范帅华, 郭伟, 林金兰, 吴圣, 郭军. 基于IBM-SPSS与机器学习的logistic回归对院内感染肺炎克雷伯菌发生多重耐药关联因素的分析方法比较[J]. 解放军医学院学报, 2023, 44(1): 11-16. DOI: 10.3969/j.issn.2095-5227.2023.01.003
引用本文: 范帅华, 郭伟, 林金兰, 吴圣, 郭军. 基于IBM-SPSS与机器学习的logistic回归对院内感染肺炎克雷伯菌发生多重耐药关联因素的分析方法比较[J]. 解放军医学院学报, 2023, 44(1): 11-16. DOI: 10.3969/j.issn.2095-5227.2023.01.003
FAN Shuaihua, GUO Wei, LIN Jinlan, WU Sheng, GUO Jun. Comparison of factors associated with multidrug-resistant Klebsiella pneumoniae hospital-acquired infection by logistic regression based on IBM-SPSS versus machine learning[J]. ACADEMIC JOURNAL OF CHINESE PLA MEDICAL SCHOOL, 2023, 44(1): 11-16. DOI: 10.3969/j.issn.2095-5227.2023.01.003
Citation: FAN Shuaihua, GUO Wei, LIN Jinlan, WU Sheng, GUO Jun. Comparison of factors associated with multidrug-resistant Klebsiella pneumoniae hospital-acquired infection by logistic regression based on IBM-SPSS versus machine learning[J]. ACADEMIC JOURNAL OF CHINESE PLA MEDICAL SCHOOL, 2023, 44(1): 11-16. DOI: 10.3969/j.issn.2095-5227.2023.01.003

基于IBM-SPSS与机器学习的logistic回归对院内感染肺炎克雷伯菌发生多重耐药关联因素的分析方法比较

Comparison of factors associated with multidrug-resistant Klebsiella pneumoniae hospital-acquired infection by logistic regression based on IBM-SPSS versus machine learning

  • 摘要:
      背景  近年来随着人工智能的发展,各种机器学习的模型开始在临床中初步应用。
      目的  分别使用基于SPSS 26.0的logistic回归与基于机器学习逻辑回归模型(logistic regression,LR)分析院内感染多重耐药肺炎克雷伯菌(multidrug-resistant Klebsiella pneumoniae,MDR-KPN)的关联因素,比较两种方法的差异。
      方法  选取2016年1月- 2020年12月清华大学附属北京清华长庚医院肺炎克雷伯菌院内感染病例254例,包括MDR-KPN 168例和非MDR-KPN(敏感菌株)86例,用两种不同的logistic回归方法对数据进行处理。
      结果  基于SPSS 26.0的logistic多因素分析显示,患有肝病、感染前3个月内有胸腔穿刺史、感染前3个月内有动脉穿刺史、感染前3个月内气管插管史是发生MDR-KPN院内感染的独立关联因素。LR模型分析显示,气管插管是发生MDR-KPN院内感染的最主要关联因素,前5位的关联因素为气管插管、基础肝病史、年龄18 ~ 49岁的人群、碳青霉烯类药物暴露史、中心静脉置管史。
      结论  两种分析方法得到的MDR-KPN院内感染关联因素具有一定的吻合度,提示了LR模型应用于MDR-KPN院内感染预测的可行性;此外LR模型有高效、便捷、准确率可量化的优势,在临床诊疗中可能有更广阔的应用前景。

     

    Abstract:
      Background  In recent years, with the development of artificial intelligence, various machine learning models have been initially applied in clinical practice.
      Objective  To study the factors associated with multidrug-resistant Klebsiella pneumoniae (MDR-KPN) infection in hospitals using traditional logistic regression and machine learning logistic regression (LR model), respectively, and compare the differences of the two methods.
      Methods  A total of 254 cases of hospital-acquired Klebsiella pneumoniae infection in a tertiary hospital were selected, including 168 cases of MDR-KPN and 86 cases of non-MDR-KPN. Two different logistic regression methods were used to process the data.
      Results  Multivariate logistic analysis based on IBM-SPSS showed that liver disease, history of puncture of chest cavity within 3 months, history of puncture of artery within 3 months, and endotracheal intubation were independent factors associated with MDR-KPN hospital-acquired infection. LR model showed that endotracheal intubation was the most important factor associated with nosoacquired MDR-KPN infection, and the top five factors were endotracheal intubation, liver disease, age of 18-49 years, carbapenems exposure history, and central venous catheterization history.
      Conclusion  The factors associated with MDR-KPN hospital-acquired infection obtained by the two methods have a certain degree of consistency. LR model has the advantages of high efficiency, convenience, and quantifiable accuracy, suggesting that it has a certain application prospect in clinical practice.

     

/

返回文章
返回