Construction and Validation of a Predictive Model for Postoperative Gastrointestinal Bleeding After Total Arch Replacement Surgery in Type A Aortic Dissection Based on Machine Learning
ZHAO Jianchao
ZHOU Lingling
PEI Yu
LIANG Shuang
Abstract：Objective To construct and validate a predictive model for postoperative gastrointestinal bleeding(GIB)after total arch replacement surgery in patients with type A aortic dissection(TAAD)using machine learning algorithms.Methods Clinical data of 736 TAAD patients who underwent total arch replacement surgery at the First Affiliated Hospital of Zhengzhou University from June 2021 to April 2024 were retrospectively analyzed.The least absolute shrinkage and selection operator(LASSO)was used to screen the influencing factors for postoperative GIB.The patients were randomly divided into a training set(515 cases)and a validation set(221 cases)at a ratio of 7∶3.Six widely used machine learning algorithms,including extreme gradient boost(XGBoost),light gradient boosting machine(LightGBM),logistic regression,Gauss Naive Bayes classification(GNB),support vector machine(SVM),and k-nearest neighbor(KNN),were used to construct predictive models for postoperative GIB in TAAD patients.The area under the receiver operating characteristic curve(AUC),calibration curve,precision-recall(PR)curve,and decision curve analysis(DCA)were used to evaluate each model.The Shapley additive interpretation(SHAP)was used to rank the importance of influencing factors.Results Among 736 patients,GIB occurred in 150 cases,with an incidence of 20.38％(150/736).There were statistically significant differences between the non-GIB group and the GIB group in terms of age,hypertension,preoperative SpO2＜95％,blood transfusion volume,low cardiac output syndrome(LCOS),ventilator weaning time＞72 h,infection,continuous renal replacement therapy(CRRT),and ICU stay(P＜0.05).Among the six predictive models constructed based on the influencing factors screened by LASSO regression,the logistic regression model demonstrated good performance,with an AUC of 0.859 in the validation set.The calibration curve and DCA showed that the logistic regression model had high accuracy and net benefit.SHAP analysis showed that the influencing factors for postoperative GIB in TAAD patients,ranked in order of importance,were blood transfusion volume,age,infection,CRRT,LCOS,and ventilator weaning time＞72 h.Conclusion The logistic regression model constructed based on blood transfusion volume,age,infection,CRRT,LCOS,and ventilator weaning time＞72 h is the optimal predictive model for postoperative GIB after total arch replacement surgery in TAAD patients,which can help clinical staff better identify the risk of GIB.
Keywords：type A aortic dissectiontotal arch replacementgastrointestinal bleedingmachine learning algorithmpredictive model
Publication Date：2026-05-28
Online Publishing Date：2026-09-12(First online date of this platform, not the publication date of the document)
Pages：6( 1770-1775 )