Some patients with undifferentiated arthritis (UA) experience spontaneous remission; however, one-third of patients progress to rheumatoid arthritis (RA) in the final process of the disease. This study evaluated clinical variables in order to find a prediction model that could predict the development of RA in patients with UA. The medical records of 164 patients, who were initially diagnosed with undifferentiated arthritis in Yonsei University Medical Center from January 2004 to December 2007, were retrospectively reviewed. They were followed up for at least 6 months. The clinical variables related to the development of RA were identified by univariate analyses. Using logistic regression analysis, the prediction model was made and the diagnostic performance of the model was evaluated. Thirty-two patients of the 164 total patients progressed to RA during the follow-up period. The prediction model was composed of clinical factors including the duration of morning stiffness, the number of tender joints, the number of swollen joints, C-reactive protein level, rheumatoid factor, anti-cyclic citrullinated peptide antibody, and erosive change on baseline X-ray. The prediction score ranged from 0 to 10. All of the patients with a higher prediction score greater than five experienced RA progression. The area under the curve value for the prediction rule was 0.976. The prediction model could predict progression to RA in patients with UA. It especially helps the clinician to decide on a management plan for patients with a high prediction score.