Machine Learning Methods for Predicting Syncope Severity in the Emergency Department: A Retrospective Analysis.

Area of Science:

• Artificial Intelligence in Medicine
• Clinical Decision Support Systems
• Machine Learning for Healthcare

Background:

• Syncope is a common emergency admission reason with challenges in risk assessment.
• Limited research exists on artificial intelligence (AI) for improving syncope patient outcomes.
• Current study focuses on predicting syncope severity using machine learning (ML).

Purpose of the Study:

• To predict the severity of syncope cases using ML algorithms.
• To analyze data collected during on-site treatment and ambulance transportation.
• To establish an experimental foundation for ML in syncope management.

Main Methods:

• Analyzed 572 patient records from five Spanish hospitals (2018-2021).
• Employed a three-phase strategy: data preprocessing, model exploration, and selection.
• Utilized ML classifiers including Random Forest (RF), Dummy Classifier (DC), and Linear Discriminant Analysis (LDA) with 10-fold cross-validation.

Main Results:

• Random Forest (RF) excelled in predicting hospitalization (accuracy 0.74, recall 0.63).
• Dummy Classifier (DC) showed better performance for ICU admission prediction (accuracy 0.58, recall 0.625).
• Linear Discriminant Analysis (LDA) was superior for predicting hospital mortality (accuracy 0.88, recall 0.6).

Conclusions:

• Machine learning models demonstrate potential for predicting syncope severity and outcomes.
• RF, DC, and LDA classifiers showed distinct strengths for different prediction tasks (hospitalization, ICU, mortality).
• Findings aim to stimulate AI research and integration into clinical workflows for syncope management.