基于吞咽造影定量评估的脑卒中后吞咽障碍并发吸入性肺炎危险因素分析与预测模型构建

关键词：吞咽造影  脑卒中后吞咽障碍  吸入性肺炎  预测模型

Keywords:videofluoroscopy swallowing study  post-stroke dysphagia  aspiration pneumonia  prediction model

摘要：

      摘要 目的：基于吞咽造影定量评估分析脑卒中后吞咽障碍患者发生吸入性肺炎的危险因素，构建预测模型并验证，以便规范管理脑卒中后吞咽障碍并发吸入性肺炎的高危患者。 方法：选择2021年7月—2023年6月北京中医药大学厦门医院收治的556例脑卒中后吞咽障碍患者为研究对象，分为建模组（436例）和外部验证组（120例）。建模组中，按照是否发生吸入性肺炎分为肺炎组113例（25.9%）和非肺炎组323例（74.1%），首先收集每位患者的临床资料与吞咽造影定量分析结果，采用logistic回归分析吸入性肺炎的危险因素，并构建列线图预测模型，然后通过绘制受试者工作特征（receiver operating characteristic，ROC）曲线、计算ROC曲线下面积（AUC），以及采用Hosmer-Lemeshow拟合优度检验评价模型的效能，最后进行外部验证，检验模型的灵敏度、特异度和一致性。 结果：logistic回归分析结果显示，年龄大、口腔清洁度差、气管切开、Rosenbek渗漏-误吸量表（PAS）评级≥5级、咽部残留及咽运送时间延长是发生吸入性肺炎的危险因素，主动咳嗽和进食指导是其保护因素（均P＜0.05）。列线图预测模型的预测风险能力指数（C-index）为0.957，AUC=0.957，灵敏度为93.8%，特异度为86.4%，Hosmer-Lemeshow拟合优度检验显示P=0.6。 结论：临床护理工作中，可采用该预测模型对脑卒中后吞咽障碍患者预测吸入性肺炎的发生，并依据分险程度制定针对性、个体化干预措施，从而提高护理质量。

Abstract:

      Abstract Objective： To analyze the risk factors for aspiration pneumonia in patients with post-stroke dysphagia based on quantitative analysis of videofluoroscopic swallowing study（VFSS）， and to construct a prediction model and validate it in order to standardize the management of high-risk patients with post-stroke dysphagia complicating aspiration pneumonia. Method： A total of 556 patients with post-stroke dysphagia admitted to Xiamen Hospital of Beijing University of Traditional Chinese Medicine from July 2021 to June 2023 were selected for the study and divided into a modeling group （n=436） and an external validation group （n=120）.The model group was divided into pneumonia group （n=113，25.9%） and non-pneumonia group （n=323，74.1%） according to the occurrence of aspiration pneumonia. Clinical data and VFSS results were collected for all participants. Logistic regression identified the risk factors for aspiration pneumonia， and a nomogram predictive model was constructed. Model performance was evaluated using receiver operating characteristic （ROC） curves， the area under the ROC curve (AUC), and the Hosmer-Lemeshow goodness-of-fit test. External validation was conducted to assess the sensitivity， specificity and consistency of the model. Result： Logistic regression analysis showed that advanced age， poor oral hygiene， tracheotomy， Penetration-Aspiration Scale (PAS) grade≥5， pharyngeal residue and prolonged pharyngeal transport time were risk factors for aspiration pneumonia， and active cough and eating instruction were protective factors （P<0.05 for both）. The predictive risk index （C-index） of the nomogram prediction model was 0.957， AUC=0.957， sensitivity was 93.8%， specificity was 86.4%， and the Hosmer-Lemeshow goodness of fit test showed P=0.6. Conclusion： The predictive model can be used to predict the occurrence of aspiration pneumonia in stroke patients with dysphagia. Its use in clinical practice allows for the development of targeted and individualized interventions according to patient risk levels， thereby improving the quality of nursing.