the usefulness of this database for estimating the incidence of thyroid cancer in France. (R.A.)