IDENTIFICATION OF SUBPHENOTYPES OF SEPSIS-ASSOCIATED LIVER DYSFUNCTION USING CLUSTER ANALYSIS

Area of Science:

• Critical Care Medicine
• Hepatology
• Data Science in Healthcare

Background:

• Sepsis-associated liver dysfunction (SALD) presents a complex clinical challenge.
• Understanding distinct SALD subphenotypes is crucial for improving patient outcomes.
• Routine clinical data offers potential for subphenotype identification.

Purpose of the Study:

• To identify and validate distinct subphenotypes of sepsis-associated liver dysfunction (SALD).
• To utilize routine clinical information for subphenotype classification.
• To compare clinical features and outcomes across identified SALD subphenotypes.

Main Methods:

• Retrospective observational cohort study using MIMIC-IV and eICU databases.
• K-means cluster analysis applied to 5234 adult patients with SALD within 48 hours of ICU admission.
• Comparison of baseline characteristics and clinical outcomes using ANOVA, Kruskal-Wallis, and chi-squared tests.

Main Results:

• Four SALD subphenotypes (α, β, γ, δ) were identified and validated.
• Subphenotype α: most severe, associated with shock and multiple organ dysfunction.
• Subphenotype β: highest bilirubin, liver disease, coagulopathy.
• Subphenotype γ: oldest patients, higher chronic kidney disease prevalence.
• Subphenotype δ: lowest mortality, unique GGT pattern.

Conclusions:

• Four distinct SALD subphenotypes with differing clinical presentations and outcomes were identified.
• These findings offer a valuable framework for understanding SALD heterogeneity.
• Improved classification can guide more precise clinical management strategies for SALD patients in the ICU.