First-Day Glycemic Exposure and 28-Day Mortality in the ICU: A Multicenter Cohort Study

Area of Science:

• Critical Care Medicine
• Metabolic Monitoring
• Artificial Intelligence in Healthcare

Background:

• Early glycemic control in ICUs is crucial but often assessed using single glucose readings.
• Developing exposure-aware metrics and predictive models can improve patient outcomes.

Purpose of the Study:

• To investigate the association between first-day time-weighted average glucose (TWAG) and 28-day mortality in ICU patients.
• To evaluate GlucoSurvAI, an interpretable AI model for predicting first-day mortality risk.

Main Methods:

• Retrospective analysis of 8,048 adult ICU patients across 7 U.S. hospitals.
• First-day glycemic exposure calculated as TWAG; mortality assessed using Cox models.
• GlucoSurvAI model performance evaluated with AUROC and Brier score; interpretability via SHAP analysis.

Main Results:

• Higher TWAG levels (≥140 mg/dL) were significantly associated with increased 28-day mortality.
• GlucoSurvAI demonstrated high predictive accuracy (AUROC 0.967) and low Brier score.
• The 100-139 mg/dL TWAG range was associated with decreased mortality, with risk rising above 140 mg/dL.

Conclusions:

• An upper TWAG boundary near 140 mg/dL on the first ICU day is linked to higher mortality.
• An interpretable AI model integrating TWAG, treatments, and physiology offers accurate risk stratification.
• These findings support individualized glycemic targets and early interventions for high-risk ICU patients.