The Cognitive Function at Work Questionnaire in memory clinic setting: a validation study

ABSTRACT

Introduction

As there is a trend toward more people seeking medical help due to cognitive symptoms, validated and targeted questionnaires are increasingly important in the clinical evaluation process. The Cognitive Function at Work Questionnaire (CFWQ) was developed to identify and rate subjective cognitive symptoms of individuals active in working life. However, its psychometric characteristics have not been previously studied in a memory clinic setting.

Method

The factorial structure, internal consistency, test-retest reliability, and convergent validity of the CFWQ were studied in a memory clinic setting (N = 113). We also investigated the instrument’s ability to identify cognitive symptoms in a cohort of early-onset dementia (EOD, N = 22), mild cognitive impairment-neurological (MCI-n, N = 18), MCI due to mood, sleep, or other physical health problems (MCI-o, N = 59), and subjective cognitive decline (SCD, N = 14) patients.

Results

Based on factor analysis, eight cognitive subscales were identified covering main cognitive domains: Memory, Language, Executive Function, Speed of Processing, Cognitive Control, Name Memory, Visuospatial/Praxis and Attention. The internal consistency (α = .93) and the test-retest reliability (ICC = .91) were high. Several correlations (r = .19 – .33, p < .05) were documented between neuropsychological impairment level and CFWQ scores. EOD, MCI-n, MCI-o, and SCD groups did not differ statistically significantly in the levels of cognitive symptoms as measured by the CFWQ Total score. EOD group scored higher (p = .009) than other patient groups on the Visuospatial/Praxis subscale, but the difference between EOD and MCI-o groups turned insignificant after correcting for multiple testing.

Conclusions

The results of the study support the validity and reliability characteristics of the CFWQ in a memory clinic setting. The instrument is easy-to-use and has clinical utility in capturing the subjective cognitive symptoms of patients active in working life and who need a referral to a more detailed evaluation.

KEYWORDS:

• Subjective cognitive decline
• early-onset dementia
• working age
• cognitive questionnaires
• memory
• self-report measures

Introduction

The health-care system is facing a two-fold challenge as the World Health Organization estimates the world-wide prevalence of dementia to rise from 55 million to 78 million within the next ten years. At the same time, increasing number of people are seeking medical evaluation because of being concerned their cognitive symptoms are caused by a neurodegenerative or other dementing disease (Jessen et al. 2020). Subjective cognitive symptoms can be the very early clinical manifestation of a neurodegenerative disorder, even before impairment in objective cognitive tests becomes detectable (Jessen, 2014). However, cognitive symptoms can also be benign or result from a great number of other conditions, such as mental health issues (Comijs et al., 2002; Hill et al., 2016), sleep disturbances (Eskildsen et al., 2017; Vaessen, Overeem, & Sitskoorn, 2015), substance abuse (Hunt et al., 2015) or physical health problems and chronic diseases (Lee, 2014; Nguyen, Killcross, & Jenkins, 2014; Valkova et al., 2019). Although neurodegenerative disorders are quite rare among the working aged people, their proportion is growing because of a trend toward extended careers. Early and accurate diagnosis is important as it enables targeted treatments and interventions, and in the case of neurodegenerative disorders, adaptation to the changes accompanied by the disease.

In the clinical evaluation process, the first step is to find out the level and type of symptoms the individual is suffering from (Rossor et al., 2010). A key objective in the assessment of subjective cognitive symptoms is to identify clinically significant symptoms that warrant more specific diagnostic evaluation. Self-reports can provide a salient source of information regarding cognitive symptoms, provided that they tap into relevant topics and are suitable for the target population and outcomes (Rabin et al., 2015). Hence, no gold standard method exists for the assessment of subjective symptoms (Molinuevo et al., 2017). When choosing a questionnaire for a working aged patient contacting health care due to cognitive symptoms, some characteristics should be taken into consideration. Firstly, the cognitive demands of everyday life are typically different for elderly than for younger patients, and for patients active in working life or outside of it. Secondly, cognitive abilities decline not only due to pathological changes but also as a part of normal aging (Salthouse, 2019), and perhaps on that account, younger dementia patients have been found to be more independent for activities of daily living as compared to elderly patients (Ryan et al., 2021). Hence, to capture the early symptoms, a questionnaire should tap into everyday situations that are cognitively demanding and relevant for the individual being assessed. Moreover, recognizing the first symptoms of dementia in working age is more challenging since early-onset dementia (EOD, i.e., symptom onset at 65 years or younger) patients have more often nonmemory cognitive presentations than late-onset dementia (LOD) patients (Koedam et al., 2010; Smits et al., 2012). Therefore, the questionnaire should also cover all main cognitive domains.

Numerous instruments have been validated for various target populations and purposes, e.g., for older adults the Everyday Cognition to assess symptoms that may point at progression to MCI or dementia (Farias et al., 2008, 2021); the Cognitive Symptom Checklist – Work for occupied cancer survivors (Ehrenstein et al., 2023; Ottati & Feuerstein, 2013); the Cognitive Change Checklist for the evaluation of early symptoms of neurodegenerative disorders from the informant viewpoint (Schinka et al., 2009), and the Workplace Cognitive Failure Scale (WCFS; Wallace & Chen, 2005) for assessing the frequency of cognitive errors at work. Of these, the WCFS may be best suited for occupied individuals as it is a specific workplace cognitive measure and it has been shown to relate to objective job performance (Wallace & Chen, 2005), but it covers only memory, attention, and behavior/action domains and it is designed for assessing occupational safety. Hence, most previous questionnaires are not specifically fitted for the health-care needs in assessing cognitive complaints of unclear etiology in individuals who are still in the workforce.

To this end, the Cognitive Function at Work Questionnaire (CFWQ) was developed to identify and rate subjective cognitive symptoms of individuals active in working life (Heikkinen et al., 2021). Instead of tapping into global functioning or narrowly defined specific tasks, the patient is asked to appraise how they manage different cognitively demanding functions in the working environment. In the original paper of Heikkinen et al., the reliability and validity characteristics were studied in a nonclinical sample of 418 participants who were on average 50 years old and employed in various occupations in a large public media service company (e.g., research, maintenance, and management). The internal consistency of the scale (α = .87) and the test-retest reliability (intraclass correlation coefficient, ICC = .84) were high. Based on factor analysis, six cognitive subscales covering 22 of the 29 CFWQ items were identified: Memory, Language, Executive Function, Speed of Processing, Cognitive Control, and Name Memory. Measurement validity was supported by a strong correlation (r = .64) between the CFWQ and the WCFS scores and a moderate correlation (r= −.40) between subjective estimate of work ability and the CFWQ score. Further, CFWQ scores were found to be associated with the severity and accumulation of mood, stress, sleep, and physical health symptoms.

The CFWQ has not been so far validated for employees with different levels of cognitive impairment and diagnosed conditions. Therefore, we set to study the properties of the CFWQ in a memory clinic setting, namely the factorial stability, internal consistency, test-retest reliability, and convergent validity. Other important research questions were to find out how the questionnaire suits for identifying cognitive symptoms of memory clinic patients who are active in working life and whether the CFWQ scores differ across memory clinic patient groups with different etiologies and severities of objective cognitive impairment.

Materials and methods

Participants of the cognitive impairment and work ability study

Patients (N = 210) were recruited 3/2019 – 3/2021 at their first visit to a specialized memory clinic at Oulu and Kuopio University hospitals in Finland. The inclusion criteria for the study were a referral to a specialized memory clinic due to cognitive symptoms and/or suspicion of a neurodegenerative disease, and the onset of symptoms at age 65 years or below. The inclusion in the current study further required that the subject had been working during the last 12 months and that the subject had filled the CFWQ without any missing data. A total of 113 patients fulfilled these criteria and requirements. Participants were all native speakers of Finnish, and they were employed in a wide variety of occupations that fall into the following three broad categories: professional/managerial-level occupations (25%), skilled workers (35%), and service, maintenance, or assistant-level occupations (40%). Only one participant did not specify his/her occupation. Other demographic data of the sample is provided in the Results section (Table 1).

Table 1. Demographics for memory clinic cohort and diagnostic patient groups.

	Total cohort (N = 113)	Diagnostic outcome
		EOD (N = 22)	MCI-n (N = 18)	MCI-o (N = 59)	SCD (N = 14)	p-Value
Age (M, SD)	56.3 (5.9)	57.5 (5.5)	60.1 (4.5)^1,2	54.9 (6.2)	55.5 (5.1)	.004
Gender (females, %)	54	50	38.9	57.6	64.3	.439
Education (%):						.361
<9 yrs	13.3	4.5	11.1	18.6	7.1
9–12rs	47.8	54.5	66.7	40.7	42.9
>12 yrs	38.9	40.9	22.2	40.7	50.0
Working (%)	65.5	40.9^1	61.1	72.9	78.6	.036

Abbreviations: EOD = Early-Onset Dementia; MCI-n = Mild Cognitive Impairment – neurological; MCI-o = Mild Cognitive Impairment – other; SCD = Subjective Cognitive Decline. ¹A significant (p < .05) pairwise group comparison to MCI-o. ²A significant (p < .05) pairwise group comparison to SCD.

Patients were diagnosed by a memory clinic neurologist per current diagnostic criteria (Gorno-Tempini et al., 2011; Mckeith et al., 2017; McKhann et al., 2011; O’Brien et al., 2003; Rascovsky et al., 2011; Winblad et al., 2004) based on a comprehensive diagnostic workup including medical history taking, neuropsychological assessment, magnetic resonance imaging (MRI), laboratory tests, and neurological examination. A cerebrospinal fluid analysis of Aβ₄₂, tau, and phosphotau was done for 34 patients. When necessary, patients were followed up for up to two years before receiving the final diagnosis. All patients received oral and written information on the Cognitive Impairment and Work Ability (CIWA) study and provided a written informed consent in accordance with the Declaration of Helsinki. The ethics committees of the Northern Ostrobothnia Hospital district and Northern Savo Hospital district approved the study (96/2018).

Four groups were formed based on the diagnostic outcome to compare CFWQ results. The early onset dementia (EOD) group (N = 22) consisted of Alzheimer’s disease (AD, N = 13), frontotemporal dementia (FTD, N = 4), vascular dementia (N = 2), and Parkinson’s disease (N = 1) patients. Two cases were diagnosed to have dementia due to a neurodegenerative disorder; however, the specific subtype of dementia could not yet be specified for them by the time of the last visit of this study. Individuals whose diagnosis was at the last follow-up visit either mild cognitive impairment (MCI, N = 9), vascular cognitive impairment (N = 7), or a suspicion of a neurodegenerative disease (N = 2) formed the MCI-neurological group (MCI-n, N = 18). Patients who exhibited cognitive impairment related to mental health conditions (N = 31), sleep disturbances (N = 16), or other health condition (N = 12) were classified as MCI-other (MCI-o) group (N = 59). In accordance with the research guidelines and recent findings related to subjective cognitive decline (SCD; Jessen et al., 2014; Molinuevo et al., 2017; Wolfsgruber et al., 2020), participants in the SCD group (N = 14) had subjective cognitive symptoms and a concern associated with the symptom, but they did not have impairment in any of the cognitive domains evaluated in neuropsychological assessment.

An occupational cohort of 418 volunteers who were on average 49.7 years old was used for two purposes in the current study. The cohort characteristics are described in more detail in Heikkinen et al. (2021). First, to calculate the number of affected cognitive domains (i.e., the number of CFWQ subscales where patient scores at least 1 SD higher than what participants from the occupational data with no depressive, anxiety, or insomnia symptoms score). This subgroup of participants (N = 117) was named “no symptoms” group in the paper mentioned above. The second purpose was to contrast the level of cognitive symptoms (i.e., CFWQ scores) of memory clinic patient cohort with population-level cohort.

Measurements

In addition to clinical neurological and neuropsychological evaluations, the CIWA study protocol included a set of questionnaires and an additional cognitive test battery at baseline, 6 months, and 12 months. The current paper focuses on the questionnaire and clinical neuropsychological data gathered at baseline visit. The 6 months follow-up data was used solely for the test-rest reliability analysis. The following information was analyzed from the questionnaires: onset of symptoms (insidious/sudden), examination seeking (whether the patient sought him-/herself or was encouraged by an informant/workplace/other to seek medical help), confirmation of cognitive symptoms by an observer, and concern associated with cognitive symptoms. The CFWQ was used to evaluate cognitive symptoms (Heikkinen et al., 2021). The examinee rates 29 items regarding cognitive difficulties experienced in the working environment on a 3-point scale (0 = works well, 1 = sometimes difficult, 2 = often difficult). The Total score, “often difficult” score, number of affected domains (i.e., number of CFWQ subscales) and CFWQ subscale scores were used in the analyses.

All participants underwent a comprehensive neuropsychological assessment as a part of the diagnostic workup. The following neuropsychological domains were used in the current paper (in brackets the primary tests used to assess the domain): verbal memory (Wechsler Memory Scale III Logical Memory and Word List (Wechsler, 2008)), visual memory (Wechsler Memory Scale III Visual reproduction, Complex Figure Memory Test (Lezak et al., 2012)), working memory (Wechsler Adult Intelligence Scale IV Digit Span and Letter-Number Sequencing (Wechsler, 2012)), attention (247 cancellation test (Lezak et al., 2012)), executive functions (Trail Making Test Part B (Reitan, 1958), Stroop (Golden, 1978), alternating s (Allison, 1966)), processing speed (Wechsler Adult Intelligence Scale IV Digit symbol coding and Symbol search subtests, Trail Making Test Part A), language skills (semantic and phonemic verbal fluency tests (Lezak et al., 2012), 15-item version of the Boston Naming (Morris et al., 1989)), visuospatial skills (flag, cube and Creek cross copy, Clock hands, and ROCFT copy (Lezak et al., 2012; Morris et al., 1989)), and behavioral/affective symptoms (Modified Frontal Behavioral Inventory (Heidler-Gary et al., 2007), Patient Health Questionnaire (Kroenke et al., 2001)). The neuropsychological testing was conducted primarily for clinical purposes and therefore some tests were not available for all participants.

We determined the level of cognitive impairment from the neuropsychological assessment report. Each cognitive domain was scored on a scale from 0 to 4 (0 = normal performance,1 = subtle impairment, 2 = mild impairment, 3 = moderate impairment, 4 = severe impairment) based on clinical judgment (i.e., neuropsychological assessment report), where the cognitive test results are interpreted considering normative data, qualitative performance features, and estimated premorbid intelligence level. Further, the performance on a single cognitive test is not evaluated strictly under one cognitive category. For example, poor performance on verbal fluency test can be due to several underlying cognitive deficiencies, including problems of language production or difficulties with strategy formation, i.e., executive functions. A general cognitive impairment level was calculated as the average of the impairment in all cognitive domains.

Data-analysis

To study the factor structure of the CFWQ in a memory clinic patient cohort, the responses of the complete sample were subjected to a factor analysis. First, the Pearson correlation of the 29 CFWQ questions was calculated and the suitability for factor analysis was evaluated with Kaiser-Meyer-Olkin measure of sampling adequacy. Communalities were calculated, and a threshold value of .30 was set as an inclusion criterion for further analysis. Principal Component Analysis (PCA) was performed using the varimax rotation. Internal consistency of the subscales and the CFWQ Total score was studied with Cronbach’s alpha coefficient. Test-retest reliability was analyzed with ICC.

Demographical and clinical characteristics as well as the CFWQ scores were studied in relation to diagnostic groups. For continuous variables, means and standard deviations were calculated. As Kolmogorov-Smirnov’s test indicated that the variables were not normally distributed, a non-parametric Kruskal-Wallis test was used to compare group differences. Spearman correlation coefficient was used for evaluating the associations between CFWQ scores and other variables. Statistically significant correlation coefficients were interpreted as follows: small (r = .10 – .29), moderate (r = .30 – .49), strong (r ≥ .50; Evans, 1996). Categorical variables are presented in percentages, and Pearson Chi-square test or Fisher’s exact test was used to evaluate group differences. For comparing the patient data results to previously published occupational population data, summary independent samples T-test with 95% confidence interval was used, and effect size (Cohen’s d) is reported.

For multiple testing, the significance values were adjusted by the Bonferroni correction. A p-value < .05 was considered statistically significant. All statistical analyses were performed using the IBM SPSS version 27.0 software.

Results

Detailed information of the demographics of the cohort is summarized in Table 1. The EOD, MCI-n, MCI-o, and SCD groups did not differ in gender (p= .439), or education level (p= .361). MCI-n patients were older than MCI-o and SCD patients (p= .004), but no other significant differences in age were found, and age did not correlate significantly with the CFWQ scores.

Factor analysis

The results of the factor analysis of the CFWQ are displayed in Table 2. Kaiser-Meyer-Olkin measure of sampling adequacy (.853) indicated that the correlation matrix is suitable for factor analysis. Communalities ranged from .52 to .76, and therefore, all the items were included in the factor model. The model produced by the PCA accounted for 64.5% of the total variance. Eight cognitive subscales were identified: Memory (Mem), Language (Lan), Executive Function (Exe), Speed of Processing (Spe), Cognitive Control (Con), Name Memory (Nam), Visuospatial/Praxis (Vis), and Attention (Att). The questions 1, 9, 12, and 19 were placed on a subscale of a lower factor loading based on the suitability of the content to the CFWQ subscale and to keep the factor structure as in the previous model (Heikkinen et al., 2021). The PCA calculated the Exe and Spe as a singular factor; however, they are kept separate to follow the original factor structure. Questions (2, 9–13, and 19) that were not included in the original factor model formed two new subscales (Vis and Att).

Table 2. The CFWQ items and the factor loadings for the cognitive domains.

CFWQ item	CFWQ subscale
	Mem	Lan	Exe^1	Spe^1	Con	Nam	Vis	Att
1. Tasks agreed with other people?	.45*							.66
4. Things you have read?	.66	.34
5. Things you have heard?	.81
6. Precise details of things?	.70
16. Fluent recall of things?	.59		.38
21. Speech comprehension?	.33	.59
22. Speech production?		.72
23. Reading?		.61				.42
24. Writing?		.73
3. Your own work plans?			.58			.35
14. Finalizing tasks?			.76
17. Forethoughts needed in tasks?	.31		.56
18. Starting/engaging in tasks?			.72
20. Making decisions and judgments at work?			.58
26. Initiative?			.49		.47
15. Completing/performing tasks fast?				.69			.33
25. Speed of actions?		.33		.65
27. Coherence of thinking?			.38		.45		.31
28. judgment in financial or problematic situations?					.77
29. Control of emotions and behavior?		.31			.68
7. Names of new people/acquaintances?	.34					.72
8. Names of familiar people?						.60		.30
9. Location of tools and objects?		.38					.37*
10. Familiar work methods or use of familiar tools?		.32					.67
11. New work methods and use of new tools?	.40						.65
2. Appointments agreed with other people?	.31							.76
12. Continuing a task after an interruption?	.49		.53					.33*
13. Monotonous repetitive tasks?								.74
19. Checking the correctness of work performance?			.67					.32*

Abbreviations: CFWQ = Cognitive Function at Work Questionnaire; Mem = Memory; Lan = Language; Exe = Executive Function; Spe = Speed of Processing; Con = Cognitive Control; Nam = Name Memory; Vis = Visuospatial/praxis; Att = Attention

Note: N = 113. Factor analysis was performed with Principal Component Analysis. Rotated factor matrix displayed (Varimax with Kaiser Normalization). Factor loadings exceeding .30 are presented. Bolding indicates the CFWQ subscale where the question is located. Questions in italics (2, 9–13, 19) were not included in the previous factor model (Heikkinen et al., 2021) due to low communality. *Items placed on a subscale of a lower factor loading based on the content and to keep the original factor structure. ¹The PCA calculated the Exe and Spe as one factor, however, they are kept separate to keep the original factor structure.

The reliability coefficients indicated high internal consistency of the CFWQ Total score (α = .93) and mainly good/acceptable for the subscales (Mem = .83, Lan = .73, Exe = .84, Spe = .75, Con = .66, Nam = .62, Vis = .66, and Att = .67). The test-retest analysis indicated high agreement between the baseline and 6 months follow-up measurements (ICC = .91). The eight CFWQ subscales were used in the following analyses.

The occupational data mean scores and standard deviations for the subgroup of participants without depressive, anxiety, or insomnia symptoms (the “no symptoms” -group, N = 177) were used to calculate the number of affected domains in the patient data. The >1 SD was used as a cutoff level on each CFWQ subscale to represent a higher-than-normal level of symptoms, while not excluding mild symptoms. The means and standard deviations for the “no symptoms” group were for Mem (M = 2.2, SD = 1.7), Lan (M = 0.3, SD = 0.6), Exe (M = 1.1, SD = 1.3), Spe (M = 0.4, SD = 0.8), Con (M = 0.3, SD = 0.5), Nam (M = 1.5, SD = 1.0), Vis (M = 0.6, SD = 0.7), and Att (M = 1.0, SD = 1.1). Therefore, when the memory clinic study participant scored on Mem ≥4, Lan ≥1, Exe ≥3, Spe ≥2, Con ≥1, Nam ≥3, Vis ≥2, or Att ≥3, the terms of showing symptoms on the cognitive domain were met.

Symptom onset and examination seeking

Clinical characteristics for the total study population and for diagnostic groups are summarized in Table 3. The EOD and MCI-n patients were prone to be encouraged by others to seek help instead of themselves seeking medical consultation. However, the differences between groups were not statistically significant. No differences were found in the concern associated with the cognitive symptoms (p = .458), or the symptom onset (p= .349). The EOD patients had the highest rates of an observer confirming the cognitive symptoms, but the difference to other groups was not significant (p = .443). The EOD patients had significantly (p < .05) higher cognitive impairment level, as compared to MCI-o and SCD groups. SCD patients were the least cognitively impaired, and the difference to all other patient groups was significant (p < .001).

Table 3. Clinical characteristics of memory clinic cohort and diagnostic patient groups.

	Total cohort (N = 113)	Diagnostic outcome
		EOD (N = 22)	MCI-n (N = 18)	MCI-o (N = 59)	SCD (N = 14)	p-Value
Examination seeking (self-referred, %)	54.3	42.9	41.2	57.4	76.9	.159
Confirmation of cognitive symptoms by observer (%)	82.0	90.9	77.8	82.5	71.4	.443
Concern associated with cognitive symptoms (%)	90.7	95.0	82.4	89.7	100.0	.458
Insidious onset of cognitive symptoms (%)	82.5	76.5	78.6	81.5	100.0	.349
General cognitive impairment level (M, SD)	1.19 (.75)	1.78 (.89)^1	1.33 (.60)	1.17 (.57)	.22 (.13)^2	<.001

Abbreviations: EOD = Early-Onset Dementia; MCI-n = Mild Cognitive Impairment – neurological; MCI-o = Mild Cognitive Impairment – other; SCD = Subjective Cognitive Decline. ¹A significant (p < .05) pairwise group comparison to MCI-o and to SCD. ²All pairwise group comparisons significant (p < .001).

Group differences in self-estimated cognitive symptoms

The CFWQ Total score, the subscales, the number of domains, and the score of “often difficult”-answers were calculated for the complete sample and for the diagnostic groups (Table 4). While there was a trend toward the EOD and the MCI-o groups reporting higher levels of cognitive difficulties than the SCD and MCI-n groups in many of the CFWQ variables, only statistically significant result was observed on the Vis scale (p = .009), where the EOD patients scored significantly higher than any of the other groups. However, with the Bonferroni correction, the difference between EOD and MCI-o groups turned out insignificant (p = .192). Combining the MCI-n and MCI-o groups did not change any of the results, as the only statistically significant (p = .009) difference between EOD, MCI, and SCD groups’ scores was observed on the Vis scale.

Table 4. CFWQ scores (M, SD) for memory clinic cohort and diagnostic patient groups.

CFWQ character	Total cohort (N = 113)	Diagnostic outcome
		EOD (N = 22)	MCI-n (N= 18)	MCI-o (N = 59)	SCD (N = 14)
29-item Total score	18.9 (10.4)	20.7 (9.8)	15.1 (10.0)	19.6 (11.2)	17.7 (7.5)
“Often Difficult” -score	3.6 (4.3)	3.7 (4.2)	1.9 (2.8)	4.2 (4.8)	3.3 (3.2)
No. of domains	4.3 (2.4)	5.0 (2.3)	3.8 (2.9)	4.2 (2.4)	4.1 (2.3)
CFWQ subscales
Mem	4.9 (2.6)	5.0 (2.2)	3.5 (2.6)	5.2 (2.8)	5.1 (2.4)
Lan	1.5 (1.7)	1.2 (1.4)	1.3 (1.5)	1.8 (1.8)	0.7 (1.1)
Exe	3.7 (2.9)	4.2 (2.8)	2.9 (2.6)	3.8 (3.2)	3.8 (2.5)
Spe	1.4 (1.2)	1.7 (1.0)	1.2 (1.2)	1.5 (1.3)	1.1 (1.0)
Con	1.3 (1.4)	1.5 (1.3)	1.3 (1.3)	1.2 (1.5)	1.2 (1.3)
Nam	2.2 (1.0)	1.9 (1.0)	2.2 (.8)	2.3 (1.0)	2.4 (1.0)
Vis	1.7 (1.4)	2.3 (1.2)	1.2 (1.0)	1.8 (1.5)	1.1 (.9)
Att	2.2 (1.8)	3.0 (1.9)	1.6 (1.5)	2.1 (1.9)	2.4 (1.4)

Abbreviations: CFWQ = Cognitive Function at Work Questionnaire; EOD = Early-Onset Dementia; MCI-n = Mild Cognitive Impairment – neurological; MCI-o = Mild Cognitive Impairment – other; SCD = Subjective Cognitive Decline¸ Mem = Memory; Lan = Language; Exe = Executive Function; Spe = Speed of Processing; Con = Cognitive Control; Nam = Name Memory; Vis = Visuospatial/praxis; Att = Attention.

When comparing the CFWQ Total score of the total patient cohort (M = 18.9, SD = 10.4) with the previously reported population-based nonclinical cohort of 418 participants (M = 10.5, SD = 6.8; Heikkinen et al., 2021), the memory clinic patients reported significantly more cognitive symptoms (t(138.9) = −8.13, p < .001, Cohen’s d= .96).

Association of neuropsychological assessment and self-estimated cognitive symptoms

In the complete clinical sample, there were mainly small but significant correlations between the neuropsychological impairment score and the amount of self-experienced cognitive difficulties. In all significant correlations, higher neuropsychological impairment score was associated with higher number of subjective cognitive difficulties (Table 5). Especially, the Vis subscale and the number of CFWQ domains showed several significant correlations to neuropsychological impairment level. There were also many significant associations between corresponding subjective and objective cognitive measures, e.g., the language domain on neuropsychological assessment and Lan subscale of the CFWQ (r = .26), behavioral/affective domain and the Con subscale (r = .27), and processing speed and Spe subscale (r = .23).

Table 5. Spearman correlation coefficients between the CFWQ and clinical cognitive impairment level for total patient cohort.

Neuropsychological domain	CFWQ measure
	Total score	No. of domains	Mem	Lan	Exe	Spe	Con	Nam	Vis	Att
General	.19*	.24*	.07	.20*	.14	.24**	.24**	−.06	.33**	.12
Verbal memory	.21**	.26**	.26**	.16	.07	.14	.13	.09	.31**	.23*
Visual memory	.16	.18	.10	.03	.14	.12	.17	−.05	.31**	.04
Working memory	.26**	.26**	.20*	.25**	.20*	.29**	.20*	−.01	.21*	.25**
Attention	.20*	.20*	.08	.21*	.16	.14	.16	.13	.18	.22*
Executive functions	.03	.09	−.11	.10	.04	.15	.12	−.08	.19*	−.04
Processing Speed	.08	.10	−.01	.05	.03	.23*	.03	−.03	.16	.12
Language skills	.20*	.22*	.06	.26**	.19*	.24*	.25**	.00	.18	.14
Visuospatial skills	.04	.06	−.06	.05	.07	.13	.15	−.18	.15	−.10
Behav/affective	.01	.02	−.10	.01	.06	.05	.27**	−.18	.11	.00

Abbreviations: CFWQ = Cognitive Function at Work Questionnaire; Mem = Memory; Lan = Language; Exe = Executive Function; Spe = Speed of Processing; Con = Cognitive Control; Nam = Name Memory; Vis = Visuospatial/Praxis; Att = Attention. * p < .05; **p < .01.

Discussion

The present study evaluated the validity and reliability characteristics of the Cognitive Function at Work Questionnaire (CFWQ) in a memory clinic setting. The factorial structure and the utility in evaluating the cognitive symptoms of individuals active in working life were assessed. The latter was done by analyzing the CFWQ scores in relation to memory clinic diagnostic outcome and the degree of cognitive impairment as defined by a comprehensive neuropsychological assessment. Regarding reliability analyses, the internal consistency of all the scales was good/acceptable, and the test-retest reliability was high.

Results of the current study are mainly in line with our previous results (Heikkinen et al., 2021) presenting the initial factor structure of the CFWQ. Despite minor differences, the factorial structure remained constant among memory clinic patient and occupational population samples. The first six subscales were identical in both samples including Mem, Lan, Exe, Spe, Con, and Nam subscales. However, the CFWQ scores in the occupational sample were expectedly lower than in the clinical sample, and possibly therefore the Vis and Att subscales did not emerge as factors in the occupational sample. Another difference between the samples was that in the patient data, the Exe and Spe loaded on one factor. The items in the Exe and Spe inarguably relate to executive functions in a wide sense, and therefore it is apprehensible that the constructs may partly overlap. However, we kept the subscales separated holding the clinical aspects in mind, as problems in planning, initiating, and finalizing tasks (i.e., Exe items) and slowed processing speed can reflect different clinical presentations and neuropathologies (e.g., Rascovsky et al., 2011; Wallin et al., 2018). Additionally, the separate Exe and Con subscales deserve further discussion, as executive functions and cognitive control are related concepts and sometimes even used as synonyms (Friedman & Robbins, 2022). Indeed, executive function is a multifaceted term and although several definitions exist, generally it is thought to consist of higher order thinking abilities like set-shifting, working memory, and inhibitory control (Diamond, 2013; Miyake et al., 2000). Executive functions can also be conceptualized by their demand on emotional processing, i.e., division to hot and cold functions (Salehinejad et al., 2021). The division does not denote that emotional processing is not present in cold functions or that cognitive processing has no role in hot functions. Instead, hot functions are involved in tasks with higher demands on self-regulation or reward appraisal, whereas cold functions associate to tasks that rely more heavily on cognitive processes. Thus, the CFWQ Exe and Con items loading on separate subscales in the present and the former (Heikkinen et al., 2021) factor analysis may be due to different demands on self-regulatory, motivational, or emotional processing of the Con items as compared to the perhaps more purely cognitive items of the Exe subscale. However, the items “Initiative?” and “Coherence of thinking?” loaded fairly similarly onto each subscale. Among other possible reasons, this may imply that the items are appraised as placing rather equivalent demands on cognitive and emotional processing, or that there was fluctuation on how much value the participants gave to emotional/motivational processing.

Previously, cognitive symptoms have been associated with lower objective cognitive performance (Cedres et al., 2019; Wolfsgruber et al., 2020), but also conflicting results have been reported (Gass et al., 2021; Mendes et al., 2008). In our study, the CFWQ scores and neuropsychological assessment variables showed many small to moderate and logical correlations (i.e., corresponding CFWQ subscales and neuropsychological domains). According to a review and a meta-analysis by Burmester et al. (2016), the association between subjective estimates of cognitive symptoms and objective cognitive performance seems to be rather weak, but significant (r= .13). In fact, they found that only a minority of papers reported higher than .20 correlations. There may be several reasons as to why the correlations are typically rather small. One possible reason is that subjective and objective measures assess partly different aspects of cognitive functioning. Questionnaires tend to target the cognitive functioning in relation to everyday life demands in a longitudinal manner (as does the CFWQ), whereas cognitive testing is a cross-sectional evaluation of the performance in relation to age and/or education corrected norms or other expectations (Jessen et al., 2014; Wolfsgruber et al., 2014). However, Burmester et al. stated that those papers that reported significant associations typically employed more profound measures of subjective symptoms, as compared to those showing no link between subjective and objective functioning. In addition to the questionnaire properties, also the neuropsychological assessment method is crucial, as low sensitivity of the chosen method to capture minor deficits might obscure the association. All the participants in the current study underwent a thorough neuropsychological evaluation, enabling us to evaluate subtle cognitive deficits and all main areas of cognition. Therefore, we conclude that small correlations between neuropsychological assessment variables and the CFWQ scores give tentative support to the validity of the CFWQ, but more validation research is needed (e.g., studying the association between the CFWQ and other cognitive and performance measures).

One of the strengths of the CFWQ is that it covers all main cognitive domains. It has been argued that memory-domain questions, albeit being the most frequently appearing questions in cognitive questionnaires (Rabin et al., 2015), might not be the most informative to clinicians as healthy volunteers also frequently report memory difficulties (Cooper et al., 2011; Weaver Cargin et al., 2008). In fact, in a large sample of elderly women, a question about how to find a way around familiar streets was associated with cognitive impairment whereas a question of whether the person has trouble remembering things from one second to the next had no such association (Amariglio et al., 2011). The former question taps into visuospatial (memory) functioning and the latter is a more general memory-related question. In our study, the memory subscale scores did not differentiate the patient groups. Instead, the Vis scale seems particularly informative in relation to EOD and cognitive impairment, as EOD patients scored significantly higher on the CFWQ Vis scale than other patient groups (apart from MCI-o after the Bonferroni correction). Additionally, the highest correlations were witnessed between cognitive domains and the Vis subscale. Indeed, questions regarding other areas of cognition might capture the early signs of cognitive decline related to neurodegenerative disorders better than memory-related questions (Rabin et al., 2015). This might be particularly true for younger patients since nonamnestic phenotypes are more common in EOD, and even the amnestic EOD phenotypes present high degrees of nonmemory domain symptoms (Mendez, 2012; Phillips et al., 2020).

One of the primary goals of the questionnaire is to identify symptoms experienced by individuals with objective cognitive impairment and neurodegenerative disorders. We found a clear difference in the mean CFWQ Total scores of nonclinical (M = 10.5) and memory clinic (M = 18.9) samples. At the same time, the EOD group scored as high on the CFWQ Total score as other memory clinic patient groups. The results are interesting, considering that the ability to realistically appraise the level of cognitive functioning (i.e., metacognitive abilities) typically decreases in neurodegenerative disorders (particularly in AD) as the cognitive decline progresses (Jessen, 2014; Sunderaraman & Cosentino, 2017; Wolfsgruber et al., 2014). However, the cognitive impairment severity in the EOD group was on average only mild (M = 1.78), although statistically significantly higher as compared to MCI-o and SCD groups. Previously, EOD patients have been found to demonstrate higher symptom awareness than LOD patients, despite having similar cognitive impairment rate (Baptista et al., 2019; Van Vliet et al., 2013). Baptista et al. (2019) proposed that it could be at least in part due to the challenges EOD patients face with having cognitive decline while coping with higher demands from environment and being the financial providers of the family. Thus, cognitive difficulties experienced in the working environment may be particularly hard to neglect. In a similar vein, we found a high percentage (95%) of EOD patients reporting concern related to the cognitive symptoms. As the concern was asked in a dichotomous manner (yes/no), we could not evaluate whether there were differences in the severity of the concerns between groups. Nevertheless, our results suggest that those EOD patients that are or have recently been working may acknowledge their symptoms rather well, at least when asked for cognitive difficulties experienced in the working environment.

One of the limitations of the current study is that we did not have access to measures of actual work performance to assess the ecological validity of the CFWQ. Additionally, as we did not have other information on the race or ethnicity of the sample than that all were fluent speakers of Finnish (and therefore, most would likely identify themselves as Finns), the generalizability of the results to other groups remains to be addressed in future studies. Other limitations are related to the sample size. To obtain a representative sample, it was important to include only patients referred to memory outpatient clinic who are active in working life and whose symptoms have begun at 65 years or younger. A relatively small sample size (N = 113) may limit the use of some statistical analyses. However, communalities appear to play a crucial role in obtaining stable and congruent factor models. MacCallum et al. (1999) showed that when communalities are above .50 (as in our factor analysis), the role of sample size becomes less important and, in that case, sample sizes of 100–200 should suffice. Further, the current CFWQ factor structure was very similar to the previous factor solution produced with a larger sample (N = 418; Heikkinen et al., 2021). Regarding patient group comparisons, it would have been interesting to study different neurodegenerative disorders and other causes to cognitive impairment separately (e.g., AD, FTD, depression, and insomnia), but the sample size should have been larger. We chose to compare EOD, MCI-n, MCI-o, and SCD groups as this level of analysis is relevant for clinical practice and making decisions of which individuals need a more thorough evaluation. Even so, sample sizes were relatively small and there is a risk of type II error, i.e., that null hypothesis is falsely accepted. Further, there were significantly more participants in the EOD group that were outside the working life than in the MCI-o group at the time of assessment, and this may have affected the results. We also did not have exact information of the duration of absence (only that it was less than 12 months). Due to the abovementioned reasons, the group comparison results are tentative and should be confirmed with larger sample sizes in future studies.

In conclusion, considering the challenges the health care is facing, short and easy-to-use cognitive assessment methods will become increasingly important as a part of the initial assessment of cognitive symptoms. Our study supports the clinical utility of the CFWQ as a tool to capture the subjective cognitive symptoms of patients active in working life, including those with EOD, who need a referral to a more detailed evaluation. Before cutoff scores or other scoring recommendations can be given, more research is needed with larger study samples and especially in relation to demographic variables, as the perception of symptom severity and the interpretation of items can differ according to education, gender, and age. More research is needed in other clinical groups and regarding the ecological validity of the scale. Future studies could also determine whether changes on objective cognitive performance are accompanied by a change in self-perceived cognition.

Acknowledgments

We are grateful to all the volunteer participants who made this study possible.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Data availability statement

The data that support the findings of this study are available from the the Finnish Institute of Occupational Health (Dr. Teemu Paajanen, [email protected]), upon reasonable request.

Additional information

Funding

This work was supported by the Academy of Finland [315460]; the University of Oulu Scholarship Foundation;the Finnish Brain Foundation [20190012]; and the Finnish Work Environment Fund [180038].