Abstract
AIMS: Heart transplantation is a critical life-saving procedure for patients with end-stage heart failure. However, predicting postoperative mortality remains challenging. The aim of this study is to examine the effectiveness of machine learning (ML) models for predicting 1-year mortality among heart transplant recipients in Saudi Arabia. METHODS AND RESULTS: A retrospective observational study was conducted using data from King Faisal Specialist Hospital & Research Centre, a large tertiary hospital in Saudi Arabia, that included all heart transplant cases from January 2007 to December 2022. We evaluate and compare the accuracy of support vector machine (SVM) and logistic regression (LR) models in predicting 1-year mortality. We also identify key predictive variables influencing mortality rates among recipients. SVM and LR models were developed and compared using accuracy, precision, recall, F1 score, and area under the receiver operating characteristic curve as performance metrics. The study analysed data from 419 patients, revealing that ischaemia time, devices like left ventricle assist device, extracorporeal membrane oxygenation, and body mass index (BMI) were significant mortality predictors. The LR model showed a testing accuracy of 96.43%, with weight and BMI having the strongest influence on mortality prediction. The SVM model had a testing accuracy of 95.24%, demonstrating consistent performance across dataset. CONCLUSION: The findings indicate that ML models, particularly SVM and LR, are effective in predicting 1-year mortality post-heart transplantation as well as identifying significant predictors of mortality. This research contributes to the global knowledge in heart transplant and highlights the importance of new technologies in tailoring healthcare strategies for the Saudi population.