The multi-dimensionality of the FIM motor items precludes an interval scaling using Rasch analysis

Abstract

Rasch analysis scaling is said to produce an interval scale of Functional Independence Measure (FIM) motor function items. Rasch analysis requires that the data to be analysed represent the influence of a single underlying unidimensional variable. A unidimensional interval scale of activities of daily living means that a person who can perform the most difficult item on a scale can also perform the easiest item. For a FIM motor function interval scale, the ability to climb stairs would imply necessarily and ability to eat normally. As this need not be the case, the FIM motor items do not constitute an adequate interval scale. Eating and walking are different activities, and a unidimensional construct linking them is unsatisfactory. A principal components analysis of the admission FIM motor function items of 515 consecutive patients admitted to an inpatient rehabilitation unit revealed that more than one significant factor was necessary to explain the variance in scores. The counter-factual and statistical evidence argues that a unidimensional construct does not underly the FIM motor function items, and the use of Rasch analysis will not lead to a description of interval properties of the FIM motor function items.