Machine learning-based identification of multidimensional predictors of quality of life in individuals with multiple sclerosis

Clinical Snapshot

Key Findings

• P Value: Not reported for machine learning model coefficients

• Effect Size: EDSS emerged as strongest negative predictor, followed by fatigue severity, sleep quality, and functional mobility

• Primary Outcome: Quality of life score on MSQoL-54 scale

• Nnt Or Sensitivity: Model explained 86% of variance in quality of life (R² = 0.86)

• Confidence Interval: Not reported for individual predictors

Clinical Application

High - uses readily available clinical assessment tools that can be implemented in routine MS care Relevant to Australian MS services; EDSS and quality of life assessments align with RACGP chronic disease management guidelines and PBS-funded MS therapies Adults with multiple sclerosis in clinical practice

Abstract

This study aimed to identify independent predictors of quality of life in patients with multiple sclerosis (MS) using machine learning approaches. One hundred and one individuals diagnosed with MS were included in this cross-sectional study. Quality of life was assessed using the MS Quality of Life-54 (MSQoL-54). Demographic variables, clinical characteristics, disability level (Expanded Disability Status Scale [EDSS]), fatigue severity, sleep quality (Pittsburgh Sleep Quality Index), depression level (Beck Depression Inventory), and functional mobility (Timed Up and Go test) were evaluated. Multiple machine learning regression models, including Linear Regression, Lasso, Elastic Net, Support Vector Machines, Random Forest, and XGBoost, were developed and compared. Five-times repeated five-fold cross-validation was applied for internal validation. Model performance was evaluated using root mean squared error (RMSE), mean absolute error (MAE), and coefficient of determination (R2). Among all models, Lasso regression demonstrated the best predictive performance (RMSE = 5.02, MAE = 4.04, R2 = 0.86). Disability level (EDSS) emerged as the strongest negative predictor of quality of life, followed by fatigue severity, sleep quality, and impaired functional mobility. Although disease duration was not significantly correlated with quality of life in univariate analyses, it showed a positive contribution in the multivariable model. Depression showed a strong bivariate association but had a relatively lower weight in the final predictive model. Quality of life in individuals with MS is strongly influenced by disability, fatigue, sleep quality, and functional mobility. Machine learning approaches provide an effective and interpretable framework for identifying key determinants of quality of life and may support personalized rehabilitation and clinical decision-making in MS.

References

1. 1.Yetiş, A., Canli, M., Canli, İ., Kocaman, H., Yildirim, H., Yildiz, N. T., & Duran, S. (2026). Machine learning-based identification of multidimensional predictors of quality of life in individuals with multiple sclerosis. Medicine, 105(22), e49025. https://doi.org/10.1097/MD.0000000000049025