中国卒中杂志 ›› 2025, Vol. 20 ›› Issue (7): 840-850.DOI: 10.3969/j.issn.1673-5765.2025.07.006

• 论著 • 上一篇    下一篇

老年急性卒中患者医院感染预测模型的构建与验证

李静1,程实1,郭军平2,胡爱香1,于鑫玮1,韩玮1,张越巍1,冀瑞俊2   

  1. 1 北京 100070 首都医科大学附属北京天坛医院感染管理处
    2 首都医科大学附属北京天坛医院神经病学中心
  • 收稿日期:2024-06-18 修回日期:2025-02-27 接受日期:2025-03-06 出版日期:2025-07-20 发布日期:2025-07-20
  • 通讯作者: 冀瑞俊 jrjchina@sina.com 张越巍 bele215@163.com
  • 基金资助:
    首都医科大学附属北京天坛医院自选科研项目(HX-B-2024011)

Construction and Validation of a Prediction Model for Nosocomial Infection in Elderly Patients with Acute Stroke

LI Jing1, CHENG Shi1, GUO Junping2, HU Aixiang1, YU Xinwei1, HAN Wei1, ZHANG Yuewei1, JI Ruijun2   

  1. 1 Infection Control Office, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, China
    2 Department of Neurology, Beijing Tiantan Hospital, Capital Medical University, Beijing 100070, China
  • Received:2024-06-18 Revised:2025-02-27 Accepted:2025-03-06 Online:2025-07-20 Published:2025-07-20
  • Contact: JI Ruijun, E-mail: jrjchina@sina.com ZHANG Yuewei, E-mail: bele215@163.com

摘要: 目的 探讨老年急性卒中患者医院感染的危险因素,建立预测模型并进行评价。
方法 回顾性选取2019年6月—2021年12月首都医科大学附属北京天坛医院神经内科住院的老年急性卒中患者作为研究对象。根据患者是否发生医院感染分为感染组和未感染组。按7∶3的比例将数据集分为训练集和测试集,训练集用来建立模型,测试集用来评估模型性能。在训练集中进行单因素和多因素分析,筛选影响因素并构建列线图预测模型。通过Hosmer-Lemeshow拟合优度检验、绘制校准曲线以及ROC曲线分析和决策曲线分析(decision curve analysis,DCA)评估模型预测能力的准确性、区分度和临床实用性。
结果 本研究共纳入2201例老年急性卒中患者,平均年龄(68.9±6.9)岁,男性占68.92%(1517/2201)。医院感染总发生率为12.22%(269/2201),其中肺部感染发生率为9.72%(214/2201),泌尿系统感染发生率为2.50%(55/2201),中枢神经系统感染发生率为0.50%(11/2201)。缺血性卒中、脑出血、蛛网膜下腔出血患者的医院感染率分别为10.31%、15.88%、29.29%,差异有统计学意义(P<0.001)。多因素分析结果显示年龄≥75岁、有慢性阻塞性肺疾病及消化性溃疡病史、入院时NIHSS评分5~15分和16~42分、入院时mRS评分≥3分、白蛋白<35 g/L、白细胞计数和中性粒细胞与淋巴细胞比值高、手术治疗是医院感染的独立危险因素,纳入列线图预测模型。训练集和测试集的Hosmer-Lemeshow拟合优度检验结果均显示P>0.05(训练集χ2=9.294,P=0.318;测试集χ2=10.173,P=0.253),说明该模型有较好的拟合度。校准曲线显示预测值与实际值的一致性较好。训练集和测试集的AUC分别为0.847(95%CI 0.819~0.876)、0.838(95%CI 0.786~0.890),表明模型有较好的预测能力和区分度。DCA显示预测模型临床实用性较高。
结论 本研究构建的列线图预测模型能较好地预测老年急性卒中患者的医院感染发生风险。

文章导读: 老年急性卒中患者发生医院感染的影响因素较多,本研究基于患者基本信息、既往病史、卒中功能评分、实验室检查、发病到入院时间、手术治疗等临床资料开发列线图预测模型。该模型有较好的识别和校准能力,有助于临床医师早期识别高风险患者,采取有效预防措施,降低医院感染的发生率。

关键词: 老年患者; 卒中; 医院感染; 预测模型; 列线图

Abstract: Objective  To explore the risk factors of nosocomial infection in elderly patients with acute stroke and to establish and evaluate a prediction model. 
Methods  Elderly patients with acute stroke who were hospitalized in the Department of Neurology, Beijing Tiantan Hospital, Capital Medical University from June 2019 to December 2021 were retrospectively selected as the study objects. Patients were divided into the infected group and the non-infected group according to whether nosocomial infection occurred. The data set was split into the training set and test set at a ratio of 7∶3, with the training set used to establish the model and the test set to evaluate the model’s performance. Univariate and multivariate analyses were performed on the training set to screen for influencing factors and construct a nomogram prediction model. The predictive capability of the model was evaluated in terms of accuracy, discrimination, and clinical utility through the Hosmer-Lemeshow goodness of fit test, calibration curve plotting, ROC curve analysis, and decision curve analysis (DCA).
Results  A total of 2201 elderly patients with acute stroke were included in this study, with an average age of (68.9±6.9) years, and 68.92% (1517/2201) were male. The overall nosocomial infection rate was 12.22% (269/2201), among which the pulmonary infection rate was 9.72% (214/2201), the urinary system infection rate was 2.50% (55/2201), and the central nervous system infection rate was 0.50% (11/2201). The nosocomial infection rates of ischemic stroke, intracerebral hemorrhage, and subarachnoid hemorrhage patients were 10.31%, 15.88%, and 29.29%, respectively, with statistically significant differences (P<0.001). Multivariate analysis results showed that age≥75 years, history of chronic obstructive pulmonary disease or peptic ulcer, NIHSS scores of 5-15 and 16-42 at admission, mRS score≥3 at admission, albumin<35 g/L, high white blood cell count and neutrophil to lymphocyte ratio, and surgical treatment were independent risk factors for nosocomial infection. These factors were incorporated into the nomogram prediction model. The Hosmer-Lemeshow goodness of fit test results for the training set and test set both showed P>0.05 (the training set χ2=9.294, P=0.318; the test set χ2=10.173, P=0.253), indicating that the model has a good fit. The calibration curve showed good agreement between predicted and actual values. The AUC values of the training set and test set were 0.847 (95%CI 0.819-0.876) and 0.838 (95%CI 0.786-0.890), respectively, indicating that the model had good predictive ability and discrimination. DCA showed that the prediction model had high clinical utility. 
Conclusions  The nomogram prediction model constructed in this study can effectively predict the risk of nosocomial infection in elderly patients with acute stroke.

Key words: Elderly patient; Stroke; Nosocomial infection; Prediction model; Nomogram

中图分类号: