儿童手足口病早期预警评分模型的构建

Establishment of an early warning score model in children with hand-foot-mouth disease

目的应用决策树构建儿童手足口病(HFMD)早期预警评分模型,为临床早期识别危重症HFMD提供科学依据。方法回顾性分析2014年1月至2016年12月青岛市妇女儿童医院住院的2488例重症及危重症HFMD患儿的病历资料,采用多因素Logistic回归分析筛选危重症HFMD的独立危险因素,利用决策树卡方自动交互检测法构建儿童HFMD早期预警评分模型,并通过受试者工作特征曲线(ROC)评价该模型的应用效果。结果Logistic回归分析筛选出年龄<3岁、血糖≥8.0 mmol/L、体温≥39.0℃、肢体抖动和震颤、手足部疱疹5个危重症HFMD的独立危险因素;通过决策树将血糖和体温2项危险因素纳入树形图,并根据树形图设立评分标准,结合儿童早期预警评分系统,最终构建的预警模型由5个维度组成,包含意识、肤色、心率、毛细血管再充盈时间、呼吸频率、呼吸肌参与情况、氧疗、血糖和体温9项评估指标,每个维度0~3分,总分为0~15分;模型的ROC曲线下面积为0.911,约登指数最大为0.684时,敏感性为0.910,特异性为0.774,截断值为6.5分。结论儿童HFMD早期预警评分模型识别危重症HFMD的准确性、敏感性和特异性显著提高。

Objective To construct an early warning score model for the identification of hand-foot-mouth disease(HFMD)in children using the decision tree,and to provide scientific evidence for early recognition of critical HFMD in clinical practice.Methods The medical records of 2488 children with severe and critical HFMD hospi-talized in Qingdao Women and Children′s Hospital from January 2014 to December 2016 were retrospectively analyzed.Multivariate Logistic regression analysis was made to screen the independent risk factors of critical HFMD.An early warning score model for identifying children with HFMD was established by using the chi-square automatic interactive detection method in the decision tree,and the effectiveness of the model was evaluated by the receiver operating characteristic curve(ROC).Results The multivariate Logistic regression analysis identified 5 independent risk factors of critical HFMD,including age<3 years old,blood sugar≥8.0 mmol/L,body temperature≥39.0℃,limb shaking and trembling,and hand-foot herpes.Two risk factors,i.e.,blood sugar and body temperature,were included in the tree chart through the decision tree,and scoring criteria were established according to the tree chart.Combining the early warning scoring system for children,the warning model were finally set up,which consisted of 5 dimensions and inclu-ded 9 evaluation indicators:consciousness,skin color,heart rate,capillary refill time,respiratory frequency,participation of respiratory muscles,oxygen therapy,blood sugar and body temperature.Each dimension ranged from 0 to 3 points,totaling 0-15 points.The area under the ROC,sensitivity,specificity,maximum Youden index,and cut-off value of the model were 0.911,0.910,0.774,0.684,and 6.5,respectively.Conclusions The early warning score model established in this paper can accurately identify critical HFMD,with high sensitivity and specificity.