BACKGROUND: DRG-based management of public hospitals in France involves the use of standardised discharge abstracts for the "Medicalisation of Information Systems Program". METHODS: To assess the quality of the medical information in these abstracts, a sample of 649 abstracts for 1994 was collected from the Hospices Civils de Lyon's data base. To validate the information in these abstracts, we reviewed the medical records of each patient. RESULTS: The results showed an error rate of 32% (CI: 28-36) for the diagnosis-related group and an error rate of 40% (CI: 36-44) for the principal diagnosis. There was no significant difference between these error rates and the calculation of "Indices Synthétiques d'Activité" (French system for attributing points to hospital stays according to DRGs categories). CONCLUSIONS: The quality of the medical information for the "Medicalisation of Information Systems Program" remains a major challenge not only for budget allocation, but also for the study of the case-mix in hospitals.
BACKGROUND: DRG-based management of public hospitals in France involves the use of standardised discharge abstracts for the "Medicalisation of Information Systems Program". METHODS: To assess the quality of the medical information in these abstracts, a sample of 649 abstracts for 1994 was collected from the Hospices Civils de Lyon's data base. To validate the information in these abstracts, we reviewed the medical records of each patient. RESULTS: The results showed an error rate of 32% (CI: 28-36) for the diagnosis-related group and an error rate of 40% (CI: 36-44) for the principal diagnosis. There was no significant difference between these error rates and the calculation of "Indices Synthétiques d'Activité" (French system for attributing points to hospital stays according to DRGs categories). CONCLUSIONS: The quality of the medical information for the "Medicalisation of Information Systems Program" remains a major challenge not only for budget allocation, but also for the study of the case-mix in hospitals.