School of Biological, Earth and Environmental Sciences, University College Cork, Cork, Ireland.
Environmental Health and Sustainability Institute, Dublin Institute of Technology, Dublin, Ireland.
Environ Pollut. 2018 Jun;237:329-338. doi: 10.1016/j.envpol.2018.02.052. Epub 2018 Feb 28.
Private groundwater sources in the Republic of Ireland provide drinking water to an estimated 750,000 people or 16% of the national population. Consumers of untreated groundwater are at increased risk of infection from pathogenic microorganisms. However, given the volume of private wells in operation, remediation or even quantification of public risk is both costly and time consuming. In this study, a hierarchical logistic regression model was developed to 'predict' contamination with E. coli based on the results of groundwater quality analyses of private wells (n = 132) during the period of September 2011 to November 2012. Assessment of potential microbial contamination risk factors were categorised into three groups: Intrinsic (environmental factors), Specific (local features) and Infrastructural (groundwater source characteristics) which included a total of 15 variables. Overall, 51.4% of wells tested positive for E. coli during the study period with univariate analysis indicating that 11 of the 15 assessed risk factors, including local bedrock type, local subsoil type, septic tank reliance, 5 day antecedent precipitation and temperature, along with well type and depth, were all significantly associated with E. coli presence (p < 0.05). Hierarchical logistic regression was used to develop a private well susceptibility model with the final model containing 8 of the 11 associated variables. The model was shown to be highly efficient; correctly classifying the presence of E. coli in 94.2% of cases, and the absence of E. coli in 84.7% of cases. Model validation was performed using an external data set (n = 32) and it was shown that the model has promising accuracy with 90% of positive E. coli cases correctly predicted. The developed model represents a risk assessment and management tool that may be used to develop effective water-quality management strategies to minimize public health risks both in Ireland and abroad.
爱尔兰共和国的私人地下水水源为大约 75 万人（占全国人口的 16%）提供饮用水。未经过处理的地下水消费者感染致病微生物的风险增加。然而，考虑到运行中的私人井的数量，修复甚至量化公共风险既昂贵又耗时。在这项研究中，建立了一个分层逻辑回归模型，以根据 2011 年 9 月至 2012 年 11 月期间私人水井的地下水质量分析结果，“预测”大肠杆菌污染情况（n=132）。评估潜在微生物污染风险因素分为三组：固有（环境因素）、特定（当地特征）和基础设施（地下水源特征），其中包括总共 15 个变量。总的来说，在研究期间，51.4%的水井检测出大肠杆菌呈阳性，单变量分析表明，在评估的 15 个风险因素中，有 11 个因素，包括当地基岩类型、当地底土类型、依赖化粪池、5 天的前期降水和温度，以及井类型和深度，都与大肠杆菌的存在显著相关（p<0.05）。分层逻辑回归用于开发私人水井易感性模型，最终模型包含 11 个相关变量中的 8 个。该模型显示出很高的效率；在 94.2%的情况下正确分类大肠杆菌的存在，在 84.7%的情况下正确分类大肠杆菌的不存在。使用外部数据集（n=32）进行模型验证，表明该模型具有很高的准确性，90%的阳性大肠杆菌病例被正确预测。所开发的模型代表了一种风险评估和管理工具，可用于制定有效的水质管理策略，以尽量减少爱尔兰和国外的公共卫生风险。
