Early postoperative endothelial activation and stress index (EASIX) as a novel biomarker for delirium and mortality after cardiac valve replacement surgery: a retrospective cohort study based on the MIMIC-IV database

Postoperative delirium (POD) following cardiac valve replacement surgery is a serious complication that impacts patient prognosis. However, effective early postoperative risk stratification tools integrating key pathophysiological mechanisms are currently lacking. Endothelial dysfunction and systemic inflammation are core mechanisms underlying POD, and the endothelial activation and stress index (EASIX), as a novel composite biomarker reflecting endothelial injury and inflammatory stress, has not yet been clearly established as a predictive tool in cardiac surgery. We conducted a retrospective cohort analysis of individuals undergoing cardiac valve replacement surgery based on the Medical Information Mart for Intensive Care IV (MIMIC-IV) database. Individuals were divided into quartile groups according to their early postoperative EASIX levels. The primary outcome was POD, and the secondary outcome was in-hospital mortality. Kaplan-Meier survival analysis and Cox proportional hazards models were adopted to appraise the association between EASIX and the outcomes, and restricted cubic spline (RCS) models were employed to ascertain the dose-response relationship. Furthermore, a random forest prediction model incorporating EASIX was constructed for POD within 28 days and internally validated using 10-fold cross-validation. Totally 1,543 individuals were finally included, with an incidence rate of POD of 36.2% and an in-hospital mortality rate of 4.3%. Cox regression analysis showed that a higher early postoperative EASIX level was an independent risk determinant for POD (adjusted hazard ratio (HR) = 1.185, 95% confidence interval (CI): 1.072–1.309; P < 0.001) and in-hospital mortality (adjusted HR = 1.521, 95% CI: 1.190–1.943; P < 0.001). The trend test revealed a significant increase in the risk of all outcomes across increasing EASIX quartiles (all P for trend < 0.05). RCS analysis demonstrated an approximately linear positive correlation between EASIX and risk. The random forest model showed moderate discrimination in 10-fold cross-validation (area under the curve (AUC) = 63.8%). In conclusion, early postoperative EASIX was substantially linked to POD and in-hospital mortality after cardiac valve replacement surgery. Being easily obtainable from routine tests, it holds potential for clinical translation and provides a novel and practical tool for early postoperative identification of high-risk patients.