Tehran University of Medical Sciences Archives of Anesthesiology and Critical Care 2423-5849 0 0 2026 11 07 1642 1642 Zaid Hikmat Kareem Albayati mmn Baraa Adnan Rahman Altaee ghh Hana Askari Department of Cybersecurity Engineering Technologies, Al-Kut University, University of Kut, Wasit, Iraq. 2026 05 23 2026 06 16 Background: General anesthesia can lead to both early and late complications. Delirium is a recognized complication of anesthesia. This study was conducted to identify the factors contributing to the onset of delirium in patients undergoing anesthesia. Methods: This three-year longitudinal study investigated the incidence and clinical determinants of delirium in a cohort of 600 patients admitted to the intensive care unit (ICU) following general anesthesia. Patient data were retrospectively extracted from medical records, encompassing demographic variables (age and gender) alongside other clinical covariates. Delirium was screened using the Confusion Assessment Method for the ICU (CAM-ICU-7), a validated tool for identifying and quantifying the severity of delirium. Data collection and preliminary statistical analyses were performed using SPSS software. Subsequently, to enhance predictive accuracy, we implemented and compared four machine learning algorithms—Random Forest (RF), eXtreme Gradient Boosting (XGBoost), Deep Neural Network (DNN), and Logistic Regression (LR)—to evaluate their performance in forecasting postoperative delirium. Results: In this study of 600 ICU patients, the prevalence of delirium was 38%. Multivariable logistic regression identified GCS score, ocular trauma, and renal disease as the strongest independent predictors of delirium. Among the machine learning models, XGBoost demonstrated superior performance with an AUC-ROC of 0.88 and 86% accuracy. Feature importance analysis revealed that GCS and mechanical ventilation duration were the primary determinants of delirium risk. These findings highlight the clinical utility of baseline neurological and comorbidity assessment for the early identification and effective management of delirium in critically ill patients. Conclusion: This study confirms the superior effectiveness of the XGBoost machine learning model in accurately predicting delirium in ICU patients. The early identification of key risk factors, such as GCS score, renal disease, and specific trauma types, offers a significant opportunity to enhance preventive strategies and improve clinical outcomes in intensive care settings. https://aacc.tums.ac.ir/index.php/aacc/article/view/1642