Association of combined stress hyperglycemia ratio and glycemic variability with 30-day and 360-day mortality after coronary revascularization: A MIMIC-IV Study with External Validation

<title>Abstract</title> <p> <bold>Background</bold> Although the roles of hyperglycemia and glycemic variability in the progression of coronary artery disease have been well established, their predictive value for adverse outcomes after coronary revascularization remains insufficiently explored. This study innovatively combines the stress hyperglycemia ratio (SHR) and glycemic variability (GV) to evaluate their ability to predict postoperative morbidity and mortality. <bold>Method</bold> We retrospectively analyzed patients undergoing percutaneous coronary intervention or coronary artery bypass grafting from the MIMIC-IV database (v3.1). Kaplan–Meier analysis, Cox regression, restricted cubic spline analysis, and machine learning models were used to assess prognostic performance. External validation was conducted in an independent cohort from Qilu Hospital, Shandong University (June 2023–June 2024). <bold>Result</bold> A total of 935 patients were included. Non-survivors had higher comorbidity burdens and GV levels. SHR alone was not significant for long-term survival, whereas GV showed predictive value. The combined SHR–GV assessment significantly stratified both 30-day and 360-day survival (both P &lt; 0.001). In Cox analyses, SHR independently predicted 30-day mortality, while both SHR and GV independently predicted 360-day mortality. Restricted cubic spline analysis showed superior prognostic performance of the combined SHR–GV model versus either marker alone. Among seven machine learning models, logistic regression achieved the best performance for 30-day mortality, while random forest performed best for 360-day mortality. External validation demonstrated AUCs of 0.765 and 0.807, respectively. <bold>Conclusions</bold> Combined assessment of SHR and GV provided superior prediction of short- and long-term mortality after coronary revascularization, with good external validity and clinical applicability. </p>