Using Symptom Frequency and Severity to Improve Machine Learning Classification of Long COVID

Composite scoring models achieved 90.12% accuracy in diagnosing Long COVID.

• Machine-learning models that consider both symptom frequency and severity show better performance than those that only focus on symptom occurrence.
• Occurrence-based models achieved 88.73% accuracy, suggesting room for improvement in diagnostic precision.
• Composite models required fewer predictive symptoms, indicating they may be more efficient for classification.
• Assessing symptom burden is linked to enhanced precision in research classifications of Long COVID.

Simplified