住院高血压脑出血患者并发呼吸机相关肺炎预测模型构建及验证

Construction and verification of prediction model of ventilator-associated pneumonia in hospitalized patients with hypertensive cerebral hemorrhage

目的 基于Logistic回归分析住院高血压脑出血(HICH)患者并发呼吸机相关肺炎(VAP)预测因素并构建预测模型。方法 回顾性分析成都大学附属医院2019年11月-2021年11月因高血压脑出血住院治疗患者160例的临床资料,根据患者住院期间是否发生VAP将其分为VAP组和非VAP组。分析患者年龄、性别、卒中史、吸烟、合并冠心病、合并糖尿病、合并慢性阻塞性肺疾病、反复吸痰、低蛋白血症、ICU住院时间和机械通气时间情况,采用多因素Logistic回归分析住院HICH患者并发VAP的影响因素,采用Hosmer-Lemeshow检验模型拟合度,采用MedCalc 11.4绘制受试者工作特征(ROC)曲线分析预测模型对住院HICH患者并发VAP风险的预测价值,获取曲线下面积(AUC)。结果 合并慢性阻塞性肺疾病、低蛋白血症、ICU住院时间和机械通气时间是住院HICH患者并发VAP的影响因素(P<0.05);根据多因素Logistic回归分析结果建立本研究预测模型,Logit(P)=-5.298+1.070×合并慢性阻塞性肺疾病+0.925×低蛋白血症+1.141×ICU住院时间+1.355×机械通气时间;采用Bootstarp重复抽样1 000次进行内部验证,得到平均绝对误差为0.0004,表明该模型具有较好的一致性。Hosmer-lemeshow检验结果显示该模型理论值与实际情况有较好的拟合度(χ~2=1.687,P=0.366);ROC分析结果显示本研究建立预测模型预测住院HICH患者发生VAP AUC为0.888,诊断敏感度为74.42%,特异度为88.89%(95%CI:0.829~0.933,P<0.001)。结论 根据合并慢性阻塞性肺疾病、低蛋白血症、ICU住院时间和机械通气时间建立预测模型对于住院HICH患者发生VAP有较好的预测价值。

OBJECTIVE Based on Logistics regression analysis, the prediction factors of hospitalized patients with hypertensive cerebral hemorrhage(HICH) complicated with ventilator-associated pneumonia(VAP) were analyzed and a prediction model was established. METHODS The clinical data of 160 hospitalized patients with hypertensive cerebral hemorrhage in our stroke center from Nov. 2019 to Nov. 2021 were retrospectively analyzed, and they were divided into VAP group and non-VAP group according to whether VAP occurred during hospitalization. The age, gender, stroke history, smoking, coronary heart disease, diabetes, chronic obstructive pulmonary disease, repeated sputum aspiration, hypoproteinemia, ICU stay and mechanical ventilation were analyzed. Multivariate Logistic regression was used to analyze the influencing factors of hospitalized patients with HICH complicated with VAP. Hosmer-Lemeshow was used to test the model fitting degree, and MedCalc 11.4 was used to draw receiver operating characteristic curve(ROC) analysis prediction model to predict the value of VAP risk in hospitalized patients with HICH and obtain the area under the curve(AUC). RESULTS Chronic obstructive pulmonary disease, hypoproteinemia, length of ICU stay and length of mechanical ventilation were all risk factors for hospitalization complicated with VAP in HICH patients(P<0.05). The prediction model of this study was established based on the results of multivariate Logistic regression analysis, Logit(P) =-5.298+1.070× chronic obstructive pulmonary disease +0.925× hypoproteinemia +1.141× length of ICU stay +1.355× mechanical ventilation time. The average absolute error was 0.0004, which indicated that the model had good consistency. The results of Hosmer-Lemeshow test showed that the theoretical value of the model was in good agreement with the actual situation(χ~2=1.687, P=0.366). ROC analysis showed that the predictive model established in this study predicted the occurrence of VAP AUC in hospitalized patients with HICH with 0.888, diagnostic sensitivity of 74.42%, specificity of 88.89%(95%CI: 0.829-0.933, P<0.001). CONCLUSION The prediction model based on chronic obstructive pulmonary disease, hypoproteinemia, length of ICU stay and length of mechanical ventilation has better predictive value for hospitalized patients with HICH complicated with VAP.