Clinical subtypes in critically ill patients with sepsis: validation and parsimonious classifier model development
Authors' conclusion
The distribution and mortality rates of clinical sepsis subtypes varied between US and European cohorts. A three-variable model could accurately identify the δ-type sepsis patients.
Study context
- Population
- Critically ill ICU patients fulfilling Sepsis-3 criteria, 4 cohorts, 2008–2023
- Participants
- 2,449 / 2,445 / 28,621 / 18,661
- Comparison
- SENECA clinical sepsis subtypes (α, β, γ, δ) across MARS, MARS study (MARS2), NICE and MIMIC-IV cohorts
- MARSn = 2,449
- MARS2n = 2,445
- NICEn = 28,621
- MIMIC-IVn = 18,661
- Outcome
- Validation of SENECA clinical sepsis subtypes and parsimonious classifier for δ-type
Methodological detailsK-means clustering, logistic regression parsimonious models, external validation
- K-means clustering
- Logistic regression parsimonious models
- External validation
Performance
As reported in the abstract
Subtype distribution (α, β, γ, δ)
MIMIC-IV
25%, 24%, 27% and 25%
- MARS, MARS2 and NICE: α-type
- 2-6%
- MARS, MARS2 and NICE: β-type
- 1-5%
- MARS, MARS2 and NICE: γ-type
- 49-65%
- MARS, MARS2 and NICE: δ-type
- 26-48%
- Original SENECA derivation cohort (α, β, γ, δ)
- 33%, 27%, 27% and 13%
Further findings
| Outcome | MARS | MIMIC-IV |
|---|---|---|
| 3-variable classifier (aspartate aminotransferase, serum lactate, bicarbonate) for δ-type vs all other types | ||
| AUC | 0.93 | 0.86 |
| AUC 95% CI | 0.92-0.94 | — not stated in the abstract |
| Accuracy | 85.5% | 82.9% |
| Accuracy CI | 84.0-86.8% | 82.4-83.4% |
| AUC CI | — not stated in the abstract | 0.85-0.87 |
— not stated in the abstract
- 3-variable classifier (aspartate aminotransferase, serum lactate, bicarbonate) for δ-type vs all other types: excellent accuracy in derivation cohort, moderate in validation cohorts
In-hospital mortality across cohorts for α, γ and δ-type
- Significantly different between the four cohorts
- P < 0.001
Parsimonious model for all four subtypes
- Accuracy (moderate to low)
- 62.2%
Parts of this Brief are withheld: the automatic checks could not confirm them against the abstract.
Original publication
Clinical subtypes in critically ill patients with sepsis: validation and parsimonious classifier model development. Crit Care. Published 4 February 2025. doi:10.
Composed by AI from the published abstract and checked against it. How Briefs are made →