脑卒中患者肺部感染风险预测模型的建立及应用研究

Establishment and application of risk prediction model for pulmonary infection in stroke patients

目的 构建脑卒中患者肺部感染的风险预测模型,并验证其临床预测效果。方法 采用便利抽样法,选择2021年1月至2023年1月在贵州省某三级甲等医院住院的750例脑卒中患者为研究对象,根据是否发生肺部感染分为肺部感染组(n=267)和非肺部感染组(n=483),对2组患者相关资料进行比较分析,应用logistic回归分析建立风险预测模型,应用Hosmer-Lemeshow(H-L)检验模型的拟合优度及受试者工作特征曲线(ROC曲线)下面积判断模型的预测效果。另选取2023年2-8月符合标准的145例患者进行模型预测效果验证。结果 经单因素及多因素分析发现,吞咽障碍[比值比(OR)=10.462]、合并肺部基础疾病(OR=6.046)、低钾血症(OR=2.266)、低钠血症(OR=3.807)、低血红蛋白(OR=4.036)、入院时美国国立卫生研究院卒中量表(NISSH)评分(OR=38.135)、入院时日常生活能力(ADL)评分(OR=12.942)、住院时间(OR=8.992)是脑卒中患者发生肺部感染的独立危险因素(P<0.05)。风险预测模型公式:Logit(P)=-4.761+2.348×(吞咽障碍的赋值)+1.799×(合并肺部基础疾病的赋值)+0.818×(低钾血症的赋值)+1.337×(低钠血症的赋值)+1.395×(低血红蛋白的赋值)+3.641×(NISSH评分的赋值)+2.560×(ADL评分的赋值)+2.196×(住院时间的赋值)。建模组ROC曲线下面积为0.953[95%可信区间(95%CI)0.940~0.967,P<0.001],约登指数为0.762,灵敏度为0.880,特异度为0.882,H-L检验的P值为0.553。模型验证结果显示:验模组ROC曲线下面积0.946(95%CI:0.927~0.987,P<0.001),灵敏度为0.898,特异度为0.875,正确率为88.3%,H-L检验P值为0.510。结论 构建的脑卒中发生肺部感染风险预测模型预测效能良好,可为临床医护人员早期识别脑卒中发生肺部感染的高危人群提供参考,为及时采取预防性管理措施提供借鉴。

Objective To construct a risk prediction model for pulmonary infection in stroke patients and verify its clinical predictive performance.Methods Using convenience sampling,750 stroke patients hospitalized in a tertiary hospital in Guizhou Province from January 2021 to January 2023 were selected as study subjects.They were divided into a pulmonary infection group(n=267)and a non-pulmonary infection group(n=483)based on whether pulmonary infection occurred.Comparative analysis of relevant data between the two groups was conducted.Logistic regression analysis was applied to establish a risk prediction model.The goodness-of-fit of the model was tested using the Hosmer-Lemeshow(H-L)test,and the predictive performance of the model was evaluated by the area under the receiver operating characteristic curve(ROC curve).Another 145 eligible patients from February to August 2023 were selected to validate the predictive performance of the model.Results Univariate and multivariate analyses revealed that dysphagia[odds ratio(OR)=10.462],coexisting pulmonary diseases(OR=6.046),hypokalemia(OR=2.266),hyponatremia(OR=3.807),low hemoglobin(OR=4.036),National Institutes of Health Stroke Scale(NIHSS)score at admission(OR=38.135),Activities of Daily Living(ADL)score at admission(OR=12.942),and length of hospital stay(OR=8.992)were independent risk factors for pulmonary infection in stroke patients(P<0.05).The risk prediction model formula was:Logit(P)=-4.761+2.348×(dysphagia score)+1.799×(coexisting pulmonary diseases score)+0.818×(hypokalemia score)+1.337×(hyponatremia score)+1.395×(low hemoglobin score)+3.641×(NIHSS score)+2.560×(ADL score)+2.196×(length of hospital stay score).The area under the ROC curve of the modeling group was 0.953[95%confidence interval(95%CI)0.940-0.967,P<0.001],with a Youden index of 0.762,a sensitivity of 0.880,a specificity of 0.882,and a P value of 0.553 in the H-L test.The validation results showed that the area under the ROC curve of the validation group was 0.946(95%CI:0.927-0.987,P<0.001),with a sensitivity of 0.898,a specificity of 0.875,an accuracy of 88.3%,and a P value of 0.510 in the H-L test.Conclusion The established risk prediction model for pulmonary infection in stroke patients has good predictive performance,providing a reference for clinical healthcare professionals to early identify high-risk groups for stroke induced pulmonary infection and facilitating the timely adoption of preventive management measures.