Abstract: ObjectiveTo study the application of control graph alert technique in early warning for hand-foot-mouth disease(HFMD) in Hubei province and to provide evidences for pre-real-time prevention and control of HFMD in the province.MethodsData on reported HFMD cases in Hubei province from 2008 to 2014 were extracted from National Information Management System for Communicable Disease Reporting.Moving percentile method was used to detect outbreak or epidemic of HFMD.After calculating sensitivity,specificity,positive predictive value and negative predictive value,and receiver operating characteristic curve(ROC),the premium alert threshold for the disease was estimated.ResultsThe premium alert threshold for HFMD was 65th percentile(P₆₅) and with the threshold,a early warning model was established,with a sensitivity of 91.11%,a specificity of 100%,a positive predictive value of 100%,and a negative predictive value of 63.64%.ConclusionApplication of a premium alert threshold selection should consider local disease report and the characteristics of infectious diseases to upgrade early detection capability.