Respiratory failure & ventilation
Development and External Validation of a Detection Model to Retrospectively Identify Patients With Acute Respiratory Distress Syndrome
Authors' conclusion
Machine-learning models analyzing EHR data can retrospectively identify patients with ARDS across different institutions.
Study context
- Population
- Patients with hypoxemic respiratory failure undergoing mechanical ventilation within two distinct health systemsthree cohorts
- Participants
- 1,845 / 556 / 199
- Exposure
- Training vs Internal validation vs External validation
- Trainingn = 1,845
- Internal validationn = 556
- External validationn = 199
- Outcome
- EHR-radiology model performance
Methodological detailsmodel development and external validation, machine learning, physician-adjudicated ARDS
- Model development and external validation
- Machine learning
- Physician-adjudicated ARDS
Main finding
As reported in the abstract
| Group | Area under receiver-operating curve (AUROC) | Integrated calibration index (ICI) | Sensitivity and specificity | PPV | Time to identification after meeting Berlin ARDS criteria, median (IQR) |
|---|---|---|---|---|---|
| Internal validation | 0.91 (95% CI, 0.88-0.93) | ||||
| External validation | 0.88 (95% CI, 0.87-0.93) | 0.13 (95% CI, 0.08-0.18) | 80% (95% CI, 75%-98%) | 64% (95% CI, 58%-71%) | 2.2 hours (0.2-18.6) |
Area under receiver-operating curve (AUROC)
Internal validation0.91 (95% CI, 0.88-0.93)
External validation0.88 (95% CI, 0.87-0.93)
Integrated calibration index (ICI)
Internal validation— not stated in the abstract
External validation0.13 (95% CI, 0.08-0.18)
Sensitivity and specificity
Internal validation— not stated in the abstract
External validation80% (95% CI, 75%-98%)
PPV
Internal validation— not stated in the abstract
External validation64% (95% CI, 58%-71%)
Time to identification after meeting Berlin ARDS criteria, median (IQR)
Internal validation— not stated in the abstract
External validation2.2 hours (0.2-18.6)
Observational data: associations, not effects.
Further findings
ARDS prevalence
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Development and External Validation of a Detection Model to Retrospectively Identify Patients With Acute Respiratory Distress Syndrome. Crit Care Med. Published 8 April 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →