Perioperative medicine & outcomes
Multimodal Machine Learning Model Predicting Postoperative Delirium Based on Heart Rate Variability: A Prospective Observational Study
Authors' conclusion
Our results indicate that integrating heart rate variability with clinical and electrocardiogram features significantly enhances the personalized predictive efficacy of postoperative delirium.
Study context
- Population
- Patients undergoing GA
- Participants
- 1,418 patients
- Exposure
- Combined clinical-ECG-heart rate variability model vs clinical-only and ECG-only models
- Outcome
- Postoperative delirium (Chinese version of the 3-Minute Diagnostic Interview for Delirium)
- Follow-up
- Within 3 days postoperatively
Methodological detailsmachine learning prediction model, external validation
- Machine learning prediction model
- External validation
Main finding
As reported in the abstract
Prediction of postoperative delirium
| Group | AUC |
|---|---|
| Clinical-ECG-HRV | 0.728 |
| Clinical-only | 0.673 |
| ECG-only | 0.679 |
AUC
Clinical-ECG-HRV0.728
Clinical-only0.673
ECG-only0.679
Observational data: associations, not effects.
Further findings
Best-discriminating machine learning algorithm
- Logistic regression showed the highest discrimination
External validation of the model
- AUC in the external validation set (robust performance maintained)
- 0.836
Key predictors (feature selection and Shapley additive explanations)
- Key predictors identified
- 17
- Core predictors highlighted by Shapley additive explanations: atrial or ventricular arrhythmia, operative time, ST-segment abnormalities, age, ASA classification, heart rate variability entropy, overall ECG abnormalities
- 7
Original publication
Multimodal Machine Learning Model Predicting Postoperative Delirium Based on Heart Rate Variability: A Prospective Observational Study. Anesth Analg. Published 13 July 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →