Sepsis Subphenotyping in ICU Patients

Introduction to Subphenotyping Approaches

Machine learning models have identified four distinct sepsis subphenotypes using more than 25 routinely measured clinical parameters available in standard electronic health data, according to the American Thoracic Society, 1, 2
These parameters include vital signs, laboratory values, organ dysfunction scores, and demographic data that are universally available, as recommended by the American College of Chest Physicians, 1
This approach requires no specialized biomarkers and can be implemented globally, as suggested by the American Journal of Respiratory and Critical Care Medicine, 1

For ARDS patients, validated 3-4 variable classifiers using clinical data and plasma inflammatory markers (excluding ferritin) can reliably identify subphenotypes with differential mortality and treatment responses, according to the American Thoracic Society, 2, 3
The Sequential Organ Failure Assessment (SOFA) score quantifies organ dysfunction severity and serves as a prognostic indicator, as stated by the Society of Critical Care Medicine, 5
Number of acquired organ system failures is the most important prognostic indicator for ICU patients, as reported by the American College of Chest Physicians, 5

Respiratory system compliance in ARDS patients identifies subgroups with differential treatment responses to low tidal volume ventilation, according to the American Thoracic Society, 2
Physiologic measurements are readily identifiable using routine clinical data without specialized biomarkers, as suggested by the American Journal of Respiratory and Critical Care Medicine, 4

Subphenotyping approaches must be readily measurable with acceptable operating characteristics to allow global implementation, as recommended by the World Health Organization, 4, 6
Clinical severity criteria alone achieve prognostic enrichment but generally fail to enrich for treatment response, according to the American College of Chest Physicians, 1, 3
Integration of clinical, biological, and physiologic data provides more robust subphenotyping than single modalities, as stated by the American Thoracic Society, 7