摘要
目的利用急性缺血性脑卒中患者诊疗数据构建基于机器学习的急性缺血性脑卒中医院感染预测模型,为临床治疗早期干预提供支持。方法选取江西某三甲医院神经内科2020年10月—2021年12月2360例脑卒中住院患者作为研究对象。按8∶2的比例随机分为训练集(1888例)与测试集(472例),纳入人口统计学数据以及入院48 h内临床诊疗数据,分析急性缺血性脑卒中患者发生医院感染的独立危险因素。使用logistic回归和3种机器学习算法[随机森林(RandomForest)、XGBoost、LightGBM]构建急性缺血性脑卒中医院感染预测模型,应用ROC曲线下面积(AUC)评估4种预测模型的预测效果。结果574例发生医院感染,医院感染发生率为24.32%。logistic回归分析结果显示,年龄>65岁、入院NIHSS评分>5分、血白细胞计数>10×10^(9)/L、血清钠≤135 mmol/L、侵入性操作、吞咽困难为医院感染的危险因素。logistic回归、RandomForest、XGBoost、LightGBM预测模型在测试集中预测缺血性脑卒中医院感染的AUC值分别为0.854、0.850、0.881、0.870。结论基于机器学习医院感染预测模型有利于早期识别急性缺血性脑卒中患者医院感染及挖掘危险因素,及时采取防控措施,可降低医院感染发生率。
Objective To construct a machine-learning-based model through diagnosis and treatment data of patients with acute ischemic stroke(AIS)to predict healthcare-associated infection(HAI)in AIS patients and provide support for early intervention of clinical treatment.Methods 2360 inpatients with stroke from October 2020 to December 2021 in department of neurology of a tertiary first-class hospital in Jiangxi Province were selected as study subjects and randomly divided at 8∶2 ratio into training set(n=1888)and testing set(n=472).Demographic data as well as clinical diagnosis and treatment data within 48 hours of admission were included to analyze the indepen-dent risk factors for HAI in AIS patients.Prediction model of HAI in AIS patients was constructed with logistic regression and three machine-learning algorithms(RandomForest,XGBoost,LightGBM).Area under the receiver operating characteristic curve(AUC)was adopted to evaluate the prediction efficacy of four prediction models.Results HAI occurred in 574 patients,incidence of HAI was 24.32%.Logistic regression analysis showed that age>65 years old,NIHSS score>5 points at admission,white blood cell count>10×10^(9)/L,serum sodium≤135 mmol/L,invasive operation and dysphagia were risk factors for HAI.AUC values of logistic regression,RandomForest,XGBoost and LightGBM prediction models for predicting HAI in AIS patients in the test set were 0.854,0.850,0.881 and 0.870 respectively.Conclusion Machine-learning-based model for predicting HAI is conducive to the early identification of HAI and relevant risk factors,and facilitates timely preventive and control measures in AIS patients,thus reduces the incidence of HAI.
作者
刘建模
罗颢文
俞鹏飞
吴一帆
韩梦琦
贾伟杰
易应萍
LIU Jian-mo;LUO Hao-wen;YU Peng-fei;WU Yi-fan;HAN Meng-qi;JIA Wei-jie;YI Ying-ping(Department of Science and Technology,The Second Affiliated Hospital of Nanchang University,Nanchang 330000,China;School of Public Health,Nanchang University,Nanchang 330000,China)
出处
《中国感染控制杂志》
CAS
CSCD
北大核心
2023年第2期129-135,共7页
Chinese Journal of Infection Control
基金
国家自然科学基金项目(81960609)
国家重点研发计划项目(2020YFC2002901)
南昌大学第二附属医院资助项目(2021efyB03)。
关键词
急性缺血性脑卒中
机器学习
预测模型
医院感染
acute ischemic stroke
machine-learning
prediction model
healthcare-associated infection
