OBJECTIVES: To derive the optimal claims-based definition of knee osteoarthritis (KOA) using a National Health Insurance (NHI) database.

METHODS: 1,200 patients were randomly selected from 2 general hospitals using the list of patients aged 50 or older with ICD-10 M17 code for one year. Actual KOA was confirmed as a case of Kellgren and Lawrence grades 2 or higher through reviewing medical records and knee radiographs by 2 rheumatologists. NHI claims data for selected patients were extracted during the same period. We constructed 23 claim-based definitions based on diagnostic codes, procedure codes, and prescribed drugs: combinations of 5 criteria (1) 2 or claims with diagnostic code M17 for a year; (2) knee X-ray with one of diagnostic codes M15, M17, or M19; (3) 30 days or more of OA medication prescription with diagnostic code M17 in a year; (4) prescription of medications approved only for OA in Korea with diagnostic code M17; (5) knee surgery with diagnostic code M17. The validity was evaluated by sensitivity, specificity, accuracy, positive predictive value (PPV), and negative predictive value (NPV).

RESULTS: Of the 1,192 patients included in this validation study after deletion the same person, actual KOA was 975. Among definitions with accuracy of 0.8 or more, sensitivity of 0.8 or more, and specificity of 0.5 or more, the ones with large values were selected in the order of accuracy, sensitivity, and specificity. The optimal definition included knee X-ray due to KOA, and specific medications (etoricoxib, polmacoxib, NSAIDs-Herbal, glucosamine, diacerein, chondroitin sulfate, or hyaluronate injection). The accuracy, sensitivity, specificity, PPV, and NPV were 75% (95% CI, 72-77%), 81% (95% CI, 78-83%), 48% (95% CI, 41-55%), 87% (95% CI, 85-90%), 36% (95% CI, 30-42%), respectively.

CONCLUSIONS: Patients with KOA can be identified using a validated claims-based definition to generate real-world evidence in the NHI database.