Mortality Risk Profiles for Sepsis: A Novel Longitudinal and Multivariable Approach

Area of Science:

• Critical Care Medicine
• Biomarkers
• Prognostic Modeling

Background:

• Sepsis poses a significant mortality risk in intensive care units (ICUs).
• Accurate and timely prediction of sepsis mortality is crucial for patient management.
• Existing prognostic tools may not fully capture the dynamic nature of sepsis progression.

Purpose of the Study:

• To evaluate time-varying biological indicators for predicting sepsis mortality risk.
• To develop a prognostic tool that generates dynamic mortality risk profiles.
• To understand the longitudinal contribution of biomarkers to sepsis mortality.

Main Methods:

• Prospective observational study in nine Canadian ICUs involving 356 septic patients.
• Longitudinal collection of clinical data and plasma biomarker levels.
• Utilized a complementary log-log model, a variant of the Cox model, to assess daily mortality risk.

Main Results:

• A composite indicator integrating six biomarkers (cell-free DNA, protein C, platelet count, creatinine, Glasgow Coma Scale score, lactate) and contextual variables achieved high predictive power (AUC, 0.90).
• The model successfully generated mortality risk profiles, highlighting the longitudinal impact of biomarkers.
• Validation in nonseptic patients who later developed sepsis and through 10-fold cross-validation confirmed high predictive accuracy.

Conclusions:

• A novel prognostic tool was developed using a modified Cox model for septic patients.
• This tool provides both predicted mortality probability and detailed mortality risk profiles.
• The profiles illustrate how six time-varying biological indicators longitudinally influence daily mortality risk in sepsis.