Researchers have developed and validated a machine learning tool to predict 28-day mortality in ICU patients with sepsis and acute respiratory failure using early clinical data.
Key Details
- 1Machine learning model predicts 28-day mortality risk for ICU patients with sepsis complicated by acute respiratory failure.
- 2Routinely collected clinical data from the first 24 hours of ICU admission were used as input.
- 3Model was trained using MIMIC-IV (v3.1) and externally validated on eICU-CRD (v2.0) databases.
- 4XGBoost outperformed other algorithms in mortality risk discrimination and generalizability.
- 5Model focused on interpretability using SHapley Additive exPlanations (SHAP) to highlight key clinical predictors.
- 6Study published in Journal of Intensive Medicine on January 10, 2026 (DOI: 10.1016/j.jointm.2025.10.010).
Why It Matters
This development demonstrates how interpretable AI can support early clinical risk stratification, guiding resource allocation and treatment in critical care settings. Such models, leveraging large clinical datasets and interpretability tools, are key for integrating AI decision support into real-world healthcare.
Source
EurekAlert
Related News
•EurekAlert
New AI Method Removes Artifacts in Super-Resolution Fluorescence Microscopy
Researchers unveil Adaptive-SN2N, a self-supervised deep learning framework that suppresses background artifacts in super-resolution fluorescence microscopy images.
•EurekAlert
Micro-CT and AI Reveal Hidden Damage in Coral Skeletons
Researchers combined micro-CT imaging and deep learning to detect subtle disease-induced changes in coral skeletons with high accuracy.
•EurekAlert
Deep Learning AI Deciphers Hidden Self-Organization in Bacterial Colonies
Rice University researchers engineered an AI system to reveal subtle organizational patterns in bacterial communities using time-lapse microscopy data.