Development and internal validation of a preoperative risk prediction model for postoperative delirium in older patients with hip fracture

Abstract Background Postoperative delirium (POD) is a common and serious complication in older patients undergoing hip fracture surgery and is associated with adverse long-term outcomes. Accurate preoperative risk stratification is essential for implementing targeted preventive strategies. Methods In this single-center retrospective cohort study, we developed and internally validated a preoperative risk prediction model for POD in patients aged ≥60 years with acute hip fracture (n = 1706; 341 POD events). The intended prediction time point was the preoperative assessment after admission but before surgery, once the delay status was known. POD was diagnosed using the Confusion Assessment Method. Ten predictors were prespecified based on prior literature and clinical relevance, and all were available before surgery. No univariable screening or stepwise selection was performed; all ten predictors were entered simultaneously into a single multivariable logistic regression model. Model performance was evaluated in terms of discrimination, calibration, and clinical utility in an internal test set. SHapley Additive exPlanations (SHAP) analysis was used as a visualization tool to illustrate individual predictor contributions. Results The final model included ten prespecified predictors. Preoperative cognitive impairment, Charlson Comorbidity Index, time from admission to surgery >48 h, age, albumin, and C-reactive protein were significantly associated with POD. The model demonstrated good discriminative ability, with an area under the receiver operating characteristic curve (AUC) of 0.84 (95% CI 0.82–0.87) in the training set (n = 1,194; 241 POD events) and 0.82 (95% CI 0.78–0.87) in the internal test set (n = 512; 100 POD events), with acceptable calibration. At the optimal cutoff determined by the Youden index (0.226), sensitivity was 76% and specificity 80% in the training set, and 68% and 76% in the internal test set. Patients in the high-risk group had an observed POD incidence exceeding 34%, compared with approximately 4–5% in the low-risk group. SHAP analysis illustrated the contribution of each predictor to individual risk estimates. Conclusion This study developed and internally validated a preoperative risk prediction model for POD in older patients with hip fracture. The model integrates ten readily accessible preoperative variables and may support preoperative risk stratification. External validation and prospective impact studies are required before clinical implementation.