To critically appraise, compare and summarize the measurement properties of self-report fatigue questionnaires validated in patients with multiple sclerosis (MS), Parkinson’s disease (PD) or stroke.
Methods
MEDLINE, EMBASE, PsycINFO, CINAHL and SPORTdiscus were searched. The COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) checklist was used to assess the methodological quality of studies. A qualitative data synthesis was performed to rate the measurement properties for each questionnaire.
Results
Thirty-eight studies out of 5,336 records met the inclusion criteria, evaluating 31 questionnaires. Moderate evidence was found for adequate internal consistency and structural validity of the Fatigue Scale for Motor and Cognitive functions (FSMC) and for adequate reliability and structural validity of the Unidimensional Fatigue Impact Scale (U-FIS) in MS.
Conclusions
We recommend the FSMC and U-FIS in MS. The Functional Assessment of Chronic Illness Therapy Fatigue subscale (FACIT-F) and Fatigue Severity Scale (FSS) show promise in PD, and the Profile of Mood States Fatigue subscale (POMS-F) for stroke. Future studies should focus on measurement error, responsiveness and interpretability. Studies should also put emphasis on providing input for the theoretical construct of fatigue, allowing the development of questionnaires that reflect generic and disease-specific symptoms of fatigue.
Patient-reported outcomes measurement information system
PS-F
Performance scale fatigue subscale
RFS
Rhoten fatigue scale
S&E
Schwab and England score
SA-SIP-30
Stroke-adapted sickness impact profile 30 item version
SD
Standard deviation
SDC
Smallest detectable change
SF-36-V
Short-form-36 vitality subscale
SF-36-V (V2.0)
Short-form-36 vitality subscale version 2.0
SOFI
Swedish occupational fatigue inventory
U-FIS
Unidimensional fatigue impact scale
VAS-1, 2 or 3
Visual analogue scale-1, 2 or 3
WEIMUS
Würzburger Erschöpfungsinventars bei Multiple sclerosis
Introduction
Fatigue is common in chronic neurological disorders [1]. Prevalence rates in conditions often seen in neurological rehabilitation, such as multiple sclerosis (MS), Parkinson’s disease (PD) and stroke, range from 58% [2] to 90% [3].
One of the challenges in assessing fatigue is the lack of a widely accepted definition [4] and with that, differentiating its many dimensions [2, 5]. Fatigue usually refers to the difficulty initiating or sustaining voluntary activity [6]. Its multidimensionality is believed to result from a complex interplay between the underlying disease process, peripheral control systems (i.e. muscle fatigability), central control systems (i.e. subjective sense of fatigue) and environmental factors [6]. This may reflect the large number of generic and disease-specific self-report questionnaires that are available to measure fatigue as either a multidimensional or a unidimensional assessment in patients considered for rehabilitation services. These questionnaires may measure different aspects or even different theoretical constructs of fatigue [7]. The clinician or researcher has to consider that each questionnaire is characterized by its own underlying concept, measurement properties and practical feasibility. A systematic review of the characteristics and measurement properties of self-report fatigue questionnaires can assist in selecting an appropriate questionnaire to evaluate fatigue in patients with MS, PD and stroke.
Several systematic reviews [7‐13] have evaluated the measurement properties of fatigue questionnaires. Three of these reviews [7, 12, 13] focused on patients with chronic disease, including samples of patients with MS and PD. Unfortunately, no recommendations were made specifically for patients with MS or PD. One review [10] focused on patients with MS. The authors recommended the Fatigue Impact Scale (FIS) and the Modified Fatigue Impact Scale (MFIS) [10]. Another review [8] recommended the Multidimensional Fatigue Inventory (MFI) and the Fatigue Severity Scale (FSS) for patients with PD. No systematic review evaluated questionnaires validated in patients with stroke.
A limitation of the aforementioned reviews is that no uniform definitions and standards for the assessment of the methodological quality of the included studies were used. Therefore, the methodological quality of these studies was not taken into account when formulating conclusions, which makes it difficult to judge the strength of the evidence underlying the formulated recommendations. Recently, the COnsensus-based Standards for the selection of health Measurement INstruments (COSMIN) checklist [14] was developed to systematically evaluate the methodological quality of studies on measurement properties. This makes it possible to appraise the methodological quality of the included studies and take this into account when formulating conclusions.
The aim of the present study was to critically appraise, compare and summarize the quality of the measurement properties of all published self-report fatigue questionnaires validated in patients with MS, PD or stroke, in order to assist clinicians and researchers in selecting a fatigue questionnaire.
Methods
Search
Five databases were searched up to November 2010 (MEDLINE (1966–2010), EMBASE (1974–2010), PsycINFO (1806–2010), CINAHL (1981–2010) and SPORTdiscus (1985–2010)). Text words and MESH terms for fatigue, MS, PD and stroke were combined with a sensitive filter (designed for PubMed) to identify studies on measurement properties of self-report questionnaires [15] (see supplementary file 1). References of the included studies were screened for additional articles.
Selection of studies
Two reviewers (RE/EvW) independently screened all titles and abstracts. The full text papers of relevant studies were obtained, and two reviewers (RE/MR) independently applied the a priori defined criteria for study selection. Studies were included if they met the following criteria: the study (1) focused on the development or evaluation of measurement properties of self-report questionnaires that assess subjective fatigue; (2) included patients with a clinical diagnosis of MS, PD or stroke and (3) included questionnaires that could be used for evaluative purposes. Studies were excluded if: the study (1) explicitly focused on the diagnostic test accuracy of the included questionnaire(s); (2) was published in a language other than Dutch, English, French or German. In case of disagreement, a third reviewer (EvW) was asked for advice to reach consensus.
Assessment of methodological quality
The methodological quality of a study was evaluated using the COSMIN checklist [14]. This checklist consists of 114 items, grouped in twelve boxes. Nine of these boxes contain standards for measurement properties (i.e. internal consistency, reliability, measurement error, content validity, structural validity, hypotheses testing, cross-cultural validity, criterion validity and responsiveness). One box contains standards for studies on interpretability, which is an important characteristic of a measurement scale [16]. In addition, two boxes contain requirements for studies in which Item Response Theory (IRT) methods are applied, and requirements for the generalizability of the results, respectively [14]. Each item was scored on a 4-point rating scale (i.e. ‘poor’, ‘fair’, ‘good’, or ‘excellent’) [17]. The methodological quality of a study was evaluated per measurement property and determined by the lowest rating of any of the items in a box. Pairs of reviewers (RE/EvW, RE/JV, RE/MR or RE/SK) independently scored the methodological quality of the included studies. Disagreement was resolved during consensus meetings.
Data extraction
A data extraction form was designed and tested before the pairs of reviewers independently extracted data on the: (1) characteristics of the study samples; (2) characteristics of the questionnaires (i.e. language version, theoretical construct of fatigue and dimensions, recall period, number of items, response options, range of scores, time to administer and ease of scoring); (3) evaluated measurement properties and (4) the interpretability and generalizability of the results.
Data synthesis
The theoretical construct of fatigue measured by a questionnaire was categorized by either ‘impact of fatigue on daily life’, ‘fatigue severity’ or ‘factors influencing fatigue’. Ease of scoring was categorized as ‘easy’ if items were simply summed, ‘moderate’ if a visual analogue scale (VAS) or simple formula was used, or ‘difficult’ if either a VAS in combination with a formula or a complex formula was used.
Measurement properties were summarized according to the COSMIN taxonomy [16]. For each study, the estimates of the investigated measurement properties were rated as ‘adequate’ (+), ‘not adequate’ (−) or ‘unclear’ (?), based on predefined criteria [18] as described below.
A qualitative data synthesis was performed to determine the overall quality of the measurement properties for each self-report questionnaire by taking into account the: (1) ratings for each measurement property; (2) consistency of results between studies; (3) methodological quality of studies and (4) the number of studies that investigated the measurement property. The possible overall quality of a measurement property was either ‘adequate’ (+), ‘not adequate’ (−), ‘conflicting’ (±) or ‘unclear’ (?). As shown in Table 1, levels of evidence were defined to express whether the strength of the evidence for the overall quality was, for example, convincing (‘strong’ level of evidence) or unconvincing (‘unknown’ level of evidence) [19].
Table 1
Levels of evidence for the overall quality of a measurement property
Level
Rating
Criteria
Strong
‘Adequate’ or ‘Not adequate’ (+ or −)
Consistent findings in multiple studies of ‘good’ methodological quality OR in one study of ‘excellent’ methodological quality
Moderate
‘Adequate’ or ‘Not adequate’ (+ or −)
Consistent findings in multiple studies of ‘fair’ methodological quality OR in one study of ‘good’ methodological quality
Limited
‘Adequate’ or ‘Not adequate’ (+ or −)
One study of ‘fair’ methodological quality
Conflicting
‘Conflicting’ (±)
Conflicting findings
Unknown
‘Unknown’ (?)
Only studies of ‘poor’ methodological quality
Criteria for the quality of measurement properties
Reliability
The domain reliability contains three measurement properties: internal consistency, reliability and measurement error [16].
Internal consistency is the degree of the interrelatedness among items, assuming the questionnaire to be unidimensional [16]. Cronbach’s α was considered an acceptable measure of internal consistency and scored adequate if it ranged between 0.70 and 0.95 [18]. If a questionnaire was multidimensional, internal consistency was considered per subscale.
Reliability was defined as the proportion of the total variance in the measurements which is because of ‘true’ differences between patients [16]. The intraclass correlation coefficient (ICC) and weighted kappa are acceptable measures for reliability and considered adequate if they were ≥0.70 [18]. If a Pearson or Spearman correlation coefficient (CC) was presented, which do not account for systematic differences between two tests [20], an estimate of ≥0.80 was considered adequate.
Measurement error, defined as the systematic and random error of a score that is not attributed to true changes in the construct to be measured [16], was scored adequate if the smallest detectable change (SDC) was smaller than the minimal important change (MIC), or if the MIC was outside the limits of agreement (LOA) [18].
Validity
Validity contains the measurement properties content validity, construct validity and criterion validity [16]. Content validity includes face validity and extends to the degree to which the content of a questionnaire is an adequate reflection of the construct to be measured [16]. It was rated adequate if the target population and experts considered all items in the questionnaire relevant and considered the questionnaire to be complete. Construct validity was defined as the degree to which scores of a questionnaire are consistent with hypothesis, based on the assumption that the instrument validly measures the construct to be measured [16]. Construct validity is divided into structural validity, hypothesis testing and cross-cultural validity. Structural validity, defined as the degree to which scores of a questionnaire are an adequate reflection of the dimensionality of the construct to be measured [16], was scored adequate if factor analysis showed that all factors together explained ≥50% of the total variance, or when IRT methods were applied to confirm unidimensionality. Hypothesis testing was scored adequate if the correlation with a questionnaire that assessed fatigue (convergent validity) was ≥0.50, or ≥75% of the results were in accordance with a priori defined hypotheses, and the correlations with other constructs (divergent validity) were lower than the correlations with fatigue. A score unclear was given if only the correlation with questionnaires measuring another construct than fatigue (divergent validity) was investigated. Cross-cultural validity was defined as the degree to which the performance of the items on a translated or culturally adapted health-related patient-reported outcomes (HR-PRO) instrument is an adequate reflection of the performance of the items of the original version of the HR-PRO instrument [16].
As no gold standard exits for fatigue questionnaires, criterion validity was not evaluated.
Responsiveness
Responsiveness was defined as the ability of a questionnaire to detect change over time in the construct to be measured [16]. Responsiveness refers to the validity of a change score [21] and scored adequate if the change score correlated ≥0.50 with the change score of an instrument assessing fatigue, or if ≥75% of the results were in accordance with a priori defined hypotheses, or if the area under the receiver operator characteristic curve (AUC) was ≥0.70 [18].
Interpretability
Interpretability was defined as the degree to which one can assign qualitative meaning to an instruments’ quantitative scores or change in scores. Authors should provide information about clinically relevant differences in scores between subgroups (mean or median with distribution of scores), floor and ceiling effects and the MIC [21]. A floor or ceiling effect was present if >15% of patients achieved the lowest or highest possible score on a questionnaire [18].
Results
Search
The search yielded 5,336 records, of which 56 studies were retrieved in full text for further assessment. This resulted in the exclusion of another 18 studies [10, 22‐38] (see Fig. 1). Thirty-eight studies were included in the review, investigating 31 different self-report fatigue questionnaires [3, 39‐75]. The FSS was most frequently investigated (n = 20) and the only questionnaire validated in patients with MS, PD and stroke. Characteristics of the included studies are presented in Table 2.
pExpressed as percentage of total score body care and movement subscale
qExpressed as percentage of total score mobility subscale
rExpressed as percentage of total score ambulation subscale
sExpressed as percentage of total score alertness behaviour subscale
×
Characteristics of questionnaires
Table 3 presents the characteristics of the included self-report questionnaires. Most questionnaires aimed to assess the impact of fatigue on activities in daily life (Fatigue Impact Scale for Daily use (D-FIS), Adapted French version of Fatigue Impact Scale (EMIF-SEP), Fatigue Assessment Scale (FAS), FIS, Fatigue Severity Scale 5 item version (FSS-5), MFI, MFIS, Modified Fatigue Impact Scale Cognitive and Physical (MFIS C-5/MFIS P-8), Parkinson Fatigue Scale 2-point scale version (PFS-16 (2)), Parkinson Fatigue Scale 5-point scale version (PFS-16 (5)), Performance Scale Fatigue subscale (PS-F), Unidimensional Fatigue Impact Scale (U-FIS), Visual Analogue Scale-1, 2 or 3 (VAS-1, VAS-2, VAS-3), Würzburger Erschöpfungsinventars bei Multiple sclerosis (WEIMUS)), whereas six questionnaires focused primarily on fatigue severity (Multidimensional Fatigue Symptom Inventory general subscale (MFSI-G), Profile Of Mood States Fatigue subscale (POMS-F), Rhoten Fatigue Scale (RFS), Short-form-36 Vitality subscale (SF-36-V), Short-form-36 Vitality subscale version 2.0 (SF-36-V (V2.0)), Swedish Occupational Fatigue Inventory (SOFI)).
Table 3
Characteristics of included questionnaires
Questionnaire
Construct assessed
Recall period
Dimensions (number of items)
Response options (range)
Range of scores
Time to administer
Ease of scoring
CIS-20R
Impact of fatigue
Fatigue severity
Last 2 weeks
Subjective experience of fatigue (8)
Reduction in motivation (4)
Reduction in activity (3)
Reduction in concentration (5)
Total (20)
7-point Likert
(1–7)
20–140
(Best–worst)
Not reported
Easy
D-FIS
Impact of fatigue
Last day
One dimension
Total (8)
5-point Likert (0–4)
0–32
(Best–worst)
Not reported
Easy
EMIF-SEP
Impact of fatigue
Last month
Cognitive (10)
Physical (13)
Psychological (4)
Social (13)
Total (40)
4-point Likert
(1–4)
0–100a
(Best–worst)
Not reported
Difficulta
FACIT-F
Impact of fatigue
Fatigue severity
Last week
One dimension
Total (13)
5-point Likert
(0–4)
0–52
(Worst–best)
Not reported
Easy
FAI
Impact of fatigue
Fatigue severity
Last 2 weeks
Psychological consequencesb
Severityb
Situation—specificb
Response to restb
Total (29)
7-point Likert
(1–7)
29–203
(Best–worst)
Not reported
Easy
FAS
Impact of fatigue
Usually…
One dimension
Total (10)
5-point Likert
(1–5)
10–50
(Best–worst)
Not reported
Easy
FIS
Impact of fatigue
Last month
Cognitive (10)
Physical (10)
Social (20)
Total (40)
5-point Likert
(0–4)
0–160
(Best–worst)
Not reported
Easy
FSMC
Impact of fatigue
Fatigue severity
Factors influencing fatigue
In general…
Cognitive (10)
Motor (10)
Total (20)
5-point Likert
(1–5)
20–100
(Best–worst)
Not reported
Easy
FSS
Impact of fatigue
Fatigue severity
Not specified
One dimension
Total (9)
7-point Likert
(1–7)
1–7c
(Best–worst)
Not reported
Moderatec
FSS-7
Impact of fatigue
Fatigue severity
Not specified
One dimension
Total (7)
7-point Likert (1–7)
1–7c
(Best–worst)
Not reported
Moderatec
FSS-5
Impact of fatigue
Not specified
One dimension
Total (5)
7-point Likert
(1–7)
0–100d
(Best–worst)
Not reported
Moderated
Easye
MFI
Impact of fatigue
Lately…
General (4)
Physical (4)
Reduced activity (4)
Reduced motivation (4)
Mental (4)
Total (20)
5-point Likert
(1–5)
20–100
(Best–worst)
Not reported
Easy
MFIS
Impact of fatigue
Last month
Cognitive (10)
Physical (9)
Social (2)
Total (21)
5-point Likert
(0–4)
0–84
(Best–worst)
Not reported
Easy
MFIS C-5/MFIS P-8
Impact of fatigue
Last month
Cognitive (5)
Physical (8)
Total (13)
5-point Likert
(0–4)
0–52
(Best–worst)
Not reported
Easy
MFSI-G
Fatigue severity
Last week
One dimension
Total (6)
5-point Likert
(0–4)
0–24
(Best–worst)
Not reported
Easy
MFSS
Factors influencing fatigue
Not specified
One dimension
Total (6)
7-point Likert
(1–7)
1–7c
(Best–worst)
Not reported
Moderatec
NFI-MS
Fatigue severity
Factors influencing fatigue
Last 2 weeks
Abnormal nocturnal sleep (5)
Cognitive (4)
Physical (8)
Relief by rest (6)
Summary scale (10)
Total (33)
4-point Likert
(0–3)
0–99e
(Best–worst)
Not reported
Moderated
Easye
NHP-E
Impact of fatigue
Fatigue severity
Not specified
One dimensional
Total (3)
Adjectival
(Weighted score per item)
0–100
(Best–worst)
Not reported
Easy
PFS-16 (2)
Impact of fatigue
Last 2 weeks
One dimension
Total (16)
2-point Likert
(0–1)
0–16
(Best–worst)
Not reported
Easy
PFS-16 (5)
Impact of fatigue
Last 2 weeks
One dimension
Total (16)
5-point Likert
(1–5)
1–5c
(Best–worst)
Not reported
Moderatec
POMS-F
Fatigue severity
Last week
One dimension
Total (6)
5-point Likert
(0–4)
0–24
(Best–worst)
Not reported
Easy
PS-F
Impact of fatigue
Last month
One dimension
Total (1)
6-point Likert
(0–5)
0–5
(Best–worst)
Not reported
Easy
RFS
Fatigue severity
Last 2 weeks
One dimension
Total (1)
11-point Likert
(0–10)
0–10
(Best–worst)
Not reported
Easy
SF-36-V
Fatigue severity
Last month
One dimension
Total (4)
6-point Likert
(1–6)
4–24
(Worst–best)
Not reported
Easy
SF-36-V (V2.0)
Fatigue severity
Last month
One dimension
Total (4)
5-point Likert
(1–5)
4–20
(Worst–best)
Not reported
Easy
SOFI
Fatigue severity
Last 6 months
Lack of energy (4)
Lack of motivation (4)
Physical discomfort (4)
Physical exertion (4)
Sleepiness (4)
Total (20)
7-point Likert
(0–6)
0–30f
(Best–worst)
Not reported
Moderatef
U-FIS
Impact of fatigue
Last week
One dimension
Total (22)
4-point Likert
(0–3)
0–66
(Best–worst)
Not reported
Easy
VAS-1
Impact of fatigue
Not specified
One dimension
Total (1)
100 mm VAS
0–100g
(Best–worst)
Not reported
Moderateg
VAS-2
Impact of fatigue
Not specified
One dimension
Total (1)
100 mm VAS
0–100g
(Best–worst)
Not reported
Moderateg
VAS-3
Impact of fatigue
Not specified
One dimension
Total (1)
100 mm VAS
0–100g
(Best–worst)
Not reported
Moderateg
WEIMUS
Impact of fatigue
Last 2 weeks
Cognitive (9)
Physical (8)
Total (17)
5-point Likert
(0–4)
0–68
(Best–worst)
Not reported
Easy
aAdjusted total score on 0–100 scale
bNot reported
cAverage of total summed items
dOrdinal-interval (Rasch) transformation
eSummed raw (ordinal) score
fSummed total of averaged domain scores
gVisual analogue scale
Fifteen unidimensional (D-FIS, Functional Assessment of Chronic Illness Therapy Fatigue subscale (FACIT-F), FAS, FSS, Fatigue Severity Scale 7 item version (FSS-7), FSS-5, MFSI-G, Multiple sclerosis-specific Fatigue Severity Scale (MFSS), Nottingham Health Profile Energy subscale (NHP-E), PFS-16 (2), PFS-16 (5), POMS-F, SF-36-V, SF-36-V (2.0), U-FIS) and eleven multidimensional questionnaires (Checklist Individual Strength (CIS-20R), EMIF-SEP, Fatigue Assessment Instrument (FAI), FIS, Fatigue Scale for Motor and Cognitive functions (FSMC), MFI, MFIS, MFIS C-5/MFIS P-8, Neurological Fatigue Index MS (NFI-MS), SOFI, WEIMUS) were identified. The total number of items per questionnaire varied from 3 (NHP-E) to 40 (EMIF-SEP, FIS). Three visual analogue scales (VAS-1, VAS-2 and VAS-3) and two single-item Likert scales (PS-F, RFS) were included. Six disease-specific questionnaires were found: the MFSS, NFI-MS, PS-F and WEIMUS for patients with MS and the PFS-16 (2) and PFS-16 (5) for patients with PD.
Most questionnaires were found easy to administer. One questionnaire (EMIF-SEP) uses a complex formula to calculate an adjusted total score from 0 to 100, and for two questionnaires (FSS-5, NFI-MS), a nomogram was provided [65, 66] for ordinal-interval (Rasch) transformation. None of the included studies reported on the time needed to complete the questionnaires.
Measurement properties and methodological quality
Details about the investigated measurement properties and the methodological quality of the included studies are summarized in Table 4. Most studies investigated reliability and construct validity, whereas results on measurement error and responsiveness were often not reported.
Table 4
Methodological quality and investigated measurement properties per study
Eight out of 31 studies that investigated hypothesis testing [41, 43, 50, 51, 61, 62, 64, 66] formulated a priori hypothesis about the expected direction or magnitude of the correlation between the investigated questionnaires. Seven studies [39, 40, 54, 59, 61, 70, 75] that translated a questionnaire scored poor methodological quality because the translated questionnaires were not pre-tested in a small sample to check interpretation, cultural relevance and ease of comprehension of the translation.
All studies [53, 56, 69, 71, 74] that reported on responsiveness scored poor methodological quality.
Overall quality of measurement properties
Table 5 presents the overall quality of the measurement properties per self-report questionnaire, accompanied by the level of evidence.
Table 5
Data synthesis, levels of evidence and overall quality of measurement properties per questionnaire
Questionnaire
Population
Measurement properties
Internal consistency
Reliability
Measurement error
Content validity
Structural validity
Hypothesis testing
Cross-cultural validity
Responsiveness
CIS-20R
MS
+
Limited
?
Unknown
−
Limited
?
Unknown
D-FIS
MS
+
Limited
+
Limited
+
Limited
−
Limited
?
Unknown
PD
+
Moderate
?
Unknown
+
Limited
−
Limited
EMIF-SEP
MS
+
Moderate
+
Limited
+
Moderate
?
Unknown
FACIT-F
PD
+
Limited
+
Limited
−
Moderate
+
Limited
FAI
MS
+
Limited
−
Limited
−
Limited
?
Unknown
FAS
Stroke
±
Conflicting
+
Limited
?
Unknown
+
Limited
−
Limited
FIS
MS
?
Unknown
±
Conflicting
?
Unknown
−
Moderate
?
Unknown
?
Unknown
FSMC
MS
+
Moderate
+
Limited
+
Limited
+
Moderate
+
Limited
FSS
MS
+
Limited
+
Moderate
−
Strong
±
Conflicting
?
Unknown
?
Unknown
PD
+
Limited
−
Moderate
+
Moderate
Stroke
?
Unknown
?
Unknown
?
Unknown
FSS-7
MS
+
Moderate
FSS-5
MS
±
Conflicting
MFI
MS
−
Limited
PD
−
Limited
MFIS
MS
−
Limited
+
Moderate
?
Unknown
±
Conflicting
+
Moderate
?
Unknown
?
Unknown
MFIS C-5/MFIS P-8
MS
+
Moderate
MFSI-G
Stroke
+
Limited
+
Limited
?
Unknown
+
Limited
−
Limited
MFSS
MS
−
Limited
?
Unknown
+
Limited
−
Limited
NFI-MS
MS
+
Limited
+
Limited
+
Limited
−
Limited
NHP-E
MS
+
Moderate
PD
+
Limited
PFS-16 (2)
PD
?
Unknown
?
Unknown
+
Limited
?
Unknown
PFS-16 (5)
PD
−
Moderate
−
Limited
+
Limited
+
Moderate
+
Moderate
?
Unknown
POMS-F
Stroke
+
Limited
+
Limited
?
Unknown
+
Limited
+
Limited
PS-F
MS
Not applicable
Not applicable
+
Limited
RFS
PD
Not applicable
Not applicable
+
Limited
SOFI
MS
−
Limited
−
Limited
−
Limited
SF-36-V
MS
+
Limited
SF-36-V (V2.0)
Stroke
+
Limited
−
Limited
?
Unknown
+
Limited
−
Limited
U-FIS
MS
−
Moderate
+
Moderate
+
Moderate
+
Moderate
+
Moderate
?
Unknown
?
Unknown
VAS-1
MS
Not applicable
−
Limited
Not applicable
?
Unknown
VAS-2
MS
Not applicable
−
Limited
Not applicable
?
Unknown
VAS-3
MS
Not applicable
−
Limited
Not applicable
?
Unknown
WEIMUS
MS
?
Unknown
?
Unknown
?
Unknown
−
Limited
+ Adequate, − Not adequate, ± Conflicting, ? Unknown
Reliability
The EMIF-SEP and FSMC showed moderate evidence for adequate internal consistency in patients with MS (Cronbach’s α = 0.82–0.93) [3, 68] and the D-FIS in patients with PD (Cronbach’s α = 0.93) [61]. Limited evidence for adequate internal consistency was found for the D-FIS and FSS in patients with MS (Cronbach’s α = 0.91–0.93) [41, 46], the FACIT-F and FSS in patients with PD (Cronbach’s α = 0.90–0.94) [49, 50], and the MFSI-G, POMS-F and SF-36-V (V2.0) in patients with stroke (Cronbach’s α = 0.76–0.93) [63].
Moderate evidence was found for adequate reliability for the FSS, MFIS and U-FIS in patients with MS (CC or ICC = 0.73–0.93) [39, 43, 52, 54, 64, 71]. Limited evidence for adequate reliability was found for the FAS, MFSI-G and POMS-F in patients with stroke (ICC = 0.74–0.77) [63] and the FACIT-F in patients with PD (ICC = 0.84–0.85) [50]. Reliability of the PFS-16 (5) was found not adequate (limited evidence, CC = 0.63) [42].
Measurement error was investigated for the CIS-20R, D-FIS, FAS, FSS, MFIS, MFSI-G, POMS-F and SF-36-V (V2.0), but only one study on the D-FIS used in patients with MS [41] reported details about the MIC. There was limited evidence for adequate measurement error of the D-FIS in patients with MS (SEM = 3.18 and MIC = 3.65) [41].
Validity
Content validity was investigated for the FAS, FIS, FSMC, MFSI-G, NFI-MS, PFS-16 (2), PFS-16 (5), POMS-F, SF-36-V (V2.0) and U-FIS. Moderate evidence was found for adequate content validity of the U-FIS in patients with MS [43, 64]. Limited evidence for adequate content validity was found for the FSMC and NFI-MS in patients with MS [66, 68], for the PFS-16 (2) and PFS-16 (5) in patients with PD [42], and for the FAS, MFSI-G, POMS-F and SF-36-V (V2.0) in patients with stroke [63].
Moderate evidence for adequate structural validity was found for the EMIF-SEP, FSMC (% total explained variance = 61.4–61.5) [3, 68] and U-FIS [43] in patients with MS and for the PFS-16 (5) in patients with PD (% total explained variance = 63.2–64.0) [42]. Four studies that applied IRT methods to assess structural validity demonstrated misfits for items in the FSS and MFIS in patients with MS [58, 65, 67] and in the FACIT-F and FSS in patients with PD [50]. Based on these analyses, new versions for the FSS (FSS-7, FSS-5) [58, 65] and for the MFIS (MFIS C-5/MFIS P-8) [67] were introduced.
Moderate evidence for convergent validity was found for the MFIS (CC = 0.54–0.89 with CIS-20R, FSMC, FSS, PS-F, WEIMUS, WEIMUS Cognitive subscale, WEIMUS Physical subscale) [46, 54, 60, 68, 71], U-FIS (CC = 0.48–0.86 with NHP-E) [43, 64] and NHP-E (CC = 0.48–0.86 with U-FIS) [43, 64] in patients with MS, and for the FSS (CC = 0.62–0.84 with FACIT-F, NHP-E, PFS-16 (5)) [49, 50] and PFS-16 (5) (CC = 0.71–0.84 with FSS, RFS) [42, 49] in patients with PD.
In 13 studies [3, 39, 40, 43, 48, 53, 54, 57, 59, 61, 70, 71, 75], questionnaires were translated. None of these studies investigated cross-cultural validity by means of confirmatory factor analysis or differential item functioning (DIF).
Responsiveness
Five studies [53, 56, 69, 71, 74] reported on responsiveness. None of these studies presented details about the correlation coefficient between change scores in the investigated questionnaires with change in an external anchor. Therefore, responsiveness was scored unknown for these questionnaires.
Interpretability
Clinically relevant differences in scores between subgroups were reported for the FIS [48], FSS [45], U-FIS [43, 64, 74] and WEIMUS [47] in patients with MS, and for the FACIT-F [50], FSS [50] and PFS-16 (5) [57] in patients with PD.
No floor or ceiling effects were found for the D-FIS [41], FSS [53], FSS-7 and FSS-5 [58], MFIS [53, 54], MFIS C-5/MFIS P-8 [67], NFI-MS [66] and U-FIS [74] in patients with MS. The SOFI showed a floor effect in patients with MS (on 12 of the 20 items, more than 25% of patients achieved the lowest possible score) [51]. The D-FIS [61], FACIT-F [50], FSS [50], PFS-16 (5) and PFS-16 (2) [57] showed no floor or ceiling effects in patients with PD.
Values for the MIC were reported for the D-FIS (MIC = 3.65) [41], FIS (MIC = 9.0–24.0) [69] and U-FIS (MIC = 2.4–7.0) [74] in patients with MS.
Discussion
To our knowledge, this review is the first that systematically appraised and summarized the evidence on the measurement properties of self-report fatigue questionnaires validated in patients with MS, PD or stroke, by taking the methodological quality of the included studies into account. Thirty-one questionnaires were evaluated. No multidimensional questionnaires were identified that were adequately validated in patients with PD or stroke. Moderate evidence was found for adequate internal consistency and structural validity of the FSMC and for adequate reliability and structural validity of the U-FIS in patients with MS. Therefore, we recommend the FSMC for the multidimensional, and the U-FIS for the unidimensional assessment of fatigue in patients with MS. The FACIT-F and FSS show promise for the assessment of fatigue in patients with PD, and the POMS-F for patients with stroke. However, reliability and validity should be confirmed in high-quality studies on the FACIT-F, FSS and POMS-F in these populations. Above recommendations should be considered with caution, given that studies investigating measurement error, responsiveness and interpretability are lacking. Second, as the level of evidence supporting the overall quality of most measurement properties was limited, future high-quality studies may change our recommendations.
Two reviews [8, 10] recommend on the use of a questionnaire. One review [10] suggested the FIS and MFIS in patients with MS. The other review [8] recommended the FSS for the unidimensional assessment of fatigue in patients with PD. Although not specifically validated in PD, the MFI was recommended for the multidimensional assessment of fatigue in patients with PD [8]. These recommendations are partially in line with our findings. However, taken the methodological quality of the studies included in our systematic review into account, most measurement properties of the FIS showed only unknown level of evidence. In addition, four studies [50, 58, 65, 67] that applied IRT methods to investigate structural validity demonstrated misfits for some items in the FSS and MFIS.
The inconsistent scores for hypothesis testing confirm that different questionnaires measure different aspects or constructs of fatigue. Unfortunately, details on the construct of fatigue measured by a questionnaire were often not reported. Furthermore, factors contributing to fatigue in patients with MS, PD or stroke are still not well known [2, 76, 77]. Translational research, bridging pre-clinical and clinical research [78], focused on physiological and clinical aspects contributing to peripheral and central fatigue [6], may provide input for more clearly defined concepts and dimensions of fatigue. As both fatigue and most clinical aspects contributing to fatigue fluctuate in time, associations between these factors may be more accurately reflected using longitudinal study designs with repeated measures in time [79]. Repeated measurement designs allow the investigation of the longitudinal construct validity of fatigue measures.
For now, we suggest that clinicians assessing fatigue carefully consider whether a questionnaire reflects the most relevant aspects of fatigue of their interest. Furthermore, a comprehensive evaluation of fatigue should be accompanied by the assessment of clinically related factors such as mood and sleep. Acknowledging that each fatigue questionnaire measures different aspects of fatigue, we recommend the simultaneous use of different questionnaires in research.
Interpretability is considered an important characteristic of a measurement scale [16], unfortunately, only a few studies reported details on clinically relevant differences in scores between subgroups [43, 45, 47, 48, 50, 57, 64, 74], floor and ceiling effects [41, 50, 51, 53, 54, 57, 58, 61, 66, 67, 74] and the MIC [41, 69, 74]. This makes it difficult to interpret scores and change scores on a fatigue questionnaire in both clinical practice and research.
Although it is believed that measurement properties are sample dependent [80], no major differences in measurement properties were found for questionnaires that were evaluated in more than one population. For example, all estimates of measurement properties for the D-FIS were consistent in patients with MS and PD. The FSS showed consistent scores for most measurement properties that were evaluated in patients with MS, PD and stroke. In addition, another review [8] concluded that the items of the disease-specific PFS-16 (5) did not differ much from other generic fatigue questionnaires and that it provided no clear advantages above a generic questionnaire for use in patients with PD. Furthermore, it is not clear whether manifestations of fatigue are different between neurological disorders [8]. These results suggest that generic fatigue questionnaires presented in this review can be used interchangeably in patients with MS, PD and stroke and favour a generic approach for the assessment of fatigue. In contrast, studies using IRT methods showed misfits on the FSS for four items in patients with MS [65], and for only one item in patients with PD [50]. This difference might have been caused by a difference in statistical power between both studies [65], but it is also possible that it was related to DIF in patients with MS and PD [65]. This emphasizes the importance of disease-specific validation for fatigue questionnaires used in patients with MS, PD and stroke. Above-mentioned findings suggest that self-report fatigue questionnaires should contain a core set of items assessing generic aspects of fatigue, whereas some additional items are more disease specific. We therefore recommend the adaptation of existing questionnaires, incorporating a uniform section on general aspects of fatigue and a section with disease-specific items. Items to assess general aspects of fatigue may be derived from the recently developed Patient-Reported Outcomes Measurement Information System (PROMIS) fatigue item bank [81].
This systematic review has some limitations. First, only studies published in Dutch, English, French or German were included. This language restriction resulted in the exclusion of six articles [22, 28‐30, 35, 38]; however, these studies evaluated a diversity of questionnaires and language versions, so it is not likely that this resulted in selection bias. Second, the COSMIN checklist has some items that require subjective judgment, which may lead to disagreement between raters. However, we tested the COSMIN checklist with all reviewers before assessing the methodological quality of the included studies, and one reviewer (RE) was involved in the assessment of all studies to improve consistency in rating across studies. Third, the quality criteria we applied for rating measurement properties heavily weighed on classical test theory (CTT). As a consequence, IRT methods were not considered for underpinning the structural validity of questionnaires. To overcome this incompleteness, we decided, post hoc, that any misfit in a questionnaire displayed by a study using IRT methods was judged as not adequate structural validity.
Conclusion
We recommend the FSMC and U-FIS for the assessment of fatigue in patients with MS. The FACIT-F and FSS show promise in patients with PD, and the POMS-F for patients with stroke. No multidimensional questionnaires were adequately validated in patients with PD or stroke. Future studies should focus on translational research in which assumed underlying physiological and clinical aspects contributing to fatigue are investigated longitudinally, as perceptions of fatigue often show fluctuations in time. Such studies may provide input for the development of the theoretical construct of self-report fatigue questionnaires. We suggest that existing questionnaires should be adapted to contain both a uniform section that reflects general aspects of fatigue, and a disease-specific section that contains items that are related with physiological and clinical aspects of underlying disease. Studies on responsiveness and the MIC of fatigue questionnaires in patients with MS, PD and stroke are needed, to establish whether an instrument can detect meaningful changes in clinical practice and research.
Acknowledgments
The authors would like to thank R. Spijker (MSc), medical information specialist and Trial Search Coordinator from the Academic Medical Center and the Dutch Cochrane Centre, Amsterdam, The Netherlands for his valuable support regarding the design of the search strategy. Financial disclosures: This research project was supported by a grant from the Royal Dutch Society for Physical Therapy, Scientific College Physical Therapy, project number: BU003-2010, and the Stichting MS Research, project number 04-553 MS.
Open Access
This article is distributed under the terms of the Creative Commons Attribution Noncommercial License which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Open Access This is an open access article distributed under the terms of the Creative Commons Attribution Noncommercial License (
https://creativecommons.org/licenses/by-nc/2.0
), which permits any noncommercial use, distribution, and reproduction in any medium, provided the original author(s) and source are credited.
Binnen de bundel kunt u gebruik maken van boeken, tijdschriften, e-learnings, web-tv's en uitlegvideo's. BSL Podotherapeut Totaal is overal toegankelijk; via uw PC, tablet of smartphone.
