Prediction of postoperative stroke in patients experienced coronary artery bypass grafting surgery: a machine learning approach
Shiqi Chen,Kan Wang,Chen Wang,Zhengfeng Fan,Lizhao Yan,Yixuan Wang,Fayuan Liu,JiaWei Shi,QianNan Guo,NianGuo Dong
DOI: https://doi.org/10.3389/fcvm.2024.1448740
IF: 3.6
2024-12-14
Frontiers in Cardiovascular Medicine
Abstract:Background: Coronary artery bypass grafting (CABG) surgery has been a widely accepted method for treating coronary artery disease. However, its postoperative complications can have a significant effect on long-term patient outcomes. A retrospective study was conducted to identify before and after surgery that contribute to postoperative stroke in patients undergoing CABG, and to develop predictive models and recommendations for single-factor thresholds. Materials and methods: We utilized data from 1,200 patients who undergone CABG surgery at the Wuhan Union Hospital from 2016 to 2022, which was divided into a training group ( n = 841) and a test group ( n = 359). 33 preoperative clinical features and 4 postoperative complications were collected in each group. LASSO is a regression analysis method that performs both variable selection and regularization to enhance model prediction accuracy and interpretability. The LASSO method was used to verify the collected features, and the SHAP value was used to explain the machine model prediction. Six machine learning models were employed, and the performance of the models was evaluated by area under the curve (AUC) and decision curve analysis (DCA). AUC, or area under the receiver operating characteristic curve, quantifies the ability of a model to distinguish between positive and negative outcomes. Finally, this study provided a convenient online tool for predicting CABG patient post-operative stroke. Results: The study included a combined total of 1,200 patients in both the development and validation cohorts. The average age of the participants in the study was 60.26 years. 910 (75.8%) of the patients were men, and 153 (12.8%) patients were in NYHA class III and IV. Subsequently, LASSO model was used to identify 11 important features, which were mechanical ventilation time, preoperative creatinine value, preoperative renal insufficiency, diabetes, the use of an intra-aortic balloon pump (IABP), age, Cardiopulmonary bypass time, Aortic cross-clamp time, Chronic Obstructive Pulmonary Disease (COPD) history, preoperative arrhythmia and Renal artery stenosis in descending order of importance according to the SHAP value. According to the analysis of receiver operating characteristic (ROC) curve, AUC, DCA and sensitivity, all seven machine learning models perform well and random forest (RF) machine model was found to perform best (AUC-ROC = 0.9008, Accuracy: 0.9008, Precision: 0.6905; Recall: 0.7532, F1: 0.7205). Finally, an online tool was established to predict the occurrence of stroke after CABG based on the 11 selected features. Conclusion: Mechanical ventilation time, preoperative creatinine value, preoperative renal insufficiency, diabetes, the use of an intra-aortic balloon pump (IABP), age, Cardiopulmonary bypass time, Aortic cross-clamp time, Chronic Obstructive Pulmonary Disease (COPD) history, preoperative arrhythmia and Renal artery stenosis in the preoperative and intraoperative period was associated with significant postoperative stroke risk, and these factors can be identified and modeled to assist in implementing proactive measures to protect the brain in high-risk patients after surgery.
cardiac & cardiovascular systems