Development and Validation of the Norfolk Quality of Life Fatigue Tool (QOL-F), a New Measure of Perception of Fatigue

Abstract

Objectives

Design a questionnaire able to evaluate and distinguish between cognitive and physical aspects of fatigue in different age groups of “non-diseased” people, to guide appropriate prevention and interventions for the impact of frailty that occurs in normative aging.

Study Design/ Participants

Develop the Norfolk QOL-Fatigue (QOL-F) with items of cognitive and physical fatigue, anxiety, and depression from validated questionnaires including items from Patient/Reported/Outcomes/Measure/Information/System (PROMIS) databank. Administer the preliminary QOL-F to 409 healthy multi-ethnic local participants (30 to 80yrs) divided into five age-groups.

Methods

We distilled the item pool using exploratory (EFA) and confirmatory factor analysis (CFA). EFA identified five latent groups as possible factors related to: problems from fatigue, subjective fatigue, reduced activities, impaired activities/of/daily/living (ADLs), and depression.

Results

CFA demonstrated good overall fit (χ2(172)=1094.23, p<.001; TLI=.978; RMSEA=.049) with factor loadings >.617 and strong inter-factor correlations (.69 to .83) suggesting that fatigue in each domain is closely related to other domains and to overall scale except for ADLs. Five-factor solution displayed good internal consistency (Cronbach’s α=.78 to .94). Total and domain scores were fairly equivalent, in all age/groups except for 40–49 year/ old group who had better overall scores. The 70–79 year/olds had better ADLs scores. In item response analysis, factor scores in the different age groups were similar, so age may not be a significant driver of fatigue scores. Fatigue scores were significantly higher in females than in males (p<.05).

Conclusions

Fatigue is a perceived cognitive phenomenon rather than an objective physical measure. The Norfolk QOL-F tool, comprised of 35 items and five domains is sensitive to aging, cognitive and emotional function and able to distinguish a unique subset of ostensibly normal 40–49 year/olds with better QOL-F scores than other age groups. Importantly, 70–79 year/olds had better ADLs scores indicating potential benefit from physical, balance and cognitive interventions.

Introduction

Background and Rationale for the Development of a New Fatigue Questionnaire

Chronic fatigue occurs in 15–45% of the American population.^1,2 Fatigue can affect all aspects of quality of life (QOL) ranging from mood, physical functioning and activities of daily living (ADLs). For older individuals in particular, fatigue has the potential to restrict their physical activity – a consequence that can have a dramatic impact on health and well-being. Reduced activity is often a precursor for a number of problems including increased falls risk, decline in muscle function, general inanition, depression and apathy. For any given individual, the level of fatigue is determined not only by the availability, utilization, and/or restoration of resources needed to perform activities,³ but also by the demands of the activities to be performed. Several approaches have been developed to measure fatigue in the community through questionnaires and surveys. However, there is no gold standard with which to compare these instruments, nor has there been a uniform approach to obtain consensus on the concept and its measurement. Some approaches assess the symptomatic complaints of distress or functional impact of fatigue.^4,5 These tools are oriented to health-related quality of life and are useful for screening as well as diagnosing fatigue, and have proven valuable in evaluating patients with medical conditions.^6–8 However, none of the current tools are sensitive to aging and its complications, including the impact of depression, somatic and autonomic nerve dysfunction, sarcopenia and frailty that occur in older, otherwise non-diseased populations. In addition, none of the existing tools assess cognitive versus physical fatigue. This tool has been designed to evaluate both cognitive and physical aspects of fatigue in different age groups.

Methods

Study Population

Development and validation of the new fatigue tool was planned in a multi-ethnicity population of 409 participants, divided into five age groups, each with 80 participants, ranging from 30 to 79 years. The patient population was recruited from the Southeastern Virginia and Northeastern North Carolina areas. The service area included a multi-ethnic city and suburban community with African American, Hispanic, Caucasian and Asian groups. There was diversity in ethnicity, education levels, literacy abilities, and socioeconomic status. The language was predominantly English including a small number of bilingual Asians.

Participants were recruited from the local community by written invitation, newspaper advertisements and flyers placed within retirement homes, with the assistance of the Glennan Center for Geriatrics at Eastern Virginia Medical School. Participant selection criteria included: no ocular or systemic disease (including diabetes), no recent or recurrent history of musculoskeletal injury, no neurological conditions, no history of vertigo, no use of an aid while walking, no difficulty standing upright, no visible tremor or uncorrected visual deficits. It was statistically determined that 400 participants would be sufficient to reach covariance matrix stability and consistent with the 10 participants per item rule of thumb. The protocol was approved by the Eastern Virginia Medical School Institutional Review Board, and was performed in accordance with the Declaration of Helsinki. Written informed consent was obtained from all individual participants included in the study.

Item Selection

To develop a comprehensive fatigue questionnaire, we examined pre-existing, validated, generic, health-related QOL questionnaires and included items associated with cognitive, mental and physical fatigue, anxiety, fear, stress, depression, and perceived health status. We added these to existing items on physical functioning, ADLs, and depression scales from our previously validated neuropathy and neuroendocrine tools, Norfolk QOL-DN and Norfolk QOL-NET,^9;10 to create a preliminary Quality of Life Fatigue questionnaire (QOL-F). We also explored the relationship between the 9 fatigue items in the Fatigue Severity Scale (FSS) and the Norfolk QOL-F. Similarly, we sought to correlate the depression questions in the Norfolk QOL-F with the questions in the CES-D and the physical functioning questions with the Modified Falls Efficacy Scale (MFES).

The original item pool consisted of over 100 items and we used the Delphi method to review the questions. The Delphi panel of experts was comprised of physicians, nurse clinicians, psychologists, education specialists, and exercise physiologists. After general consensus by the Delphi panel (unweighted), a draft questionnaire was created with responses on a 5-point Likert scale. A focus group of 10 patients, representative of the intended study population, was developed to assess readability and content as well as question redundancy and suggestions for additional questions. After group discussion, agreement was reached; the changes were made and the preliminary item pool comprising 42 items was arranged so that each item related to a priori sub-scales of Symptoms of General Fatigue (lack of energy or tiredness), General Health, ADLs, Physical Functioning, Cognitive Scale, and Feelings Scale.

During this development process, we explored another psychometric methodology - item response theory - and the use of computerized adaptive testing, using the PROMIS (Patient Reported Outcomes Measurement Information System) inventory, in which items are generated from an item bank. The aim of this tool, promoted by NIH, is to improve and standardize the measurement of patient reported outcomes. We examined the 96-item PROMIS inventory of fatigue questions,^11;12 and again using the Delphi technique with our panel of experts, we chose 75 items to be combined with the 42 in the preliminary QOL-F, giving a total of 117 questions as a new basis for the combined Norfolk/PROMIS QOL-F tool. We believed that the addition of the items from PROMIS would strengthen these correlations and improve the structure of the ultimate tool. This sequence of events is shown diagrammatically in Figure 1. All 117 questions were answered by all participants.

Figure 1. Item Development.

Norfolk QOL-DN: Norfolk quality of life patient reported outcome measures for diabetic neuropathy:

SF-36: (Short Form 36) self-reported functional and well-being measures

CES-D: Center for Epidemiologic Studies Depression Scale

Norfolk QOL-NET: Norfolk quality of life patient reported outcome measures for Neuroendocrine Tumors

PROMIS: (Patient Reported Measurement Information System) Item bank for measuring quality of life with an item-response theory approach

N is number of items

Item-Level Diagnostics

All items in the initial pool were examined for distribution, skewness, and kurtosis to identify potential floor or ceiling effects, invariance, and other problems that would interfere with the planned factor analysis. Some items measuring impairment in activities of daily living showed expected low base rates, but these items were judged to have sufficient variance and content importance to be carried forward to the next stage of analysis.

Identifying Dimensionality between Items from PROMIS and Norfolk QOL-F

Exploratory Factor Analysis (EFA) was performed on the data from the full sample (n=409) to identify dimensionality among items from the PROMIS and Norfolk QOL-F. To assess the factor structure, a principal axis factor analysis with a rotation method of Promax with Kaiser Normalization was used. Parallel analysis was used to select the number of factors to extract. Items that were identified as having low communalities (<.30), loading <.40, or ambiguous loadings were dropped. This yielded a tool with refined items. A Confirmatory Factor Analysis (CFA) model was fit in SAS version 9.4 using the five scales and 35 items that were derived via EFA. This analysis was fit using a conservative fit algorithm, Weight measured as least squares with Mean and Variance adjustment.

Results

Exploratory Factor Analysis

All EFA analyses were conducted using principal axis factoring with Promax rotation. We used an oblique rotation strategy because we hypothesized that the underlying dimensions in all factor analyses would be correlated. Due to the large number of combined PROMIS and Norfolk-QOL items relative to our sample size, we first sought to reduce the size of the overall combined item pool. Accordingly, we examined the factor structure of the 42 preliminary Norfolk QOL-F items and 75 PROMIS items separately using an SPSS program developed by O’Connor to evaluate the number of factors to retain.^13,14 For each parallel analysis we generated 1,000 random data sets based on permutations of our raw data. We initially retained factors for the PROMIS and Norfolk-QOL analyses if their Eigen values were greater than the 95^th percentile of the randomly generated data. Within each instrument, items that were identified as having low communalities (below .30), pattern coefficients below .40, or high cross loadings were dropped.

From this procedure, a total of 56 items were retained for further analysis. Because the Norfolk-QOL and PROMIS scales were developed independently, the combined item pool was examined for duplicative item content. Here, an EFA was again conducted and the lowest-loading item of a pair with similar content was dropped. This step resulted in a reduced pool of 42 items. We subsequently conducted an EFA on the actual data and a parallel analysis using 1,000 randomly generated data sets based on permutations of the actual data. The parallel analysis suggested five factors. As previously described, items with low communalities, pattern coefficients below .40, or ambiguous loadings were dropped. Table 1 shows item loadings into the five factors. All items with loading scores <.4 were dropped.

Table 1.

Exploratory Factor Analysis: Item Loadings

	Factor
	1	2	3	4	5
1. How often did you feel run-down?	.836
4. How often were you sluggish?	.787
n2 Lack Energy	.783
n1 Tired During the Day	.781
13. How often did you find yourself getting tired easily?	.769
5. How often did you run out of energy?	.764
n3 Sleepy During Day	.739
7. How often were you bothered by your fatigue?	.656
6. How often were you physically drained?	.617
34. How often was it an effort to carry on a conversation because of your fatigue?		.843
31. How often were you too tired to take a bath or shower?		.818
32. How often did you fatigue make it difficult to organize your thoughts when doing things at home?		.707
37. How often were you too tired to think clearly?		.669
43. I am too tired to eat?		.643
36. How often were you too tired to leave the house?		.607
44. I need help doing my usual activities		.598
46. I have to limit my social activity because I am tired		.574		.319
27. How often did your fatigue make it difficult to make decisions?		.552
69. To what degree did your fatigue interfere with your physical functioning?		.551
40. How often were you too tired to take a short walk?		.414
n28 Could not shake off blues			.738
n30 Tired, depressed, crying			.706
n37 Easily Annoyed			.685
n31 Lonely when other people around			.640
n26 Bothered by things that don’t usually bother			.608
n27 Over/undereating			.531
n29 Trouble keeping mind on what doing			.524
n10 Limited in Work or Activities				.825
n11 Difficulty Performing Work				.729
n9 Accomplished Less				.699
n8 Cut Down on Time Work/Activities				.695
n16 Getting on/off Toilet					.795
n14 Dressing					.701
n17 Getting out of Chair					.635
n13 Bathing/Showering		.324			.546

KEY: “n” represents items derived from the Norfolk scales, the remainder are derived from PROMIS.

Color coding matches the five factors which emerge after confirmatory factor analysis in Figure 2

• Problems from Fatigue

• Subjective Fatigue

• Reduced Activities

• Impaired Activities of Daily Living

• Dysphoria (Depression)

Confirmatory Factor Analysis

A CFA model was fit in SAS version 9.4 using the five scales and 42 items that were derived via EFA. This CFA model was fit using Weighted Least Squares with Mean and Variance adjustment (WLSMV), which better accounts for the ordinal response format of items. This model demonstrated good overall fit (χ² (172) = 1094.23, p < .001; TLI = .978; RMSEA = .049) with standardized factor loadings for all factors being .617 or higher. Strong inter-factor correlations were observed, ranging from .690 to .830, suggesting that experience of fatigue in one domain is closely related to experience of fatigue in other domains and overall. See Figure 2 for CFA model loadings and Table 2 for factor inter-correlations.

Figure 2.

Confirmatory Factor Analysis: Model Loadings

Table 2.

Correlations between Confirmed Factors

Factor	1	2	3	4	5
1. Subjective Fatigue	-	.83	.77	.77	.64
2. Problems Due to Fatigue	-	-	.83	.78	.70
3. Dysphoria	-	-	-	.69	.61
4. Reduced Activities	-	-	-	-	.75
5. Activities of Daily Living	-	-	-	-	-

Presents the final scale names and factor correlations for each scale of the Norfolk QOL-F

Items that loaded clearly on each of five factors were designated as constituents of five scales:

1. Subjective Fatigue, 2. Problems Due to Fatigue, 3. Dysphoria, 4. Reduced Activities, 5. Activities of Daily Living.

Strong inter-factor correlations were observed, ranging from .690 to .830, suggesting that experience of fatigue in one area is closely related to experience of fatigue in other areas and overall. “Activities of Daily Living” is an exception and does not correlate with the other factors but is the major driver of fatigue when analyzed using Item Response analysis (see below).

Reliability Analysis

We calculated internal consistency for each of the five resulting scales using Cronbach’s alpha. Scales 1–4 showed good internal consistency (Cronbach’s α = .939, .932, .872, .890, respectively), and scale 5 showed acceptable internal consistency (Cronbach’s α = .783). The CFA showed good overall fit (Tucker-Lewis index = .978, root mean square error of approximation = .049), with standardized factor loading for all factors being .617 or higher.

Items that loaded clearly on each of the five factors were designated as constituents of preliminary subscales. Each of these item sets was examined for internal consistency using item-total correlations and scale-level Cronbach’s alpha.

• Scale #1 (Subjective Fatigue). It is comprised of 9 items and has good internal consistency (Cronbach’s α = .939). Corrected Item total correlations range from .680 to .802.

• Scale #2 (Problems due to Fatigue). It is comprised of 11 items and has good internal consistency (Cronbach’s α = .932). Corrected Item total correlations range from .608 to .797.

• Scale #3 (Dysphoria). It is comprised of 7 items and has good internal consistency (Cronbach’s α = .872). Corrected Item total correlations range from .579 to .699.

• Scale #4 (Reduced Activities). It is comprised of 4 items and has good internal consistency (Cronbach’s α = .890). Corrected Item total correlations range from .744 to .767.

• Scale #5 (Activities of Daily Living). It is comprised of 4 items and has acceptable internal consistency (Cronbach’s α = .783). Corrected Item total correlations range from .556 to .655.

Item Response Theory Analysis

EFA identified five latent groups as possible factors related to: problems from fatigue, subjective fatigue, reduced activities, impaired activities of daily living, and dysphoria. After determining the latent groups from EFA, we performed CFA using an item response theory model to estimate the factor scores related to each of these latent groups. Factor scores for the general model and other latent groups are presented in Table 1. As the responses were ordinal ranging from 0–4, we implemented the graded response model and marginal maximum likelihood method to calculate these factor scores. Low factor scores indicate the lower latent response while higher scores indicate higher latent scores.

Table 3 shows scores for the Total QOL-F and effects of aging in all five groups and all domains. It also graphically displays the scores in the subset of 40–49 year olds shown to have better quality of life and less fatigue and depression than the other groups, except for the 70–79y group, which surprisingly displayed the least fatigue and best quality of life related to the activities of daily living domain.

Table 3:

Scores for the Total QOL-F Fatigue Score and the five domains for all Age Groups

	Entire Group	Group 1 (30–39)	Group 2 (40–49)	Group 3 (50–59)	Group 4 (60–69)	Group 5 (70–79)
Total Fatigue Score	26.44 ± 2.27	33.41 ± 6.18	20.66 ± 3.51*	27.40 ± 4.61	26.43 ± 5.91	24.56 ± 4.80
Factor 1 – Subjective Fatigue	10.08 ± 0.80	12.29 ± 2.04	8.66 ± 1.50	10.07 ± 1.60	9.81 ± 2.04	9.63 ± 1.77
Factor 2 – Problems with Fatigue	7.21 ± 0.78	9.88 ± 2.05	5.50 ± 1.20*	7.53 ± 1.62	6.75 ± 1.98	6.44 ± 1.80
Factor 3 –Dysphoria	5.66 ± 0.48	7.24 ± 1.40	4.06 ± 0.80*	6.13 ± 1.05	6.06 ± 1.27	4.94 ± 0.62
Factor 4 – Reduced Activities	3.03 ± 0.32	3.11 ± 0.71	2.27 ± 0.55	3.26 ± 0.69	3.18 ± 0.80	3.44 ± 0.86
Factor 5 – ADLs	0.44 ± 0.14	0.88 ± 0.36	0.16 ± 0.12	0.40 ± 0.33	0.63 ± 0.50	0.13 ± 0.13+

Data are presented as mean ± SEM; Comparisons between groups were analyzed using Wilcoxon’s Rank Sum test; ADL=activities of daily living; SBP=Systolic

^*P<.05 – Group 1 vs. Group 2

⁺P<.05 – Group 1 vs. Group 5

A summary of all factor scores related to each latent group is presented in Table 4.

Table 4:

Summary of latent groups

Latent Groups	Age Groups (yrs.)					All (pooled) (n=409)	p-value (H0: Mean=0)
	≤39 (n=80)	40–49 (n=83)	50–59 (n=88)	60–69 (n=84)	≥70 (n=74)
General Fatigue Factor	0.23 (1.01)	−0.07 (0.9)	0.07 (0.99)	−0.05 (1.09)	−0.14 (0.79)	0.01 (0.97)	0.838
Problems from Fatigue	0.24 (0.97)	−0.06 (0.88)	0.07 (1.00)	−0.05 (1.04)	−0.12 (0.76)	0.02 (0.94)	0.888*
Subjective Fatigue	0.36 (0.97)	0.02 (0.89)	0 (0.86)	−0.07 (0.93)	−0.25 (0.75)	0.02 (0.90)	0.982*
Dysphoria	0.17 (0.98)	0.03 (0.89)	0.16 (0.84)	−0.11 (1)	−0.13 (0.71)	0.03 (0.9)	0.835*
Reduced activities	0.07 (0.87)	−0.12 (0.78)	0.09 (0.85)	0.18 (0.89)	0.07 (0.86)	0.06 (0.85)	0.627*
Impaired activities of daily living	0.08 (0.49)	0.04 (0.4)	0.19 (0.57)	0.31 (0.76)	0.3 (0.68)	0.18 (0.6)	<0.001*

Note: All data presented as mean (SD). p-value is for testing the null hypothesis that average score is ‘0’.

^*p-value based on Non-parametric (Wilcoxon Signed Rank Test).

SD=Standard Deviation

Noticeably, the average scores of all but one related to impaired ADLs are not different from “0” thus any negative score can be interpreted as less than average and positive scores can be considered more than average score. Also, Shapiro-Wilk test shows that scores for the general fatigue model is normally distributed (p=.093) while factor scores for other latent groups are not normally distributed (p<.05). The differences among age groups on factor scores are compared by using one-way ANOVA, with Tukey’s adjustment for the general fatigue score and pairwise Kruskal-Wallis test for the other latent groups. The results of multiple comparisons are presented in Table 5. In most of the cases, there were not significant differences between the different age groups on the average factor scores of all latent outcomes.

Table 5:

Multiple comparisons of factor scores of different latent groups by age group

Latent Group	Age Groups (yr.)		Average difference	Standard Error	p-value unadjusted	p-value (Adjusted)
General Fatigue	≤39	40–49	0.30	0.15	0.037	0.226
	≤39	50–59	0.16	0.16	0.245	0.772
	≤39	60–69	0.28	0.16	0.076	0.389
	≤39	≥70	0.37	0.15	0.007	0.057
	40–49	50–59	−0.14	0.14	0.362	0.892
	40–49	60–69	−0.02	0.15	0.900	1.000
	40–49	≥70	0.07	0.13	0.546	0.974
	50–59	60–69	0.12	0.16	0.484	0.956
	50–59	≥70	0.21	0.14	0.132	0.556
	60–69	≥70	0.08	0.15	0.506	0.964
	Gender (F vs M)		0.37	0.09	<0.001	<0.001
Problems from Fatigue	≤39	40–49	0.30	0.15	0.041	0.243
	≤39	50–59	0.17	0.15	0.273	0.808
	≤39	60–69	0.29	0.16	0.067	0.352
	≤39	≥70	0.36	0.14	0.010	0.075
	40–49	50–59	−0.13	0.14	0.361	0.891
	40–49	60–69	−0.01	0.15	0.951	1.000
	40–49	≥70	0.06	0.13	0.628	0.989
	50–59	60–69	0.12	0.16	0.432	0.935
	50–59	≥70	0.20	0.14	0.159	0.620
	60–69	≥70	0.07	0.14	0.613	0.987
Subjective Fatigue	≤39	40–49	0.34	0.15	0.020	0.134
	≤39	50–59	0.36	0.14	0.011	0.079
	≤39	60–69	0.43	0.15	0.004	0.032
	≤39	≥70	0.61	0.14	<.0001	0.000
	40–49	50–59	0.02	0.13	0.868	1.000
	40–49	60–69	0.09	0.14	0.533	0.971
	40–49	≥70	0.27	0.13	0.040	0.237
	50–59	60–69	0.07	0.14	0.631	0.989
	50–59	≥70	0.25	0.13	0.050	0.285
	60–69	≥70	0.18	0.13	0.171	0.647
Dysphoria	≤39	40–49	0.14	0.15	0.327	0.864
	≤39	50–59	0.01	0.14	0.922	1.000
	≤39	60–69	0.28	0.15	0.067	0.352
	≤39	≥70	0.30	0.14	0.030	0.192
	40–49	50–59	−0.13	0.13	0.327	0.864
	40–49	60–69	0.14	0.15	0.340	0.875
	40–49	≥70	0.15	0.13	0.229	0.749
	50–59	60–69	0.27	0.14	0.056	0.311
	50–59	≥70	0.29	0.12	0.020	0.137
	60–69	≥70	0.01	0.14	0.916	1.000
Reduced activities	≤39	40–49	0.18	0.13	0.159	0.621
	≤39	50–59	−0.03	0.13	0.832	1.000
	≤39	60–69	−0.11	0.14	0.407	0.921
	≤39	≥70	0.00	0.14	0.991	1.000
	40–49	50–59	−0.21	0.12	0.092	0.442
	40–49	60–69	−0.30	0.13	0.023	0.152
	40–49	≥70	−0.18	0.13	0.162	0.627
	50–59	60–69	−0.09	0.13	0.519	0.967
	50–59	≥70	0.03	0.13	0.843	1.000
	60–69	≥70	0.11	0.14	0.419	0.928
Impaired ADLs	≤39	40–49	0.04	0.07	0.559	0.977
	≤39	50–59	−0.12	0.08	0.152	0.605
	≤39	60–69	−0.24	0.10	0.018	0.123
	≤39	≥70	−0.22	0.10	0.023	0.152
	40–49	50–59	−0.16	0.08	0.035	0.217
	40–49	60–69	−0.28	0.09	0.003	0.027
	40–49	≥70	−0.26	0.09	0.004	0.033
	50–59	60–69	−0.12	0.10	0.249	0.777
	50–59	≥70	−0.10	0.10	0.308	0.846
	60–69	≥70	0.02	0.11	0.882	1.000

p-value < 0.05 is considered significant.

We also included ethnicity and gender in the model to assess whether they have an impact on the general fatigue score. While ethnicity was not a significant factor in predicting the general fatigue score, gender certainly was. The average fatigue score of females was significantly higher than the average fatigue score of males (p-value <.05) (Table 5).

Discussion

We report here on the development of the Norfolk QOL-F tool based upon the factor structure of the Norfolk QOL-F items distilled from 100 questions in a Delphi iterative process and 75 PROMIS items. After EFA and CFA the tool was further distilled to 35 items with Cronbach’s α loading of > .74 resulting in 5 scales: #1 (Subjective Fatigue) with 9 items, #2 (Problems due to Fatigue) with 11 items, #3 (Dysphoria) with 7 items, #4 (Reduced Activities) with 4 items and #5 (Activities of Daily Living) with 4 items. Four hundred and nine healthy subjects, aged 30 to 79, completed the questionnaire and their resulting scores in the five domains were not significantly different, except for those in the 40–49y group who had better total fatigue scores than all the other groups (as shown in Table 3) and also better scores in all other domains than the other groups. The 70–79y group had the best ADLs scores. The Norfolk QOL-F tool has been shown here to quantify cognitive fatigue, the impact of depression, problems due to fatigue, and activities of daily living in 30–79 year olds.

The literature reveals that there have been several attempts to relate fatigue to the quantified perception of an individual’s response to normal activities. The Dutch Exertion Fatigue Scale (DEFS) has measured exertion fatigue by specifically describing fatigue inducing situations, such as walking, shopping or visiting,⁵ while the Dutch Fatigue Scale (DUFS) measures patients’ general fatigue. Both of these instruments targeted patients with heart disease, post-partum women, and patients living in homes for the elderly.¹⁵ However, they do not examine the impact of increasing duration of time nor do the tools contain items that address the impact of cognitive activity on fatigue. The Chalder Fatigue Questionnaire (CFQ)¹⁶ is used to measure physical and mental fatigue in patients with chronic fatigue syndrome (CFS). More recently Yang and Wu³ constructed a 13-point subjective rating scale referred to as the Situational Fatigue Scale (SFS) to determine fatigue, with questions specifically addressing physical and mental fatigue. Principal components analysis revealed two underlying constructs of physical and cognitive activity with good internal consistency for the total scale as well as these two domains. They examined the relationship with the Fatigue Assessment Instrument (FAI) and found weak correlations, mainly due to the fact that the FAI targets were pathological fatigue whereas the SFS was designed to examine normative fatigue. None of the above-mentioned tools are designed or are appropriate for examining normative fatigue of aging and the conditions to which the target population is exposed which the Norfolk QOL-F fulfills. The strong inter-factor correlations, ranging from .690 to .830, suggest that experience of fatigue in one domain is closely related to experience of fatigue in other domains and in the overall scale except for ADLs. In a later stress paradigm study, we assessed the short-term effects of performing fatiguing walking activity (on a treadmill) on falls risk, balance and general physiological function on seventy five healthy adults randomly drawn from this larger cohort of 409 subjects. Significant age-related differences were observed before the walking activity. Increasing age was associated with declines in gait speed, lower limb strength, slower reaction times, and increases in overall falls risk. Following the treadmill task, older adults (60 to 79 year olds) exhibited increased sway, worst postural coordination and declines in lower limb strength. However, significant decrease in reaction times and increase in overall falls risk were only seen in the oldest group (70 to 79 years old). For all other persons (30 – 69 years), changes resulting from the treadmill-walking task did not lead to a significant increase in falls risk.¹⁷ Paradoxically, the oldest group had significantly better ADLs than all the younger subjects. It seems that aging in a younger group (40–49y) is compatible with a reduction in fatigue until one reaches the age of >70y when cognitive fatigue decreases dramatically, yet aspects of physical function deteriorate separating the cognitive from physical components of fatigue.¹⁷ Thus, it is mandatory that cognitive as well as physical measures are included in the evaluation of people as they age and the two may not be parallel.

Because these quantitative instrument development procedures relied on a single sample that is modest in size (n=409) for work of this nature, in order to fully assess the structural validity of the five selected factors, a replication using a large independent sample will be needed. Additionally, the approach used to identify the five selected factors used primarily empirical methods that started with the item pools from the Norfolk fatigue measures and PROMIS. Although this approach can produce meaningful and replicable factors, it may also neglect other theoretically-driven areas of fatigue, such as fatigue related to sleep problems, medical conditions, psychopathology, and others.

Conclusions

We have developed the QOL-F tool that is sensitive to aging, gender, cognitive and emotional function, and can be used to evaluate the impact of health-related problems on fatigue. (The completed tool is included as Appendix 1). Our empirical approach produced meaningful and replicable factors, but we concede that it may have (purposefully) neglected other areas of fatigue related to sleep problems, underlying medical conditions, and psychopathology in this “non-diseased” population. Our preliminary observations indicate that cognitive function plays an important role in fatigue. In fact, we suggest that fatigue is a cognitive rather than a physical phenomenon or possibly cognitive fatigue influences the experience of physical fatigue. QOL-F also displays sensitivity to the effects of aging in all five groups and all domains and is capable of distinguishing a subset of 40–49 year olds with better quality of life and less fatigue and depression than even their younger counterparts (30 to39 years). These unique findings warrant further exploration.

Abbreviations

QOL

Quality of Life

EFA

Exploratory Factor Analysis

CFA

Confirmatory Factor Analysis

PROMIS

Patient –Reported Outcomes Measure Information System

RMSEA

Mean Square Error of Approximation (Ranges from 0–1 Smaller numbers denote better fit)

ADL

Activities of Daily Living