OBJECTIVES: Rheumatoid arthritis (RA) causes significant disability and health care expenditures.  Whilst RA disability RA is characterized, related cost data are sparse. An RA disability model from a UK biologics register was mapped to the General Practice Research Database (GPRD) (UK) and total health care costs estimated. METHODS: Step 1:   General linear modeling was used to predict HAQ score from RA cases from the biologics register.  Step 2: healthcare costs were estimated from RA cases with one year’s observation selected from the GPRD.  Costs (GBP2011) included prescriptions, consultations and investigations from primary and secondary care. Step 3: the association between healthcare costs and predicted HAQ was explored using Generalized linear modeling (Poisson distribution & log-link) and predicted costs were fitted to predicted disability. RESULTS: Step 1&hibar;Disability was modeled using 6129 cases, (75% female, mean age 57 years (sd 12), mean baseline HAQ  1.81 (0.72)). Six optimal HAQ models included: gender, age, smoking, BMI, prior DMARDs, joint replacements, systemic RA features, current DMARD and steroid treatment, ESR &/or CRP; current non-RA drugs, and co-morbidities. R2 values ranged from 0.277 (no inflammatory marker) to 0.331 (with CRP), while mean HAQ prediction error was 0.13 (0.27). Step 2&hibar;Costs were assessed in 8,423 GPRD cases with similar gender balance (69.8% female) but older mean 64 years). Predicted HAQ (pHAQ) scores ranged from 0.45 to 3.00, with a mean of 1.78 (0.31). Mean annual cost of care in this population was £2792 (£5633), with 46% from hospital admissions, 24% prescriptions, 20% GP consultations, 11% outpatient attendances, and 11% investigations. Step 3&hibar;Total costs (pCOST) were optimally predicted from EXP^(5.613+(Age*0.005)+(pHAQ*1.185)+(Age*pHAQ*-0.001)). pCOST has an exponential correlation to pHAQ (R2=0.986; EXP^(342+(pHAQ*1.188)). CONCLUSIONS:  Age-adjusted predicted total health care costs for RA patients increased exponentially across the range of estimated disability. Although higher than published estimates, exclusion of support-service costs suggests these values may be conservative.