MACHINE LEARNING-BASED ONE-YEAR OUTCOME PREDICTION AFTER PERCUTANEOUS CORONARY INTERVENTION

Abstract

Cardiovascular disease (CVD), a leading non-communicable disease, remains one of the primary causes of mortality worldwide, with its prevalence increasing annually. In Thailand, CVD accounts for approximately 35% of total deaths, imposing a significant burden on the healthcare system (World Health Organization, 2021). Among the available treatment options, percutaneous coronary intervention (PCI) is widely performed to alleviate blood vessel blockages. However, the procedure carries notable risks, including in-hospital and post-discharge mortality. Accurate prediction of adverse outcomes following PCI is therefore critical for improving patient care and supporting clinical decision-making. Recently, machine learning techniques have emerged as powerful tools for predictive analysis and decision support in healthcare. This study aims to evaluate and compare the predictive performance of three widely used machine learning algorithms—Support Vector Machine (SVM), Random Forest (RF), and Extreme Gradient Boosting (XGBoost)—alongside traditional logistic regression in assessing in-hospital and one-year mortality outcomes in PCI patients. The results indicate that the Random Forest model demonstrated the highest predictive performance for both in-hospital and one-year mortality. Specifically, the in-hospital mortality model achieved an accuracy of 0.900 (95% CI: 0.890–0.909), while the one-year mortality model attained an accuracy of 0.778 (95% CI: 0.764–0.790). These findings highlight the potential of machine learning algorithms, particularly Random Forest, in providing reliable predictions and assisting clinicians in identifying high-risk patients undergoing PCI. In conclusion, this study underscores the effectiveness of machine learning in improving mortality outcome predictions and emphasizes its utility in clinical decision-making to enhance patient outcomes within Thailand's healthcare system.

-