Zhongguo cuzhong zazhi (May 2024)
脑出血患者微创颅内血肿清除术后肺部感染的影响因素分析及预测模型构建 Analysis of Influencing Factors and Construction of Prediction Models for Pulmonary Infection after Minimally Invasive Intracranial Hematoma Removal in Patients with Intracerebral Hemorrhage
Abstract
摘要： 目的 探讨脑出血（intracerebral hemorrhage，ICH）患者微创颅内血肿清除术后住院期间肺部感染的影响因素并构建预测模型，以期为医务人员评估ICH患者术后肺部感染风险和制订预防策略提供指导。 方法 连续纳入2018年1月—2023年7月在苏州大学附属苏州九院神经外科接受微创颅内血肿清除术治疗的ICH患者，依据ICH患者术后住院期间是否出现肺部感染分为肺部感染组和无肺部感染组。比较2组患者的临床资料，采用多因素logistic回归分析探讨ICH患者微创颅内血肿清除术后肺部感染的影响因素并构建logistic回归模型。同时，使用R软件通过梯度提升机（gradient boosting machine，GBM）算法构建ICH患者微创颅内血肿清除术后肺部感染的GBM预测模型。采用ROC曲线分析2个模型的预测效能，并通过Delong检验进行比较。 结果 研究共纳入297例接受微创颅内血肿清除术治疗的ICH患者，其中术后住院期间发生肺部感染者52例（17.5%）。多因素logistic回归分析显示，营养不良（OR 2.737，95%CI 1.249～5.998，P=0.012）、气管切开（OR 2.716，95%CI 1.296～5.690，P=0.008）和留置胃管（OR 3.521，95%CI 1.724～7.193，P＜0.001）是ICH患者微创颅内血肿清除术后肺部感染的独立危险因素；术前GCS评分高（OR 0.622，95%CI 0.515～0.752，P＜0.001）和口腔处理（OR 0.105，95%CI 0.028～0.390，P＜0.001）则是其保护因素。ROC曲线分析显示，logistic回归模型的AUC为0.837，GBM预测模型的AUC为0.861。Delong检验表明GBM预测模型的效能优于logistic回归模型（Z=2.318，P=0.021）。 结论 营养不良、气管切开和留置胃管是ICH患者微创颅内血肿清除术后肺部感染的危险因素，术前GCS评分高和口腔处理则是其保护因素，基于上述指标构建的GBM预测模型效能优于logistic回归模型。 Abstract: Objective To investigate the influencing factors of pulmonary infection during hospitalization after minimally invasive intracranial hematoma removal in patients with intracerebral hemorrhage (ICH) and construct a prediction model, in order to provide guidance for medical staff to evaluate the risk of postoperative pulmonary infection in ICH patients and develop prevention strategies. Methods ICH patients who underwent minimally invasive intracranial hematoma removal at the Department of Neurosurgery, Suzhou Ninth Hospital Affiliated to Soochow University from January 2018 to July 2023 were continuously included. The ICH patients were divided into pulmonary infection group and non-pulmonary infection group according to whether they had pulmonary infection during hospitalization. The clinical data of the two groups were compared. Multivariate logistic regression analysis was used to explore the influencing factors of pulmonary infection after minimally invasive intracranial hematoma removal in ICH patients, and the logistic regression model was constructed. At the same time, the gradient boosting machine (GBM) algorithm of R was used to construct a GBM prediction model for pulmonary infection after minimally invasive intracranial hematoma removal in ICH patients. The prediction efficiency of the two models was analyzed by ROC curve and compared by Delong test. Results A total of 297 ICH patients who underwent minimally invasive intracranial hematoma removal were included in the study. Among them, 52 cases (17.5%) occurred pulmonary infection during postoperative hospitalization. Multivariate analysis showed that malnutrition (OR 2.737, 95%CI 1.249-5.998, P=0.012), tracheotomy (OR 2.716, 95%CI 1.296-5.690, P=0.008), and indwelling gastric tube (OR 3.521, 95%CI 1.724-7.193, P<0.001) were independent risk factors for pulmonary infection in ICH patients after minimally invasive intracranial hematoma removal. High preoperative GCS score (OR 0.622, 95%CI 0.515-0.752, P<0.001) and oral care (OR 0.105, 95%CI 0.028-0.390, P<0.001) were protective factors. According to ROC curve analysis, the AUC of the logistic regression model was 0.837, and the AUC of the GBM prediction model was 0.861. Delong test showed that the efficiency of GBM prediction model was better than that of logistic regression model (Z=2.318, P=0.021). Conclusions Malnutrition, tracheotomy and indwelling gastric tube were the risk factors of pulmonary infection in ICH patients after minimally invasive intracranial hematoma removal, while high preoperative GCS score and oral care were protective factors. The efficiency of GBM model based on the above indicators was better than logistic model.
Keywords
