Postoperative delirium risk in elderly high-risk patients undergoing hip and knee arthroplasty (ASA Ⅲ-Ⅳ): construction and validation of a nomogram model based on frail frailty index
TL;DR
The nomogram, which is based on six clinical factors, successfully predicts POD risk in elderly arthroplasty patients, exhibiting good discrimination, calibration, and utility for early risk identification and customized prevention.
Abstract
Objective To create and verify a nomogram for predicting postoperative delirium (POD) in elderly patients having knee or hip replacements based on the FRAIL frailty index. Methods 582 elderly patients with American Society of Anesthesiologists (ASA) grades III-IV who had elective primary total hip or knee arthroplasty at Yinchuan Guolong Orthopedic Hospital between January 2022 and June 2025 were the subject of a retrospective study. The patients were randomly assigned to a training set (n = 407) and a validation set (n = 175) at a 7:3 ratio. Least absolute shrinkage and selection operator regression was used to screen candidate variables, followed by multivariable logistic regression to identify independent risk factors for POD. A nomogram was constructed based on these factors. Area under the receiver operating characteristic curve, calibration curves, and decision curve analysis (DCA) were used to assess model discrimination, calibration, and clinical utility in both cohorts. SHapley Additive exPlanations (SHAP) analysis was used to improve model interpretability. Results Preoperative serum albumin (OR = 0.901, 95% CI: 0.830–0.978), age (OR = 1.092, 95% CI: 1.023–1.166), preoperative frailty index (OR = 2.021, 95% CI: 1.612–2.534), ASA grade IV (OR = 2.105, 95% CI: 1.131–3.917), preoperative Chinese Mini Mental Status <27 (OR = 3.608, 95% CI: 1.878–6.931), and early postoperative Numerical Rating Scale (NRS) score ≥6 (OR = 5.998, 95% CI: 3.028–11.880) emerged as six independent risk factors for POD. With AUCs of 0.864 (95% CI: 0.819–0.909) in the training set and 0.875 (95% CI: 0.800–0.949) in the validation set, the nomogram demonstrated good discrimination. Additionally, calibration curves demonstrated good agreement with Brier scores of 0.107 and 0.094, respectively. Clinical net benefit was verified by DCA over a broad range of threshold probabilities (8%–45%). With directional effects in line with clinical knowledge, SHAP analysis identified the frailty index as the most significant predictor, followed by postoperative NRS and preoperative albumin. Conclusion The nomogram, which is based on six clinical factors, successfully predicts POD risk in elderly arthroplasty patients, exhibiting good discrimination, calibration, and utility for early risk identification and customized prevention.