公共场所喷泉水嗜肺军团菌污染贝叶斯网络预警模型研究  

Study on the early warning model based on Bayesian network for the contamination of Legionella pneumophilia in Fountain water in public areas

作　　者：牟敬锋[1] 段永翔[1] 严宙宁[1] 严燕[1] 罗文亮[1] 袁梦[1] 曾宪冬[1] MOU Jing-feng;DUAN Yong-xiang;YAN Zhou-ning;YAN Yan;LUO Wen-liang;YUAN Meng;ZENG Xian-dong(Nanshan District Center for Disease Control and Prevention, Shenzhen, Guangdong 518054, China)

机构地区：[1]深圳市南山区疾病预防控制中心环境卫生科,广东深圳518054

出　　处：《中国预防医学杂志》2017年第10期742-746,共5页Chinese Preventive Medicine

基　　金：深圳市南山区卫生科技计划立项项目(2015071)

摘　　要：目的构建喷泉水嗜肺军团菌污染贝叶斯网络预警模型,为预防和控制喷泉水军团菌污染提供科学依据。方法选择2015年深圳市公共场所正运行的70座喷泉作为研究对象,采用问卷调查、现场监测及实验室检测等方法收集相关数据,构建贝叶斯网络预警模型。选择2016年深圳市公共场所正运行的30座喷泉作为研究对象,收集相关数据对模型预警结果进行验证。结果深圳市喷泉水嗜肺军团菌检出率为42.00%(42/100)。从构建的贝叶斯网络模型结构看,对喷泉水中嗜肺军团菌最具有影响的因素有定期清洗、浊度、游离性余氯和溶解性总固体。该模型ROC曲线最佳诊断临界点为0.475,利用该临界点进行嗜肺军团菌阳性诊断ROC曲线下面积为0.941(95%CI:0.893~0.998),诊断灵敏度为92.90%,特异度为90.90%,预测准确率为93.33%(28/30)。结论本研究构建的贝叶斯网络模型预测准确率较高,可以满足喷泉水嗜肺军团菌污染预警要求,对公共场所喷泉水嗜肺军团菌污染的判定具有一定的参考价值。Objective To establish an early warning model based on Bayesian network for the contamination of Legionella pneumophilia in fountain water, and provide scientific basis for prevention and control of legionella pollution in fountains. Methods 70 fountains in public places were randomly selected for the questionnaire survey, field tests and laboratory tests in Shenzhen in 2015, the relative data was used to establish the early warning model based on Bayesian network. Then, 30 fountains were randomly selected in 2016 to verify and e- valuate the model. Results 42.00~ （40/100） of fountains were tested positive for Legionella pneumophilia in Shenzhen. Based on the constructed model in this study, the most common influencing factors for the con tamination of Legionella pneumophila in fountains were periodic cleaning, turbidity of water, free chlorine re- sidual and total dissolved solids. The critical point of ROC curve of the model was 0. 475 with the area under the ROC curve of 0. 941 （95%CI.. 0. 893-0. 998）. The sensitivity and specificity of the model on the detection of Legionella pneumophila were 92.90% and 90.90%, respectively, and the predicted accuracy rate of the model was 93.33%. Conclusions The model based on the Bayesian network is able to predict the contamina- tion of Legionella pneumophila in fountains with high accuracy, and can meet the early warning requirements.

关 键 词：喷泉水 嗜肺军团菌 预警 贝叶斯网络 卫生学调查 实验室检测 

分 类 号：R126.4[医药卫生—环境卫生学]