A nomogram for predicting postoperative delirium in patients with post-radical gastrectomy: a retrospective cohort study

Background postoperative delirium (POD), a common complication of post-radical gastrectomy, is an important risk factor for postoperative recovery and prognostic survival. Therefore, early identification of patients at high risk for POD is clinically important. Methods A total of 1337 gastric cancer patients with post-radical gastrectomy were enrolled in this study, with 54 parameters collected, including demographic characteristics, preoperative, intraoperative, postoperative and pathological data. Patients were randomly divided into a training cohort (n = 937) and a validation cohort (n = 400) in a 7:3 ratio. Univariate and multivariate logistic regression analyses were performed to identify predictors and construct a visual nomogram for POD risk prediction. Model performance was evaluated using area under curve (AUC) by receiver operating characteristic (ROC), calibration curve and decision curve analysis (DCA). Bootstrap resampling was further performed for internal validation and optimism correction. Results Among all patients, 189 (14.1%) developed POD. Multivariate logistic regression identified age, American Society of Anesthesiologists (ASA) III/IV, smoking, history of cerebrovascular disease, history of pulmonary disease, emergency surgery, and postoperative pain as independent risk factors ( p < 0.05). The nomogram demonstrated good discrimination, with an AUC of 0.779 in training cohort and 0.809 in validation cohort. Calibration curves and DCA demonstrated favorable agreement and clinical benefit in both cohorts. Bootstrap internal validation showed an optimism-corrected AUC of 0.817. Conclusions The internally validated nomogram based on seven independent predictors showed good predicted performance for POD risk assessment in patients with post-radical gastrectomy.