A framework and analytical exploration for a data-driven update of the Sequential Organ Failure Assessment (SOFA) score in sepsis.

Summary

Across three international ICU cohorts, a data-driven SOFA variant that optimizes domain biomarkers and adds a metabolic domain improved mortality discrimination over the current SOFA. Performance gains were consistent by AUROC and precision–recall metrics.

Key Findings

• A data-driven SOFA with an added metabolic domain outperformed the current SOFA across US, Dutch, and Austrian ICU cohorts.
• AUROC improvements: US 0.766 vs 0.727; Netherlands 0.700 vs 0.653; Austria 0.704 vs 0.665 (all p<0.01).
• Precision–recall analyses corroborated superior performance, demonstrating robustness across metrics.

Clinical Implications

Hospitals can pilot the enhanced SOFA as a parallel score to identify high-risk patients earlier and refine enrollment criteria and covariate adjustment in sepsis trials.

Limitations

• Retrospective design without prospective clinical validation.
• Endpoint limited to mortality; implementation feasibility and clinical impact not yet tested.

Future Directions

Prospective validation, EHR integration for real-time scoring, and evaluation of whether improved classification alters management or outcomes.