Prediction model for the risk of postoperative death in patients with acute type A aortic dissection / 中华胸心血管外科杂志

ABSTRACT

Objective:Using different machine learning methods to construct and screen the best prediction model for predicting the risk of death within 30 days after surgery in patients with acute type A aortic dissection.Methods:Five hundred and twenty-one patients with acute type A aortic dissection who underwent surgery between 2015 and 2022 were included, after collecting their perioperative date and screening them, 329 patients were retained. two different groups of predictor variables were generated by using Lasso regression and principal component analysis, after that, logistic regression, support vector machine algorithm, random forest algorithm, gradient boosting algorithm, and super learning algorithm were used to develop prediction models for the risk of death within 30 days after surgery. Finally, we compare the models and select the best one. Results:The AUC values for all models rangrd from 0.791-0.959. The model using Lasso regression to determine the predictor variables and built by the super learning algorithm had the best prediction with an AUC value of 0.959. Conclusion:The super learning algorithm better than other algorithms in predicting death within 30 days after acute type A aortic dissection.