The Prediction Model for Postoperative Delirium in Stanford Type A Aortic Dissection Based on Random Forest Algorithm
LIU Chunyan
LI Jingjing
WU Guiqin
ZHANG Pingzhen
LI Songjun
ZHOU Jinling
Abstract：Objective:To analyze the risk factors of postoperative delirium in patients with Stanford type A aortic dissection and to establish a random forest model.Methods:A total of 185 patients with Stanford type A aortic dissection in our hospital from January 2016 to October 2023 were divided into the delirium group and the non-delirium group according to whether delirium occurred after the operation.The clinical datas of all patients were collected.Multivariate Logistic regression was used to screen the risk factors of postoperative delirium.A random forest model for predicting postoperative delirium in Stanford type A aortic dissection was established using R software.Results:Among 185 patients with Stanford type A aortic dissection,postoperative delirium occurred in 33 patients with Stanford type A aortic dissection,while no postoperative delirium occurred in 152 patients with Stanford type A aortic dissection.The incidence of postoperative delirium was 17.84％(33/185).There were statistically significant differences in hypoxemia,D-dimer,blood lactate,operation time,deep hypothermic circulatory arrest time and mechanical ventilation time between the delirium group and the non-delirium group(P＜0.05).Hypoxemia,D-dimer＞15.46 mg/L,blood lactate＞2.87 mmol/L,operation time＞265.81 min,deep hypothermia circulatory arrest time＞44.32 min,and mechanical ventilation time＞64.83 h were independent risk factors for postoperative delirium in patients with type A aortic dissection(P＜0.05).The ranking of relatively important predictors for predicting postoperative delirium were operation time,mechanical ventilation time,deep hypothermia circulatory arrest time,blood lactate,D-dimer,and hypoxemia.The average reduction in the variable Gini value was proportional to its importance in the model.The area under the receiver operating characteristics(ROC)curve of the random forest algorithm for predicting postoperative delirium was higher than that of the multivariate Logistic regression model(Z=2.296,P=0.022).Conclusion:Hypoxemia,D-dimer＞15.46 mg/L,blood lactate＞2.87 mmol/L,operation time＞265.81 min,deep hypothermia circulatory arrest time＞44.32 min,and mechanical ventilation time＞64.83 h were the influencing factors for postoperative delirium in patients with Stanford type A aortic dissection.The random forest model was constructed based on the above factors with better risk prediction efficiency.
Keywords：Stanford type A aortic dissectionpostoperative deliriuminfluencing factorsrandom forest model
Publication Date：2025-07-25
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 2168-2174 )