Respiratory failure & ventilation
Evaluating the SpO2 Waveform ICU Forecasting Technique (SWIFT) algorithm's efficacy in predicting hypoxemia across multiple critical care datasets
Authors' conclusion
The study highlights challenges in generalizing prediction models across diverse ICU populations, emphasizing need for external validation.
Study context
- Population
- ICU patients, ventilated and non-ventilated
- Comparison
- EICU Collaborative Research Database (eICU-CRD) vs Medical Information Mart for Intensive Care IV (MIMIC-IV) vs University Medical Centers Database (UMCdb)
- Outcome
- Performance of SWIFT (LSTM algorithm) in predicting SpO2 5 and 30 min in advance
Methodological detailsmulticentre, retrospective databases, Sampling procedure resulted in substantial population size reduction for MIMIC-IV and UMCdb due to differences in SpO2 measurement frequency
- Multicentre
- Retrospective databases
- Sampling procedure resulted in substantial population size reduction for MIMIC-IV and UMCdb due to differences in SpO2 measurement frequency
Performance
As reported in the abstract
SWIFT predictive performance
- SWIFT performed well on eICU-CRD data
- Reduced performance on MIMIC-IV data, particularly for SWIFT-30
- UMCdb: comparable performance to eICU-CRD for ventilated patients
- All datasets exhibited high specificity and NPV
Original publication
Evaluating the SWIFT algorithm's efficacy in predicting hypoxemia across multiple critical care datasets. J Crit Care. Published 19 May 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →