Background. In France, cancer registries cover 18% of the population. Certain cancers, such as central nervous system (CNS) tumors, which may be linked to environmental factors, warrant national surveillance to detect abnormal fluctuations in incidence. The PMSI database is of interest due to its comprehensive, standardized nature and mandatory reporting by all healthcare facilities. The objective of this study was to define methods for identifying incident CNS tumors using PMSI data. Methods. A selection of patients from the Gironde region was made from the 2004 PMSI database of the Bordeaux University Hospital, based on CNS tumor codes. These were validated against the 2004 Gironde registry of primary CNS tumors, used as a reference, or via medical records. Several combinations of criteria were defined and tested. Results. The initial selection, based on diagnoses, identified patients with a sensitivity of 84% and a positive predictive value (PPV) of 34%. Patients incorrectly identified by the PMSI were non-incident cases (49%) or patients without a CNS tumor (45%). Patients not identified by the PMSI had been hospitalized in 2005 (44%) or did not have a CNS tumor code (42%). Depending on the algorithms, sensitivity ranged from 64% to 84%, and PPV from 34% to 69%. The best compromise, with a sensitivity of 67% and a PPV of 69%, was achieved using CNS tumor codes from 2004, with no code in previous years and no CNS metastasis code in 2004, combined with a diagnostic or therapeutic procedure in patients under 70 years of age. Conclusion. These results do not support the use of PMSI data alone to calculate the incidence of these various tumors. However, in conjunction with other sources, the PMSI plays an important role in cancer surveillance, complementing the expertise of cancer registries. This role may grow as further analysis is conducted and data quality improves. (R.A.)