AI-assisted identification of disability patterns within identical EDSS grades

The Neurostatus-Expanded Disability Status Scale (EDSS) is the most frequently used measure of disability in multiple sclerosis (MS) trials. However, EDSS scores ⩾4.5 are mainly based on ambulation and may fail to capture relevant disability patterns in other functional domains.

The objective was to determine how assessments categorized with the same EDSS score may reflect distinct disability patterns.

We analysed 13,103 assessments from 1636 people with secondary progressive MS, from the EXPAND trial. The data set is composed of Functional System scores (FSS) and their corresponding subscores, Ambulation scores and EDSS scores. We performed a descriptive analysis to define the relevant Functional Systems (FS). The subscores were then binarized based on the Neurostatus definition and grouped by respective EDSS scores. Finally, we applied two consecutive machine learning algorithms, to cluster the data. New subscore patterns were then created by aggregating clusters based on their dominant features.

The clustering algorithm yielded numerous clusters, grouping assessments with similar patterns. In patients with EDSS ⩾4.0, our approach allowed differentiation into four subscore patterns within the same EDSS score.

Applying Artificial Intelligence (AI) to large data sets of high-quality clinical assessments allows for distinguishing among different subscore patterns within identical EDSS scores.