Sun Qiao, Lai Sheng-Jie, Li Zhong-Jie, Lan Ya-Jia, Zhang Hong-Long, Zhao Dan, Jin Lian-Mei, Yang Wei-Zhong
Shanghai Pudong New Area Center for Disease Control and Prevention, Shanghai 200136, China.
Zhonghua Liu Xing Bing Xue Za Zhi. 2011 May;32(5):450-3.
To compare the different thresholds of 'moving percentile method' for outbreak detection in the China Infectious Diseases Automated-alert and Response System (CIDARS).
The thresholds of P(50), P(60), P(70), P(80) and P(90) were respectively adopted as the candidates of early warning thresholds on the moving percentile method. Aberration was detected through the reported cases of 19 notifiable infectious diseases nationwide from July 1, 2008 to June 30, 2010. Number of outbreaks and time to detection were recorded and the amount of signals acted as the indicators for determining the optimal threshold of moving percentile method in CIDARS.
The optimal threshold for bacillary and amebic dysentery was P(50). For non-cholera infectious diarrhea, dysentery, typhoid and paratyphoid, and epidemic mumps, it was P(60). As for hepatitis A, influenza and rubella, the threshold was P(70), but for epidemic encephalitis B it was P(80). For the following diseases as scarlet fever, typhoid and paratyphoid, hepatitis E, acute hemorrhagic conjunctivitis, malaria, epidemic hemorrhagic fever, meningococcal meningitis, leptospirosis, dengue fever, epidemic endemic typhus, hepatitis C and measles, it was P(90). When adopting the adjusted optimal threshold for 19 infectious diseases respectively, 64 840 (12.20%) signals had a decrease, comparing to the adoption of the former defaulted threshold (P(50)) during the 2 years. However, it did not reduce the number of outbreaks being detected as well as the time to detection, in the two year period.
The optimal thresholds of moving percentile method for different kinds of diseases were different. Adoption of the right optimal threshold for a specific disease could further optimize the performance of outbreak detection for CIDARS.
比较中国传染病自动预警与应对系统（CIDARS）中“移动百分位数法”用于疫情检测的不同阈值。
分别采用P(50)、P(60)、P(70)、P(80)和P(90)阈值作为移动百分位数法预警阈值的候选值。通过2008年7月1日至2010年6月30日全国19种法定传染病的报告病例检测异常情况。记录疫情数量和检测时间，并将信号量作为确定CIDARS中移动百分位数法最佳阈值的指标。
细菌性和阿米巴性痢疾的最佳阈值为P(50)。非霍乱性感染性腹泻、痢疾、伤寒和副伤寒以及流行性腮腺炎的最佳阈值为P(60)。甲型肝炎、流感和风疹的阈值为P(70)，而乙型脑炎的阈值为P(80)。猩红热、伤寒和副伤寒、戊型肝炎、急性出血性结膜炎、疟疾、流行性出血热、流行性脑脊髓膜炎、钩端螺旋体病、登革热、流行性斑疹伤寒、丙型肝炎和麻疹的最佳阈值为P(90)。与采用前默认阈值（P(50)）相比，分别采用调整后的19种传染病最佳阈值时，两年内有64840个（12.20%）信号减少。然而，在这两年期间，它并没有减少检测到的疫情数量以及检测时间。
不同疾病的移动百分位数法最佳阈值不同。针对特定疾病采用正确的最佳阈值可进一步优化CIDARS疫情检测的性能。
