ICU sedation, delirium & recovery
Machine learning-based immunocompromise and severity score for early risk stratification of critically ill patients
Authors' conclusion
Derived from common immunocompromising conditions and biomarkers, the ICS and SICS scores and their risk stratification system provide an accurate and timely solution for identifying and stratifying immunocompromise in critically ill patients.
Study context
- Population
- Adults admitted to intensive care unit (ICU) for at least three days, two ICUs of an academic tertiary care center in Chinaderivation cohort
- Participants
- 1,863 adults
- Exposure
- Immunocompromise and Severity (ICS) score and SICS score vs APACHE II and SOFA for 28-day mortality prediction
- Outcome
- 28-day all-cause mortality
- Follow-up
- 28 days
Methodological detailssingle-centre, machine learning prediction model development, 10-fold cross-validation, temporal validation
- Single-centre
- Machine learning prediction model development
- 10-fold cross-validation
- Temporal validation
Main finding
As reported in the abstract
Prediction of 28-day all-cause mortality
| Group | AUC | Sensitivity | Specificity | Accuracy |
|---|---|---|---|---|
| ICS score | 0.887 | 0.896 | 0.723 | 0.787 |
| SICS score | 0.851 | 0.83 | 0.72 | 0.765 |
AUC
ICS score0.887
SICS score0.851
Sensitivity
ICS score0.896
SICS score0.83
Specificity
ICS score0.723
SICS score0.72
Accuracy
ICS score0.787
SICS score0.765
- ICS score vs APACHE II and SOFA: P < 0.001 — ICS performance superior
Observational data: associations, not effects.
Secondary outcomes
| Outcome | ICS score | SICS score |
|---|---|---|
| Temporal validation (n = 216) | ||
| AUC | 0.854 | 0.845 |
Mortality by risk stratum (ICS vs SICS)
- Low risk mortality, ICS vs SICS
- 2.6% vs. 2.8%
- Moderate risk mortality, ICS vs SICS
- 24.2% vs. 24.4%
- High risk mortality, ICS vs SICS
- 69.9% vs. 69.6%
- Proportion classified low risk, ICS vs SICS
- 31.4% vs. 19.6%
- Proportion classified high risk, ICS vs SICS
- 42.6% vs. 36.6%
Secondary endpoints (septic shock, special antimicrobials, peak IL-6)
- Significantly correlated with higher risk stratification
Original publication
Machine learning-based immunocompromise and severity score for early risk stratification of critically ill patients. Anaesth Crit Care Pain Med. Published 6 November 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →