OBJECTIVES: The purpose of this study was to develop a mapping model for estimating EuroQol 5D (EQ-5D) utility values from Short Form 36 (SF-36) scores in Japanese patients with rheumatoid arthritis (RA), with or without clinical characteristics. METHODS:  Linear regression models were applied to a cross-sectional data set of 112 patients with RA collected from a regional hospital in Niigata prefecture, Japan. Four model specifications were estimated, where EQ-5D was regressed on 1) eight SF-36 scores, 2) as per 1) plus squared and pair-wise interaction terms, 3) as per 1) plus clinical characteristics, and 4) as per 3) plus squared and pair-wise interaction terms, respectively. Model 2 and 4 were developed by using stepwise regression analyses. Model goodness of fit was examined by using Akaike information criterion (AIC), R2, and adjusted R2. Predictive performance was evaluated by using root mean square error (RMSE). RESULTS: Model 1 with eight SF-36 scores explained more than 59% of the variation in EQ-5D utility values. The best-performing model based on goodness of fit and predictive performance was model 4 (AIC=-200.4, R2=0.767, adjusted R2=0.709, RMSE=0.090). The model included four SF-36 scores (GH, MH, PF, and RP), five squared terms, twelve pair-wise interaction terms, and log transformed simplified disease activity index (SDAI). Also model 2, which included no clinical characteristics, had similar predictive ability (AIC=-195.0, R2=0.764, adjusted R2=0.699, RMSE=0.093). CONCLUSIONS: EQ-5D utility values can be predicted from SF-36 scores and SDAI with Japanese patients with RA. The mapping model can be applied to SF-36 datasets to produce utility scores for economic evaluation of RA treatments from the perspective of Japan healthcare system.