Machine Learning-Based Random Forest to Predict 3-Year Survival after Endovascular Aneurysm Repair.

Purpose: Endovascular aneurysm repair (EVAR) is widely used to treat abdominal aortic aneurysms (AAAs), but mid-term survival remains a concern. This study aims to develop a machine learning-based random forest model to predict 3-year survival after EVAR.

Methods: A random forest model was trained using data from 176 EVAR patients, of whom 169 patients were retained for analysis, incorporating 23 preoperative and perioperative variables. Model performance was evaluated using 5-fold cross-validation.

Results: The model achieved an area under the receiver-operating characteristic curve (AUC) of 0.91, with an accuracy of 81.1%, a sensitivity of 81.6%, a specificity of 80.9%, and an F1 score of 0.66. Feature importance analysis identified poor nutritional status (geriatric nutritional risk index <101.4), compromised immunity (neutrophil-to-lymphocyte ratio >3.19), chronic kidney disease (CKD), octogenarian status, chronic obstructive pulmonary disease (COPD), small aneurysm size, and statin use as the top predictors of 3-year mortality. The average values of the AUC, accuracy, sensitivity, specificity, and F1 score across the 5-folds were 0.76 ± 0.10, 73.9 ± 5.8%, 60.4 ± 1.9%, 77.8 ± 0.7%, and 0.59 ± 0.17, indicating consistent performance across different data subsets.

Conclusions: The random forest model effectively predicts 3-year survival after EVAR, indicating key factors such as poor nutritional status, compromised immunity, CKD, octogenarian status, COPD, small aneurysm size, and statin use.