Particle.news
Download on the App Store

NutriSighT Predicts Underfeeding Risk in Ventilated ICU Patients

Researchers plan prospective trials to test whether using its alerts improves care.

Overview

  • The Icahn School of Medicine at Mount Sinai reported the findings in Nature Communications on December 17, 2025.
  • The interpretable transformer model analyzes routine ICU data and updates risk predictions every four hours for days three through seven of ventilation.
  • Training and validation used large deidentified ICU datasets from Europe and the United States.
  • The study found underfeeding was common early in care, affecting about 41–53% of patients by day three and 25–35% by day seven.
  • Investigators describe NutriSighT as an early-warning aid for clinicians, with plans for electronic health record integration and multi-site trials to evaluate impact.