ICU sedation, delirium & recovery
External Validation of Mortality Prediction Models for Critical Illness Reveals Preserved Discrimination but Poor Calibration
Authors' conclusion
External validation showed that the discriminative ability of APACHE II, APACHE IV, and SAPS II was acceptable to excellent, whereas calibration was poor.
Study context
- Population
- Critically ill patients2 cohorts
- Participants
- 1,075 vs 2,901
- Comparison
- 43 published mortality prediction models for critically ill patients, externally validated
- Simple Intensive Care Studies-I (SICS-I)n = 1,075
- Finnish Acute Kidney Injury (FINNAKI)n = 2,901
- Outcome
- Model performance on external validation
Methodological detailsmulticentre, two independent cohorts, recalibration, Models recalibrated using SICS-I data and validated using FINNAKI data
- Multicentre
- Two independent cohorts
- Recalibration
- Models recalibrated using SICS-I data and validated using FINNAKI data
Performance
As reported in the abstract
- Models that could be externally validated
- 11 (26%)
- Best scaled Brier score, Acute Physiology And Chronic Health Evaluation (APACHE) II
- 0.11
- Best scaled Brier score, APACHE IV
- 0.10
- Best scaled Brier score, SAPS-R
- 0.10
- Best scaled Brier score, Simplified Mortality Score for the ICU
- 0.06
- Discrimination range of models, AUC (lowest)
- 0.63 (0.61-0.66)
- Discrimination range of models, AUC (highest)
- 0.83 (0.81-0.85)
Further findings
Decision curve analysis
- Positive net benefit for APACHE IV and SAPS-R, threshold probability range
- 0-60%
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
External Validation of Mortality Prediction Models for Critical Illness Reveals Preserved Discrimination but Poor Calibration. Crit Care Med. Published 15 November 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →