A temporal deep learning algorithm for prediction of extubation failures in critical care patients
Authors' conclusion
Furthermore, it demonstrated an advantage over SBT procedures in promptly identifying patients ready for extubation.
Study context
- Population
- ICU patients undergoing extubation (planned post-extubation NIV or tracheostomy excluded); MIMIC-III and LOCAL-Ext datasetsMIMIC-III 5,895
- Comparison
- TrAcE vs other models; TrAcE extubation timing vs spontaneous breathing test (SBT)
- Outcome
- Prediction of extubation failure (AUROC, AUPRC)
Methodological detailsdeep learning (Transformer-based, temporal fusion), external test dataset, explainable (Captum occlusion)
- Deep learning (Transformer-based, temporal fusion)
- External test dataset
- Explainable (Captum occlusion)
Performance
As reported in the abstract
Extubation failure prediction
| Test or model | AUROC, validation (95% CI 0.808-0.838) | AUROC, testing (95% CI 0.836-0.881) | AUPRC, validation | AUPRC, testing |
|---|---|---|---|---|
| TrAcE | 0.823 | 0.859 | 0.734 | 0.757 |
TrAcE
AUROC, validation (95% CI 0.808-0.838)0.823
AUROC, testing (95% CI 0.836-0.881)0.859
AUPRC, validation0.734
AUPRC, testing0.757
Further findings
| Outcome | TrAcE | SBT | P value |
|---|---|---|---|
| Days to suitable extubation identified (LOCAL-Ext)TrAcE identified suitable extubation days earlier | 4.32 (IQR 0–20) | 5.65 (IQR 0–23) | P < 0.01 |
Original publication
A temporal deep learning algorithm for prediction of extubation failures in critical care patients. J Clin Monit Comput. Published 7 May 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →