Objectives: The primary objective of this study was to evaluate the feasibility of using the national PMSI database to estimate the incidence of stroke in France by measuring its positive predictive value (PPV). Methods: A retrospective multicenter study involving 31 hospitals, each of which randomly selected 56 hospital stays from among those caused by a stroke or transient ischemic attack (TIA), coded according to the International Classification of Diseases, 10th Revision (ICD-10), and occurring during 2009 and 2010. Three algorithms were evaluated: - Algorithm 1 selected hospital stays for which one of the medical unit summaries (MUS) listed as the primary diagnosis one of the following ICD-10 codes: I60 (subarachnoid hemorrhage); I61 (intracerebral hemorrhage); I62.9 (non-traumatic intracranial hemorrhage, unspecified); I63 (cerebral infarction); I64 (stroke, unspecified mechanism); G45 (transient ischemic attack and related syndromes); G46 (cerebral vascular syndromes); G81 (hemiplegia); - Algorithm 2 selected hospital stays with a primary diagnosis of stroke in the anonymous discharge summaries (ADS); - Algorithm 3 applied the same logic as Algorithm 1, but considered the code assigned by the most specialized department that treated the patient. All records were reviewed, in each hospital, by an investigator neurologist. The PPV of each algorithm was measured using the diagnosis confirmed by the investigator neurologist as the reference diagnosis. Results: Ultimately, 1,669 randomly selected records were included in the analysis, of which 196 were false positives. The overall PPV was 88.3% for Algorithm 1, 90.0% for Algorithm 2, and 92.7% for Algorithm 3. To achieve maximum completeness, Algorithm 1 appears to be the most appropriate. Conclusion: It is possible to develop an algorithm that improves the PMSI’s ability to identify stroke and TIA with a PPV greater than 88%. The PMSI could therefore serve as a tool for measuring the burden of stroke in France, provided that the measurement of PPV is combined with a measurement of sensitivity, which will be the subject of future work using the three specialized stroke population registries in France as a reference.