Pediatric anesthesia & critical care
Interoperable Models for Identifying Critically Ill Children at Risk of Neurologic Morbidity
Authors' conclusion
This prognostic study of prediction models for detecting neurologic morbidity in critically ill children demonstrated a well-performing ensemble of models with biomolecular corroboration.
Study context
- Population
- Critically ill children admitted to a quaternary PICUencounters
- Participants
- 18,568 vs 6,825
- Comparison
- Development site vs Validation site
- Development siten = 18,568
- Validation siten = 6,825
- Outcome
- Neurologic morbidity (computable composite definition at development site; neurocritical care consultation order at validation site)
Methodological detailsmachine-learning model development, external validation
- Machine-learning model development
- External validation
Performance
As reported in the abstract
Performance of generalizable extreme gradient boosted model (24-hour time horizon, 48-hour feature window) at validation site
| Test or model | F1 score | AUC | Number needed to alert | Brier score after recalibration |
|---|---|---|---|---|
| Validation site | 0.37 (95% CI, 0.33-0.40) | 0.81 (95% CI, 0.78-0.83) | 4 | 0.04 |
Validation site
F1 score0.37 (95% CI, 0.33-0.40)
AUC0.81 (95% CI, 0.78-0.83)
Number needed to alert4
Brier score after recalibration0.04
Further findings
Correlation of serum glial fibrillary acidic protein with model output (convenience cohort)
- Significant correlation (rs)
- rs = 0.34; P = .007
Original publication
Interoperable Models for Identifying Critically Ill Children at Risk of Neurologic Morbidity. JAMA Netw Open. Published 3 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →