Identification of rheumatoid arthritis in German claims data using different algorithms: Validation by cross-sectional patient-reported survey data.

OBJECTIVE

To evaluate different algorithms for the identification of rheumatoid arthritis (RA) in claims data using patient-reported diagnosis as reference.

METHODS

Within longitudinal data from a large German statutory health insurance, we selected a random sample of persons with ICD-10 code for RA (M05/M06) in ≥2 quarters in 2013. The sample was stratified for age, sex, and M05/M06. Persons were asked to confirm RA diagnosis (gold standard), which was linked to claims data given consent. Analyses were weighted to represent the total RA population of the database. Positive predictive values (PPVs) and discriminative properties were calculated for different algorithms: ICD-10 code only, additional examination of inflammatory markers, prescription of specific medication, rheumatologist appointment, or combination of these.

RESULTS

Of 6193 persons with a claims diagnosis of RA, 3184 responded (51%). Overall, PPV for the ICD-10 code was 81% (95% confidence interval 79%-83%) with 94% (92%-95%) for M05 and 76% (73%-79%) for M06. PPVs increased (with loss of case numbers) if inflammatory markers (82% [80%-84%]), rheumatology visits (85% [82%-87%]) or specific medication (89% [87%-91%]) had been used in addition. Specific medication had the best discriminative properties (diagnostic odds ratio of 3.0) among persons with RA diagnosis.

CONCLUSIONS

The ICD-10 codes M05 and (less optimal) M06 have high PPVs and are valuable to identify RA in German claims data. Depending on the respective research question, researchers should use different criteria for identification of RA.