Association between EuroSCORE II and postoperative delirium in cardiac surgery patients: a prospective cross-sectional study.

Background: Postoperative delirium (POD) is a significant neuropsychiatric issue after cardiac surgery, linked to increased morbidity, lengthened hospitalization, and adverse outcomes. EuroSCORE II is a validated tool for predicting surgical mortality, but its predictive utility for POD remains underexplored. This study focused on assessing the relationship between preoperative EuroSCORE II and the incidence of POD in adult patients undergoing elective cardiac surgery. Materials and methods: A prospective cross-sectional study was conducted, including 270 patients aged between 18 and 75 years scheduled for elective coronary artery bypass grafting, mitral valve replacement, or aortic valve replacement. Preoperative EuroSCORE II was calculated, and patients were stratified into four risk categories. POD was assessed using the Confusion Assessment Method for intensive care unit every 12 hours for 48 hours postoperatively. Models using logistic regression were employed to determine independent predictors of POD. Results: POD incidence was 4.3%. Patients in the moderate-to-high EuroSCORE II category had a significantly higher incidence of POD as compared to low-risk groups (P < 0.05). A higher EuroSCORE II was associated with greater delirium severity at all time points. Age was an independent predictor of POD (adjusted odds ratio = 1.19; 95% confidence interval: 1.09-1.32; P < 0.001), while diabetes and hypertension, although prevalent in higher-risk groups, were not independently predictive. Intraoperative variables such as cardio-pulmonary bypass, aortic cross-clamp times, and duration of surgery showed no significant association with POD. Conclusion: EuroSCORE II is significantly associated with both the incidence and severity of POD after cardiac surgery. It may be a useful component of preoperative risk stratification models for cognitive outcomes. However, its standalone predictive power is limited, and future models should integrate cognitive screening and frailty assessment to enhance accuracy.