Rasch Analysis of Disability Measures

Because adding ordinal items assumes that all items are equally difficult and that response categories like none-mild-moderate-severe are evenly spaced, assumptions that are rarely true and seldom tested. A summed total can then distort measurement, making a given change mean different things in different parts of the scale. Rasch analysis tests whether the items actually conform to a single interval-level trait and, when they do, provides a conversion from the raw total to an interval estimate, so that summing remains convenient but the resulting score is measurement-justified rather than merely an index.

Differential item functioning (DIF) occurs when an item has a different difficulty for two groups who are otherwise at the same level of the underlying trait — for example, a self-care item that is systematically harder to endorse for one gender, diagnosis, or culture. DIF biases comparisons: differences in total scores may reflect item bias rather than true differences in disability. Detecting DIF is essential when a measure is used across groups or translated for cross-cultural research, and Rasch analysis can flag biased items so they can be split, removed, or interpreted with caution.

Ordinal scores tell you only the order of categories, not the distances between them, so arithmetic on them — averaging, computing change scores, running parametric tests — is technically invalid. When data fit the Rasch model, the model places persons and items on a common interval scale measured in logits, where equal differences mean equal amounts of the trait. This transformation, often delivered as a raw-score-to-logit conversion table, lets clinicians keep summing items while analysts treat the converted measure as interval-level, which is the practical reason Rasch analysis is valued in outcome measurement.