JMIR Publications

ABSTRACT

Background:

As a major health hazard, the incidence of coronary heart disease has been increasing year by year. Although coronary revascularization, mainly percutaneous coronary intervention (PCI), has played an important role in the treatment of coronary heart disease, major adverse cardiovascular events (MACE) such as recurrent or persistent angina pectoris after coronary revascularization remain a very difficult problem in clinical practice.

Objective:

While there is a high probability of MACE after coronary revascularization, this study aims to develop and validate a predictive model for MACE occurrence within six months based on machine learning algorithms.

Methods:

A retrospective study was performed that contains 1004 patients who have undergone coronary revascularization at The People’s Hospital of Liaoning Province and Affiliated Hospital of Liaoning University of Traditional Chinese Medicine from June 2019 to December 2020. According to the characteristics of available data, an over-sampling strategy was adopted for pre-processing at first, and then we employed four machine learning algorithms, including decision tree (DT), random forest (RF), support vector machine (SVM), and extreme gradient boosting (XGBoost), depended on clinical information and six-month follow-up information to develop prediction models for MACE. Among all samples, 70% were randomly selected for training and the remaining 30% were used for model validation. Metrics like the accuracy, recall, F1_socre, area under curve (AUC), and receiver operating characteristic curve (ROC) were taken to assess the performance of models.

Results:

Univariate analysis showed that a total of 21 characteristic variables were statistically significant with a p-value less than 0.05 between the group without MACE and with MACE. Coupled with these significant factors, four machine learning algorithms were developed respectively, among which XGBoost stood out with the accuracy of 0.7788, recall of 0.7345, F1_score of 0.7685, and AUC of 0.8599. Further explorations of the built models identified factors affecting the occurrence of MACE and revealed that the use of anticoagulant drugs and course of the disease ranked top 2 in three developed prediction models without fail.

Conclusions:

The machine learning risk models constructed in this study can achieve the acceptable performance of MACE prediction with XGBoost performing best, which is valuable to provide a reference for the pointed intervention and clinical decision-making in MACE prevention.