基于随机森林算法建立非急诊大手术后延迟拔管的预测模型

Predictive model for extubation delay undergoing non-emergency major surgery based on random forest algorithm

中文关键词: 随机森林  大手术  延迟拔管  危险因素  预测模型

英文关键词: Random forest  Major surgery  Extubation delay  Risk factors  Prediction model

中文摘要:

      
目的：基于随机森林算法分析非急诊大手术后延迟拔管的影响因素，建立并验证术后延迟拔管的预测模型。
方法：回顾性收集2018年1月至2022年12月全麻下行非急诊大手术的7 528例患者的临床资料。根据术后2 h内是否拔管，将患者分为两组：非延迟拔管组（≤2 h）和延迟拔管组（>2 h）。将患者按照7∶3分为训练集和验证集，通过LASSO回归、Logistic回归筛选术后延迟拔管的预测因素，采用随机森林算法建立并验证预测模型。
结果：有123例（1.6%）出现术后延迟拔管。ASA分级、科室、术中使用氟比洛芬酯、右美托咪定、激素、术中出现低钙血症、重度贫血、术中输血、气道痉挛是术后延迟拔管的独立预测因素。基于随机森林算法建立的预测模型在验证集中的曲线下面积（AUC）为0.751（95%CI 0.742~0.778），敏感性98.1%，特异性41.9%。
结论：基于随机森林算法建立的非急诊大手术后拔管延迟的预测模型具有较好的预测性能，利用该模型有助于预防非急诊大手术后延迟拔管。

英文摘要:

      
Objective: To construct and validate a clinical prediction model for delayed extubation undergoing non-emergency major surgery based on the random forest algorithm.
Methods: Clinical data of 7 528 patients undergoing non-emergency major surgery under general anesthesia from January 2018 to December 2022 were retrospectively collected. The patients were divided into two groups according to whether extubation was performed within 2 hours after surgery: non-delayed extubation group (≤ 2 hours) and delayed extubation group (> 2 hours). All the patients were randomly divided into a training set and a validation set in a ratio of 7∶3. The predictive factors for delayed extubation after surgery were screened through LASSO regression and Logistic regression. The random forest model was established and verified by random forest algorithm.
Results: There were 123 patients (1.6%) experienced delayed extubation after surgery. ASA physical status, department, intraoperative use of flurbiprofen ester, dexmedetomidine, glucocorticoid, hypocalcemia, severe anemia, intraoperative blood transfusion, and airway spasm were identified as independent predictive factors for delayed extubation. The area under curve (AUC) value of the random forest prediction model in the validation set was 0.751 (95% CI 0.742-0.778), and the sensitivity was 98.1%, and the specificity was 41.9%.
Conclusion: The predictive model of delayed extubation undergoing non-emergency major surgery based on random forest algorithm has a good predictive value, which may be helpful to prevent delayed extubation undergoing non-emergency major surgery.