Abstract: Objective To explore appropriate models for making early warning of hand-foot-mouth disease (HFMD) in Qianjiang city of Hubei province and to provide technical support for the automatic warning of disease surveillance system.Methods Data on HFMD from April 1,2013 to December 31,2014 in Qianjiang and simulated outbreaks were used to test the warning efficacy of Shewhart,moving average (MA),exponential weighted moving average (EWMA),cumulative sums (CUSUM) and early abreation reporting system (EARS'C1,C2,C3) time-series models.The warning efficacy of each warning model was compared using average detection time and the Youden's index with different simulated outbreak data.Results With an increase in intensity of simulated outbreak,the Youden's index of warning model increases but the average detection time decreases; the Youden's index and average detection time increase with the increment of duration of simulated outbreaks.Based on comprehensive analyses on 9 patterns of simulated outbreaks with various intensity and duration,every model could reach the best efficacy when adopting 0.5 as the warning threshhold.The Youden's index and the average detection time were 0.57 and 1.12 days for Shewhart model,0.58 and 2.62 days for MA,0.44 and 3.70 days for EWMA,0.46 and 3.56 days for CUSUM,0.59 and 2.43 days for EARS'C1,0.57 and 2.47 days for EARS'C2,and 0.55 and 2.20 days for EARS'C3,respectively.Conclusion Among the seven warning models evaluated,Shewhart model is more appropriate for early warning of HFMD in Qianjiang city.