Objective To develop and validate a predictive model of postoperative delirium (POD) for patients undergoing cardiac surgery. Methods 935 patients, admitted to the cardiovascular surgery intensive care unit between 2023.3.1 and 2023.12.31, were enrolled into this study. They were divided into the training set and the internal validation set at a ratio of 7:3. The least absolute shrinkage and selection operator (LASSO) cox regression analysis was used to detect potential risk factors for POD and multivariable cox regression analysis was then conducted to construct the final nomogram model in the training set. Finally, we assessed the nomogram performance in terms of discrimination and calibration, including area under the curve (AUC), calibration curve and decision curve analysis (DCA) both in the training set and the internal validation set. Results 108 (16.34 %)patients in the training group and 47 (16.51 %) patients in the validation group developed POD respectively. Four independent risk factors were integrated into the nomogram including postoperative AKI, hypoalbuminemia, PSQI and CPOT score. The nomogram showed a good discrimination efficacy with an AUC of 0.882 (95 % CI: 0.843–0.922) and 0.921 (95 % CI: 0.871–0.972) in the training and internal validation set respectively. Both calibration curve and DCA demonstrated that this nomogram was clinically useful. Conclusions The nomogram based on four clinically available factors has good ability to identify POD risk in patients undergoing cardiovascular surgery. This tool may help physicians to prevent the appearance of delirium in advance using various methods and improve prognosis.

Unlabelled TablePODpostoperative deliriumBISbispectral indexCAM-ICUConfusion Assessment Method for the Intensive Care UnitRASSRichmond Agitation-Sedation ScaleLVEFleft ventricular ejection fractionPSQIPittsburgh Sleep Quality IndexCPBCardiopulmonary BypassMAPmean arterial pressureMVmechanical ventilationAKIacute kidney injuryKDIGOKidney Disease: Improving Global OutcomesCPOTCritical-Care Pain Observation ToolIQRinterquartile rangesLASSOleast absolute shrinkage and selection operatorAUCarea under the curveDCAdecision curve analysis