Abstract

Background: Cardiovascular diseases, particularly Coronary Artery Disease (CAD), remain the leading cause of death worldwide, imposing significant health and economic burdens. It is crucial to emphasize early diagnosis of CAD to prevent complications and improve patient outcomes. This study aims to predict the likelihood of CAD recurrence within 6 months post-treatment.
Methods: The Medical Information Mart for Intensive Care (MIMIC-III) database was used to perform a retrospective study. Predictive features include demographic data and laboratory test results. A 6-month CAD recurrence was set as the study outcome. We used the Machine Learning (ML) Methods Of Logistic Regression (LR), Random Forest (RF) and Extreme Gradient Boosting (XGBoost) to develop a predictive model for CAD recurrence. The prognostic capacity and clinical utility of these three models were compared using the Area Under the Receiver Operating Characteristic Curves (AUROC), precision, sensitivity, specificity, f1 measure and Area Under Precision -Recall (AUPR) curve.
Results: Of 7,583 CAD patients in this study population, 2,361 (31%) had CAD recurrence during 6-month follow-up. Out of 38 features selected and extracted from the MIMIC III database, 15 variables were chosen using stepwise regression. The RF model performed best with an AUC of 0.83. The top 6 significant features in our model were platelet, WBC, RBC, INR, chloride, and creatinine.
Conclusion: Our study shows that the random forest model outperforms the XGBoost and LR models in predicting CAD recurrence within 6 months post-treatment. The study suggests a connection between certain lab indices (platelet count, WBC, RBC, INR, chloride, calcium, creatinine) and CAD recurrence, bridging knowledge gaps and guiding future research on preventive strategies and treatments for CAD.

Keywords: Prediction; Machine learning; Coronary artery disease; Recurrence; MIMIC-III database