Development of An Explainable Machine Learning Model for Predicting Fatigue in Rheumatoid Arthritis

Area of Science:

• Rheumatology
• Medical Informatics
• Machine Learning

Background:

• Fatigue is a prevalent symptom in rheumatoid arthritis (RA), significantly impacting patient quality of life.
• Few predictive models exist for RA-related fatigue using clinical indicators.
• This study explores machine learning (ML) for developing such predictive models.

Purpose of the Study:

• To develop and evaluate machine learning models for predicting fatigue in rheumatoid arthritis (RA) patients.
• To identify key clinical predictors of RA-related fatigue.

Main Methods:

• Retrospective analysis of clinical data from 271 RA patients.
• Feature selection using Lasso, Boruta, and RF-RFE algorithms.
• Development and evaluation of ML models including SVM, XGBoost, LightGBM, ANN, KNN, and RF, with external validation.

Main Results:

• Six key predictors identified: CCP, ESR, lymphocyte count, MDGA, VAS for pain, and TG.
• ML models, particularly SVM, XGBoost, and Random Forest, demonstrated superior predictive performance compared to logistic regression in external validation.
• The SVM model showed potential in identifying RA patients at risk of fatigue.

Conclusions:

• Machine learning models can capture complex relationships between clinical variables for fatigue prediction in RA.
• The developed SVM model shows potential for identifying RA patients at risk of fatigue.
• Results are preliminary, requiring validation in larger, multicenter studies.