Open access Original research

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
Validation of Visual Analogue Scales of
job demand and job control at the
workplace: a cross-­sectional study
Frederic Dutheil ‍ ‍,1,2 Bruno Pereira,3 Jean-­Baptiste Bouillon-­Minois ‍ ‍,4
Maëlys Clinchamps ‍ ‍,1 Georges Brousses,5,6 Samuel Dewavrin,7
Thomas Cornet,7 Martial Mermillod,8,9 Laurie Mondillon,10,11 Julien S Baker,12
Jeannot Schmidt,4 Farès Moustafa,13 Charlotte Lanhers14

To cite: Dutheil F, Pereira B, ABSTRACT

Bouillon-­Minois J-­B, et al. Strengths and limitations of this study
Objectives Work-­related stress is a major concern.
Validation of Visual Analogue One of the best performing models is the Job Content
Scales of job demand and job ► The use of Visual Analogue Scales (VASs) for job de-
Questionnaire (JCQ) of Karasek, assessing job demand
control at the workplace: a mand and job control are valid, quick, easy to use
and job control using 18 items. However, the JCQ is long
cross-­sectional study. BMJ Open and reliable tools compared with the Job Content
2022;12:e046403. doi:10.1136/ and complex. Visual Analogue Scales (VASs) are easy to
Questionnaire (JCQ) of Karasek, considered as the
bmjopen-2020-046403 use and quick to implement. VASs have been validated
historical model to assess job demand and job
to assess pain and occupational stress; however, VASs
► Prepublication history and control.
demand and control have not been evaluated. Therefore,
additional supplemental material ► The findings are based on a test–retest completed
we aimed to validate the use of VAS demand and control
for this paper are available by 129 participants. The number of participants who
compared with the 18 items of the JCQ.
online. To view these files, responded to both the test and retest was higher
please visit the journal online Design We implemented a cross-­sectional observational
than commonly reported in the literature.
(http://dx.doi.org/10.1136/​ study, by administering a self-­reported questionnaire to
► Despite the widespread use and relevance of VAS,
bmjopen-2020-046403). the users of Wittyfit software, with a second test (retest)
VAS provide global first-­line evaluations that do not
proposed 1 week later. In addition to JCQ, VAS demand and
Received 15 February 2021
accurately assess putative explanations of stress
control, we measured sociodemographic outcomes, as
Accepted 27 February 2022 factors such as those provided using the JCQ of
well as characteristics of work, sleep, well-­being, stress,
Karasek.
depression and anxiety.
► We promote the use of VAS for both demand and
Participants 190 volunteers French workers using
control domains as an easy tool to detect abnormal
WittyFit software participated in the study, and 129
situations. If abnormal situations are detected; oc-
completed the test–retest.
cupational physicians should further assess work-
Results VAS demand and VAS control correlated with
place stress with the use of the JCQ.
the two Karasek domains from the JCQ, respectively, at
0.59 and 0.57 (p<0.001). Test–retest reliability highlighted
concordance coefficients higher than 0.70. Sensitivity
was higher than 70% for each VAS. External validity was
any threat to an organism’s homoeostasis.2
acceptable. For both demand and control, VAS cut-­offs Stress is a condition that affects people daily
were 75/100. Compared with other workers, senior and is influenced by multiple environmental
executives and individuals with master’s degrees had factors, especially factors related to job such
higher levels of job control but did not differ in job demand as deleterious working conditions.3 Indeed,
using the VAS and JCQ. work-­related stress is now a major concern.4
Conclusions VAS demand and VAS control are valid, Workplace stress encompasses a range of
quick, easy to use, and reliable tools for the assessment psychological and physical perturbations that
of job demand and job control. They can be used in daily negatively affect health, relationships, quality
clinical practice for primary prevention and diagnosis. of life and well-­being.5 Stress at work is a risk
© Author(s) (or their
employer(s)) 2022. Re-­use However, when results are over 75 mm on VAS, we
factor for chronic diseases such as cardiovas-
permitted under CC BY-­NC. No promote the use of JCQ to be more discriminant and
specific to initiate action plans to help workers.
cular or mental events.6–8 However, with most
commercial re-­use. See rights
and permissions. Published by Trial registration number NCT02596737. adults spending half of their waking hours at
BMJ. work, the workplace is an important setting
For numbered affiliations see to promote health and well-­being.9 Over the
end of article. INTRODUCTION past decades, the concept of occupational
Stress refers to the consequences of the failure stress has increased in popularity and several
Correspondence to
Maëlys Clinchamps; of a living organism to respond appropriately questionnaires, scales, measurements or
​mclinchamps@​chu-​ to emotional or physical threats whether evaluation tools have been developed and
clermontferrand.f​ r actual or imagined.1 Stress can be defined as validated.

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403 1

Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
Despite further evolution and other models,10–12 one aims to improve well-­being at work with an epidemiological
of the historical concepts of occupational stress is the design, and a collaborative partnership with researchers
job demand-­control model (JDC) created and validated (public–private partnership between WittyFit and the
by Karasek et al.13 This model recognises the importance University Hospital of Clermont-­ Ferrand).30 Workers
of daily environmental stressors on the long-­term expe- using WittyFit were proposed to answer validated self-­
rience of stress.14 It is based on two main hypotheses: questionnaires on behavioural data for personalised eval-
the strain hypothesis predicts that job demands, arede- uation. Workers can answer WittyFit questionnaires using
fined as a high workload, time pressures and increased a dedicated application, available for computers or smart-
employee’s stress. The buffer hypothesis predicts that phones, at any time. The concept of WittyFit is to provide
increasing control (autonomy and skill use) can alleviate health profiling with an individualised feedback based on
the negatives effects of high demands. The active job evidence-­based medicine. It aims to support behavioural
quadrant is defined by high demands and high control, change over time, and to assess the relationships between
and the strain high stress quadrant is defined by high changes in knowledge, practices, and health outcomes.
demands with low control.13 15 The Job Content Ques- Exclusion criteria of participants were the inability to read,
tionnaire (JCQ) derived from Karasek’s model16 has been understand or answer on-­line questionnaires. Answering
developed and validated in several languages. Studies Wittyfit questionnaires implied consent. The study was
have highlighted its psychometric properties, especially conducted on workers using WittyFit software between 1
in a French population of 24 486 workers.17 18 However, June 2016 and 30 June 2016. All data were anonymous.
this tool is difficult to use in daily medical consultations The name of the employee was never revealed. The data-
because of its length (18 items) and complexity. Limits base was implemented from a human resource generated
of self-­
reported questionnaires include their low level number, automatically converted in another number into
of completion and participation,19 20 and their low level the WittyFit database.30 Data provided by employers (eg,
of representativeness21 inducing a high level of missing occupation, department, sick leave) were automatically
data.22 Participation rate decreases with the length of the associated with the human resource-­generated number.
questionnaire, in addition to a decrease in attention and
concentration.23
Patient and public involvement
Occupational physicians have limited time because
This was an exploratory study in an ecological situation.
of the number of workers and worksites.24 As stress at
The WittyFit users were informed of a forthcoming ques-
work is a major public health concern,25 occupational
tionnaire on the platform, explaining the purpose of the
physicians have to include consequences of psychosocial
study (validation study) and the need to complete the
risks in their clinical diagnosis. They need a simple and
questionnaire twice (test, and retest 1 week later). Workers
fast instrument to use daily. The Visual Analogue Scale
were not involved in the design, conduct, reporting or
(VAS) is well known and valid in the clinical assessment
dissemination plans of this research.
of pain,26 for occupational stress27 28 and satisfaction at
work.29 The VAS is a suitable tool for clinical activity and
has good psychometric characteristics.26–28 Therefore, we Primary outcome
hypothesised that VAS for job demand and job control The primary objective was to validate the two VAS
would be appropriate tools to distinguish workers at risk (demand, control) in replacement of Karasek’s ques-
of workplace-­related stress versus the job-­demand control tionnaire. Job demand and latitude decision making was
questionnaire of Karasek. evaluated by the 18 items of the Karasek’s questionnaire
The main purpose of the present work was to validate (JCQ)16 and by VAS control and VAS demand. The JCQ
VAS demand and VAS control replacing the 18 items measures nine items of job control (questions Q1–Q9),
of the two domains of Karasek’s model (demand and nine items of job demand (Q10–18). Items of JCQ were
control). We evaluated the external validity of the two scored on a four-­point Likert-­type scale, ranging from one
VAS (control and demand) by highlighting their relation- (strongly disagree) to four (strongly agree). As suggested
ships with various parameters such as sociodemographics, by the authors, the score for each dimension was calcu-
professionals (VAS stress at work or working hours) or lated using the following equations: Q10 +Q11+Q12 +
well-­being (sleep). (5-­Q13)+Q14+Q15+Q16+Q17+Q18 for job demand, and
4*Q4 +4*(5-­ Q6)+4*(Q8) +2*(5-­ Q2)+2*(Q5) +2*(Q7)
+2*(Q1) +2*(Q3) +2*(Q9) for job control. According
METHODS to Karasek,13 14 combining scores of job demand and job
Participants control allowed to define four situations at work (active,
We implemented a cross-­ sectional observational study. passive, relaxed, tense/job strain) corresponding to four
Self-­reported questionnaires were proposed to voluntary quadrants, with an allocation between various sociopro-
French workers using WittyFit software (https://wittyfit.​ fessional groups (figure 1). From French data,18 the job
com/). Users of WittyFit were disseminated through strain threshold was set for a demand score higher than
several national French companies, mostly from the 21 and a control score less than 70. VAS assessed the
service sector (tertiary sector). WittyFit is software that perceived control and demand of individuals at work, on

2 Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403

 Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
Figure 1 The job demand-­control model of Karasek and its four quadrants: active, passive, relaxed and tense (job strain).
NSW, non-­skilled worker; SW, skilled worker.17 18

a horizontal, non-­calibrated line of 100 mm, ranging from variable, with a minimum number of 100 subjects to
very low (0) to very high (100). ensure stability of the variance-­covariance matrix’ and (2)
‘Often 0.70 is recommended as a minimum standard for
Secondary outcomes reliability. We gave a positive rating for reliability when
We retrieved sociodemographic data such as age, gender, the ICC (Intraclass correlation coefficients) or weighted
education level and marital status. We also collected Kappa was at least 0.70 in a sample size of at least 50
characteristics of work: occupation, number of hours at patients.’
work per week, and seniority within the company. Sleep All analyses were performed using Stata software
quantity was assessed by the number of hours of sleep per (V.13, StataCorp) for a two-­sided type I error of α=5%.
night.31 Sleep quality, well-­being, stress at work and stress Participant’s characteristics were expressed as means±SD
at home were evaluated using VAS ranging from very low or median (IQR) for continuous data (assumption of
(0) to very high (100).24 28 We evaluated depression (D) normality assessed using the Shapiro-­Wilk test) and as
and anxiety (A) with the use of the Hospital and Anxiety numbers (percentages) for categorical parameters. We
Depression Scale (HAD).32 The HAD is a 14-­item self-­ followed the usual steps of validation of a new ques-
reported measure with two subscales of seven items, one tionnaire.33 Internal validity allows verification that the
assessing anxiety (HAD-­A) and the other assessing depres- independent variable (Karasek score) and is responsible
sion (HAD-­D), rated from zero to three. As suggested by for variation of the dependent variable (VAS control
the authors, the HAD-­A score is calculated by adding and VAS demand). The internal validity of the two VAS
the scores for questions 1, 3, 5, 7, 9, 11 and 13; and the was assessed according tointernal consistency based on
HAD-­D score with questions 2, 4, 6, 8, 10, 12 and 14. The correlation coefficient (Pearson or Spearman) and Cron-
scores for each subscale range from 0 to 21. A score less bach alpha coefficient (adequate expected values higher
than seven means the absence of disease; between eight than 0.64)34—we also explored more deeply the rela-
and ten a doubtful disease; a score higher than 11 is a tionship between each VAS and items of Karasek ques-
confirmed disease. tionnaire using correlation coefficients and principal
Time of measurements component analysis.33 Test–retest reproducibility was
Participants completed the questionnaires when time was assessed using Lin concordance coefficient and Bland
convenient to them. They were automatically asked to and Altman plots.35 External validity takes into account
complete the two VAS (VAS demand and VAS control) generalisability of the new scales, that is, relation with
and the JCQ 1 week after the first completion of question- other variables or groups. External validity was assessed
naires, to perform test–retest. Total completion time was using correlation coefficients (Pearson or Spearman)
approximately 30 min (20 min first session and 10 minutes between VAS and sociodemographic or other psycho-
second session). logical measures, such as sleep, well-­being, stress, anxiety
or depression scores. Then, a receiver operating charac-
Statistics teristic (ROC) curve analysis was proposed to determine
Sample size was determined according to COSMIN the best thresholds of VAS to predict a gold-­standard
(COnsensus-­based Standards for the selection of health Karasek, according to clinical relevance and usual indexes
status Measurement INstruments) recommendations32 33: reported in the literature (Youden, Lin and efficiency).
(1) ‘Rules-­of-­thumb vary from four to ten subjects per Sensitivity, specificity, positive and negative predictive

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403 3

Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
values were calculated and presented with 95% CIs.
The concordance between Karasek quadrants and their
equivalents from VAS, according to cut-­offs determined
by ROC curve analysis, was evaluated using agreement
rate and Kappa concordance coefficient. Finally, quan-
titative variables were compared between independent
groups by analysis of variance (ANOVA) or Kruskal-­Wallis
test if ANOVA conditions were not met (normality and
homoscedasticity analysed using the Bartlett test). When
appropriate, post hoc tests were performed considering
multiple comparisons (Tukey-­Kramer post ANOVA and
Dunn after Kruskal-­ Wallis). The comparisons between
groups were carried out using the χ2 or Fisher's exact test
for categorical variables. When appropriate, a post hoc
test was used (Marascuillo procedure).

RESULTS
Participants
Among the 1580 workers using WittyFit, 190 (12.0%) agreed
to participate. We had no missing data (n=190; 100%)
for primary outcomes that is, Karasek questionnaire, VAS
demand and VAS control. Secondary outcomes were fulfilled
by 163 (85.8%)of respondents. Among them, 89 (54.1%)
were women. The test–retest approach was performed on
the 129 participants who answered twice to the Karasek ques-
tionnaire and the two VAS (demand and control) (figure 2).
At baseline, the 190 and 163 workers did not differ according
to sensibility analysis. Mean age was 41.9±11.7 years. Seniority
within the company was 11.0±10.8 years. Most of the workers
were senior managers (68.1%) and had a master’s degree
(74.9%) (table 1).

JCQ of Karasek
According to Karasek’s Model, 53.2% of participants were
‘active’ (demand score ≥21 and control score ≥70) with a
mean demand score of 25.0±4.0 and a mean control score Figure 2 Flow chart and study design. JCQ, Job Content
of 83.8±35.0; 30.0% were ‘relaxed’ (demand score <21 and Questionnaire; VAS, Visual analogue scale.
control score ≥70) with a mean demand score of 22.0±5.0 and
a control score higher of 83.8±35.0; 14.7% were ‘tense/job control ≥75 (table 2). As for JCQ, there were also no influ-
strain’ (demand score ≥21 and control score <70) with a ence of age and sex on both VAS demand and VAS control.
demand score of 25.0±4.0 and a control score of 61.9±5.6; Married workers perceived a higher job demand (p=0.03).
10.5% were ‘passive’ (demand score <21 and control Workers with a higher seniority in the company had a higher
score <70) with a demand score of 22.0±5.0 and a control VAS control (p=0.02). Workers with a seniority within the
score of 61.9±5.6 (table 2). There were no statistical differ- company above 15 years had a higher VAS control than
ences regarding age, sex and marital status, at the dimen- workers with a seniority below 2 years (90.0±5.3 vs 10.0±5.2,
sions evaluated by the JCQ of Karasek (demand and control) p<0.001). Similarly, to JCQ, senior executives and master’s
or between quadrants (active, passive, tense/job strain, degree holders had or tended to have higher levels of job
relaxed). Workers with longer seniority/experience within control (p<0.10) but did not differ in job demand (online
the company tended to have higher job control (p=0.09). supplemental appendix 1).
Compared with other workers, senior executives and master’s
degree staff had higher levels of job control (p<0.05) but did Validation of VAS demand and VAS control: internal validity
not differ in job demand (online supplemental appendix 1). Internal consistency
Significant correlations were emphasised: 0.59 (p<0.001)
VAS demand and VAS control between the demand score from the JDC of Karasek and
Mean VAS demand was 75.2±20.6 and mean VAS control the VAS demand and 0.57 (p<0.001) between the control
was 81.9±21.9. Using the retrieved cut-­off of 75 (see below), (latitude decision) score from the JDC of Karasek and
56.3% of workers had a VAS demand ≥75% and 72.6% a VAS the VAS control. The Cronbach alpha coefficient was

4 Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403

 Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
Table 1 Characteristics of participants Table 2 Job-­demand and job-­control assessed with the
Characteristics of participants Sample size (n=163) % use of Visual Analogue Scale (VAS) or the Job Content
Questionnaire of Karasek
Sex
Sample size
 Women 89 54.6 Variables n=190 (%) Mean±SD
 Men 74 45.4
VAS
Age, year (mean±SD) 41.9±11.7  VAS job-­demand 75.2±20.6
Education level  VAS job-­demand <75 83 (43.7) 53.8±19.2
 Bachelor degree or less 8 4.9  VAS job-­demand ≥75 107 (56.3) 88.3±8.3
 Undergraduate 33 20.3  VAS latitude decision 81.9±21.9
 Master’s degree 122 74.9  VAS latitude decision <75 52 (2.4) 52.3±22.2
Marital status  VAS latitude decision ≥75 138 (72.6) 91.1±8.6
 Single 33 20.3 Modèle de Karasek
 De Facto 45 27.6  Dimensions
 Married 84 51.5  Job-­demand 22.0±5.0
 Widow 1 0.6  Job-­demand <21 61 (32.1) 16.0±3.0
Occupational categories  Job-­demand ≥21 129 (67.9) 25.0±4.0
 Senior executives 111 68.1  Latitude decision 76.7.±8.5
 Mid-­level workers 10 6.1  Latitude decision <70 48 (25.3) 61.9±5.6
 Skilled workers 16 9.8  Latitude decision ≥70 142 (74.7) 83.8±35.0
 Unskilled workers 26 26.9
 Quadrants
 Active 101 (53.2)
0.68. The relationships between each VAS and items of  Passive 20 (10.5)
Karasek questionnaire are presented in online supple-  Tense/job strain 28 (14.7)
mental appendix 2. Briefly, for each dimension (control  Relaxed 41 (21.6)
and demand), all items of the Karasek questionnaire were
correlated with the corresponding VAS, except item 14,
and the highest correlations were found between items quadrants and their equivalents. Groups according to
4, 6 and 8 and VAS control, and between items 11 and 15 VAS emphasised more ‘actives’ and ‘strain’ workers than
and VAS demand. JCQ. However, ‘relaxed’ and ‘passives’ workers were
overall more highlighted with JCQ (table 3).
Test–retest reproducibility
Lin concordance coefficient was 0.73 (95% CI 0.64 to Validation of VAS demand and VAS control: external validity
0.81) for job demand between the demand score from VAS demand
the JDC of Karasek and the VAS demand and 0.71 (95% According to cut-­offs for VAS, a high job demand was
CI 0.63 to 0.80) for job control the control score from the linked with a higher stress at work (p<0.001) and at home
JDC of Karasek and the VAS control (figure 3). (p=0.03), and a higher time spent at work (p<0.001). The
higher job control, the higher VAS well-­being (p=0.04).
Cut-off’s determination and concordance Relationships were similar using the JCQ, as well as for
For both VAS demand and VAS control, we emphasised other quantitative or qualitative secondary outcomes
a significant cut-­off set at 75, with a good sensitivity (age, sex, education level, seniority within the company,
and specificity. For VAS demand, the retrieved cut-­off occupation, HAD-­D, HAD-­A, number of hours at work
(p<0.001) of 75 had a sensitivity of 71%, a specificity of per week, duration of sleep) that were non-­significant
74%, an area under ROC curve of 0.79±0.04 (95% CI according to cut-­offs both for VAS demand and JCQ,
0.72 to 0.85), a positive predictive value of 70.5% (95% CI except the relationship between quality of sleep and job
61.9% to 78.2%) and a negative predictive value of 73.8% demand that was retrieved mainly using the JCQ and
(95% CI 60.9% to 84.2%). For VAS control, the retrieved between marital status and job demand that was retrieved
cut-­off (p<0.001) of 75 had a sensitivity of 81%, specificity mainly using the VAS demand (table 4 and online supple-
of 56%, an area under ROC curve of 0.76±0.04 (95% CI mental appendix 1).
0.68 to 0.84), a positive predictive value of 91% (95%CI
73.6% to 87.1%) and a negative predictive value of 54.2% VAS control
(95% CI 39.2% to 68.6%). According to cut-­offs for VAS, the higher job control is
The concordance between Karasek quadrants and their linked to the higher VAS well-­being (p=0.04) and the
equivalents from VAS was relatively low (k=0.37) with higher education level (p=0.009). Using the JCQ, the
percentages of agreement of 54.9% between Karasek same relationship was shown (p<0.02). Relationships

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403 5

Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
Figure 3 Bland and Altman plot or representation of agreement between both series of Visual Analogue Scales (VAS) of job
demand and job control.

were similar—non-­ significant—for other quantitative acceptability, internal validity, reproducibility and
secondary outcomes (age, sex, marital status, HAD-­ D, external validity.
HAD-­A, number of hours of work per week, duration of
sleep, VAS stress at work, VAS stress at home) both for Acceptability
VAS control and JCQ, except the relationship between Workplace stress—that is, jobstrain—is a concern for
quality of sleep, occupation and job control who was only both employers and workers because it can be the cause
retrieved using the JCQ and between seniority and job of absenteeism36 and various pathologies.37 38 Directives
control that was retrieved mainly using the VAS control stated that occupational physicians needed to assess
(table 4, online supplemental appendix 1). psychosocial risks to fight against work-­ related stress,
and improve occupational health and safety concerns.39
VAS quadrant
Although the JDC model of Karasek is the gold standard13
Using quadrants from our retrieved cut-­ offs for VAS,
job strain (tense) workers had higher job stress and to assess psychosocial risks at work, its length makes it
time spent at work, and lower well-­being compared with difficult to use routinely in daily clinical practice by occu-
relaxed workers (p<0.001, p<0.001 and p=0.05, respec- pational practitioners.24 However, the fact that VAS are
tively). Relationships were also similar using the JCQ, easy to implement, time-­efficient in execution and easy
except the relationship between quality of sleep, stress to understand by the patient, offers a unique possibility
at home, education level, occupation and job demand for a consistent standardised use in common practice.
that were more pronounced using the JCQ and between Moreover, the VAS is already a common tool used by
marital status and job demand that was retrieved mainly occupational physicians to assess other conditions such as
using the VAS demand (table 5 and online supplemental stress.24 Occupational physicians have limited time to deal
appendix 1). Comparisons between other quadrants also with considerable numbers of workers and worksites, and
retrieved similar findings. considerable numbers of occupational risks such as phys-
ical, chemical, biological and psychosocial risks.40 Even if
a questionnaire is implemented before the consultation,
DISCUSSION a limited number of questions are necessary to be able to
This study allowed the validation of VASs of job demand assess all occupational risks factors.41 Therefore, VAS of
and job control at the workplace and focused on its job demand and job control should be used frequently in

Table 3 Concordance between quadrants retrieved from Job-­Content Questionnaire (JCQ) of Karasek and their equivalents
retrieved from Visual Analogue Scales (VAS) of demand and control
No of workers within each quadrant retrieved
From JCQ of From VAS of demand and control Total
Karasek Passive Relaxed Tense Active
 Passive 11 6 1 2 20
 Relaxed 4 24 2 11 41
 Tense/job strain 4 5 10 9 28
 Active 8 21 13 59 101
Total 27 56 26 81 190

6 Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403

 Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
Table 4 Agreements between Visual Analogue Scales (VAS) and Job Content Questionnaire (JCQ) of Karasek and differences
in outcomes according to cut-­offs for the two dimensions of the job-­demand control model of Karasek
Demand Control
VAS JCQ of Karasek VAS JCQ of Karasek
Variables <75 ≥75 <21 ≥21 <75 ≥75 <70 ≥70

HAD-­Depression                
 Mean±SD 9.4±1.5 9.5±1.7 9.4±1.5 9.5±1.7 9.7±1.7 9.4±1.6 9.9±2.0 9.3±1.5
 P value 0.28 0.28 0.47 0.31
 Agreement Yes Yes
HAD-­anxiety                
 Mean±SD 14.8±3.7 13.7±4.3 15.5±3.7 13.7±4.1 13.2±4.2 14.7±4.0 13.0±4.2 14.6±4.0
 P value 0.28 0.30 0.18 0.56
 Agreement Yes Yes
No of hours of work per week              
 Mean±SD 38.7±10.0 44.8±12.9 37.4±12.1 44.3±11.5 42.3±1.7 42.1±1.2 37.5±1.3 43.5±1.2
 P value <0.001 <0.001 0.90 0.07
 Agreement Yes Yes
VAS quality of sleep                
 Mean±SD 60.6±25.4 54.2±29.2 64.9±25.2 53.3±28.2 53.0±3.3 58.5±2.5 49.6±27.9 59.5±27.3
 P value 0.11 0.01 0.18 0.04
 Agreement Moderate Moderate
Duration of sleep                
 Mean±SD 429.2±54.8 423.9±56.9 437.5±55.7 420.9±55.5 426.2±53.7 426.3±56.9 418.1±56.5 429.0±55.6
 P value 0.74 0.76 0.06 0.18
 Agreement Yes Yes
VAS stress at work                
 Mean±SD 45.9±24.1 64.1±22.7 43.4±25.8 62.2±22.2 56.8±3.5 55.9±2.1 56.0±27.8 56.2±24.1
 P value <0.001 <0.001 0.83 0.95
 Agreement Yes Yes
VAS stress at home                
 Mean±SD 32.5±24.9 37.1±25.3 28.2±3.1 38.3±2.2 39.4±3.5 33.4±2.1 38.9±3.6 33.8±3.6
 P value 0.03 0.01 0.15 0.22
 Agreement Yes Yes
VAS well-­being                
 Mean±SD 64.7±21.4 61.4±22.0 66.5±3.0 61.1±1.8 57.3±3.1 65.0±2.1 54.0±24.5 65.8±19.1
 P value 0.30 0.13 0.04 0.01
 Agreement Yes Yes

Mean±SD.
Bold for significant relationship (p<0.05)
HAD, Hospital and Anxiety Depression.

common routine clinical practice of occupational physi- model appeared to be valid, reliable and precise instru-
cians.42 Physicians do not have time to go to the workplaces ments for the assessment of job demand and job control.
of every stressed worker to investigate working conditions, Lin concordance coefficients were greater than 0.70 for
to individually clarify the situation with managers, and to both VAS of job demand and job control.35 VAS demand
find solutions.43 44 Identifying workers under the highest and control detection had acceptable sensitivity and spec-
pressure with improved accuracy would assist by reducing ificity factors of job strain. Positive and negative predictive
the numbers in this group to a manageable sample size. values are acceptable. The thresholds are determined to
Moreover, from a statistical point of view, data ranged
be 75/100 for both VAS demand and control. SUMER
from minimal to maximal values, with a reasonable SD,
(SUrveillance Médicale des Expositions aux Risques
as suggested by variation coefficients around 0.25–0.30.
professionnels) study demonstrated satisfactory psycho-
Internal consistency and reproducibility metric properties of Karasek’s JCQ scales for the French
According to this study, VAS developments to assess working population.18 Internal consistency was satis-
psychosocial job characteristics based on the Karasek factory as Cronbach’s alpha coefficients were observed

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403 7

8
Table 5 Agreements between Visual Analogue Scales (VAS) and Job Content Questionnaire of Karasek and differences in outcomes according to quadrants (active,
passive, tense, relaxed)
VAS quadrants Karasek quadrants
Variables Active Passive Tense Relaxed Active Passive Tense Relaxed

HAD-­Depression
Open access

 Mean±SD 9.4±1.6 9.6±1.7 9.8±1.8 9.3±1.5 9.3±1.6 9.6±2.1 10.1±1.9 9.3±1.2

 P value 0.66 0.33
 Agreement Yes
HAD-­anxiety
 Mean±SD 14.1±4.4 13.8±4.2 12.5±4.2 15.4±3.4 14.0±4.0 14.1±3.8 12.2±4.3 16.2±3.6
 P value 0.36 0.14
 Agreement Yes
No of hours of work per week
 Mean±SD 45.4±12.8 41.5±7.0 43.2±13.4 37.5±10.8 45.6±12.3 34.6±9.7 39.3±5.9 38.3.±12.7
 P value <0.001 <0.001
 Agreement Yes
VAS quality of sleep
 Mean±SD 55.6±30.5 55.0±23.1 49.9±24.9 62.8±26.0 55.6±27.9 56.1±27.3 45.0±27.9 68.7±23.1
 P value 0.21 <0.001
 Agreement Moderate
Duration of sleep
 Mean±SD 424.1±58.6 426.9±55.8 423.5±52.5 430.3±54.8 424.8±55.9 433.5±59.6 407.1±52.5 439.4±54.3
 P value 0.94 0.10
 Agreement Yes
VAS stress at work
 Mean±SD 63.6±22.5 49.4±24.8 65.8±23.6 44.3±23.8 61.7±22.0 44.8±30.2 63.9±23.3 42.7±23.8
 P value <0.001 <0.001
 Agreement Yes
VAS stress at home
 Mean±SD 34.5±25.4 35.1±25.6 45.1±23.4 31.4±25.6 37.6±25.3 36.0±27.8 41.0±23.3 24.6±21.9
 P value 0.13 0.01
 Agreement Moderate
VAS well-­being
 Mean±SD 64.0±21.5 60.5±23.1 53.4±22.2 66.6±20.5 62.9±19.9 52.9±30.2 54.8±24.0 73.1±15.5
 P value 0.05 <0.001
 Agreement Yes

Mean±SD.
Bold for significant relationship (p<0.05)
HAD, Hospital and Anxiety Depression.

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
 Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
as being higher than 0.64. Convergent validity tests of participants who responded to both the test and retest
confirmed the expected association with key variables was higher than commonly reported in the literature.42 43
such as age, occupation, well-­being, stress, depression or However, VAS for job demand and job control have some
anxiety. In line with the literature,45 46 we demonstrated limits concerning discrimination of at-­ risk workers.
that workers with a master’s degree had greater job Despite its use for screening at-­risk workers, JCQ was not
control both using VAS and JCQ, as well as for workers initially conceived as a discriminative tool for clinical
with higher seniority. However, the relationship between practice,16 such as other questionnaires, that is, burn-­out,
seniority and job control was found significant only using for example.50 The validation of a single cut-­off point for
the VAS. Even if there was no relevant literature on senior both job demand and job control, by using easy, quick,
executives, it seems logical that they would have higher and reliable VAS, may improve the decision-­ making
job control as they are often more educated and quali- process of occupational health practitioners. Even if clin-
fied, with greater seniority within the company. Although ical examination and occupational physicians’ observa-
no difference was apparent using the JCQ and was similar tions remain essential for assessing job stressors and their
to findings reported in the literature,47 we showed that repercussions, the two validated VAS demand and VAS
married workers had higher levels of job demand than control tools will help mass screening—especially consid-
single workers using the VAS, which may also favour the ering that an occupational physician may have >10 000
discriminant use of VAS. Family obligations may increase workers to follow. We did not assess social support because
a perception of overload at work.48 of an informatics issue regarding this item; however, the
JCQ does not include the social support and is still widely
External validity used.13 45–47 Sensibility, specificity, and Lin concordance
We reported a prevalence of job strain of 15%. We did coefficients were only moderate. Correlations between
not show relationships between VAS scores and quadrants VAS and the Karasek groups was low. Moreover, we did
and age, sex, education level, family situation, concerning not have many precisions regarding each domain using
Karasek domains, Karasek groups or VAS. However, VAS. Despite the widespread use and relevance of VAS,
the SUMER study, based on a larger sample of 24, 486 VAS are global first-­line evaluations that do not accurately
workers, reported a higher prevalence of job strain assess putative explaining stress factors that are possible
than our results (24% vs 15%) with a predominance of with the use of the JCQ of Karasek. We used some other
women (30% vs 21%).18 The SUMER study emphasised non-­validated VAS to evaluate well-­being and quality of
a sharp gradient in socio-­professional levels. Job control sleep, previously used in literature.33 51 52 The selective use
increased with high socioprofessional level as well as job of VAS demand and VAS control may limit occupational
demand. Consequently, job strain increased for low socio- physicians in their capacity to understand work-­related
professional level. In addition, the SUMER study demon- stress, and thus to give easy counselling to workers. There-
strated a higher job control among men and a higher fore, we promote the use of VAS demand and control as
job demand among women.17 18 We hypothesise that our an easy tool to detect abnormal situations, and if detected,
lower sample size and the representativity of all sociopro- occupational physicians should further assess workplace
fessional level in our sample may have precluded some stress with the use of the JCQ. Moreover, VAS for job
significant statistical results. Nonetheless, we demon- demand and job control could also be easily used there-
strated that job control and job demand had accept- after to follow workers at regular intervals, possibly using
able external validity. According to the literature,12 18 we dedicated software.24 Further studies should encompass
showed that stress at work and time spent at work were the subsequent modifications of the JCQ, specifically,
correlated with a strong job demand, for both VAS and adding social support—job-­demand-­control-­support
JCQ. VAS well-­being increased with high job control. Job (JDCS) model; as well as the forthcoming effort–reward
strain was linked with low VAS well-­being, a high job stress imbalance (ERI) model of Siegrist10–12; and finally, the
and a high time spending at work. Job Demands-­Resources model that is, a sort of umbrella
model including both JDCS and ERI concept.53 54
Limitations
The study has some limitations. The response rate may
seem low compared with other studies using question-
naires in French populations.16 39 The majority of partic- CONCLUSION
ipants were high seniority precluding generalisability of Although it performs less than JCQ, VAS demand and
our results.45 However, we included a substantial sample control are simple and rapid tools for screening patients
size of workers allowing us to carry out statistical anal- with putative work-­related stress. VAS demand and control
yses with the number of subjects required, determined a can be used by the occupational practitioner in daily clin-
priori. Moreover, we had a few missing data points and ical practice for primary prevention and diagnosis. We
the number of respondents followed recommendations determined a cut-­off of 75 mm for both VAS job demand
for the validation of questionnaires.33 40 Despite the liter- and job control to discriminate at-­risk workers. However,
ature reports that a high drop-­out rate is inherent to this when difficulties are highlighted by overreaching cut-­
type of study with several questionnaires,49 the number offs of 75 mm, we promote the use of JCQ to be more

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403 9

 Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
discriminant and specific because greater accuracy is permits others to distribute, remix, adapt, build upon this work non-­commercially,
needed to establish action plans and help workers and license their derivative works on different terms, provided the original work is
properly cited, appropriate credit is given, any changes made indicated, and the use
is non-­commercial. See: http://creativecommons.org/licenses/by-nc/4.0/.
Author affiliations
1
Occupational and Environmental Medicine, University Hospital Centre, F-­63000 ORCID iDs
Clermont-­Ferrand, France Frederic Dutheil http://orcid.org/0000-0002-1468-6029
2
Physiological and Psychosocial Stress, LAPSCO, F-­63000 Clermont-­Ferrand, Jean-­Baptiste Bouillon-­Minois http://orcid.org/0000-0002-6556-4187
France Maëlys Clinchamps http://orcid.org/0000-0003-4756-6928
3
Biostatistics Unit, Clinical Research and Innovation Direction, University Hospital
Centre, F-­63000 Clermont-­Ferrand, France
4
Emergency Medicine, University Hospital Centre, F-­63000 Clermont-­Ferrand,
France
5
Psychiatry, University Hospital Centre, F-­63000 Clermont-­Ferrand, France REFERENCES
6
NPSY-­SYDO, University Clermont Auvergne, F-­63000 Clermont-­Ferrand, France 1 Chrousos GP, Gold PW. The concepts of stress and stress system
7 disorders. overview of physical and behavioral homeostasis. JAMA
Wittyfit, F-­92600 Asnières-­sur-­Seine, France 1992;267:1244–52.
8
Psychology and NeuroCognition Laboratory, University Grenoble Alpes, F-­38000 2 Chrousos GP. Stress and disorders of the stress system. Nat Rev
Grenoble, France Endocrinol 2009;5:374–81.
9
Institut Universitaire de France, F-­75000 Paris, France 3 Nieuwenhuijsen K, Bruinvels D, Frings-­Dresen M. Psychosocial
10
Psychology, University Clermont Auvergne, F-­63000 Clermont-­Ferrand, France work environment and stress-­related disorders, a systematic review.
11 Occup Med 2010;60:277–86.
LAPSCO, F-­63000 Clermont-­Ferrand, France 4 Backé E-­M, Seidler A, Latza U, et al. The role of psychosocial stress
12
Hong Kong Baptist University, Kowloon, HK, Hong Kong at work for the development of cardiovascular diseases: a systematic
13
Pôle SAMU-­SMUR-­Urgences, Service Urgences Adultes, University Hospital Centre, review. Int Arch Occup Environ Health 2012;85:67–79.
F-6­ 3000 Clermont-F­ errand, France 5 Eisenach JH, Sprung J, Clark MM, et al. The psychological
14
General medicine, University Hospital Clermont-­Ferrand, F-­63000 Clermont-­ and physiological effects of acute occupational stress in
new anesthesiology residents: a pilot trial. Anesthesiology
Ferrand, France
2014;121:878–93.
6 Silva-­Junior JSda, Fischer FM. Long-­Term sickness absence due
Acknowledgements We express our sincere gratitude to Marie-­Anne Cousseau to mental disorders is associated with individual features and
who established the public-­private partnership between WittyFit and the University psychosocial work conditions. PLoS One 2014;9:e115885.
Hospital of Clermont-­Ferrand, France 7 Cohen S, Janicki-­Deverts D, Miller GE. Psychological stress and
disease. JAMA 2007;298:1685–7.
Contributors FD conceptualised and designed this study and developed the 8 Proietti R, Mapelli D, Volpe B, et al. Mental stress and ischemic
primary aims and hypotheses. FD was responsible for the submission of the heart disease: evolving awareness of a complex association. Future
protocol to the ethics committee. FD and BP primarily conceptualised the data Cardiol 2011;7:425–37.
analysis strategy to respond to objectives. CL was responsible for the clinical 9 Ruotsalainen JH, Verbeek JH, Marine A. Preventing occupational
stress in healthcare workers. Cochrane Database Syst Rev 2014;13.
trial. SD and TC was responsible of the data collection and extraction. BP was
10 Clinchamps M, Auclair C, Prunet D, et al. Burnout among hospital
responsible for the statistical analysis. CL and FD interpreted the data and wrote Non-­Healthcare staff: influence of job Demand-­Control-­Support, and
the first draft of this manuscript which was critically revised and upgraded by Effort-­Reward imbalance. J Occup Environ Med 2021;63:e13–20.
BP, JB-­BM, MC, GB, SD, TC, MM, LM, JSB, JS and FM. All authors were actively 11 Sérole C, Auclair C, Prunet D, et al. The forgotten health-­care
involved in the planning and conduct of the study described and all were involved occupations at risk of Burnout-­A burnout, job Demand-­Control-­
in interpretation and reporting the findings of this study. They all approved the final Support, and Effort-­Reward imbalance survey. J Occup Environ Med
manuscript. FD is responsible for the overall content as the guarantor 2021;63:e416–25.
12 Trousselard M, Dutheil F, Naughton G, et al. Stress among nurses
Funding The study is integrally funded by WittyFit, Paris, France working in emergency, anesthesiology and intensive care units
depends on qualification: a job Demand-­Control survey. Int Arch
Competing interests SD and TC are founders of WittyFit. Other authors have
Occup Environ Health 2016;89:221–9.
declared that no competing interests exist. 13 Karasek R, Baker D, Marxer F, et al. Job decision latitude, job
Patient and public involvement Patients and/or the public were involved in the demands, and cardiovascular disease: a prospective study of
design, or conduct, or reporting, or dissemination plans of this research. Refer to Swedish men. Am J Public Health 1981;71:694–705.
14 Theorell T, Karasek RA. Current issues relating to psychosocial job
the Methods section for further details.
strain and cardiovascular disease research. J Occup Health Psychol
Patient consent for publication Consent obtained directly from patient(s) 1996;1:9–26.
15 Pisanti R, van der Doef M, Maes S, et al. Job characteristics,
Ethics approval This exploratory study was undertaken in an ecological situation organizational conditions, and distress/well-­being among Italian
and received approval from the Ethics Committee of the University Hospital of and Dutch nurses: a cross-­national comparison. Int J Nurs Stud
Clermont-­Ferrand, France and from the National Commission for Information 2011;48:829–37.
Technology and Civil Liberties (CNIL). 16 Karasek R, Brisson C, Kawakami N, et al. The job content
questionnaire (JCQ): an instrument for internationally comparative
Provenance and peer review Not commissioned; externally peer reviewed. assessments of psychosocial job characteristics. J Occup Health
Data availability statement All data relevant to the study are included in the Psychol 1998;3:322–55.
17 Niedhammer I, Chastang J-­F, Levy D, et al. Study of the validity of
article or uploaded as online supplemental information. All relevant data are within a job-­exposure matrix for psychosocial work factors: results from
the paper. the National French SUMER survey. Int Arch Occup Environ Health
Supplemental material This content has been supplied by the author(s). It has 2008;82:87–97.
not been vetted by BMJ Publishing Group Limited (BMJ) and may not have been 18 Niedhammer I, Chastang JF, Gendrey L, et al. [Psychometric
properties of the French version of Karasek's "Job Content
peer-­reviewed. Any opinions or recommendations discussed are solely those
Questionnaire" and its scales measuring psychological pressures,
of the author(s) and are not endorsed by BMJ. BMJ disclaims all liability and decisional latitude and social support: the results of the SUMER].
responsibility arising from any reliance placed on the content. Where the content Sante Publique 2006;18:413–27.
includes any translated material, BMJ does not warrant the accuracy and reliability 19 Dutheil F, Delaire P, Boudet G, et al. [Cost/effectiveness comparison
of the translations (including but not limited to local regulations, clinical guidelines, of the vaccine campaign and reduction of sick leave, after
terminology, drug names and drug dosages), and is not responsible for any error vaccination against influenza among the Clermont-­Ferrand University
and/or omissions arising from translation and adaptation or otherwise. Hospital staff]. Med Mal Infect 2008;38:567–73.
20 Kelly C, Dutheil F, Haniez P, et al. [Analysis of motivations for antiflu
Open access This is an open access article distributed in accordance with the vaccination of the Clermont-­Ferrand University Hospital staff]. Med
Creative Commons Attribution Non Commercial (CC BY-­NC 4.0) license, which Mal Infect 2008;38:574–85.

10 Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403

 Open access

BMJ Open: first published as 10.1136/bmjopen-2020-046403 on 17 March 2022. Downloaded from http://bmjopen.bmj.com/ on December 7, 2023 by guest. Protected by copyright.
21 Carlsson F, Merlo J, Lindström M, et al. Representativity of a 38 Richardson S, Shaffer JA, Falzon L, et al. Meta-­Analysis of perceived
postal public health questionnaire survey in Sweden, with special stress and its association with incident coronary heart disease. Am J
reference to ethnic differences in participation. Scand J Public Health Cardiol 2012;110:1711–6.
2006;34:132–9. 39 Iavicoli S, Natali E, Deitinger P, et al. Occupational health and safety
22 Reilly EE, Brown TA, Wierenga CE. Evaluating patterns of policy and psychosocial risks in Europe: the role of stakeholders'
inconsistent and missing data on the eating disorders examination-­ perceptions. Health Policy 2011;101:87–94.
questionnaire in a sample of treatment-­seeking adults and 40 Lesage F-­X, Berjot S, Altintas E, et al. Burnout among occupational
adolescents. Eat Disord 2019:1–10. physicians: a threat to occupational health systems?--A nationwide
23 Rolstad S, Adler J, Rydén A. Response burden and questionnaire cross-­sectional survey. Ann Occup Hyg 2013;57:913–9.
length: is shorter better? A review and meta-­analysis. Value Health 41 Koitsalu M, Eklund M, Adolfsson J, et al. Effects of pre-­notification,
2011;14:1101–8. invitation length, questionnaire length and reminder on participation
24 Dutheil F, Pereira B, Moustafa F, et al. At-­Risk and intervention rate: a quasi-­randomised controlled trial. BMC Med Res Methodol
thresholds of occupational stress using a visual analogue scale. 2018;18:3.
PLoS One 2017;12:e0178948. 42 Choi BK, Schnall P, Landsbergis P, et al. Recommendations for
25 Li J, Riedel N, Barrech A, et al. Long-­Term effectiveness of a stress individual participant data meta-­analyses on work stressors and
management intervention at work: a 9-­year follow-­up study based health outcomes: comments on IPD-­Work Consortium papers.
on a randomized Wait-­List controlled trial in male managers. Biomed Scand J Work Environ Health 2015;41:299–311.
Res Int 2017;2017:18. 43 Psenka TM, Freedy JR, Mims LD, et al. A cross-­sectional study
26 Boonstra AM, Schiphorst Preuper HR, Reneman MF, et al. Reliability of United States family medicine residency programme director
and validity of the visual analogue scale for disability in patients with burnout: implications for mitigation efforts and future research. Fam
chronic musculoskeletal pain. Int J Rehabil Res 2008;31:165–9. Pract 2020;37:772–8.
27 Lesage FX, Berjot S. Validity of occupational stress assessment 44 Rao CR, Darshan B, Das N, et al. Practice of physical activity
using a visual analogue scale. Occup Med 2011;61:434–6. among future doctors: a cross sectional analysis. Int J Prev Med
28 Lesage F-­X, Berjot S, Deschamps F. Clinical stress assessment using 2012;3:365–9.
a visual analogue scale. Occup Med 2012;62:600–5. 45 Giorgi G, Arcangeli G, Ariza-­Montes A, et al. Work-­Related stress
29 Ratinaud MC, Chamoux A, Glace B, et al. Job satisfaction evaluation in the Italian banking population and its association with recovery
in low back pain: a literature review and tools appraisal. Ann Phys experience. Int J Occup Med Environ Health 2019;32:255–65.
Rehabil Med 2013;56:465–81. 46 Moreno-­Pimentel AG, Meneses Monroy A, Martín-­Casas P, et al.
30 Dutheil F, Duclos M, Naughton G, et al. WittyFit-­Live your work Impact of social and occupational factors over job control. Med Lav
differently: study protocol for a Workplace-­Delivered health 2019;110:226–33.
promotion. JMIR Res Protoc 2017;6:e58. 47 Blumenthal JA, Thyrum ET, Siegel WC. Contribution of job strain,
31 Dutheil F, Trousselard M, Perrier C, et al. Urinary interleukin-­8 is job status and marital status to laboratory and ambulatory blood
a biomarker of stress in emergency physicians, especially with pressure in patients with mild hypertension. J Psychosom Res
advancing age--the JOBSTRESS* randomized trial. PLoS One 1995;39:133–44.
2013;8:e71658. 48 Komlenac N, Stockinger L, Vogler T, et al. Psychometric analysis of a
32 Bocéréan C, Dupret E. A validation study of the hospital anxiety and German-­Language version of the Work-­Family conflict and Family-­
depression scale (HADS) in a large sample of French employees. Work conflict scale. Front Psychol 2021;12:782618.
BMC Psychiatry 2014;14:354. 49 Guo Y, Kopec JA, Cibere J, et al. Population survey features and
33 Ravoux H, Pereira B, Brousse G, et al. Work addiction test response rates: a randomized experiment. Am J Public Health
questionnaire to assess Workaholism: validation of French version. 2016;106:1422–6.
JMIR Ment Health 2018;5:e12. 50 Choron G, Dutheil F, Lesage FX. Are nurses burned out? Int J Nurs
34 Taber KS. The Use of Cronbach’s Alpha When Developing and Stud 2016;58:80–1.
Reporting Research Instruments in Science Education. Res Sci Educ 51 Dutheil F, Boudet G, Perrier C, et al. JOBSTRESS study: comparison
2018;48:1273–96. of heart rate variability in emergency physicians working a 24-­
35 Knell G, Gabriel KP, Businelle MS, et al. Ecological Momentary hour shift or a 14-­hour night shift--a randomized trial. Int J Cardiol
assessment of physical activity: validation study. J Med Internet Res 2012;158:322–5.
2017;19:e253. 52 Dutheil F, Marhar F, Boudet G, et al. Maximal tachycardia and high
36 Nielsen ML, Rugulies R, Smith-­Hansen L, et al. Psychosocial work cardiac strain during night shifts of emergency physicians. Int Arch
environment and registered absence from work: estimating the Occup Environ Health 2017;90:467–80.
etiologic fraction. Am J Ind Med 2006;49:187–96. 53 Bakker AB, Demerouti E. Job demands-­resources theory: taking
37 Rusli BN, Edimansyah BA, Naing L. Working conditions, self-­ stock and looking forward. J Occup Health Psychol 2017;22:273–85.
perceived stress, anxiety, depression and quality of life: a structural 54 Demerouti E, Bakker AB, Nachreiner F, et al. The job demands-­
equation modelling approach. BMC Public Health 2008;8:48. resources model of burnout. J Appl Psychol 2001;86:499–512.

Dutheil F, et al. BMJ Open 2022;12:e046403. doi:10.1136/bmjopen-2020-046403 11