Objective: Identifying incident rheumatoid arthritis (RA) is desirable in order to create inception cohorts. We evaluated an approach to identify incident RA in health plan claims data. Methods: Both Medicare and commercial claims data were linked to Corrona, a US RA registry. We evaluated the accuracy of year of RA onset in the registry (gold standard) versus different claims algorithms, varying International Classification of Diseases, Ninth Revision codes for RA/arthritis, duration of health plan enrollment preceding diagnosis (minimum of 1 versus 2 years), and use of RA medications. Results were reported as positive predictive values (PPVs) of the claims-based algorithm for incident RA. Results: Depending on the algorithm tested and whether patients were enrolled in Medicare or the commercial health plan, the PPVs for incident RA ranged from 68–81%. A 2-year clean period free of all RA-related diagnoses and medications was somewhat more optimal although, by comparison, a 1-year clean period yielded similar PPVs and retained approximately 90% more RA patients for analysis. Conclusion: Claims-based algorithms can accurately identify incident RA.