牟敬锋, 段永翔, 严宙宁, 严燕, 罗文亮, 袁梦, 曾宪冬. 构建及评价公共场所喷泉水嗜肺军团菌污染影响因素的logistic回归模型[J]. 环境与职业医学, 2017, 34(2): 112-117. DOI: 10.13213/j.cnki.jeom.2017.16538
引用本文: 牟敬锋, 段永翔, 严宙宁, 严燕, 罗文亮, 袁梦, 曾宪冬. 构建及评价公共场所喷泉水嗜肺军团菌污染影响因素的logistic回归模型[J]. 环境与职业医学, 2017, 34(2): 112-117. DOI: 10.13213/j.cnki.jeom.2017.16538
MOU Jing-feng, DUAN Yong-xiang, YAN Zhou-ning, YAN Yan, LUO Wen-liang, YUAN Meng, ZENG Xian-dong. Establishment and evaluation of logistic regression model of influencing factors for Legionella pneumophila contamination in public fountains[J]. Journal of Environmental and Occupational Medicine, 2017, 34(2): 112-117. DOI: 10.13213/j.cnki.jeom.2017.16538
Citation: MOU Jing-feng, DUAN Yong-xiang, YAN Zhou-ning, YAN Yan, LUO Wen-liang, YUAN Meng, ZENG Xian-dong. Establishment and evaluation of logistic regression model of influencing factors for Legionella pneumophila contamination in public fountains[J]. Journal of Environmental and Occupational Medicine, 2017, 34(2): 112-117. DOI: 10.13213/j.cnki.jeom.2017.16538

构建及评价公共场所喷泉水嗜肺军团菌污染影响因素的logistic回归模型

Establishment and evaluation of logistic regression model of influencing factors for Legionella pneumophila contamination in public fountains

  • 摘要: 目的 探讨公共场所喷泉水嗜肺军团菌污染的影响因素,拟合logistic回归方程用于喷泉水嗜肺军团菌污染的预测,为预防和控制喷泉水军团菌污染提供科学依据。

    方法 选择深圳市公共场所正常运行的80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式收集相关数据。2015年7-12月收集50座喷泉(其中包括30座住宅小区喷泉和20座公共建筑周围的喷泉)相关资料,利用logistic回归法对可能影响喷泉水嗜肺军团菌污染的因素进行分析并建立回归模型,绘制受试者工作特征(ROC)曲线,确定喷泉水嗜肺军团菌污染的回归模型的最佳临界值。采用2016年1-7月收集的30座喷泉(其中包括20座住宅小区喷泉和10座公共建筑周围的喷泉)相关数据验证上述模型的准确率。

    结果 单因素logistic回归分析发现,游离性余氯含量升高是嗜肺军团菌污染的保护因素(OR=0.986);运行年限的增加、喷泉水浊度升高、溶解性总固体含量升高是嗜肺军团菌污染的危险因素(OR=1.096、1.261、1.100)。经多因素logistic回归分析筛选出游离性余氯(OR=0.952,P<0.05)、浊度(OR=1.314,P<0.05)和溶解性总固体(OR=1.098,P<0.05)3个因素用于模型的拟合,其ROC曲线下面积为0.877(95% CI:0.808~0.997),该模型预测准确率为80.0%。

    结论 logistic回归模型发现喷泉水游离性余氯、浊度和溶解性总固体是影响嗜肺军团菌污染的主要因素。该模型预测准确率较高,对公共场所喷泉水嗜肺军团菌污染的判断有一定的参考价值。

     

    Abstract: Objective To study the risk factors of Legionella pneumophila contamination in public fountains, establish a logistic regression model for predicting the contamination, and provide a scientific basis for related prevention and control.

    Methods Eighty functioning public fountains were selected in Shenzhen. Data were collected using questionnaire survey, field test, and laboratory test. The data collected from 50 fountains (including 30 fountains in residence communities and 20 fountains surrounding public buildings) from July to December 2015 were used to establish a logistic regression model to assess the associations between Legionella pneumophila contamination and selected influencing factors, and then a receiver operating characteristic curve (ROC curve) was drawn to determine the optimal threshold of Legionella pneumophila contamination for the model. The data from 30 fountains (including 20 fountains in residence communities and 10 fountains surrounding public buildings) were collected from January to July 2016 to validate the accuracy of the model.

    Results The results of univariate logistic regression analysis showed that the increased level of free residual chlorine was a protective factor for Legionella pneumophila contamination in fountains (OR=0.986); whereas the increased operating years, turbidity, and total dissolved solids were risk factors (OR=1.096, 1.261, 1.100). According to multivariate logistic regression analysis, the independent factors predicting Legionella pneumophila contamination were free residual chlorine (OR=0.952, P<0.05), turbidity (OR=1.314, P<0.05), and total dissolved solids (OR=1.098, P<0.05). The area under the ROC curve was 0.877 (95%CI:0.808-0.997) and the prediction accuracy rate of the model was 80.0%.

    Conclusion Free residual chlorine, turbidity, and total dissolved solids are the main risk factors of Legionella pneumophila contamination in the logistic regression model. The model with a high accuracy could provide reference for predicting Legionella pneumophila contamination in fountains in public places.

     

/

返回文章
返回