Integrating Clinical Information and Electrocardiographic Signal Features to Develop a Prediction Model for...

Postoperative delirium (POD) is a frequent and serious complication in older surgical patients. While conventional prediction models rely heavily on subjective clinical indicators, this study aimed to develop a simple, non-invasive preoperative risk prediction model by integrating objective physiological signals (electrocardiographic [ECG] features and heart rate variability [HRV]) with conventional demographic and clinical data.

In this single-center prospective observational study, 767 patients aged ≥65 years who underwent elective non-cardiac surgery were included. Preoperative clinical characteristics, conventional ECG abnormalities, and HRV parameters were extracted. Independent predictors were identified via multivariate logistic regression. A combined clinical-ECG/HRV model was developed, internally validated, and compared against single-domain baseline models.

POD occurred in 185 patients (24.1%). Advanced age, lower educational level, prolonged operative duration, non-sinus rhythm, ST-segment abnormalities, atrial/ventricular arrhythmias, and higher HRV fuzzy entropy were identified as independent risk factors for POD. The integrated clinical-ECG/HRV model achieved areas under the receiver-operating curves of 0.842 in the training set and 0.783 in the testing set, demonstrating higher predictive accuracy and improved risk reclassification compared with the baseline clinical model.

Preoperative ECG anomalies and increased HRV complexity (fuzzy entropy) are independently associated with an increased risk of POD. Integrating these objective, non-invasive physiological signals with routine clinical data yields a simple and practical prediction model. This multimodal approach enhances early preoperative risk stratification, providing valuable evidence to support comprehensive perioperative management for older surgical patients.