Decoding risk: a machine learning model for predicting major adverse cardiovascular events post-acute type A aortic dissection surger

Abstract Background Acute type A aortic dissection (ATAAD) is a disease with high mortality. The occurrence of postoperative major adverse cardiovascular events (MACE) leads to a worse prognosis. Purpose This study aimed to identify the risk factors associated with postoperative MACE and develop a prediction model for MACE. Methods Between October 2018 and March 2023, 1779 patients diagnosed with ATAAD who underwent surgical treatment were included in this study. A total of 301 patients (16.9%) experienced postoperative MACE. Eight machine learning models were developed to predict postoperative MACE. The variables of the model were selected employing LASSO, XGBoost, and Random Forest algorithms. Furthermore, the model was validated using the 10-fold cross-validation method. The SHapley Additive explanation (SHAP) values were employed for elucidating the predictive model. Results The identified characteristic variables included pre-operation shock, duration of surgery, coronary artery involvement, cerebrovascular disease, and age. The results of the multi-model comparison show that the Logistics Regression model has the best prediction performance with an AUC of 0.729 (95% CI: 0.6955-0.763). The Hosmer–Lemeshow test indicated that the model demonstrated a good fit (P = 0.126), and the calibration curve closely approximated the ideal diagonal line. The decision curve analysis revealed a significantly greater net benefit associated with the model. Conclusion This study developed an accurate machine learning model to predict postoperative MACE in ATAAD patients. Based on the machine learning model, an online prediction platform was constructed to help clinicians identify the high-risk group of postoperative MACE in ATAAD patients and conduct early management.