Wang Y, Zhang L, Wu S S, Duan W, Sun Y, Zhang M, Zhang X X, Zhang Y, Ma C N, Wang Q Y, Yang P
Institute for Infectious Disease and Endemic Disease Control, Beijing Center for Disease Prevention and Control, Beijing Research Center for Preventive Medicine, Beijing 100013, China.
Institute for Infectious Disease and Endemic Disease Control, Beijing Center for Disease Prevention and Control, Beijing Research Center for Preventive Medicine, Beijing 100013, China; School of Public Health, Capital Medical University, Beijing 100069, China.
Zhonghua Liu Xing Bing Xue Za Zhi. 2020 Feb 10;41(2):201-206. doi: 10.3760/cma.j.issn.0254-6450.2020.02.012.
To calculate both the epidemic and intensity thresholds for different levels in Beijing and to establish a tiered alert system in the 2018-2019 influenza season as well as to evaluate the performance of calculated thresholds. Weekly count of influenza-like illness and percentage of influenza-like illness (ILI) of the last five influenza seasons were modeled by 'moving epidemic method' (MEM) to calculate the influenza epidemic and intensity thresholds at different levels. A cross-validation procedure was used to evaluate the performance. Indicators of Matthew correlation coefficient, Youden's index, sensitivity and specificity were calculated. For weekly count of influenza-like illness, data showed that the epidemic threshold for 2018-2019 influenza season was 12 984 and the medium, high and very high intensity thresholds were 22 503, 37 589, 47 157, respectively. Matthew correlation coefficient of the epidemic threshold was 62 and youden's index as 60 , sensitivity as 69, specificity as 91. Data on weekly ILI, the epidemic threshold for 2018-2019 influenza season was 1.66, with medium, high and very high intensity thresholds as 2.46, 3.84 and 4.66, respectively. The overall Matthew correlation coefficient of the epidemic threshold was 59, with 54 for the Youden's index, sensitivity as 60 and specificity as 94. MEM produced a good specific signal for detecting the influenza epidemics and the accuracy of the method was acceptable. The early warning performance regarding the application of weekly count on influenza-like illness was slightly better than ILI. This method could be applied in the practical influenza epidemic alert "work in Beijing" .
计算北京不同级别流感的流行阈值和强度阈值，在2018 - 2019流感季建立分级预警系统，并评估所计算阈值的性能。采用“移动流行法”（MEM）对过去五个流感季的流感样病例每周计数和流感样病例百分比进行建模，以计算不同级别的流感流行阈值和强度阈值。使用交叉验证程序评估性能，计算马修相关系数、约登指数、敏感性和特异性等指标。对于流感样病例每周计数，数据显示2018 - 2019流感季的流行阈值为12984，中、高和极高强度阈值分别为22503、37589、47157。流行阈值的马修相关系数为62，约登指数为60，敏感性为69，特异性为91。对于每周流感样病例百分比数据，2018 - 2019流感季的流行阈值为1.66，中、高和极高强度阈值分别为2.46、3.84和4.66。流行阈值的总体马修相关系数为59，约登指数为54，敏感性为60，特异性为94。MEM在检测流感流行方面产生了良好的特异性信号，该方法的准确性可以接受。关于应用流感样病例每周计数的预警性能略优于流感样病例百分比。该方法可应用于北京实际的流感流行预警工作。
