金华市流行性感冒监测预警曲线的设计及应用

Design and application of the early warning curves for influenza surveillance in Jinhua,Zhejiang province

目的为科学有效地预警流感的暴发与流行,制定符合金华市的流感流行早期预警曲线。方法使用全市3所哨点监测医院每周流感样病例(ILI)占门诊病例总数百分比的历史监测数据建立数据库,采用控制图法建立预警模型;在综合2009-2011监测年度流感流行特征和ILI就诊病例中流感估计感染率的基础上,确定判定流感流行的指标及其参考标准;通过计算、比较灵敏度、特异度、阳性预测值、阳性似然比和绘制ROC曲线,选择合适的预警界值。结果选择P75作为流感流行基准曲线预警界值,预警效果相对较好,灵敏度为87.5%,特异度为62.3%;选择P85作为流感流行警戒曲线预警界值,预警效果相对较好,灵敏度为88.9%,特异度为91.0%。结论控制图法可以作为流感预警方法,周ILI%的P75点可以作为流感流行基准预警曲线;周ILI%的P85点可以作为流感流行警戒预警曲线。

Objective To design an early warning curves for influenza surveillance, and to assess its application value in order to alert influenza epidemic in Jinhua. Methods Alert database was established based on the percentage of historical weekly influenza-like illnesses in total outpatients （ILIa） in three sentinel hospitals in Jinhua. Control chart was used to detect outbreaks or epidemics of influenza. Reference criterions of influenza epidemic were defined based on epidemiological characteristics of influenza and estimated influenza infection rate of ILI cases from 2009 to 2011. After calculating sensitivity, specificity, positive predictive value, posi- tive likelihood ratio and describing receiver-operating characteristic curve （ROC）, the optimal model and thresholds were chosen. Results 75 percentile was chosen as cutoff value for early warning of influenza epi- demic with the sensitivity and specificity of 87.5% and 62.3%. Meanwhile, 85 percentile was chosen as threshold for early warning of influenza pandemic with the sensitivity and specificity of 88.9 % and 90.0 %. The efficiencies of early warning of two percentiles were both high. Conclusions Control chart can be as an es- sential method for early warning system of influenza. 75 percentile and 85 percentile are optimal alert thresholds for influenza epidemic early warning.