Wang Ruiping, Long Yun, Liu Huihui, Yao Xiaqing, Zeng Guang
Chinese Field Epidemiology Training Program(CFETP), Chinese Center for Disease Control and Prevention, Beijing 100015, China. Email:
Zhonghua Yu Fang Yi Xue Za Zhi. 2014 Jan;48(1):53-7.
To select the premium alert threshold for major communicable disease by using the control graph alert technique based on the local disease information.
8 major communicable diseases in Songjiang district were ascertained by analysis of the national early warning detection information system which include the other diarrhea, mumps, chickenpox, scarlet fever, rubella, hand foot mouth disease, influenza and dysentery; weekly reported cases from 2008 to 2011 were used to establish the early detection model (PERCENTILE (array, x), array (4×5), x = 0.05, 0.10…0.95) by moving percentile method, next applying the established early detection model and the golden standard (AKX(-)D ± 2s) to predict the expected weekly cases in 2012 respectively, and then ascertain the predict results by comparison with the actual weekly cases in 2012 respectively, finally the premium threshold was selected by comparison of the model predicted results with the golden standard predicted results after comprehensive consideration of the sensitivity, specificity, positive predictive value and negative predictive value and receiver operating characteristic (ROC) curve.
The premium alert threshold for mumps, other diarrhea and rubella was P90, dysentery was P75, scarlet fever and chickenpox was P80, and the premium threshold for hand-foot-mouth disease (HFMD) and influenza was P95, the sensitivity of 8 major communicable diseases were 100%, 100%, 86%, 100%, 100%, 100%, 94%, 100%, respectively; the specificities were 92%, 73%, 72%, 77%, 73%, 92%, 66%, 80%, respectively; the positive predictive values were 43%, 40%, 32%, 8%, 24%, 20%, 59%, 47%, respectively; and the negative predictive values were 100%, 100%, 97%, 100%, 100%, 100%, 96%, 100%, respectively. The national recommended alert thresholds for the 8 major communicable diseases were P80, except for chickenpox (P50) and HFMD (CUSUM).
6 out of 8 major communicable diseases' early detection thresholds in Songjiang district should be adjusted according to the analysis results. Premium alert threshold selection need to consider the local disease report and the characteristics of infectious diseases to upgrade the early detection capability.
基于当地疾病信息，运用控制图预警技术选择重大传染病的最佳预警阈值。
通过分析国家预警监测信息系统确定松江区8种主要传染病，包括其他腹泻病、流行性腮腺炎、水痘、猩红热、风疹、手足口病、流行性感冒和痢疾；利用2008年至2011年的每周报告病例，采用移动百分位数法建立早期检测模型（PERCENTILE(数组,x)，数组(4×5)，x = 0.05, 0.10…0.95），接着分别应用建立的早期检测模型和金标准（AKX(-)D ± 2s）预测2012年的每周预期病例数，然后分别与2012年的实际每周病例数进行比较确定预测结果，最后在综合考虑灵敏度、特异度、阳性预测值、阴性预测值和受试者工作特征（ROC）曲线的基础上，通过比较模型预测结果和金标准预测结果来选择最佳阈值。
流行性腮腺炎、其他腹泻病和风疹的最佳预警阈值为P90，痢疾为P75，猩红热和水痘为P80，手足口病和流行性感冒的最佳阈值为P95，8种主要传染病的灵敏度分别为100%、100%、86%、100%、100%、100%、94%、100%；特异度分别为92%、73%、72%、77%、73%、92%、66%、80%；阳性预测值分别为43%、40%、32%、8%、24%、20%、59%、47%；阴性预测值分别为100%、100%、97%、100%、100%、100%、96%、100%。8种主要传染病的国家推荐预警阈值除水痘（P50）和手足口病（累积和法）为P80外。
根据分析结果，松江区8种主要传染病中的6种早期检测阈值应进行调整。最佳预警阈值的选择需要考虑当地疾病报告情况和传染病特征，以提升早期检测能力。
