Background: Clinical outcomes of undifferentiated arthritis (UA) are diverse, and only 40% of patients with UA develop rheumatoid arthritis (RA) after 3 years. Discovering predictive markers at disease onset for further intervention is critical. Therefore, our objective was to analyze the clinical outcomes of UA and ascertain the predictors for RA development. Methods: We performed a prospective, multi-center study from January 2013 to October 2016 among Chinese patients diagnosed with UA in 22 tertiary-care hospitals. Clinical and serological parameters were obtained at recruitment. Follow-up was undertaken in all patients every 12 weeks for 2 years. Predictive factors of disease progression were identified using multivariate Cox proportional hazards regression. Results: A total of 234 patients were recruited in this study, and 17 (7.3%) patients failed to follow up during the study. Among the 217 patients who completed the study, 83 (38.2%) patients went into remission. UA patients who developed RA had a higher rheumatoid factor (RF)-positivity (42.9% vs. 16.8%, chi(2) = 8.228, P = 0.008), anti-cyclic citrullinated peptide (CCP) antibody-positivity (66.7% vs. 10.7%, chi(2) = 43.897, P < 0.001), and double-positivity rate of RF and anti-CCP antibody (38.1% vs. 4.1%, chi(2) = 32.131, P < 0.001) than those who did not. Anti-CCP antibody but not RF was an independent predictor for RA development (hazard ratio 18.017, 95% confidence interval: 5.803-55.938; P < 0.001). Conclusion: As an independent predictor of RA, anti-CCP antibody should be tested at disease onset in all patients with UA.