Abstract
This talk examines the limitations of the Expanded Disability Status Scale (EDSS), highlighting its ordinal nature, insensitivity, domain imbalance, and conceptual inconsistencies in multiple sclerosis assessment. We introduce Item Response Theory (IRT) as a principled alternative that models disability on a continuous latent scale and leverages item-level information. The presentation explains how graded response models work, the meaning of difficulty and discrimination parameters, and the key assumptions required for reliable inference. Finally, we discuss why standard unidimensional IRT may struggle with FS scores in MS, how this affects parameter estimation, and how multidimensional extensions may better capture the structure of the disease.