OBJECTIVES: Health care claims databases do not contain information about disease severity. The goal of this study was to develop a severity index for rheumatoid arthritis (SIFRA) for private health care claims data using a previously developed claims-based index from the Veteran’s Administration (VA) Health System and rheumatoid arthritis medical records-based index of severity (RARBIS). METHODS: We extracted the following variables related to rheumatoid arthritis from the claims data: total number of synthetic disease-modifying anti-rheumatic drugs (DMARDs), total number of biological DMARDs, tests for C-reactive protein (CRP) and erythrocyte sedimentation rate (ESR) ordered, rehabilitation visits, rheumatology visits, Felty’s syndrome and Sjorgen’s syndrome, pulmonary, soft tissue nodules, joint surgery, number of platelet counts and chemical panels ordered, and rheumatoid factors testing. A linear regression model was used to create the severity score. The severity score was compared with RARBIS and currently-used comorbidity scores to proxy severity in outcomes research studies related with rheumatoid arthritis. RESULTS:   According to the Akaike Information Criterion (AIC), Bayesian Information Criterion (BIC), log likelihood function, R-squared values and average squared prediction error, SIFRA performed better than RARBIS, Charlson Comorbidity Score (CCI),  Elixhauser comorbidity score and Chronic disease score. Spearman correlation with RARBIS was 0.65 and significant. However, the correlation with the Charlson Comorbidity Index (0.1, p=0.6521), Elixhauser Index (0.15, p=0.5312) and Chronic disease score (0.13, p=0.6011) were low and insignificant. CONCLUSIONS: Controlling disease severity is crucial in retrospective studies.  Comorbidity scores are inadequate to be proxy variable. SIFRA, at least for rheumatoid arthritis, controls for disease severity better than any other commonly used measure.