Respiratory failure & ventilation
Risk heterogeneity within hypoinflammatory acute respiratory failure: continuous probabilities identify high-risk patients masked by binary classification
Authors' conclusion
Binary classification obscures substantial prognostic heterogeneity within hypoinflammatory AHRF patients.
Study context
- Population
- Critically ill adults with acute hypoxemic respiratory failure (AHRF) (Pittsburgh Acute Lung Injury Registry); validation in EDEN trial, COVID-19 cohorts and RoCI registry1,134 validation
- Participants
- 575 adults
- Exposure
- Continuous subphenotype probability tertiles within hypoinflammatory and hyperinflammatory subphenotypes
- Hypoinflammatoryn = 498reference
- Hyperinflammatoryn = 77
- Outcome
- Mortality
- Follow-up
- 90 days
Methodological detailsregistry-based, external validation
- Registry-based
- External validation
Main finding
As reported in the abstract
90-day mortality across probability tertiles
Hyperinflammatory
55%
- Hypoinflammatory: mortality increased across probability tertiles
- 19 to 31% to 40%; P < 0.001
- Hyperinflammatory: homogeneous mortality across tertiles
- P = 0.72
Observational data: associations, not effects.
Further findings
Continuous probability vs mortality (restricted cubic spline)
- Strong non-linear relationship, steepest risk increases below the 0.5 threshold
Clinical severity scores and immune activation biomarkers across hypoinflammatory tertiles
- Increased progressively
- all P < 0.001
Mortality by probability trajectory (longitudinal sampling, n = 330)
- Hypoinflammatory rising trajectories vs stable or declining
- 50-100% vs. 16-40%; all P < 0.001
- Hyperinflammatory: poor outcomes regardless of trajectory
External validation
- Heterogeneity and non-linear probability-mortality patterns confirmed across cohorts; similar with procalcitonin-based model
Original publication
Risk heterogeneity within hypoinflammatory acute respiratory failure: continuous probabilities identify high-risk patients masked by binary classification. Intensive Care Med. Published 27 April 2026. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →