Development and Validation of a Prediction Model for Coronary Artery Disease in Chest Pain Patients: A Real-World Multicenter Study Based on Machine Learning

Coronary artery disease (CAD) is a prevalent condition among chest pain patients, and accurate prediction of the disease is crucial to ensure timely interventions and improve patient outcomes. We aim to elaborate a prediction model for CAD in chest pain patients using machine learning approaches. A retrospective analysis was performed using electronic health records of patients who presented with chest pain at seven hospitals. A total of 8474 patients were included in the study, where 63.25% were diagnosed with CAD. The data included demographic information, medical history, and laboratory results. Machine learning algorithms, including Random Forest, CatBoost, XGBoosting, Gradient Boosting, Light Gradient, AdaBoost, Ridge Classifier, Linear Discriminant, Logistic Regression, Decision Tree, SVM, Quadratic Discriminant, K Neighbors, Naive Bayes, and Dummy Classifier were trained and evaluated to predict the presence of CAD.The prediction model achieved an overall accuracy of 0.766 in identifying CAD in chest pain patients. The sensitivity and precision were 0.938 and 0.746, respectively. Important predictors for CAD included age, pulse rate, monocyte, and red cell distribution width SD. The eXtreme Gradient Boosting showed the best performance (area under the receiver operating characteristics, AUROC, 0.820, and 95% CI, 0.801–0.839) Additionally, the model demonstrated robust performance in the validation group. This study successfully developed and validated a prediction model for CAD in chest pain patients using machine learning techniques. The model exhibited good predictive ability and could aid in the early identification of CAD in clinical practice, potentially leading to appropriate interventions and improved patient outcomes.