OBJECTIVES: To compare the two risk adjustment models when estimating healthcare costs of rheumatoid arthritis (RA) patients. METHODS: Continuously eligible adult patients with confirmed diagnoses of RA between June 2004 and June 2009 were included.  Patients were new to tumor necrosis factor (TNF) therapy and subsequently either switched to another anti-TNF or escalated their dose. The difference in total health care costs and RA-related health care costs between switchers and escalators 1 year after the switch/escalation were estimated using the propensity score matching and instrumental variable methods. When using propensity score matching, the differences in patient, clinical, cost and utilization characteristics during the baseline period were controlled. When using the instrumental variable method, patients' copayment, distance from patient to provider, and doctors’ prescribing patterns were used as instruments to estimate the outcomes, while other differences in patient and clinical characteristics were controlled as well. RESULTS: After risk adjustment using propensity score matching, the difference between switchers and escalators is $648 in total healthcare costs, and $245 in RA-related health care costs. After using the instrumental variable method, the difference between switchers and escalators is $2054 in total health care costs and $2809 in RA-related health care costs. CONCLUSIONS:  After adjusting for patient, clinical and demographic characteristics, the choice of risk adjustment method affected the results. In this study, the cost burden is higher for switchers when using the instrumental variable method as the risk adjustment technique.