Measuring fatigue in multiple sclerosis: using psychometrics to evaluate the MAF, FSS, MFIS.
|Title||Measuring fatigue in multiple sclerosis: using psychometrics to evaluate the MAF, FSS, MFIS.|
|Publication Type||Journal Article|
|Authors||Amtmann D, Boesflug E, Bamer AM, Johnson KL|
Fatigue is one of the most common symptoms in persons with multiple sclerosis (MS). Several scales are available for measuring fatigue. Modern psychometric methods can be useful in selecting the most appropriate scale for the given purpose and for examining the scales’ psychometric properties. A large cross sectional sample (N = 1014) of community dwelling individuals with MS completed a mail survey that included the Multidimensional Assessment of Fatigue (MAF), the Fatigue Severity Scale (FSS), and the Modified Fatigue Impact Scale (MFIS). Item Response Theory (IRT) was used to evaluate the psychometric properties of the 3 scales. The test information function, ceiling/floor effects, and reliability were examined at the scale level. At the item level, the category characteristic function and item fit statistics were examined. MAF provided the greatest level of precision at the mean level of fatigue for the MS sample with some ceiling (3.3% of sample) and floor effects (1.3%). MFIS provided the greatest level of precision slightly below the mean, with floor and ceiling effects similar to MAF (3.7% & 1.3% respectively). For both scales 95% of the respondents fell within the effective range of measurement. FSS measured with the greatest precision just below the mean, at about the same level as MFIS, but displayed a much larger ceiling effect (21%), with 78% of the respondents within the effective range of measurement. Person reliability (similar to Cronbach’s alpha) was slightly higher than recommended for MAF (.92) and MFIS (.95), suggesting some redundancy in items, and was adequate for FSS (.84). The response categories functioned well for most items of MFIS and MAF. The categories for 6 out of 14 items of MAF did not function as expected and most appeared to function as dichotomous (yes/no) questions. One item of 21 items in MFIS and 1 of 14 items of MAF showed significant misfit. Three of 9 items of FSS had the item fit statistics outside of the acceptable range. Psychometric properties of all 3 analyzed scales could be improved by addressing the issues identified in IRT analyses. The 21 item MFIS appears to be most appropriate for measuring the broad range of fatigue levels in individuals with MS because of a small ceiling effect and only one misfitting item. IRT scores were used to design short forms of the scale to address respondent burden. Differential item functioning needs to be examined for the short forms.