Table 4.
Comparison of model performance using discrimination and reclassification metrics.
| Study | Outcome | Best model | Validation | AUC (95%CI) | Calibration metrics | Reclassification metrics | Additional performance metrics |
|---|---|---|---|---|---|---|---|
| Early mortality | |||||||
| Chen, 2025 | 30-day mortality | SVM | Hold-out | 0.782 (0.698, 0.860) | Brier = 0.058; Good calibration curve fitting; HL P = 0.949 | NRI = 0.143; IDI = 0.062 | NI |
| Guo, 2021 | In-hospital mortality | XGBoost | 10-fold CV | 0.927 (0.860, 0.968) | NI | NI | Acc: 91.8%; Prec: 68.3% |
| Jiang, 2023 | 30-day mortality | BSS | 5-fold CV | 0.841 (0.771, 0.911) | NI | NI | Acc: 86.6%; Prec: 76.9% |
| Lei, 2024 | In-hospital mortality | RF | Hold-out + External | 0.870 (0.744, 0.996) | NI | NI | NI |
| Li, 2024 | 30-day mortality | SL | NI | 0.959 (0.927, 0.992) | Brier = 0.046 | NI | Sens: 98.9%, Spec: 75.9% |
| Song, 2024 | In-hospital mortality | LR | Hold-out | 0.856 (0.791, 0.922) | Good calibration curve fitting; HL P = 0.804 | NI | NI |
| Wu, 2023 | In-hospital mortality | XGBoost | Hold-out | 0.926 (0.855, 0.997) | Good calibration curve fitting | NI | NI |
| Zhang, 2025 | 30-day mortality | PSO-ELM-FLXGBoost | Hold-out | 0.869 | NI | NI | NI |
| Zhang, 2025 | 30-day mortality | XGBoost | Hold-out + External | 0.941 (0.915, 0.967) | NI | NI | Acc: 90.1%, Sens: 81.4%, Spec: 92.1%, Prec: 70.6% |
| Long-term mortality | |||||||
| Cai, 2025 | Mortality at follow-up | SVM | Hold-out + 10-fold CV + External | 0.877 (0.870, 0.898) | Brier = 0.132 | NI | Acc: 87.7%, Prec: 82.4% |
| Guo, 2022 | 1-year mortality | NN | 10-fold CV | 0.899 (0.824, 0.975) | Brier = 0.120 | NI | Acc: 87.1%, Sens: 65.0%, Spec: 92.3% |
| He, 2025 | Mortality and aortic-related composite endpoint at follow-up | StepCox[backward] + RSF | Hold-out | 0.826 (0.760, 0.891) | Good calibration curve fitting | NI | NI |
| Sun, 2024 | 3-year mortality | RSF | Hold-out | 0.756 | Brier = 0.137 | NI | NI |
| Zhang, 2024 | 1-year mortality | Treebag | 5-fold CV | 0.910 (0.841, 0.962) | Brier = 0.128 | NI | Sens: 81.6% |
| Acute kidney injury | |||||||
| Chen, 2025 | AKI | LightGBM | 10-fold CV | 0.874 (0.831, 0.918) | Brier; Good calibration curve fitting | NI | NI |
| Dai, 2023 | AKI | XGBoost | 10-fold CV | 0.800 (0.683, 0.917) | Brier | NI | Sens: 71.7% |
| Li, 2022 | AKI | RF, LightGBM | Bootstrap | 0.760 (0.630, 0.881) 0.734 (0.623, 0.847) | Brier = 0.160 and 0.150; Good calibration curve fitting | NI | NI |
| Liu, 2024 | AKI | ANN | 10-fold CV + Bootstrap | 0.916 | Good calibration curve fitting | NI | Sens: 94.0%, Spec: 81.0% |
| Wei, 2025 | AKI | CatBoost | 10-fold CV | 0.712 (0.604, 0.791) | NI | NI | Acc: 72.1%, Sens: 60.0%, Spec: 75.0%, Prec: 36.8% |
| Neurological complications | |||||||
| Chen, 2023 | Cerebral complications | RF | Hold-out | 0.828 (0.585, 0.902) | Brier = 0.139 | NI | Acc: 88.4%, Sens: 83.3%, Spec: 89.2% |
| Zhao, 2021 | Pre-operation AIS | DNN | Hold-out | 0.964 (0.932, 0.997) | NI | NI | Acc: 92.1%, Sens: 96.0%, Spec: 90.6% |
| Gastrointestinal bleeding | |||||||
| Li, 2025 | postoperative GIB | RF | 5-fold CV | 0.933 (0.840, 0.997) | Good calibration curve fitting | NI | NI |
| Composite outcome | |||||||
| Lu, 2024 | Postoperative adverse outcomes | XGBoost | Hold-out + External | 0.985 (0.965, 1.000) | Good calibration curve fitting | NI | Acc: 92.0%, Sens: 94.0%, Spec: 91.0%, Prec: 92.0% |
| Luo, 2025 | Major adverse outcomes | RSF + GBM | Hold-out | 0.882 | Brier = 0.138 | NI | NI |
| Pei, 2021 | MAEs | LR | Hold-out | 0.766 (0.718, 0.734) | NI | NRI = 0.653; IDI = 0.654 | Acc: 77.0%, Prec: 65.3% |
| Wang, 2023 | Adverse aortic events | CART | 10-fold CV | 0.852 | Good calibration curve fitting; HL P = 0.586 | NI | Acc: 88.9%, Sens: 77.8%, Spec: 92.6% |
| Xie, 2024 | Postoperative adverse outcomes | XGBoost | 10-fold CV | 0.716 (0.668, 0.854) | Brier = 0.192; Good calibration curve fitting | NI | NI |
| Xie, 2024 | MAEs | RF | 10-fold CV | 0.975 (0.940, 1.000) | Brier = 0.078; Good calibration curve fitting | NI | Acc: 88.7% |
| Other specified complications | |||||||
| Chen, 2021 | LOS | RF | 5-fold CV | 0.837 (0.766, 0.908) | NI | NI | NI |
| Dong, 2021 | reintervention | LR | 5-fold CV + Bootstrap | 0.802 (0.689, 0.915) | Good calibration curve fitting | NI | Acc: 78.1%, Sens: 75.0%, Spec: 82.5% |
| Jin, 2025 | MMP | RF | 5-fold CV | 0.797 (0.794, 0.800) | Brier = 0.133; Good calibration curve fitting | NI | Acc: 74.0%, Sens: 81.1%, Spec: 72.2% |
| Jin, 2025 | MMP | Deep learning model | 5-fold CV + External | 0.780 (0.777, 0.785) | Brier = 0.143 | NI | Acc: 76.0%, Sens: 66.7%, Spec: 78.3% |
| Li, 2025 | Postoperative CRRT | XGBoost | 5-fold CV | 0.960 (0.930, 0.990) | Good calibration curve fitting | NI | Acc: 96.0%, Sens: 93.0%, Spec: 96.0% |
| Li, 2025 | Prolonged LOS (exceeding 30 days) | XGBoost | Hold-out | 0.710 (0.620, 0.800) | Good calibration curve fitting | NI | Acc: 71.0%, Sens: 76.0%, Spec: 84.0% |
| Lin, 2023 | Death due to the dissection rupturing within 72 h after the CTA | CNN | Bootstrap | 0.990 (0.950, 0.990) | Good calibration curve fitting; HL P = 0.91 | NI | Acc: 90.0%, Spec: 90.0%, Prec: 90.0% |
| Ma, 2024 | AGI | RF | NI | 0.990 (0.966, 1.000) | NI | NI | NI |
| Pang, 2024 | Complicated TBAD | LR | Hold-out | 0.818 (0.641, 0.995) | NI | NI | NI |
| Wang, 2023 | MS-ARDS within 24 h | LR | Hold-out | 0.797 | NI | NI | Acc: 80.0%, Prec: 66.3% |
| Wei, 2025 | postoperative pulmonary complications | LightGBM | 10-fold CV | 0.763 (0.667, 0.859) | HL P = 0.32 | NI | Acc: 75.3%, Sens: 65.0%, Spec: 82.0%, Prec: 70.3% |
| Wen, 2025 | Postoperative reintubation | XGBoost | 10-fold CV | 0.969 | Good calibration plots fitting | NI | Acc: 86.5%, Sens: 88.9%, Spec: 86.1%, Prec: 49.0% |
AUC, area under the curve; ANN, Artificial Neural Network; BSS, Best Subset Selection; BPNN, Back Propagation Neural Networks; CART, Classification and Regression Trees; CNN, Convolutional Neural Network; CNB, Categorical Naive Bayes; DT, Decision Tree; DNN, deep neural network; ELM, Extreme Learning Machine; GBDT, Gradient Boosting Decision Tree; GBM, Gradient Boosting Machine; GNB, Gaussian Naive Bayes; IDI: integrated discrimination improvement; KNN, K-Nearest Neighbors; K-SVM, Kernel Support Vector Machine; LASSO, Least Absolute Shrinkage and Selection Operator; LR: logistic regression;LightGBM, Light Gradient Boosting Machine; MLP, Multilayer Perceptron; MLR, Multiple Logistic Regression; NB, Naive Bayes; NI, Not Provided; NN, Neural Network; NNET, Neural Network; NRI, Net Reclassification Improvement; PCA, Principal Component Analysis; PMM, Predictive Mean Matching; PSO, Particle Swarm Optimization; RF, Random Forest; RFE, Recursive Feature Elimination; RSF, Random Survival Forest; Ridge, Ridge Regression; SL, Super Learner; SVM, Support Vector Machine; SVC, Support Vector Classifier; SuperPC, Supervised Principal Components; XGBoost, Extreme Gradient Boosting.