Objective: To investigate the risk factors for severe hand, foot and mouth disease(HFMD) by using univariate and multivariate logistic regression model combined with receiver operating characteristic(ROC) curve,and to assess its predictive value in the early identification and treatment of HFMD.Methods: One hundred and two HFMD cases were selectedfor a retrospective analysis,including 46 ordinary patients as general group,54 severe cases as severe group and 2 critically ill patients as critical group. Univariate analysis was used to analyze the difference of clinical manifestations, signs, and laboratory test results between the two groups. Multivariate logistic regression analysis was used to analyze the independent risk factors of severe HFMD. ROC curve was used to evaluate their predictive value. Results: The differences were statistically significant(P<0.05) between the two groups in body temperature, fever, mental status, lethargy, easily frightened, limb shaking, vomiting, respiratory rate, heart rate and fasting blood glucose. Binary logistic regression analysis showed that easily frightened, limb shaking, fast heart rate were the independent risk factors for severe HFMD. Combination of the three risk factors was the better prediction for severe HFMD than any single risk factor.Conclusion: Appearance of easily frightened, limb shaking and fast heart rate in HFMD patients may be the symptoms of severe HFMD and timely interventions should be taken.