OBJECTIVE: The aim of the present study was to examine the validity of local and national electronic databases using medical records as gold Standard. STUDY DESIGN AND SETTING: All hospital admissions with ICD 9-code 820.X (hip fracture) in a 1-year period were identified in the electronic discharge registers of the hospitals in Oslo and in the national electronic database (The Norwegian Patient Register). Medical records for all patients identified by the discharge registers and the logbooks of the operating theater of the hospitals were retrieved, and the diagnosis was verified. RESULTS: Compared with the total number of fractures confirmed in medical records, the electronic discharge register of one of the hospitals underestimated the number of fractures by 46%, whereas the two other overestimated the number by 17% and 19%. For the national electronic database, an overall overestimation of 19% was found. CONCLUSION: The present findings question the validity of electronic databases and thus have implications for epidemiologic studies.
OBJECTIVE: The aim of the present study was to examine the validity of local and national electronic databases using medical records as gold Standard. STUDY DESIGN AND SETTING: All hospital admissions with ICD 9-code 820.X (hip fracture) in a 1-year period were identified in the electronic discharge registers of the hospitals in Oslo and in the national electronic database (The Norwegian Patient Register). Medical records for all patients identified by the discharge registers and the logbooks of the operating theater of the hospitals were retrieved, and the diagnosis was verified. RESULTS: Compared with the total number of fractures confirmed in medical records, the electronic discharge register of one of the hospitals underestimated the number of fractures by 46%, whereas the two other overestimated the number by 17% and 19%. For the national electronic database, an overall overestimation of 19% was found. CONCLUSION: The present findings question the validity of electronic databases and thus have implications for epidemiologic studies.