Perioperative medicine & outcomes
Deep learning models for Intensive Care Unit (ICU) readmission prediction: a systematic review and meta-analysis
Authors' conclusion
DL models showed promising performances in predicting ICU readmissions, but exhibited several shortcomings, including low reproducibility, over-reliance on a few US-based datasets, and limited explainability.
Study context
- Population
- Studies developing or validating deep learning (DL) models for ICU readmission prediction24 studies · 49 DL models
- Outcome
- Model performance (pooled AUROC)
Pooled result
As reported in the abstract
Pooled AUROC (11 studies)
Mean AUROC
0.78
95% CI 0.72–0.84
Heterogeneity I2
99.9%
Further pooled results
AUROC in models targeting disease-specific ICU subpopulations (subgroup)
- Mean AUROC, significantly higher
- 0.92 (95% CI 0.89-0.95), p = 0.002
- Heterogeneity I2
- 17.1%
Study characteristics
- Predominantly US-based datasets, rarely externally validated; considerable variability in outcome definition and DL architecture; substantial risk of bias; technical reproducibility and model interpretation rare
Original publication
Deep learning models for ICU readmission prediction: a systematic review and meta-analysis. Crit Care. Published 17 October 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →