基于移动百分位数法的深圳市南山区社区级肺结核疫情预警

摘要

目的 探寻深圳市南山区社区一级肺结核疫情预警最优阈值.方法 采用移动百分位数法,以2010年1月-2015年5月为基线,2015年3月-2016年3月为预警期,以社区为单位,以月为观察期,建立两种5年基候选阈值.根据ROC曲线下面积选择基线历史同期摆动周期,根据约登指数结合特异度优选阈值.结果 以25个历史数据(前后摆动2个月)与15个历史数据(前后摆动1个月)建立肺结核移动百分位数预警,其ROC曲线下面积分别为0.936,0.945,差异无统计学意义(P=0.131).15个历史数据的候选阈值中,P80的约登指数与P60接近(最大者),特异度>90%.结论 综合考虑操作性及后续处置成本,建议深圳市南山区采用5年基前后摆动1个月的P80作为社区一级肺结核预警暴发的阈值.%Objectives To explore the optimal threshold for outbreak early warning of pulmonary tuberculosis(PTB)at community level in Nanshan District,Shenzhen.Methods We used the moving percentile method(MPM)to calculate two types of five-year based candidate thresholds for monthly PTB outbreaks detection,with data from January 2010 to May 2015 as a baseline and data from March 2015 to March 2016 as a testing data.Then we decided reference months at baseline based on areas under receiver operating characteristic curve(AUC)and chose the threshold based on Youden index.Results The AUCs were 0.936 and 0.945,respectively,for the 25 historical data(data at corresponding month,two preceding months,and the two following months)and 15 historical data(data at corresponding month,the preceding month,and the following month),and the differences between them were not statistically significant(P=0.131).For the 15 baseline data,the P80 got a close value of Youden index to that of P60,which was the maximum,with a specificity > 90%.Conclusions Considering the practicality and cost,it is suggested that data at corresponding month,the preceding month,and the following month for the previous 5 years should be the reference data,and P80 should be a warning threshold of the MPM for community-level PTB outbreak detection at Nanshan District,Shenzhen City.