Background: Rigorous testing of the original Fatigue Severity Scale (FSS-9) with modern psychometric methods is warranted.
Objective: To determine the psychometric properties of the FSS-9 in multiple sclerosis (MS): internal scale validity; person response validity; unidimensionality; uniform differential item functioning; temporal stability of response patterns; and ability to separate people into distinct groups of fatigue.
Methods: Rasch analyses were conducted on data from a Norwegian and a Swedish MS cohort followed for two years.
Results: Item estimations in the FSS-9 did not differ between sex or levels of education but between the cohorts with regard to disability, disease course and time for evaluation, however, items 1 and 2 demonstrated unacceptable high outfit mean-square values in both cohorts. In an FSS-7 item version, items 3 and 4 in the Norwegian and 4 in the Swedish cohort demonstrated unacceptable goodness of fit but high separation indexes. In the FSS-7, the first unidimensional factor explained 87.5% (Norwegian cohort) and 86.4% (Swedish cohort) of the total variation.
Conclusions: In MS, the FSS-7 demonstrates better psychometric properties than the FSS-9; items 1 and 2 neither empirically nor conceptually fit with the other seven items.