Patient-reported outcomes as predictors of disability evolution in Multiple Sclerosis: An interpretable machine...

Area of Science:

• Neurology
• Artificial Intelligence
• Clinical Research

Background:

• Disability progression in multiple sclerosis (MS) is unpredictable, hindering personalized patient care.
• Machine learning (ML) offers potential for integrating patient-reported outcomes (PROs) and clinician-assessed outcomes (CAOs) to predict MS trajectories.
• Developing predictive models is crucial for tailoring interventions in MS management.

Purpose of the Study:

• To develop and validate an interpretable ML model for predicting disability accrual trajectories in MS patients.
• To assess the model's performance in forecasting disability changes over time.
• To identify key predictors of disability progression in MS.

Main Methods:

• A multicenter dataset of 1,176 MS patients with up to 8 years of follow-up was utilized.
• A random forest model was trained using baseline clinical variables, PROs, and initial risk class to predict disability accrual at 2, 3, 4, and 5 years.
• Model performance was evaluated using accuracy, area under the curve, and survival analysis.

Main Results:

• The final cohort included 437 patients.
• The ML model achieved an accuracy of 0.82 at 2 years and 0.73 at 5 years for predicting disability changes.
• Initial risk class and baseline Expanded Disability Status Scale (EDSS) were the most significant predictors of disability accrual.

Conclusions:

• The developed ML model provides robust and interpretable predictions for MS disability progression.
• This tool can support clinical decision-making by leveraging routine clinical data.
• The model's ability to capture time-dependent patterns aids in understanding disease trajectories.