Abstract
BACKGROUND: The Norwegian Patient Registry is a valuable source for nationwide data about different diseases and disorders, hence it is used frequently by researchers for epidemiological studies. However, the data quality varies greatly between different conditions/diseases, and how the data are combined and handled is therefore not reliable without knowledge about data validity. The aim of the current study was to validate the medical diagnosis code for forearm fracture (ICD-10: S52.X) in patients over 18 years at Oslo University Hospital (OUS) Ullevål and Skadelegevakten from 2nd February through 1st March 2015. METHOD: Registrations in DIPS and x-ray descriptions were collected independently. The number of actual fractures captured in DIPS were calculated. Then, we looked at every patient diagnosed with a forearm fracture medical diagnose code and compared them with the corresponding X-ray description to verify the fracture. Afterwards, we calculated the validity (completeness and correctness) of the diagnose codes. RESULTS: There were 250 registration who met the inclusion criteria. They constituted of 194 (77.6%) true positives, 48 (19.2%) false positives, and 8 (3.2%) false negative. This resulted in a completeness (sensitivity) and correctness (positive predictive value) of 96.0% and 82.4%, respectively. Furthermore, the results showed that there were 30 patients with two registrations, which were due to coding error or operations. As a result, the correctness increased to 86.8% when only allowing one registration per patient. However, when restricting to only distal forearm fracture codes (S52.5 and S52.6), the correctness decreased to 75.2% (83.0% when only allowing one registration per patient). CONCLUSION: This study found that the completeness and correctness of the S52 codes is high, and the correctness became even higher when only allowing one S52.x registration per patient for a defined time. Additionally, restricting to distal forearm fracture codes (S52.5 and S52.6) will result in a slightly lower correctness.