Related Experiment Video
Updated: Jun 13, 2025
08:10
Estimating Bilateral Atrial Function by Cardiovascular Magnetic Resonance Feature Tracking in Patients with Paroxysmal Atrial Fibrillation
Published on: July 20, 2022
1.6K
A machine learning-based risk prediction model for atrial fibrillation in critically ill patients
Laith Alomari1, Yaman Jarrar2, Zaid Al-Fakhouri3
1Department of Medicine, Jefferson Einstein Philadelphia Hospital, Thomas Jefferson University, Philadelphia, Pennsylvania.
Heart Rhythm O2
|June 11, 2025
Summary
Machine learning accurately predicts atrial fibrillation (AF) in intensive care unit (ICU) patients. A new compact model identifies high-risk individuals for early intervention, improving outcomes.
Area of Science:
- Critical Care Medicine
- Cardiology
- Artificial Intelligence in Healthcare
Background:
- Atrial fibrillation (AF) significantly increases morbidity and healthcare costs in critically ill patients.
- Current AF prediction tools lack efficacy in intensive care unit (ICU) settings.
Purpose of the Study:
- To develop and validate a machine learning model for early identification of AF risk in ICU patients.
- To create a novel, compact model incorporating unique risk factors for improved prediction.
Main Methods:
- Retrospective analysis of 46,266 adult ICU patients from the MIMIC-IV database.
- Training and evaluating multiple machine learning models, including CatBoost, for AF prediction within 48 hours of admission.
- Developing a compact model using 15 variables and two novel features, validated with SHAP analysis.
Main Results:
- 4.6% of patients developed AF within 48 hours.
- The CatBoost model achieved an AUROC of 0.850; the compact model achieved an AUROC of 0.820.
- Key predictors included serum magnesium, age, and novel composite risk features.
Conclusions:
- Machine learning models show significant potential for predicting AF in the ICU.
- The developed compact model offers a practical tool for early risk stratification and intervention in high-risk ICU patients.
