40

Epidemiologic data concerning colorectal cancer in France mainly comes from local registries that cover about 15% of the population. The Programme of medicalisation of information systems (PMSI) database records and archives the diagnosis and treatment provided during a patient’s hospital stay. In France, the chaining of different hospital stays for a given patient recently became possible and permits epidemiological studies to be carried out any given pathologies at the national level. However, evaluation of the quality of chained data has been limited to local experiences. This study hence aimed to evaluate the quality of chained data for colorectal cancer patients for the purpose of conducting epidemiological studies. All hospital stays, totalling 556,206, at the national level for the year 2003 of patients diagnosed with colorectal cancer (principal or secondary) were obtained and analysed. Chaining was defective for 6.47% of the stays. Those which were not chained came exclusively from public hospitals. The number of stays without a principal diagnosis was 134 (0.2%); 82 were from public hospitals (61.2%) and 52 from private ones (38.8%). Stays with at least one error totalled 6.58%. The chained data appears reliable; nevertheless, due to the stays which are not chained, the data should be used with caution.