Haemodynamics, fluids & transfusion
Deep learning algorithms for predicting intraoperative hypotension: A systematic review and meta-analysis
Authors' conclusion
DL models for IOH prediction achieve moderate diagnostic accuracy, with hybrid CNN-RNN architectures and 5-min prediction windows showing the most favourable performance. Universal absence of formal calibration assessment, scarce external validation, and geographic concentration constrain immediate clinical translation.
Study context
- Population
- Studies of deep learning (DL)-based intraoperative hypotension (IOH) prediction modelsstudies included, nine contributed 22 validation datasets
- Participants
- Twelve studies
- Comparison
- DL-based IOH prediction models: pooled diagnostic accuracy
- Outcome
- Pooled diagnostic accuracy
Methodological detailsbivariate random-effects model, random-effects meta-regression, PROBAST+AI, GRADE, PRISMA-DTA
- Bivariate random-effects model
- Random-effects meta-regression
- PROBAST+AI
- GRADE
- PRISMA-DTA
Pooled result
As reported in the abstract
| Group | Sensitivity | Specificity | SROC-AUC | Diagnostic odds ratio | Positive likelihood ratio | Negative likelihood ratio |
|---|---|---|---|---|---|---|
| DL models | 0.78 (95% CI 0.73-0.81) | 0.88 (0.82-0.92) | 0.87 (0.83-0.90) | 24.7 (16.1-37.9) | 6.31 | 0.26 |
DL models
Sensitivity0.78 (95% CI 0.73-0.81)
Specificity0.88 (0.82-0.92)
SROC-AUC0.87 (0.83-0.90)
Diagnostic odds ratio24.7 (16.1-37.9)
Positive likelihood ratio6.31
Negative likelihood ratio0.26
Further pooled results
| Outcome | β (95% CI) | P value |
|---|---|---|
| Meta-regression: specificity by DL architectureHybrid CNN-RNN vs pure CNN | 1.77 (0.45–3.09) | P = 0.02 |
- Meta-regression: specificity by DL architecture: hybrid CNN-RNN exceeding pure CNN; DL architecture the only significant moderator of specificity
Heterogeneity
- τ2(Se)
- 0.25
- τ2(Sp)
- 1.04
- ρ
- −0.28
5-min prediction window (subgroup)
- Sensitivity
- 0.81
- Specificity
- 0.91
- Highest performance
Publication bias (Deeks' test)
- Non-significant
- P = 0.06
GRADE certainty
- GRADE certainty
- Low
Original publication
Deep learning algorithms for predicting intraoperative hypotension: A systematic review and meta-analysis. J Clin Anesth. Published 20 October 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →