ICU sedation, delirium & recovery
Predicting Readmission or Death After Discharge From the ICU: External Validation and Retraining of a Machine Learning Model
An erratum to this article was published (Crit Care Med, 1 Apr 2023 ↗)
Authors' conclusion
In this era of expanding availability of ML models, external validation and retraining are key steps to consider before applying ML models to new settings.
Study context
- Population
- Adults admitted to the ICU for longer than 12 hours, 2 tertiary care ICUs in The NetherlandsICU admissions at new site
- Participants
- 10,052 adults
- Comparison
- Original machine learning (ML) model vs retrained and recalibrated model at a new site
- Outcome
- Model performance (discrimination, AUC)
Methodological detailstemporal validation, retraining and recalibration, cross-validation, two tertiary care centres
- Temporal validation
- Retraining and recalibration
- Cross-validation
- Two tertiary care centres
Performance
As reported in the abstract
| Test or model | AUC |
|---|---|
| External validation | 0.72 (95% CI 0.67-0.76) |
| Final validation model | 0.79 (95% CI 0.75-0.82) |
AUC
External validation0.72 (95% CI 0.67-0.76)
Final validation model0.79 (95% CI 0.75-0.82)
Further findings
Calibration
- Poor initially; good after recalibration via isotonic regression
Readmission or death within 7 days after ICU discharge, new site
- Readmission or death within 7 days
- 577 of 10,052 (5.7%)
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Predicting Readmission or Death After Discharge From the ICU: External Validation and Retraining of a Machine Learning Model. Crit Care Med. Published 16 December 2022. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →