Heart rate dynamics predict anaesthetic depth: a compact machine learning model
Authors' conclusion
High-dimensional descriptors of heart rate dynamics enabled accurate prediction of BIS >60 events.
Study context
- Population
- Surgical patients with ECG data
- Participants
- 3,338 patients
- Comparison
- 0 min vs 5 min vs 10 min vs 15 min
- Outcome
- Prediction of inadequate anaesthesia (BIS >60) from HR dynamics
Methodological detailsgradient boosting models, nested 10-fold cross-validation, feature reduction
- Gradient boosting models
- Nested 10-fold cross-validation
- Feature reduction
Performance
As reported in the abstract
Discrimination for BIS >60 events
0 min
0.953
AUC · 95% CI 0.949–0.958
5 min
0.917
AUC · 95% CI 0.909–0.925
10 min
0.910
AUC · 95% CI 0.905–0.916
15 min
0.903
AUC · 95% CI 0.894–0.912
Further findings
Compact model (27 features)
- Improvement in computational speed
- 110-fold
- High performance maintained across all time points
- Features predominantly captured fractal HR dynamics
Original publication
Heart rate dynamics predict anaesthetic depth: a compact machine learning model. Br J Anaesth. Published 20 November 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →