ICU sedation, delirium & recovery
Real-time machine learning model to predict short-term mortality in critically ill patients: development and international validation
Authors' conclusion
Our real-time machine learning model to predict short-term mortality in critically ill patients showed excellent performance in both internal and external validations.
Study context
- Population
- Critically ill patients; internal cohort from a single academic hospital in South Korea (2007–2021); external validation in Medical Information Mart for Intensive Care (MIMIC), eICU-Collaborative Research Database (eICU-CRD) and AmsterdamUMCdb
- Comparison
- iMORS vs National Early Warning Score
- Outcome
- Short-term mortality prediction (AUROC)
Methodological detailsensemble machine learning (deep learning and light gradient boosting machine), internal validation, international external validation, Internal validation on last two years of internal cohort; external validation on full MIMIC, eICU-CRD and AmsterdamUMCdb
- Ensemble machine learning (deep learning and light gradient boosting machine)
- Internal validation
- International external validation
- Internal validation on last two years of internal cohort; external validation on full MIMIC, eICU-CRD and AmsterdamUMCdb
Performance
As reported in the abstract
AUROC for short-term mortality
| Test or model | Value | AUROC internal | AUROC MIMIC | AUROC eICU-CRD | AUROC AmsterdamUMCdb |
|---|---|---|---|---|---|
| iMORS | 0.964 (95% CI 0.963–0.965) | 0.964 | 0.890 | 0.886 | 0.870 |
| National Early Warning Score (NEWS) | 0.866 | 0.866 | 0.746 | 0.798 | 0.819 |
Value
iMORS0.964 (95% CI 0.963–0.965)
National Early Warning Score (NEWS)0.866
AUROC internal
iMORS0.964
National Early Warning Score (NEWS)0.866
AUROC MIMIC
iMORS0.890
National Early Warning Score (NEWS)0.746
AUROC eICU-CRD
iMORS0.886
National Early Warning Score (NEWS)0.798
AUROC AmsterdamUMCdb
iMORS0.870
National Early Warning Score (NEWS)0.819
- iMORS vs NEWS: P < 0.001 — higher AUROCs in internal and external validation
- iMORS AUROC MIMIC 95% CI
- 0.889-0.891
- iMORS AUROC eICU-CRD 95% CI
- 0.885-0.887
- iMORS AUROC AmsterdamUMCdb 95% CI
- 0.868-0.873
Original publication
Real-time machine learning model to predict short-term mortality in critically ill patients: development and international validation. Crit Care. Published 14 March 2024. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →