Introduction. This study aims to assess the quality of diabetes diagnosis identification during hospital stays related to diabetes complications by cross-referencing the PMSI database with survey data on the treated diabetic population. Methods. For the 2007 Entred survey, 8,926 adults with diabetes who were reimbursed for at least three antidiabetic medications between August 2006 and July 2007 were randomly selected from health insurance databases. Hospital stays were extracted from the PMSI for 7,534 individuals (84%) who did not refuse to participate in Entred and reconstructed for the period from August 2006 to July 2007. Depending on the type of complication, diagnoses were sought among the primary, related, and/or associated diagnoses, and diabetes was examined across all diagnoses. Results. Nearly one-third of people with diabetes in Entred (mean age 65 years) were hospitalized at least once during the year. Heart disease (hospitalization rate: 30 patients, including 10 with heart failure and 22 with ischemic heart disease), chronic kidney disease (13), and cancer (22) were the most common diagnoses. Diabetes was mentioned in 53% of hospital stays and more frequently during stays for foot ulcers (95%), lower limb amputation (94%), stroke or transient ischemic attack (79%), ischemic heart disease (71%), and chronic kidney disease (42%). These reasons accounted for 15% of all hospital stays and included a diagnosis of diabetes in 62% of cases. Furthermore, diabetes was mentioned in 78% of hospital stays for heart failure and 18% for cancer. Discussion and conclusion. Among treated diabetic patients, diabetes does not appear, erroneously, as a diagnosis in 38% of hospital stays that nevertheless mention a complication of diabetes, which highlights the limitations of the PMSI for the epidemiological surveillance of certain complications. Cross-referencing data from the diabetic population survey (Entred) with PMSI data therefore provides real value compared to PMSI data alone. (R. A.)