Prediction and early warning for HCV:comparison of three methods

Objective To apply and compare different models in prediction of the trend of hepatitis C virus(HCV)epidemic in Jiangsu province and to provide bases for the control of HCV epidemic.Methods The data on infectious diseases report in Jiangsu province form 1997 to 2009 were collected.Markov chain model and Hole-Winters exponential smoothing forecasting model were used to predict epidemics and control chart was used to detect outbreaks or epidemics of HCV infection.Results Markov chain model predicted that the incidence of HCV from September to December in 2009 would be more than 120 cases and the prediction was consistent with the case number reported.Hole-Winters-additive model(α=0.46,β=0.001,γ=0.001)had good performance in predicting HCV epidemic,with the root mean square error(RMSE)and mean absolute percentage error(MAPE)of 12.6 and 19.1%,respectively.A five-year control chart was chosen as an essential method in early warning system.The early warning threshold for HCV epidemic was 70 percentile and at the threshold level,the sensitivity and specificity for HCV epidemic prediction were over 90%.Conclusion Markov chain model and Hole-Winters exponential smoothing forecasting model are good method for prediction of HCV epidemic and control chart can be used in the early warning.