关键词: 气管切开, 拔管, 危险因素, 预测模型

Abstract: Objective To identify the risk factors for extubation failure in patients after tracheotomy and to construct a risk prediction model based on a decision tree algorithm. Methods A total of 215 patients with tracheotomy who were hospitalized in the First Affiliated Hospital of Nanjing Medical University from February 2022 to April 2024 were retrospectively selected as the research subjects. They were divided into the successful extubation group(176 cases)and the failed extubation group(39 cases)according to whether the tracheotomy cannulas were successfully removed after the operation. Logistic regression was used to analyze and screen the related factors of postoperative extubation failure in patients with tracheotomy, and the decision tree model was constructed using R language software. Results The extubation failure rate was 18.14%; Logistic regression analysis showed that age ≥60 years, Glasgow Coma Scale(GCS)score ≤8, cough effectiveness grade 0~2, absent swallowing reflex, swallowing dysfunction, and abnormal hemoglobin levels were all independent risk factors for extubation failure after tracheotomy(P<0.05); Decision tree modeling identified swallowing reflex as the most influential factor for extubation failure in tracheotomy patients, with an information gain of 0.32; Both the logistic regression model and the decision tree model demonstrated good predictive performance for postoperative extubation failure risk, with areas under the receiver operating characteristic(ROC)curve of 0.848 and 0.823, respectively. Conclusion The construction of a risk prediction model, based on decision trees, has proven effective in the prediction ofextubation failure subsequent to tracheotomy.

Key words: Tracheotomy, Pull the tube, Risk factors, Prediction model