Parameters Setting On Hand-foot-and-mouth Cluster Detection In Prospective Space-time Scan

Objective： To analyze the prevalence of hand, foot and mouth diseasecharacteristics in Qingdao in2009, provide the basis for the development ofprevention and control measures. the space-time statistic was tested by analysis ofHFMD data, the length of baseline and size of scan window and number of thewarning signal were taken as variable parameter, we use statistical methods tocheck the warning effect of each parameter of different values and select theoptimal parameter, provide technical support for the early warning of outbreak ofHFMD in other areas.Methods：All the HFMD cases and real outbreak in Shandong during2009, werecollected from the “Disease surveillance information reporting system” and“public health emergency events surveillance system”. Taking sensitivity andpositive prediction ratio and detection time interval as evaluation indices，takingbaseline and scan window and number of warning single as parameter, theprospective space-time scan statistic was tested by mimicking daily prospectiveanalyses of HFMD data of Qingdao city in June2009.We compared thetwenty-four groups performances under four kinds of different parametersResults: In2009,the total number of hand, and mouth disease cases was15323inQingdao. The overall incidence was23.30per1000.It occurred from April toaugust.All cases were distributed wide over12counties of Qingdao. HFMD hasobviously seasonality with unimodal distribution. The incidence of each countywas different. The high incidence areas mainly located in Jimo, Licang and Laixi,incidence was36.91per10000,34.82per10000and34.14per10000respectively.The high number of cases areas mainly located Jomo,Pingdu and Laixi, thenumber of cases were3318,2549and1968.Taking time window of three days and the length baseline of cases in the last21days，we could get the best performance. At the same，we found that the settingparameters of the candidate areas had no statistically significant effect onprospective space-time scan statistics method in early warning.The warning effect of different time window values: Time window for3daysand5days of the positive predictive value, sensitivity and detecting time intervalfor matching T test. Results show that the positive predictive value (T=2.641, P=0.012) between different time window values was statistical significance， sensitivity (T=0.167, P=0.167) and detecting time interval (T=1.568, P=1.568)were no statistical significance, the positive predictive value of timewindow for3days is greater than the time window for5days.The warning effect of different space window values: space window for3percentof the total population and5percent of the total population of the positivepredictive value, sensitivity and detecting time interval for matching T test.Results show that three evaluation indices were no statistical significancerespectively.The warning effect of different length of baseline: the length baseline ofcases in the last14days and the length baseline of cases in the last21days ofpositive predictive value, sensitivity and detecting time interval for matching Ttest. Results show that the positive predictive value (T=9.86, P <0.001),sensitivity (T=8.704, P <0.001) was statistical significance, there was nostatistical significance in positive predictive value and detecting time interval.Thelonger length of baseline, the higher sensitivity.The warning effect of different maximum number of early warning signal:Maximum early warning signal number of1,2,3respectively of the positivepredictive value, sensitivity and detection of time interval for compatibility ofvariance analysis.Results show that positive predictive value (F=40.050, p <0.001), sensitivity (F=40.351, p <40.351) and detecting time interval (F=26.538, p <0.001) were statistical significance.The maximum number of warningsignal is inversely proportional to the positive predictive value, and proportionalto the sensitivity and detection time interval.Conclusion：1.The hand，foot and disease has obvious aggregation of space andtime. HFMD has obviously seasonality with unimodal distribution. HFMD iswidely distributed in Qingdao and mainly occur from April to August. In April,hand, foot and mouth disease cases began to increase sharply, peaked in July, andthen gradually decline. The characteristics of hand, foot and mouth disease areTravels faster, strong infective, and spread to a wide range.2. When applying prospective space-time scan statistics method in earlywarning of infectious disease outbreak，we ought to choose different parametersof infectious disease in order to get the best results． In next stage, we shouldchoose the best parameters based on the different regions of the same infectiousdisease.