Respiratory failure & ventilation
Development and External Validation of Models to Predict Persistent Hypoxemic Respiratory Failure for Clinical Trial Enrichment*
Authors' conclusion
Parsimonious, interpretable models that predict persistent HRF may improve enrichment of trials testing HRF-targeted therapies
Study context
- Population
- Adults with acute hypoxemic respiratory failure (new invasive mechanical ventilation and hypoxemia on first day after ICU admission), medical and surgical ICUs at two US centresderivation n = 630, validation n = 511
- Comparison
- Least absolute shrinkage and selection operator (LASSO) clinical model and LASSO + IL-6 model vs PaO2/FiO2 alone
- Outcome
- Persistent HRF (ongoing invasive mechanical ventilation and hypoxemia on the third calendar day after admission); model discrimination (AUC)
- Follow-up
- Third calendar day after admission
Methodological detailsprospective cohorts, two-centre
- Prospective cohorts
- Two-centre
Performance
As reported in the abstract
Discrimination (AUC) in validation
| Test or model | AUC | 95% CI |
|---|---|---|
| LASSO | 0.68 | 0.64-0.73 |
| LASSO + IL-6 | 0.71 | 0.66-0.76 |
| PaO2/FiO2 | 0.64 | 0.59-0.69 |
AUC
LASSO0.68
LASSO + IL-60.71
PaO2/FiO20.64
95% CI
LASSO0.64-0.73
LASSO + IL-60.66-0.76
PaO2/FiO20.59-0.69
Further findings
Incidence of persistent hypoxemic respiratory failure (HRF)
- Derivation set
- 49%
- Validation set
- 69%
Calibration
- Both models underestimated risk in lower risk deciles but showed better calibration at relevant risk thresholds
Practical utility (decision curve and enrichment analysis)
- LASSO and LASSO + IL-6 showed greater net benefit and greater sample size savings vs PaO2/FiO2
- Added utility of LASSO + IL-6 over LASSO modest
Original publication
Development and External Validation of Models to Predict Persistent Hypoxemic Respiratory Failure for Clinical Trial Enrichment*. Crit Care Med. Published 10 January 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →