A machine learning-based prediction model for in-hospital mortality in male patients with acute type A aortic dissection

This study aimed to develop and validate a machine learning (ML)-based model for predicting in-hospital mortality after surgical treatment in male patients with acute type A aortic dissection (ATAAD). We retrospectively collected clinical data from patients who underwent ATAAD surgery at the Affiliated Union Hospital of Fujian Medical University between January 2017 and September 2025. Eligible patients were randomly allocated into training and testing sets at an 8:2 ratio. Univariate and multivariate logistic regression analyses were used to identify significant predictors. Three machine learning algorithms—Logistic Regression (LR), Neural Network, and Support Vector Machine (SVM)—were used to construct prediction models. Model performance was evaluated using the area under the receiver operating characteristic curve (AUC). Calibration curves and decision curve analysis (DCA) were applied to assess calibration and clinical utility, respectively. Model interpretability was explored using SHAP (Shapley Additive exPlanations) analysis. A total of 700 male ATAAD patients were included. Univariate and multivariate logistic regression identified four independent predictors for model construction. The AUC values in the training and test sets, respectively, were 0.784 and 0.684 for LR, 0.795 and 0.778 for the neural network, and 0.785 and 0.728 for SVM. The neural network yielded the numerically highest AUC in both sets, although the test-set AUC differences did not reach statistical significance after Holm adjustment. Furthermore, calibration curves and DCA suggested reasonable calibration and potential clinical utility within the present cohort. SHAP analysis revealed that surgery time, fibrinogen (FIB), age, and pan-immune-inflammation value (PIV) were the most influential predictors. The neural network showed promising discrimination for postoperative in-hospital mortality in male patients with ATAAD, but its comparative performance and clinical applicability require further evaluation in larger, independent cohorts. Not applicable.