Disentangling self-reported fatigue, depression, and cognitive impairment in people with multiple sclerosis

Abstract

BackgroundFatigue is a common problem among people with multiple sclerosis (MS) and can have a negative effect on mental, physical, and social function. Self-reported measures of MS fatigue are often operationalized as a multi-dimensional symptom. However, questions remain about how best to account for the multi-dimensional aspects of self-reported fatigue and whether these aspects are distinct entities. Thus, the purpose of this study was to explore the overlap and distinctions between self-reported measures of the severity and impact of fatigue, between mental and physical fatigue, and between mental fatigue, depressive symptoms, and cognitive impairment. MethodsAn observational study was conducted with 289 participants with MS . The questionnaires were the Unidimensional Fatigue Impact Scale (UFIS), the Chalder Fatigue Scale (CFS), the Fatigue Scale for Motor and Cognitive Functions (FSMC), the Multiple Sclerosis Neuropsychological Screening Questionnaire (MSNSQ), and the Quality of Life in Neurological Disorders short form for depression (Neuro-QoL). Spearman's correlation coefficient was used to examine the bivariate correlations between composite and subscale scores. Exploratory structural equation modeling (ESEM) was used to determine the factor structure under a pre-specified number of factors to retain in the modeling of multiple items across questionnaires and examine model fit. Subsequently for poor fitting models in an iterative procedure to determine a better fitting multidimensional model, we posited a bifactor confirmatory factor analysis model. ResultsThe bivariate correlation analysis revealed that subscales from the same questionnaire measuring different aspects of fatigue had the highest correlations (r = 0.61–0.68), subscales from different questionnaires measuring the same aspect of fatigue had the next highest correlations (r = 0.43–0.60), and subscales from different questionnaires measuring different aspects of fatigue had the lowest correlations (r = 0.34–0.40). Bifactor models with a general fatigue factor and subdomains pertaining to impact, severity, and mental and physical fatigue had relatively good model fits compared to models omitting the subdomains. However, an ESEM model using subscales from the CFS and FSMC fit poorly and did not adequately identify separate factors for mental and physical fatigue. An ESEM model with separate factors for self-reported mental fatigue, depressive symptoms, and cognitive impairment was a good fit. ConclusionsThe working study hypothesis that fatigue constructs would be moderately correlated yet distinct entities was generally supported by the results of the study. However, we found that our hypothesized separation into a latent dimension existed only when the items or subscales came from the same questionnaire, in which case their level of specificity in terms of target, action, context, and time elements for measuring fatigue were consistent. The implications for the principle of compatibility in measuring self-reported MS fatigue are discussed.