基于机器学习的急性缺血性脑卒中医院感染预测模型建立与评价
作者:
作者单位:

1.南昌大学第二附属医院科技处;2.南昌大学公共卫生学院

作者简介:

通讯作者:

易应萍  E-mail: yyp66@126.com

中图分类号:

+2]]>

基金项目:

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)


Construction and evaluation of a machine-learning-based model for predicting healthcare-associated infection in patients with acute ischemic stroke
Author:
Affiliation:

1.Department of Science and Technology, The Second Affiliated Hospital of Nanchang University, Nanchang 330000, China;2.School of Public Health, Nanchang University, Nanchang 330000, China

Fund Project:

  • 摘要
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 文章评论
    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48h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>10×109/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。

    Abstract:

    Objective To construct a machine-learning-based model through diagnosis and treatment data of patients with acute ischiic 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 Deciber 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). Diographic 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 >10×109/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.

    参考文献
    相似文献
引用本文

刘建模,罗颢文,俞鹏飞,等.基于机器学习的急性缺血性脑卒中医院感染预测模型建立与评价[J]. 中国感染控制杂志,2023,(2):129-135. DOI:10.12138/j. issn.1671-9638.20233300.
Jian-mo LIU, Hao-wen LUO, Peng-fei YU, et al. Construction and evaluation of a machine-learning-based model for predicting healthcare-associated infection in patients with acute ischemic stroke[J]. Chin J Infect Control, 2023,(2):129-135. DOI:10.12138/j. issn.1671-9638.20233300.

复制
分享
文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2022-08-30
  • 最后修改日期:
  • 录用日期:
  • 在线发布日期: 2024-04-28
  • 出版日期: