CSWG-SCAI Stages Combined With Machine Learning-Based Phenotypes for Serial Risk Stratification in Cardiogenic Shock

Area of Science:

• Cardiology
• Medical Informatics
• Critical Care Medicine

Background:

• Cardiogenic shock (CS) severity is currently classified using Society for Cardiovascular Angiography and Interventions (SCAI) stages (A-E) or machine learning (ML)-based phenotypes (I: noncongested, II: cardiorenal, III: cardiometabolic).
• These classification systems aim to define and stratify patients based on their clinical status.

Purpose of the Study:

• To evaluate the sequential applicability and prognostic relevance of combining SCAI stages and ML-based phenotypes for risk stratification in CS patients.
• To determine if a combined classification approach offers improved accuracy in predicting outcomes compared to individual systems.

Main Methods:

• Retrospective analysis of 7,716 patients from the Cardiogenic Shock Working Group (CSWG) registry.
• Application of both SCAI staging and ML-based phenotypes at 6- to 12-hour intervals for the first 72 hours.
• Primary outcome assessed was in-hospital mortality.

Main Results:

• Within 6 hours of admission, 78% of patients showed changes in SCAI stages and 77% in phenotypes, with relative stability thereafter.
• Combining ML-based phenotypes with SCAI stages significantly improved risk stratification, with distinct mortality rates for combined categories (e.g., C-I: 12%-14% vs. D-III: 37%-40%).
• Phenotypes II and III were strong predictors of CS progression and in-hospital mortality (Phenotype III: OR 11.4, P < 0.0001).

Conclusions:

• Most CS patients reach phenotype I and stage D within 6 hours of admission.
• Integrating ML-based phenotypes with SCAI staging provides a more mechanistic classification of CS.
• This combined approach may enhance clinical decision-making by better reflecting the heterogeneity within CS populations.