Psychometric properties of the Fatigue Severity Scale—Rasch analyses of individual responses in a Norwegian stroke cohort

Abstract

Background

The psychometric properties of the Fatigue Severity Scale (FSS) have not been tested in people with stroke, despite its being the most frequently used instrument measuring fatigue in this population.

Objective

The objective of our study was to assess the psychometric properties of the FSS using both traditional and modern test theory approaches in people with stroke.

Design and setting

A total of 119 patients admitted to two hospitals in Norway with their first stroke were recruited to a longitudinal study of poststroke fatigue. Data on socio-demographics, fatigue and other clinical variables were collected by face-to-face interviews using standardized questionnaires or by mailed questionnaires at four time points six months apart. A total of 428 responses on the FSS were analysed. The internal scale validity, person response validity, unidimensionality, and uniform differential item functioning were assessed by applications of a Rasch model and by Cronbach's alpha. Additionally, concurrent validity was assessed using bivariate correlation analysis.

Results

Items #1 and #2 in the original FSS demonstrated unacceptable high mean-square values. When these two items were removed, the psychometric properties in FSS-7 demonstrated better evidence of validity. No systematic differential item functioning of the items in FSS-7 was found in relation to any of the analysed variables.

Conclusion

In patients with stroke, the FSS-7 showed better psychometric properties and had better potential to detect changes in fatigue over time than the FSS-9 version, suggesting satisfactory grounds for removal of items #1 and #2 for its application.

Introduction

Fatigue is among the most prevalent symptoms after stroke (Carlsson et al., 2007, Ingles et al., 1999), with prevalence rates ranging from 36 to 70% during the first two years after stroke (Lynch et al., 2007, Appelros, 2006, Choi-Kwon et al., 2005, Schepers et al., 2006). Since fatigue is a subjective phenomenon, researchers and clinicians rely on subjective measurements. It is therefore important to have reliable and valid instruments to measure fatigue.

The Fatigue Severity Scale (FSS) is the most frequently used measurement used in stroke studies (Lerdal et al., 2009). It has also been used in the general population (Lerdal et al., 2005) and in several different populations of people with chronic illnesses (Johansson et al., 2008, Aouizerat et al., 2010). In a published review of measurements of fatigue in chronic illnesses (Whitehead, 2009), the FSS was rated with the highest scores on robust psychometric properties among the 18 fatigue measurements evaluated. Despite the strong psychometric properties of the FSS in other populations, an extensive evaluation of the psychometric properties of FSS in stroke patients has not been published, and it was not considered in a recent published evaluation of fatigue scales in stroke patients (Mead et al., 2007). As any aspect of validity and reliability of an assessment is contextual, it is crucial to evaluate the psychometric properties of the FSS in the stroke population for a specific application. Furthermore, fatigue is an understudied concept without any clinically validated definition. Certain characteristics of fatigue might be disease specific, such as heat sensitive fatigue in patients with multiple sclerosis (MS) (Lerdal et al., 2007). We therefore believe that testing an instrument measuring fatigue in different patient populations are warranted, and might reveal generic characteristics of fatigue across different diagnostic groups as well as more disease specific information about fatigue. The FSS is a nine-item unidimensional questionnaire developed by Krupp et al. (1989) (Table 1). Each item is scored on a seven-point Likert scale ranging from 1 (“disagree”) to 7 (“fully agree”). The mean score of the nine items is used as the FSS score. Except for three studies that reported satisfactory internal reliability assessed with Cronbach's alpha (Schepers et al., 2006, Valko et al., 2008, van de Port et al., 2007), we have not found studies assessing the psychometric properties of the FSS in stroke using either traditional or modern test theory approaches.

Two recent published studies that assessed FSS using Rasch models (Mills et al., 2009, Lerdal et al., 2010b) report inconsistent responses in the FSS in people with MS. Items #1 and #2 of the FSS had high positive residual statistics, while other items showed uniform differential item functioning (DIF) by age, disability, and disease course (Lerdal et al., 2010b, Mills et al., 2009). Thus, they recommended that items #1 and #2 should be removed when computing the FSS mean score.

The aim of this study was to examine the psychometric properties of the FSS in people with stroke from a longitudinal study in Norway by testing (a) the validity of the FSS-measure generated by Rasch analysis for internal scale validity, person response validity, and unidimensionality, (b) concurrent validity of the FSS scale, (c) the reliability and precision of the FSS measure, and (d) the presence of uniform DIF in relation time since stroke/time of measurement, socio-demographic, and clinical variables.