Predicting Multiple Sclerosis Outcomes: A Machine Learning Approach Integrating Patient-Reported Outcomes and Clinical Data

Multiple sclerosis (MS) is a chronic, immune-mediated disease with variable progression that complicates clinical management(1). Traditional assessments, such as the Expanded Disability Status Scale (EDSS), have limitations, prompting the integration of patient-reported outcome measures (PROMs) alongside clinician-reported outcomes (CROs) to capture a comprehensive view of disease impact(2). This study investigates the use of machine learning (ML) techniques to predict three key clinical outcomes in MS: EDSS level, spasticity severity, and the development of new lesions detected through magnetic resonance. Data were collected from 240 MS patients over 18 months, including baseline demographics, CROs, and PROMs. The ML pipeline involved feature encoding, data splitting (80:20), oversampling for underrepresented classes, and hyperparameter tuning via grid search and cross-validation. Regression models (evaluated with MSE and R²) and classification models were trained depending on predicted variable characteristics. Results indicate that models incorporating both PROMs and CROs achieved superior performance in predicting EDSS and spasticity, while ML models reliably identified new lesions with 100% true positive rate. PROMs collected before clinical assessment combined with basal demographics also lead to predictive models with acceptable performance. These findings support the potential of integrating PROMs into clinical decision-making and telemedicine for improved MS management.