ArticlePDF Available

The Demand-Control-Support Model: Methodological Challenges for Future Research

Authors:
  • Task-Consult

Abstract and Figures

The demand-control-support model was developed by R. Karasek and his colleagues during the 1980s. The model operates with three main dimensions: job demands, job decision latitude and job social support. According to the model, workers with jobs characterized by high demands, low decision latitude and low social support (so-called iso-strain) have a higher risk of poor psychological well-being and cardiovascular diseases (CVD). The model has been tested in relation to CVD in 16 epidemiological studies of which only two have been negative. Moreover, the model has been used in studies with a large number of other endpoints. Recently the model and the methods used in connection with the model have been criticized by several research workers. In this article the studies on iso-strain and CVD are reviewed and a number of methodological problems are discussed. It is recommended that future studies are prospective and use non-representative population samples of well-chosen occupations. Each of the key variables -job stressors, stress and sickness - should be measured in three independent ways in order to increase the methodological and theoretical strength of the research performed.
Content may be subject to copyright.
The distinction between work pace and working
hours in the measurement of quantitative demands
at work
TAGE S. KRISTENSEN
w
, JAKOB B. BJORNER,
KARL B. CHRISTENSEN and VILHELM BORG
National Institute of Occupational Health, Lerso Parkalle 105, DK-2100
Copenhagen O, Denmark
Keywords: Job demands; Differential item functioning (DIF); Psychosocial questionnaire; Methodological
issues; Validity; Work pace; Working hours.
During recent years many researchers have criticized the widely used scales on psychological job
demands. For instance, they comment that in most cases different types of demand seem to be mixed
in one measure. In this paper we analyse the scale on quantitative job demands in the recently
developed Copenhagen Psychosocial Questionnaire (COPSOQ), with special emphasis on
Differential Item Functioning (DIF). DIF refers to basic differences between groups of respondents,
which may affect how they respond to questionnaire items. The data material for our study comprised
a representative sample of Danish employees. The respondents were categorized into 32 specific jobs
according to the International Standard Classification of Occupations (ISCO 1968). We analysed DIF
with respect to the respondents’ jobs with logistic regression analyses. These analyses showed that the
items used in the original demand scale functioned very differently for different jobs in the population.
The conclusion is that scales on quantitative demands are very sensitive to the choice of specific items.
If many items on fast work pace and tempo are included in a scale, a number of blue-collar jobs will be
identified as high-demand jobs. If, on the other hand, many questions on long working hours and
overtime are included, the use of the scale will result in an entirely different picture. This issue has so
far received little attention in occupational health psychology. The results have wide theoretical and
methodological implications for research on quantitative job demands.
1. Introduction
Psychological job demands probably constitute the most important single factor in the field
of psychosocial work environment research. This seems to be the case whether we look at
the theories of person-environment fit, the demand-control model, the burnout research,
or the model of effort-reward imbalance. All theories or models seem to focus on the
balance*/or lack of balance*/between demands at work and something else, be it personal
resources, decision latitude, social support, coping strategies, or rewards. While the
literature on job demands is abundant, there have been very few contributions focusing on
theoretical and methodological issues relating to the concept of psychosocial job demands. This
Work & Stress ISSN 0267-8373 print/ISSN 1464-5335 online #2004 Taylor & Francis Ltd
http://www.tandf.co.uk/journals/
DOI: 10.1080/02678370412331314005
w
Author for correspondence. e-mail: tsk@ami.dk
WORK
&
STRESS
,
OCTOBER
2004,
VOL
. 18,
NO
. 4, 305/322
seems to be one of the major paradoxes of recent research in psychosocial factors at work.
In the present paper we will focus on the concept of quantitative demands and we will base
our discussion of validity and measurement on an analysis of differential item functioning (DIF)
in a newly-developed scale on quantitative demands at work.
1.1. The concept of quantitative demands
In the international literature on psychosocial factors at work the discussion about demands
at work has been going on for some years. One of the most striking features of the whole
literature on job strain/job demands is the almost complete lack of clear conceptual
definitions. In spite of the fact that the job strain model has been the dominating model in
occupational health psychology for about 25 years, the basic question about a clear
definition remains unanswered. In the now classical paper of 1979, Karasek loosely defines
demands as ‘work load demands, conflicts or other stressors which place the individual in a
motivated or energized state of ‘‘stress’’’ (p. 287) and as ‘the psychological stressors involved
in accomplishing the work load, stressors related to unexpected tasks, and stressors of job-
related personal conflict’ (p. 291). In the book by Karasek and Theorell (1990), job
demands are briefly defined as ‘how hard you work’ (p. 63), but the authors also add the
following prudent remark: ‘Indeed, psychological demands on the job remain difficult to
conceptualize and measure because of the diversity of subcomponents and because of some
theoretical problems that are as yet unresolved’ (Karasek & Theorell, 1990, p. 63).
During the last 5/6 years a number of researchers have expressed concern that
the demand concept and the measurement of demands at work need to be refined
and improved (de Jonge & Kompier, 1997; Hallqvist, Diderichsen, Theorell, Reuterwall, &
Ahlbom, 1998; Johnson, Stewart, Hall, Fredlund & Theorell, 1996; Steenland, Johnson, &
Nowlin, 1997; Theorell et al ., 1998; van der Doef & Maes, 1999). Among the points raised
by these and other writers are that the demand concept is poorly defined, that several types
of demands seem to be mixed in one measure, that the same questionnaire items may have
different meanings for different respondents, and that some of the most widely used scales
have low reliability and validity. Most of these comments have been made in relation to
research based on the demand-control model, but the general theoretical and methodo-
logical issues are relevant for other models and measures as well. A number of widely used
scales for measuring (quantitative) job demands are shown in Figure 1.
The basic definition of validity is that we measure what we want to measure . This
immediately makes it clear that it is meaningless to discuss the validity of a scale if it is
unclear what the scale is intended to measure. A number of authors have commented on
this problem in recent papers. A good example is the comments by Steenland et al . (1997)
on a negative study of job strain and heart disease: ‘The variable for job demands may be
measuring something else than what was originally intended */demands for fast-paced
performance. In our data, psychological demands are positively correlated with both
income and education. Occupations with high demand scores may be those which require
more challenging and mentally active work */components of the work process that are
more health enhancing than otherwise’ (Steenland et al ., 1997, p. 259). In that study a job
exposure classification based on the five items of the Job Content Questionnaire (JCQ) was
used. The study showed that high demands were associated with lower risk of heart disease.
The comment by Steenland et al . (1997) is interesting because it states that the JCQ may be
measuring something other (challenging work) than it was originally intended to measure.
According to Steenland et al. (1997) the scale was intended to measure fast-paced work, but
this statement is hardly correct since it assumes that a strict definition exists in the literature.
306 T. S. Kristensen et al.
The point made by Steenland et al . (1997) */that job demands correlate positively with
socio-economic status */is, however, a very important one. A large number of studies have
found the same pattern, most notably the Whitehall II study (Marmot, Bosma,
Hemingway, Brunner, & Stansfeld, 1997). Also, many recent studies on job strain and
cardiovascular diseases (CVD) have been negative with regard to the expected association
A. Alfredson, Karasek & Theorell, 1982:
‘Unemployed at least once during the past 5 years’
‘Working at least 10 h every day’
‘At least one third of total work income from piece wage’
‘Hectic work’
‘Continuously changing day and night work schedule’
‘Rather or very risky work’
B. Karasek, 1985. (The Job Content Questionnaire, 5-item version of the demand scale):
‘My job requires working very fast’
‘My job requires working very hard’
‘I am not asked to do an excessive amount of work’
‘I have enough time to get the job done’
‘I am free from conflicting demands that others make’
C. Johnson & Stewart, 1993; Johnson, Stewart, Hall, Fredlund & Theorell, 1996. (Job exposure
matrix):
‘Is your job psychologically demanding?’
‘Is your job hectic?’
D. Bosma, Marmot, Hemingway, Nicholson, Brunner, & Stansfeld 1997. (The Whitehall II job
demands scale):
‘Do you have to work very fast?’
‘Do you have to work very intensively?’
‘Do you have enough time to do everything?’
‘Do different groups at work demand things from you that you think are hard to combine?’
E. Lynch et al, 1997a, b.
The respondents were asked to rate how much mental strain or stress each of the following things
caused them at work:
‘Excessive supervision of time schedules’
‘Troublesome supervisors’
‘Troublesome fellow workers’
‘Job responsibility’
Poorly defined tasks and responsibilities
Risk of accidents
Risk of unemployment
Irregular work schedules
Mental strain of work
Work deadlines
Physical strain of work
F. van Yperen & Snijders, 2000.
Do you have to work very fast?
Do you have too much work to do?
Do you have to work extra hard to finish a task?
Do you work under time pressure?
Do you have to rush?
Can you do your work in comfort?
Do you have to deal with a backlog at work?
Do you have too little work?
Do you have problems with the pace of work?
Do you have problems with the workload?
Do you wish you could work at an easier pace?
Figure 1. Questions on psychological job demands from a number of scales used in studies on job
demands and health, 1979 /2002.
307Distinction between Work pace and Working hours
between high demands and increased risk of CVD (Kristensen, 1999). The explanation for
both of these empirical findings may be found by taking a closer look at the definition and
measurement of the demand concept, which is the central point of this paper.
1.2. Differential item functioning
The concept of differential item functioning */which is defined and discussed in Section
1.3*/will be the main methodological cornerstone of this paper. This makes it necessary to
explain the methodological and theoretical implications of DIF for analyses of scales such as
the scale for quantitative demands. It will be clear why DIF is important when we consider
the reason for using scales instead of individual items (questions). Let us use a test in
arithmetic as an example: if we want to test peoples’ ability to solve arithmetical problems,
we could choose to use only one item, such as 67
/14/? A test based on only one item
would, however, be quite crude, since it would result in a scale with only two values:
correct and incorrect. The reliability would also be low since there would be a large
element of chance (random error). Therefore we usually choose to have many items in a
test since this will give more reliable and precise results. In principle we can ask about a
large number of possible arithmetical problems, but for practical reasons we only use a very
small fraction of these in a given test. The selection of items from the large item pool of
those potentially possible is a process that is very poorly described by most researchers. This
is a process where the intuition of the researcher and random circumstances play a very
large*/and often ignored*/role. Now, suppose that women are better than men when
adding and subtracting, while men are best when it comes to multiplying and dividing (a
strictly hypothetical example!). This would have the consequence that a test involving
many calculations with addition and subtraction would show that women are superior to
men in the field of arithmetics, while a test involving many calculations with multiplication
and division would show the opposite. That would be a clear case of DIF. Analyses of DIF
have played a major role in psychometric testing in the field of intelligence and entrance
tests since bias against racial groups, women, or disadvantaged socio-economic groups has
been an important issue. For this reason DIF has also been called item bias . In connection
with analyses of scales in the field of psychosocial factors at work, DIF has so far largely been
ignored (a rare exception is Ørhede & Kreiner, 2000).
1.3. A scale for demands at work without differential item functioning?
Quantitative job demands are directly related to the amount of work to be done , and the basic
source of stress is the possible mismatch between the amount of work and the time available
to do it. One of the main research problems seems to be that the mismatch between time
and task demands seems to manifest itself differently in different jobs. For example, an
assembly line worker with fixed working hours may try to respond to increasing demands
by working faster , while his boss will have little to win by talking faster on the phone. He
will probably try to meet high demands by working longer hours. Thus, our main
hypothesis is as follows:
Hypothesis : People in different jobs will have different possibilities for resolving (or trying to resolve)
the mismatch between time and amount of work to be done, and this will lead to differential item
functioning in scales on quantitative demands.
308 T. S. Kristensen et al.
DIF is present in a scale if the response to an item for respondents at a given level on
the scale differs for different groups of respondents (such as, for example, different age
groups, men and women, or employees with different jobs); (Ørhede & Kreiner, 2000;
Swaminathan & Rogers, 1990; Zumbo, 1999). DIF is a sign of low construct validity: a
scale with a high degree of DIF may not measure what it is intended to measure in all parts
of the population. If our hypothesis is confirmed, the conclusion will be that the results of
the research depend on the type of questions that the investigator chooses to ask, not on the
structures and processes that they want to study. It is quite clear that such a conclusion has
far reaching consequences for the job demand literature.
In connection with the development of the Copenhagen Psychosocial Questionnaire
(COPSOQ*/see below), by two authors of the present paper (Kristensen and Borg), one
of our aims was to operationalize the concept of quantitative demands in a ‘pure’ scale. By
this we mean a scale without questions on role conflicts, emotional or other qualitative
demands, and without ‘psychological’ items (such as ‘my work is hectic’). However, in
connection with our use of this new scale, we discovered that we had run into a new and
unanticipated problem: DIF. Thus, our aims in this paper are: (1) to present a new scale for
the measurement of quantitative demands at work and to compare this with other scales
in the literature; (2) to analyse this scale with special attention to the problem of DIF; and
(3) to propose solutions for measuring quantitative demands in future studies in the light of
our results.
2. Method
2.1. Sample
Data for our analyses were obtained through a questionnaire study of a representative
sample of all adult working Danes (age 20 /60 years): The National Danish Psychosocial
Work Environment Study. Names and addresses of a random sample were received from
the national population registry. Respondents (N
/1858) were asked about a broad range
of psychosocial work environment factors. The response rate was 62%; 49% of the
respondents were women (Kristensen, Borg, & Hannertz, 2002).
2.2. Measures of job demands
One of the purposes of the original study was to develop a new questionnaire, the
Copenhagen Psychosocial Questionnaire (COPSOQ; Kristensen, 2002). This question-
naire includes 30 scales on different aspects of the psychosocial work environment, health,
stress, and well-being. Five of the scales measure different demands at work: quantitative
demands (7 items, a
/.80); emotional demands (3 items, a/.87); demands for hiding
emotions (2 items, a
/.59); cognitive demands (8 items, a/.86); and sensory demands (5
items, a/.70). In the present paper we only analyse the scale on quantitative demands.
2.3. Measures of health and well-being
Health, fatigue, and psychological well-being were measured with three scales from the
Short Form 36 (SF-36) questionnaire: General Health (5 items, a
/.75), Mental Health
(5 items, a
/.80), and Vitality (4 items, a/.80) (Ware, Snow, Kosinski, & Gandek, 1993)
and three scales from the Setterlind Stress Profile: Behavioural Stress (8 items, a
/.79),
Somatic Stress (7 items, a/.76), and Cognitive Stress (4 items, a/.85) (Setterlind &
309Distinction between Work pace and Working hours
Larson, 1995). Furthermore, the respondents were asked about their use of painkillers and
sleeping pills during the last 3 months.
2.4. Analyses
DIF has often been analysed with regard to race, gender, socio-economic status, or age
(Cammilli & Shepard, 1994; Holland & Wainer, 1993). In this paper we will concentrate
on DIF with regard to the job of the respondent (in the following referred to as job-DIF).
The 1858 respondents were categorized according to their main job at the time of the
interview. We used the 5-digit coding system of the Danish version of the ISCO 1968
(International Standard Classification of Occupations). On the basis of this classification we
grouped the respondents into 32 fairly homogeneous jobs such as driver, nurse, elementary
school teacher, architect, etc. In this way we succeeded in classifying 1222 persons into
32 jobs with at least 20 respondents, while the remaining 636 had jobs with less than
20 respondents. In the job analyses below we only include the 1222 respondents with the
32 jobs.
All scales in COPSOQ have been given scale values of between 0 and 100. All items in
the demand scales had five response options. (Wording of questions and response options
are given in Table 1.) The items in the scales have been given equal weights and the
response categories have been treated as equal interval categories. High levels on the scales
mean high levels of the property being measured (such as, for example, high level of
vitality, high stress level, or high quantitative demands). If more than half of the items in a
scale were not answered, the person was considered missing on the scale. Hence, we used
the conventions of the SF-36 questionnaire (Ware et al ., 1993).
We used factor analyses and analyses of internal reliability (Cronbach’s a) according to
normal standards and procedures in psychometric research (Nunnally & Bernstein, 1994;
Streiner & Norman, 1998). Furthermore we used multivariate linear and logistic regression
analyses and analyses of rank correlations (Spearman). According to the definition of DIF,
response to a particular item should not depend on the person’s job for respondents
Table 1. Basic characteristics of the 7-item scale on quantitative demands at work (Copenhagen
Psychosocial Questionnaire).
Questions
Average
score
Rwith
total scale
Factor
loading
a. Do you have to work very fast? (‘work fast’) 58.7 .44 .56
b. Is your work unevenly distributed so it piles up? (‘work piles up’) 49.8 .56 .65
c. How often do you not have time to complete all your work tasks?
(‘unfinished tasks’)
34.2 .59 .72
d. Do you get behind with your work? (‘behind’) 30.0 .62 .73
e. How often can you take it easy and still do your work? (‘take it easy’) 52.7 .55 .71
f. Do you have enough time for your work tasks? (‘enough time’) 40.2 .60 .74
g. Do you have to do overtime? (‘overtime’) 44.8 .42 .47
N
/1791.
The response options: ‘Always’, ‘Often’, ‘Sometimes’, ‘Seldom’, ‘Never/hardly ever’ were given the scores of: 100,
75, 50 25 and 0 respectively. Reversed scores for items eand f.
Cronbach’s afor the whole scale: .80.
Inter-item correlations: .22 /.66.
Items aand fare from the Whitehall II Study (Bosma et al. , 1997), b is from the Finnish OSQ (Elo, Leppa
¨nen,
Lindstro
˜in, Roppnen, 1992), cand eare from the Danish Agervold questionnaire (Agervold, 1998), dis from the
Dutch QEEW questionnaire (van Veldhoven & Meijman, 1994), and gis from the QPSNordic (Dallner et al. ,
2000).
310 T. S. Kristensen et al.
experiencing the same level of quantitative demands (i.e. at the same level of the sum
score). We tested this assumption using logistic regression of each item score on jobs where
the sum score was entered as a covariate (Swaminathan & Rogers, 1990; Zumbo, 1999).
The assumption of no DIF is then equivalent to independence between job and item score
(i.e. a regression coefficient of zero) when the sum score is taken into account. For each
item, we did an overall test of DIF across all jobs (entered as a class variable) and we
evaluated the impact of DIF using explained variation (R
2
) as suggested by Zumbo (1999).
For the analyses reported here, we considered an item to exhibit severe DIF if the job
explained more than 5% of the item variation, once the sum score was taken into account.
We also examined DIF for each job, using the beta parameters from the logistic regression.
3. Results
We had originally included eight questions on quantitative demands in our test
questionnaire when we developed the COPSOQ. One of the questions (‘Do you have
too little to do at work?’) was excluded from the scale because of an extremely skewed
response pattern, very low loading on the latent factor, and very low correlation with the
other items. The remaining seven items seemed to function very satisfactorily as one scale
on quantitative demands at work. The basic characteristics of this scale and the origin of the
items are shown in Table 1. This scale lived up to our general ‘rules of thumb’ for scales in
COPSOQ: theoretically meaningful, confirmed by factor analysis with all loadings above
.40, Cronbach’s aabove .70, inter-item correlations between .20 and .70, all item
correlations with total scale above .40, no strong ceiling or floor effects, and less than 5%
missing values on all items. Furthermore, the average scores on the seven items were
different, but still not too close to 0 or 100. The reliability of the whole scale (a
/.80) was
clearly better than for most demand scales in the literature (Karasek et al ., 1998; Kawakami,
Kobayashi, Araki, Haratani, & Furui, 1995; Schwartz, Pieper, & Karasek, 1988).
A more critical look at the scale reveals a few minor problems already at this stage. Items
a(‘work fast’) and g(‘overtime’) clearly have lower correlations (.42 and .44) with the
whole scale than the other five items (.55
/) and the same two items have rather strong
loadings on other demand scales. Item a(‘work fast’) has a loading of .40 on the scale of
sensory demands, while item g(‘overtime’) has a loading of .37 on the cognitive demands
scale. We were aware of these ‘warning signs’ when we constructed the scale but we chose
to keep the two items in the scale. Our main reason was that we wanted to have a broad
scale that would be able to capture many different signs of quantitative demands. We felt
that important indicators of quantitative demand would be missing if these items were
removed from the scale.
Table 2 shows the distribution of the 32 jobs on the scale of quantitative demands. The
job averages go from 26 to 56 points on the 100-point scale. Most of the high-demand jobs
appear to have been white collar jobs (managers, clerks, technicians, architects, etc.). Blue
collar jobs known to have piece rate work or strict time limits such as food industry
workers, cleaners, drivers, and cooks had low or rather low scores. Jobs in the human
service sector also had low scores (nurses, kindergarten teachers, and home helps). Thus, the
7-item scale tended to paint a picture of high demands in a number of office jobs and low
demands in the production sector, the service sector, and the human service sector. This
picture is counter-intuitive and probably has low face validity for most researchers in the
field of psychosocial work environment.
Our next step was to analyse job-DIF with regard to the seven items in the scale. All
items showed significant DIF (all seven p-values were below .001). The most severe DIF
311Distinction between Work pace and Working hours
was seen for item a(‘work fast’) where job explained 6.3% of the variance and item g
(‘overtime’) where job explained 8.3% of the variance. The rest of the items had explained
variance between 2.4% and 4.9%. With regard to jobs we also found a fairly clear pattern:
based on the logistic regression, three of the jobs showed strong DIF on four of the seven
items*/cleaners, food industry workers, and drivers. Figure 2 shows the pattern of DIF for
these three jobs and also for three jobs with a low degree of DIF: mechanics, metal workers,
and nurses.
In Figure 2 the response patterns for the six jobs are compared using the results of the
logistic regression. A complete lack of DIF would result in bars that were close to zero.
Figure 2 shows that cleaners answered very differently (taking the overall score into
account) on the questions of the scale. The analyses showed job-DIF for five of the items
for this group. On items band gthe scores were significantly lower than the overall job
score, while they were higher on items a,e, and f. For the food industry workers there was
significant job-DIF on four items, while for the drivers there was job-DIF on five items.
Thus, for these three occupations the overall score on the scale on quantitative demands
depended heavily on the selection of items. For mechanics, metal workers, and nurses we
see no significant job-DIF.
Table 2. The mean score on the 7-item quantitative demands scale of 32 jobs in the National Danish
Psychosocial Work Environment Study, (Range 0 /100).
Rank Job Score
1 Managers 56.4
2 Systems planners 56.3
3 Secretaries 53.0
4 Bank clerk 52.2
5 High school teachers 51.3
6 Store managers 50.8
7 Warehouse assistants 49.7
8 Head clerks 48.9
9 Book keepers 48.5
10 Technicians 48.0
11 Architects 47.4
12 Office clerks 47.2
13 Electricians 46.3
14 Vocational school teachers 45.5
15 Nurses 45.0
16 Food industry workers 44.9
17 Elementary school teachers 44.6
18 Nurses’ aids 43.7
19 Foremen 43.3
20 Cooks 43.1
21 Home helps 42.5
22 Mechanics 41.8
23 Drivers 40.9
24 Metal workers 40.6
25 Kindergarten teachers 39.9
26 Salesmen 39.8
27 Construction workers 39.3
28 Shop assistants 39.0
29 Agricultural workers 37.1
30 Cleaners 31.9
31 Foster parents 30.2
32 Kindergarten assistants 26.2
All groups: N]
/20.
312 T. S. Kristensen et al.
The issue of job-DIF can also be illustrated by looking at the distribution of the 32 jobs
on combinations of the seven items in the scale. This is illustrated in Figure 3, where we
show two different combinations of items from the scale. The first part of the figure (3a)
shows the distribution of the 32 jobs on items c(‘unfinished tasks’) and d(‘behind’). The
pattern is quite clear: jobs with high average scores on one of these items also have high
average scores on the other (rank correlation
/.84). All combinations of items b,c, and d
gave similar pictures. This corresponds to a situation with low job-DIF. Figure 3(b)
illustrates the opposite situation. In this case the two items are item a(‘work fast’) and item
b(‘work piles up’). Here the rank correlation between job averages was as low as .24.
We also wanted to elucidate possible differences with regard to the potential impact on
health. In Table 3 we show the correlations (adjusted for age and gender) between the
seven items and a number of health-related variables: the three scales from the SF-36
(General Health, Mental Health, and Vitality; Ware et al ., 1993), the three scales from the
Stress Profile (Cognitive Stress, Somatic Stress, and Behavioural Stress; Setterlind & Larson,
1995), and the use of two kinds of medication (pain-killers and sleeping pills). The seven
items showed almost the same pattern with two exceptions: Item g(‘overtime’) was
strongly associated with only one of the scales: behavioural stress. The correlations between
item a(‘work fast’) and the scales were also comparatively low. The 7-item scale was
strongly associated with all six scales measuring health and well-being. In contrast, all
correlations with the use of medication were low (.00 /.06).
Table 4 shows the same combination of variables, but here they are analysed in a
multivariate linear regression analysis where gender and age were included as ‘independent
Cleaners
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
Work fast
Work piles u p
Unfinished tasks
Behind
Tak e i t e a s y
Time enough
Overtime
Beta
** *
***
*
***
Drivers
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
Work fast
Work piles u p
Unfinished tasks
Behind
Tak e i t e a s y
Time enough
Overtime
Beta
**
***
***
***
*
Metal workers
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
Work fast
Work piles u p
Unfinished tasks
Behind
Tak e i t e a s y
Time enough
Overtime
Beta
Mechanics
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
Work fast
Work piles u p
Unfinished tasks
Behind
Tak e i t e a s y
Time enough
Overtime
Beta
* p<.05; ** p<.01; *** p<.001
Food industry worker s
–2 –1.5 –1 –0.5 0 0.5 1 1.5 2
Work fast
Work piles u p
Unfinished tasks
Behind
Tak e i t e a s y
Time enough
Overtime
Beta
**
**
*
*
Nurses
–2 –1.5 –1 0.5 0 0.5 1 1.5 2
Work fast
Work piles u p
Unfinished tasks
Behind
Tak e i t ea sy
Time enough
Overtime
Beta
(a) (b)
Figure 2. Illustration of differential item funtion (DIF) for a number of selected jobs: (a) three jobs
with high job-DIF and (b) three jobs with low job-DIF.
313Distinction between Work pace and Working hours
variables’. Item g(‘overtime’) was associated with Behavioural Stress in the expected
direction, but an inverse association is seen with the scale for Vitality (high vitality is
associated with high levels of overtime). For items a(‘work fast’), c(‘unfinished tasks’), and
e(‘take it easy’) we only see one or no significant associations. Most of the significant
associations in the table are seen for the three items b(‘work piles up’), d(‘behind’), and f
(‘enough time’).
At this point in the analyses it was decided to construct a new and shorter scale with five
items by omitting items aand g(‘work fast’ and ‘overtime’). This decision was based on
five different considerations: (1) Theoretical (see below). (2) Reliability . As indicated above,
items aand ghad the lowest correlations with the total scale, and the overall awas not
reduced by removing the two items in spite of the fact that shorter scales usually have
Behind
Unfinished tasks
r = 0.84
100
100
00
Work piles up
Work fast
r = 0.24
100
100
00
(a) (b)
Figure 3. The distribution of average scores of 32 jobs on two difference combinations of items on
quantitative demands. (a) The distribution of average scores of 32 jobs on the two dimensions:
‘Unfinished tasks’ and ‘Behind’; (b) The distribution of average scores of 32 jobs on the two
dimensions: ‘Work fast’ and ‘Work piles up’.
Table 3. Correlations between the seven items on quantitative demands, the 7-item scale, the 5-
item scale of the COPSOQ and a number of health-related variables. Adjusted for age and gender.
Health-related variables
SF-36 scales Stress scales Use of medication
Items and scales GH MH VT Cognitive Somatic Behavioural
Pain-
killers
Sleeping
pills
a. ‘work fast’
/.06* /.13*** /.11*** .08** .14*** .16*** .03 .04
b. ‘work piles up’
/.08** /.18*** /.16*** .17*** .14*** .24*** .05* .05*
c. ‘unfinished
tasks’
/.05* /.18*** /.16*** .18*** .11*** .19*** .00 .01
d. ‘behind’
/.10*** /.23*** /.17*** .21*** .12*** .22*** .00 .04
e. ‘take it easy’
/.10*** /.20*** /.18*** .14*** .15*** .19*** .03 .01
f. ‘enough time’
/.13*** /.26*** /.28*** .24*** .21*** .26*** .06** .03
g. ‘overtime’ .00
/.08** /.05 .09*** .07** .17*** .01 .05*
7-item scale
/.10*** /.26*** /.23*** .23*** .19*** .30*** .03 .05*
5-item scale
/.12*** /.27*** /.25*** . 25*** .19*** .30*** .03 .04
N]
/1791.
*pB
/.05; **pB/.01; ***pB/.001.
Correlations are Spearman rank correlations.
Abbreviations of the SF-36 scales: GH: General Health. MH: Mental Health. VT: Vitality.
314 T. S. Kristensen et al.
smaller as. (3) Factor analyses . The same two items had the lowest loadings and had
relatively high loadings on other scales. (4) Job-DIF . We found strong job-DIF for items a
and g. (5) Differential association pattern with potential endpoints . We found that items a,c,e,
and gshowed weak or even ‘opposite’ associations with a number of health-related
endpoints. While none of these points is sufficient alone, we felt that the overall evidence
against keeping items aand gin our scale was very convincing.
The new and shorter 5-item scale had a Cronbach’s aof .80 (the same as the longer
scale, in spite of having fewer items), and inter-item correlations between .36 and .66.
The correlations with the total scale varied from .53 to .64. In Table 5 we show the rank of
the 32 jobs on the new 5-item scale and on the two omitted items. Comparisons between
the three columns of the figure show some striking differences. For example, cooks rank as
number 23 on the 5-item scale, as number 1 on item a, and as number 23 on item g. For
food industry workers we find a very similar pattern. Architects rank as number 8 on the
5-item scale, as number 26 on item a, and as number 7 on item g. For other jobs we also
find very different ranks: high school teachers (ranks 5 /25 /2), agricultural workers
(29 /22 /7), home helps (17 /12 /31), head clerks (6 /27 /12), and drivers
(28 /8/5).
We then ran the job-DIF analyses on the new 5-item scale. In these analyses item d
(‘behind’) did not show significant job-DIF when the number of tests was taken into
account (p
/.011, R
2
/.017) while the other items still showed significant job-DIF (all
four test had pB
/.005, R
2
between .021 and .041). Fifteen jobs had no significant regression
coefficients (against nine jobs for the longer scale). Thus, the level of DIF was substantially
reduced but the new and shorter scale was not free from job-DIF.
The new scale showed strong associations with the six scales for health and
psychological well-being (Table 3). Four of these correlations were 1 /2 points higher
with the 5-item scale than with the original 7-item scale, while two of the correlations
remained unchanged. The correlations with use of medication remained practically
unchanged. In a multivariate analysis with the new scale and the two single items as
independent variables together with gender and age, the 5-item scale showed strong
independent associations with the same six scales, while ‘work fast’ was independently
associated with Somatic Stress, and ‘overtime’ with Vitality (inverse), Behavioural Stress,
and the use of sleeping pills (Table 6).
Table 4. Multivariate analyses of associations between the seven items of the quantitative demands
scale of the COPSOQ and a number of health-related variables. Adjusted for age and gender.
Health-related variables
SF-36 scales Stress scales Use of medication
Items GH MH VT Cognitive Somatic Behavioural
Pain-
killers Sleeping pills
a. ‘work fast’ .09***
b. ‘work piles up’
/.05* .07* .08** .29*
c. ‘unfinished tasks’
d. ‘behind’
/.11*** /.06* .12*** .07**
e. ‘take it easy’
/.06*
f. ‘enough time’
/.13*** /.16*** /.27*** .14*** .17*** .16*** .12**
g. ‘overtime’ .06* .05*
N]
/1791.
*pB
/.05; **pB/.01; ***pB/.001.
All parameters in the table are standardized betas.
315Distinction between Work pace and Working hours
4. Discussion
In this section of the paper we will first discuss a couple of general issues related to the
concept and measurement of demands at work. Then we will discuss the results of our own
analyses and continue with an attempt to develop a model for quantitative demands at
work. On the basis of our theoretical and methodological considerations we will then make
a number of recommendations regarding the measurement of demands in future studies.
4.1. The issue of DIF: Different meanings or different conditions?
As stated above, the present paper is, to our knowledge, one of the first systematic papers
analysing job demand scales with regard to DIF (see also Ørhede & Kreiner, 2000). A
number of authors have, however, been discussing the issue of DIF without referring to the
concept. In the 1998 paper on the Job Content Questionnaire (JCQ), Karasek et al . wrote:
‘The variability in the association of the psychological demands scale across samples supports
the interpretation that its meaning may differ by population group’ (Karasek et al., 1998, p.
347), and Theorell et al. (1998) concluded along the same lines in another paper: ‘The
measures of psychological job demands need to be refined. Analyses in the case of
Table 5. The distribution of 32 jobs on the 5-item scale on quantitative demands and on the two
items: a(‘work fast’) and g(‘overtime’).
Points on the
5-item scale
Rank on the
5-item scale
Rank on item
a‘work fast’
Rank on item
g‘overtime’
53.5 Systems planners 1 14 6
50.8 Managers 2 5 1
50.7 Secretaries 3 9 22
47.9 Bank clerk 4 5 11
46.7 High school teachers 5 25 2
46.1 Head clerks 6 27 12
44.2 Book keepers 7 15 17
44.1 Architects 8 26 7
44.0 Office clerks 9 11 25
43.4 Technicians 10 13 9
43.3 Store managers 11 3 3
42.7 Warehouse assistants 12 4 4
41.7 Elementary school teachers 13 22 20
41.1 Electricians 14 15 9
40.8 Vocational school teachers 15 20 12
40.5 Foremen 16 22 24
40.0 Home helps 17 12 31
39.4 Nurses’ aids 18 15 19
39.3 Nurses 19 5 16
37.6 Mechanics 20 28 15
37.1 Food industry workers 21 1 14
37.0 Kindergarten teachers 22 28 27
37.0 Cooks 23 1 23
35.9 Shop assistants 24 19 30
35.8 Metal workers 25 18 21
35.2 Salesmen 26 30 18
33.8 Construction workers 27 10 28
32.0 Drivers 28 8 5
30.8 Agricultural workers 29 22 7
28.4 Cleaners 30 20 32
25.9 Foster parents 31 32 25
21.1 Kindergarten assistants 32 31 29
N]
/20 for all jobs.
316 T. S. Kristensen et al.
psychological demands should perhaps be made separately in different social classes, since
these questions may be perceived differently. Inferred psychological demands are
particularly vulnerable to differences in social class’ (Theorell et al ., 1998, p. 388).
While Karasek and Theorell write about differential meaning and perception by
population groups and social classes, Hallqvist et al . (1998) go one step further:
In our group we have been concerned with the findings of lower prevalence of high demands among
manual workers in cross-sectional surveys, and the main hypothesis has been that the measures are
insensitive to demands experienced by manual workers. Questions such as ‘Did you have enough time
to complete your work tasks?’ could for example be expected to be more apt for individuals who sell
their competence rather than their time, an aspect of the theoretical foundation for separation of blue-
collars and white-collars. The results here point in another direction. The measures of demands may if
anything, be too unspecific with regard to non-manuals. And perhaps it is not only a measurement
problem in that the concept of demands may be too loosely defined to capture only what is really
demanding in this social stratum (Hallqvist et al ., 1998, p. 1414).
We reprint this rather long quote from the paper by Hallqvist et al . because the line of
thought presented here is very close to the main points of this paper: DIF in scales on
demands at work is not only a question of perception and meaning but reflects the basic
structural and material conditions of different classes and occupational groups. The
comments by Hallqvist et al . (1998) have been an important source of inspiration for this
paper.
A temporary conclusion is that the job demand concept seems to be both poorly
defined and measured. A number of the main researchers in the field seem to agree with us
on this issue: ‘Clearly, one practical conclusion of this study is that the basic measurement
of the demand construct should be improved’ (Johnson et al ., 1996, p. 329); ‘Exposure
measurement should be improved, especially for the psychological demand variable’
(Steenland et al., 1997, p. 260), and ‘The measures of psychological job demands need to be
refined’ (Theorell et al ., 1998, p. 388).
4.2. Discussion and integration of the results of the present paper
Two of the main advantages of our study are that the study base is a large heterogeneous
sample, and that so many different demand dimensions were included in the questionnaire.
We would, however, also like to mention two of the weaknesses of the study before we
proceed to the results. First, the job classification was rather crude. We wanted to consider
as many jobs as possible, and we also wanted the number of respondents for each job to be
Table 6. Multivariate analyses of associations between the 5-item scale, two single items and a
number of health-related variables. Adjusted for age and gender.
Health-related variables
SF-36 scales Stress scales Useofmedication
Items and scale GH MH VT Cognitive Somatic Behavioural
Pain-
killers
Sleeping
pills
a. ‘work fast’ .09***
g. ‘overtime’ .07* (inverse) .05* .28*
5-item scale
/.12*** /.25*** /.28*** .25*** .15*** .24***
N]
/1791.
*pB
/.05; **pB/.01; ***pB/.001.
All parameters in the table are standardized betas.
317Distinction between Work pace and Working hours
at least 20 in order to limit the random variation. The consequence was that we formed a
number of job specifications that were rather broad. For instance, ‘drivers’ included bus
drivers, truck drivers, and taxi drivers. A larger study base would have made it possible to
distinguish between such subgroups. We believe that the crude job classification tends to
blur the degree of job-DIF. The other problem is that we only had cross-sectional data.
This means that our analyses in Tables 3, 4 and 6 should not be interpreted as analyses of
causal associations. We bring these tables in order to show the differential picture, not to
demonstrate causality.
The main conclusion of our analyses seems to be the following: The (7-item) scale for
quantitative demands at work was confirmed by analyses of internal reliability and by factor
analyses, but the scale is still unsatisfactory due to job-DIF. Our analyses of job-DIF have
clearly demonstrated that the rank of a job on a scale depends strongly on the selection of
items for the scale. If we had constructed a scale with many questions of the ‘overtime’
type, jobs such as high school teachers, drivers, store managers, and architects would have
received high job demand scores. Had we, on the other hand, constructed a scale with
many questions on ‘fast work pace’, the result would have been quite different. In this case
cooks, food industry workers, and nurses would have received high scores, while high
school teachers and architects would have been close to the bottom. Clearly, these items
cannot measure the same dimension in spite of the fact that we find satisfactorily high
correlations between them at the individual level.
It might be argued that job-DIF would be a minor problem if all the items in the scale
measured job factors with the same health effects. In that case it would be like adding
different types of fruits in a questionnaire on dietary habits. The problem is that we have no
evidence for this. Our cross-sectional analyses in Tables 3 and 4, and 6 seem to indicate that
the health effects are different, and our knowledge of the literature on fast-paced work
versus long working hours seems to suggest the same. For instance, it does make sense that
‘work fast’ is related to somatic symptoms, ‘work piles up’ to the use of sleeping pills, and
‘behind’ and ‘enough time’ to (poor) mental health and cognitive stress (Table 4).
Prospective studies are needed to elucidate this issue further.
When we decided to split the concept of job demands into five different dimensions in
the COPSOQ we thought that we had solved the problem of heterogeneity of the demand
construct. We can now see that we need to refine our measures further. The two main
dimensions of quantitative demands at work seem to be intensity (work pace) and extensity
(number of working hours). We will proceed with this issue in the next paragraph.
4.3. How should quantitative demands be measured?
We think that having two scales for quantitative demands at work could be a possible
solution to the problems discussed in this paper. The first could be a basic scale for
quantitative demands such as the 5-item scale presented in this paper. This scale measures
different aspects of the mismatch between task and time, and it is our impression that it is
valid and has good psychometric qualities. Needless to say, it has to be tested in more
studies, in particular prospective studies. With regard to intensive quantitative demands we
think that our item a(‘work fast’) should be supplemented with other questions on work
pace and tempo. This is a task for future development of the COPSOQ or other
questionnaires. With regard to the other basic dimension*/extensity*/our solution is
different. We believe that the issue of working hours should be elucidated by asking a
number of specific questions and not by forming a scale.
318 T. S. Kristensen et al.
Pressure for
increased
productivity
Mismatch between
the amount of work
to be done and the
time available
to do it
Pressure for
Faster work pace
(Intensification)
Longer working hours
(Extensification)
Scale for intensive
quantitative demands
Basic scale for
quantitative demands at work
Questions on formal
and actual working hours
Processes at the
(global) labour
market
Processes at the workplace and job level Measures of quantitative
job demands
Figure 4. Model for the measurement of quantitative job demands.
319Distinction between Work pace and Working hours
With regard to the problems of measuring demands, the ‘three-factor solution’ that we
are suggesting is illustrated in Figure 4: the pressure for increased productivity leads to a
mismatch between the amount of work and the time available to do it. This mismatch leads
to pressure for both longer working hours and faster work pace. These two ways of
increasing productivity will be applied differently in different segments of the labour market
depending on laws, agreements, wage structure, professional norms, etc. Consequently,
quantitative demands could be measured in three ways: a scale for intensive demands, a scale
for the basic mismatch between task and time , and specific questions on working hours . We are not
claiming that this solution solves all the problems with job-DIF, but we do think that our
suggestion represents an improvement relative to the scales that are being used today (such
as those shown in Figure 1). Most of these scales include a mixture of intensity, extensity,
and items that measure mismatch between task and time. For instance, the JCQ (scale B in
Figure 1) includes an item on intensity (‘working very fast’), an item on extensity
(‘excessive amount of work’), and an item on mismatch (‘enough time to get the
job done’). Furthermore, this scale also has an item (‘working very hard’), which
seems to refer to physical as well as psychological strain, and an item on role conflict
(‘conflicting demands’). The other scales shown in Figure 1 also appear to contain
internal heterogeneity, which is a strong indicator of potential problems with DIF in these
scales.
We also recommend that analyses of DIF in the important scales of a study should
always be included in research relying on questionnaire scales. If DIF is present */with
regard to age, gender, class or occupation */knowledge about it will always give more
insight than overlooking this important information.
5. Conclusions
In this paper we have been discussing one of the main dimensions of modern research
on psychosocial working conditions: quantitative job demands. This dimension has been in
focus for at least 50 years of research and it is our belief that job conditions in the global
economy will be characterized by increasing quantitative job demands for many years to
come. Hence, the issue of quantitative job demands is important and relevant.
One of our main conclusions is that (quantitative) job demands have been poorly
conceptualized, defined and measured in the literature so far. In most cases researchers have
used demand scales with poor quality and unknown validity. Most researchers have been
satisfied with using the same scales as others have used. Very few have taken a close look at
the very items of the scales employed.
A number of studies on job strain and cardiovascular disease and other studies on
job demands and health, have shown negative or contradictory results during the last
10/15 years. We believe that one of the main explanations may be the use of job demand
measures with low validity. These measures have to a large degree measured white collar
demands associated with job conditions of higher socio-economic classes. This type of
measurement bias has been mentioned by a few authors but the proper ways of analysing
the issue and suggestions for solving the problem have been lacking. In this paper we have
used analyses of differential item functioning in order to elucidate the issue and we have
suggested the use of refined scales in the hope of improving future research in this
important field.
320 T. S. Kristensen et al.
References
A
GERVOLD
, M. (1998). Spørgeskema til psykosocialt arbejdsmiljø */kortlægning og ændring. [Questionnaire
for psychosocial work environment */survey and change.] Copenhagen: Arbejdsmiljøfondet.
A
LFREDSON,
L
.
,K
ARASEK,
R
.
,&T
HEORELL,
T
.
(1982). Myocardial infarction risk and psychosocial
work environment: An analysis of the male Swedish working force. Social Science and Medicine ,
16, 463 /467.
B
OSMA,
H
.
,M
ARMOT,
M
.
G
.
,H
EMINGWAY,
H
.
,N
ICHOLSON,
A
.
C
.
,B
RUNNER,
E
.
,&S
TANSFELD,
S
.
A
.
(1997). Low job control and risk of coronary heart disease in Whitehall II (prospective
cohort) study. British Medical Journal ,314 , 558 /565.
C
AMMILLI,
G
.
&S
HEPARD,
L
.
A
.
(1994). Methods for Identifying Biased Test Items . Thousand Oaks:
Sage.
D
ALLNER,
M
.
,E
LO,
A
.-
L
.
,G
AMBERALE,
F
.
,H
OTTINEN,
V
.
,K
NARDAHL,
S
.
,L
INDSTRO
¨M,
K
.
,
S
KOGSTAD,
A
.
,&Ø
RHEDE,
E
.
(2000). Validation of the General Nordic Questionnaire (QPSNordic)
for Psychological and Social Factors at Work. Nord 2000:12 . Copenhagen: Nordic Council of
Ministers.
DE
J
ONGE,
J
.
&K
OMPIER,
M
.
A
.
J
.
(1997). A critical examination of the demand-control-support
model from a work psychological perspective. International Journal of Stress Management ,4,
235/258.
E
LO,
A
.-
L
.
,L
EPPA
¨NEN,
A
.
,L
INDSTRO
¨M,
K
.
,&R
OPPONEN,
T
.
(1992). OSQ. Occupational Stress
Questionnaire: User’s Instructions . Helsinki: Finnish Institute of Occupational Health.
H
ALLQVIST,
J
.
,D
IDERICHSEN,
F
.
,T
HEORELL,
T
.
,R
EUTERWALL,
C
.
,A
HLBOM,
A
.
,&S
HEEP
S
TUDY
G
ROUP.
(1998). Is the effect of job strain on myocardial infarction risk due to interaction
between high psychological demands and low decision latitude? Results from Stockholm Heart
Epidemiology Program (SHEEP). Social Science and Medicine ,46 , 1405 /1415.
H
OLLAND
,P.W.&W
AINER
, H. (Eds.). (1993). Differential Item Functioning . Hillsdale, NJ: Lawrence
Erlbaum.
J
OHNSON,
J
.
V
.
&S
TEWART,
W
.
F
.
(1993). Measuring work organization exposure over the life
course with a job-exposure matrix. Scandinavian Journal of Work, Environment & Health ,19 ,
21/28.
J
OHNSON,
J
.
V
.
,S
TEWART,
W
.
,H
ALL,
E
.
M
.
,F
REDLUND,
P
.
,&T
HEORELL,
T
.
(1996). Long-term
psychosocial work environment and cardiovascular mortality among Swedish men. American
Journal of Public Health ,86 , 324 /331.
K
ARASEK,
R
.
A
.
(1979). Job demands, job decision latitude, and mental strain: Implications for job
redesign. Administrative Science Quarterly ,24 , 285 /308.
K
ARASEK,
R
.
A
.
(1985). Job Content Questionnaire and User’s Guide . Lowell: Department of Work
Environment, University of Massachusetts Lowell.
K
ARASEK,
R
.
A
.
,B
RISSON,
C
.
,K
AWAKAMI,
N
.
,H
OUTMAN,
I
.
,B
ONGERS,
P
.
,&A
MICK,
B
.
(1998).
The Job Content Questionnaire (JCQ): An instrument for internationally comparative
assessments of psychosocial job characteristics. Journal of Occupational Health Psychology ,3,
322/355.
K
ARASEK,
R
.
&T
HEORELL,
T
.
(1990). Healthy Work. Stress, Productivity, and the Reconstruction of
Working Life. New York: Basic Books.
K
AWAKAMI,
N
.
,K
OBAYASHI,
F
.
,A
RAKI,
S
.
,H
ARATANI,
T
.
,&F
URUI,
H
.
(1995). Assessment of job
stress dimensions based on the job demands-control model of employees of telecommunication
and electric power companies in Japan: Reliability and validity of the Japanese version of the Job
Content Questionnaire. International Journal of Behavioral Medicine ,2, 358 /375.
K
RISTENSEN,
T
.
S
.
(1999). Challenges for research and prevention in relation to work and
cardiovascular disease. Scandinavian Journal of Work, Environment & Health ,25 , 550 /557.
K
RISTENSEN,
T
.
S
.
(2002). A new tool for assessing psychosocial factors at work: The Copenhagen
Psychosocial Questionnaire. TUTB Newsletter [Newsletter of the European Trade Union
Technical Bureau for Safety and Health] , 19 /20(September), 45 /47.
K
RISTENSEN,
T
.
S
.
,B
ORG,
V
.
,&H
ANNERTZ,
H
.
(2002). Socioeconomic status and psychosocial
work environment. Results from a national Danish study. Scandinavian Journal of Public Health ,
30,41/48.
L
YNCH,
J
.
,K
RAUSE,
N
.
,K
APLAN,
G
.
A
.
,S
ALONEN,
R
.
,&S
ALONEN,
J
.
T
.
(1997a). Workplace
demands, economic rewards, and progression of carotid atherosclerosis. Circulation ,96 ,
302/307.
321Distinction between Work pace and Working hours
L
YNCH,
J
.
,K
RAUSE,
N
.
,K
APLAN,
G
.
A
.
,T
UOMILEHTO,
J
.
,&S
ALONEN,
J
.
T
.
(1997b). Workplace
conditions, socioeconomic status, and the risk of mortality and acute myocardial infarction: The
Kuopio Ischemic Heart Disease Risk Factor Study. American Journal of Public Health ,87 , 617/
622.
M
ARMOT,
M
.
G
.
,B
OSMA,
H
.
,H
EMINGWAY,
H
.
,B
RUNNER,
E
.
,&S
TANSFELD,
S
.
(1997).
Contribution of job control and other risk factors to social variations in coronary heart disease
incidence. Lancet,350 , 235 /239.
N
UNNALLY,
J
.
C
.
&B
ERNSTEIN,
I
.
H
.
(1994). Psychometric Theory . New York: McGraw /Hill.
Ø
RHEDE,
E
.
&K
REINER,
S
.
(2000). Item bias in indices measuring psychosocial work environment
and health. Scandinavian Journal of Work, Environment and Health ,26 , 263 /272.
S
CHWARTZ,
J
.
E
.
,P
IEPER,
C
.
F
.
,&K
ARASEK,
R
.
A
.
(1988). A procedure for linking psychosocial job
characteristics data to health surveys. American Journal of Public Health ,78 , 904 /909.
S
ETTERLIND,
S
.
&L
ARSON,
G
.
(1995). The Stress Profile: A psychosocial approach to measuring
stress. Stress Medicine ,11,85/92.
S
TEENLAND,
K
.
,J
OHNSON,
J
.
,&N
OWLIN,
S
.
(1997). A follow-up study of job strain and heart disease
among males in the NHANES1 population. American Journal of Industrial Medicine ,31 , 256 /260.
S
TREINER,
D
.
L
.
&N
ORMAN,
G
.
R
.
(1998). Health Measurement Scales. A Practical Guide to their
Development and Use, 3rd edn. Oxford: Oxford University Press.
S
WAMINATHAN,
H
.
&R
OGERS,
J
.
H
.
(1990). Detecting differential item functioning using logistic
regression procedures. Journal of Educational Measurement ,27 , 361 /370.
T
HEORELL,
T
.
,T
SUTSUMI,
A
.
,H
ALLQVIST,
J
.
,R
EUTERWALL,
C
.
,H
OGSTEDT,
C
.
,F
REDLUND,
P
.
,
E
MLUND,
N
.
,J
OHNSON,
J
.
V
.
,&
THE
SHEEP S
TUDY
G
ROUP
(1998). Decision latitude, job
strain, and myocardial infarction: A study of working men in Stockholm. American Journal of
Public Health ,88, 382 /388.
VAN DER
D
OEF,
M
.
&M
AES,
S
.
(1999). The job demand-control (-support) model and psychological
well-being: A review of 20 years of empirical research. Work & Stress,13 ,87/114.
VAN
V
ELDHOVEN,
M
.
&M
EIJMAN,
T
.
F
.
(1994). Questionnaire on the Experience and Assessment of
Work. Amsterdam: The Foundation for Quality in Occupational Health Care.
VAN
Y
PEREN,
N
.
W
.
&S
NIJDERS,
T
.
A
.
B
.
(2000). A multilevel analysis of the demand-control
model: Is stress at work determined by factors at the group level or the individual level? Journal of
Occupational Health Psychology ,5, 182 /190.
W
ARE,
J
.
E
.
,S
NOW,
K
.
K
.
,K
OSINSKI,
M
.
,&G
ANDEK,
B
.
(1993). SF-36 Health Survey. Manual and
Interpretation Guide . Boston: New England Medical Center, The Health Institute.
Z
UMBO,
B
.
D
.
(1999). Handbook on the Theory and Methods of Differential Item Functioning (DIF): Logistic
Regression Modeling as a Unitary Framework for Binary and Likert-type (Ordinal) Item Scores. Ottawa,
ON: Department of National Defense, Directorate of Human Resources and Evaluation.
322 T. S. Kristensen et al.
... The new third dimension of the model has been the job social support. According to the extended model, the demand-control-support model, the highest risk of ill-health is to be expected in the iso-strain group with high demands, low decision latitude and low social support (Kristensen, 1995). High Strain X B ...
... The Demand-Control-Support model provides a large body of literature, both theoretical and empirical findings that strongly suggest a causal association between job strain and health, especially job strain and CVD (Karasek & Theorell, 1990). However, this model receives theoretical and methodological criticism (for a review see Kristensen, 1995;van der Doef & Maes, 1998) that points out the need for more research and empirical studies. ...
Thesis
p>Increases in women's labour force-participation, including that of women with children, have led researchers to study the health effects of women's multiple roles. Recent investigations have shown that the specific qualities of women's social roles as well as other psychosocial variables such as social support, distribution of household responsibilities and work-family relationship can affect women's health. The present study aims to explore the relationships between sociodemographic and psychosocial variables on self-perception of physical and mental health in Venezuelan working women at different occupational levels. Four studies were carried out using different methodologies, designs and testing additive and interactive models. Study 1 considered only secretaries (n=122), Study 2 included working women from different occupational levels and a group of housewives (n=417). Both studies were cross-sectional. In order to test the potential causal influence of psychosocial variables on self-reported physical and mental health, a longitudinal study was carried out (n=130). To enrich the results obtained from the three quantitative studies, the final study was qualitative (n=32). Only the level of education and having pre-school children turned out to be health predictors. In general, women with partners reported better health than women without partners. The characteristics of women's social roles were the most important psychosocial variables considering the main and interactive effects that either put a strain on, or enhance women's health and well-being. Supervisor and co-workers support as well as the perception of job control were the most important protective aspects. Social relations at work played a fundamental role on mental health, moderating important work-related stressing conditions such as dissatisfaction with the salary/lack of recognition and promotion. Marital satisfaction was directly related to women's perception of health. Job control and social integration protected working women from stress related to marital conflict. None of the psychosocial factors studied for the mother role aced as protectors against the effects of this role on women's health. The results showed that integral approaches, which simultaneously consider work and family spheres, are required for the comprehension of working women's health.</p
... While many work-related characteristics had a bivariate association with the start of bullying, the prediction model based on these variables did not reach a satisfactory predictive power (C-index >0.7) for individualized risk assessment. This may be due to the items being associated with a shared (latent) phenomenon, such as occupationspecific work characteristics (Kristensen, 1995). As bullying is more common within the health care sector (Zapf et al., 2003), the prediction model consists of characteristics that describe work in the health care sector, such as high proportion of women, repeated tasks, and shift work, whereas work characteristics linked with white-collar work, such as working very hard is seen as a protective factor. ...
Article
Full-text available
Aim To determine the extent to which change in (i.e., start and end of) workplace bullying can be predicted by employee responses to standard workplace surveys. Methods Responses to an 87-item survey from 48,537 Finnish public sector employees at T1 (2017–2018) and T2 (2019–2020) were analyzed with least-absolute-shrinkage-and-selection-operator (LASSO) regression. The predictors were modelled both at the individual- and the work unit level. Outcomes included both the start and the end of bullying. Predictive performance was evaluated with C-indices and density plots. Results The model with best predictive ability predicted the start of bullying with individual-level predictors, had a C-index of 0.68 and included 25 variables, of which 6 remained in a more parsimonious model: discrimination at work unit, unreasonably high workload, threat that some work tasks will be terminated, working in a work unit where everyone did not feel they are understood and accepted, having a supervisor who was not highly trusted, and a shorter time in current position. Other models performed even worse, either from the point of view of predictive performance, or practical useability. Discussion While many bivariate associations between socioeconomic characteristics, work characteristics, leadership, team climate, and job satisfaction were observed, reliable individualized detection of individuals at risk of becoming bullied at workplace was not successful. The predictive performance of the developed risk scores was suboptimal, and we do not recommend their use as an individual-level risk prediction tool. However, they might be useful tool to inform decision-making when planning the contents of interventions to prevent bullying at an organizational level.
... En effet, Niedhammer, Chastang, Gendrey et collègues (2006b) ajoutent que ce modèle ne considère pas tous les facteurs de risque du stress professionnel. De même, d'après Kristensen (1995), ce modèle simple et trop général ne pourrait pas être adapté aux diverses situations de travail. Selon Lacomblez et Vézina (2008, p. 3) : ...
Thesis
Full-text available
In Canada, 27% of workers consider most of their days extremely stressful. The consequences of this stress are heavy for individuals and organisations. The strong growth (nearly 400%) in requests for assistance addressed to the PAMBA in recent years demonstrates that the legal profession in Quebec is no exception to this trend. In this respect, lawyers are three times more likely to suffer from depression compared to the rest of the employed population. Faced with this alarming portrait, this doctoral thesis aims to answer the following main question: what role do the field of practice and the sector of activity play in the explanation of psychological distress at work (PDW) among Quebec lawyers? Two main objectives are pursued 1) to validate an abbreviated scale of PDW; and 2) identify the specific contribution of the field of practice and the sector of activity on PDW among Quebec lawyers, using a multidimensional approach. To do this, factorial analyses and hierarchical multiple regressions are carried out, from a secondary database (2086 participants) collected via a self-reported questionnaire comprising 44 key variables allowing a relevant conceptual coverage to the study of PDW in the context studied. Overall, the results confirm the importance of considering determinants coming from several spheres of the individual’s life when analyzing PDW and support the theoretical model adopted. Contrary to the literature, our results show that most of the variance of PDW is explained by working conditions. Also, it appears that the effect of the latter on the PDW differs according to the field of practice in which the lawyers work. Finally, it seems that each field of practice is exposed to different risks arising from the working conditions that are specific to each of these fields, and which are likely to play a role in developing PDW.
... Research indicates that a work-life imbalance can drive employees to burnout and a deteriorating mental and physical health, such as increased instances of cardiovascular diseases (Kristensen,1996), autoimmune diseases (Friedlander, 2008), musculoskeletal disorders, diabetes, psychiatric illness, gastro-intestinal illness, alcohol-related diseases, absenteeism from work, (over)use of medications, sleeping problems, and reproductive problems (Kristensen, 1995). Also, the work-life imbalance decreases individuals' wellbeing and leads them to substance abuse, cynicism, and suicidal ideation (Mavor et al., 2014;Bisschoff et al., 2019). ...
Thesis
Full-text available
This thesis explores interpretations of Turkish Islamic bankers with regard to the Islamic Banking phenomenon in Turkey. It attempts to understand religion's impact on financial institutions in special banks. The case of Turkey distinguishes it from other countries in two ways: first, Turkey is a secular country; it is ruled by secular laws and has never been governed in line with religious law. Second, even though the majority of the population claims to be Muslim, at around 99 per cent, the share of Islamic banking is limited just to 5 per cent of the total banking sector. The research leans towards an interpretivist approach. It adopts case study as a methodology and benefitted from the use of a qualitative approach. Research material was collected via semi-structured interviews, observation, and documental analysis. Thirty-one Islamic banking professionals participated in interviews in five different major Turkish cities. The data collected during the field study was analysed through thematic analysis and the results represented in three separate chapters. The study revealed that there appear to be three levels-the macro, meso, and micro-in the Islamic banking phenomenon. Institutional theory helped to understand the business environment of the phenomenon at the macro-level. This included legitimacy, institutional logics, isomorphism, and other concepts underlying the challenges to the Turkish Islamic banking sector. It also covered future projections of Islamic banking. The meso-level unveiled organisational explorations of corporate culture, HRM, and miscellaneous issues. Finally, the micro-level was concerned with the individual identities of Islamic bankers. The findings explained according to themes such as belongingness, spirituality, and work-life balance. ii "Count no man happy until he is dead" Solon (c.630-c.560 BC) iii
... It is assumed in the DCS model that a combination of high job demand and low job control leads to physical and psychological strain. This is the worst case scenario for employees who have high job demand but little job control, as well as a lack of social support [37]. The findings revealed that the greater the job demands and demands for nurses in the workplace, as well as the less support they receive, the greater the pressure and stress they face. ...
Article
Full-text available
Background Nowadays, counter-productive work behaviors (CWBs) have turned into a common and costly position for many organizations and especially health centers. Therefore the study was carried out to examine and compare the demand-control-support and effort-reward imbalance models as predictors of CWBs. Materials and Methods The study was cross-sectional. The population was all nurses working in public hospitals in ....., ..... of whom 320 were selected as the sample based on simple random sampling method. The instruments used were Job Content Questionnaire (JCQ), effort-reward imbalance questionnaire and counter-productivity work behaviors questionnaire. Data was analyzed using correlation and regression analysis in SPSS18. Results The findings indicated that both ERI and DCS models could predict CWB (P≤0.05); however, the DCS model variables can explain the variance of CWB-I and CWB-O approximately 8% more than the ERI model variables and have more power in predicting these behaviors in the nursing community. Conclusion According to the results, job stress is a key factor in the incidence of CWBs among nurses. Considering the importance and impact of each component of ERI and DCS models in the occurrence of CWBs, Corrective actions can be taken to reduce their incidence in nurses.
... For one, Karasek highlights his findings' implications for job redesign, arguing that productivity and 'job-related mental health' (1979, p. 303) could be increased through job redesign, that is, by increasing workers' autonomy to decide what tasks to do and how to accomplish them (1979,1989). Furthermore, his 'Demand-Control Model' has been extended into a 'Demand-Control (Support) Model', which argues that social support from colleagues may significantly help reduce stress (e.g., Dawson et al., 2016;Jalilian et al., 2019;Johnson & Hall, 1988;Karasek & Theorell, 1990;Kristensen, 1995;Luchman & Gonzales-Morales, 2013;Theorell, 2020). ...
Article
Full-text available
The worldwide spread of work‐related mental unhealth suggests that this is a major problem affecting organizations and employees on a global scale. In this paper, we therefore provide a thematic review of the literatures that address this issue in management and organization studies (MOS) and related fields. While these literatures examine how employee mental health is affected by organizational and occupational structures and managed by organizations and employees, they have paid relatively little attention to the capitalist labour relations which underpin the unhealthy conditions of contemporary working life. They have paid even less attention to how these conditions may be resisted. To help future scholarship in MOS challenge this state of affairs, we draw on some of the most basic but central notions of exploitation, alienation and resistance in classic and current critiques of capitalism, optimistic that this may help strengthen the field's capacity to confront mental unhealth in settings of work and organization.
... En considérant les deux premiers axes de Karasek, la latitude décisionnelle et la demande psychologique, il est apparu que 70,40% de la population d´étude se trouve dans une situation de job strain contre 21,52% en situation de passivité ( Figure 1). Le modèle de Karasek présente deux diagonales: un axe de tension ou de pression, qui s´échelonne du travail détendu au travail sous pression et un axe d´apprentissage allant du travail passif au travail actif [27]. La Figure 1 décrit la situation des salariés des entreprises sur ces diagonales. ...
Article
Introduction: in Burkina Faso, work-related stress is a public health problem. The purpose of this study is to evaluate factors of stress among formal private sector employees in the city of Ouagadougou. Methods: we conducted a survey using the 26-item scale derived from Karasek and the 23-item scale derived from Siegrist. Validated Karasek and Siegrist´s models as well as SPSS software were used to analyze data. Results: we surveyed 223 employees (186 men and 37 women) with an average age of 36.70 years ± δ = 33.25. In addition, 70,40% of employees had job strain; 50,22% iso strain and 52,02% effort-reward imbalance. Post-hoc analyses showed the following stress factors: great efforts and poor decision-making ability. Conclusion: this study confirms the presence of stress among private sector employees and highlights the importance of combining Karasek and Siegrist's questionnaires in the study of stress factors.
Article
Full-text available
Objetivo: Identificar la relación entre Factores de Riesgo Psicosocial y Satisfacción Laboral en el personal administrativo de una universidad privada. Material y Métodos: Diseño metodológico no experimental de corte transversal y de tipo correlacional. Se utilizó el cuestionario SUSESO-ISTAS 21 versión breve y el cuestionario de satisfacción laboral S20/23, dirigido al personal administrativo de una universidad privada. Para evaluar la correlación entre las variables mencionadas se calculó el coeficiente de correlación de Rho de Spearman. Resultados: Se evidencian que los riesgos psicosociales con riesgo alto fueron el apoyo social en la empresa y calidad de liderazgo (54%), seguido de la doble presencia (46%) y las compensaciones (42%). Se encontraron relaciones negativas significativas entre los resultados totales de las dos variables (p<0,05), excepto entre los riesgos psicosociales en la dimensión de doble presencia y las dimensiones de satisfacción con el trabajo (excepto la satisfacción con la supervisión) y el trabajo activo con la satisfacción con los servicios recibidos (p>0,05). Conclusiones: Existe relación negativa significativa entre la mayoría de los factores de riesgos psicosociales y la satisfacción laboral en el personal administrativo. Esto significa que a medida que se intensifican los riesgos psicosociales se reduce la satisfacción laboral en los trabajadores por lo que se recomienda establecer protocolos que definan claramente los roles y funciones de los trabajadores administrativos para evitar la sobre exigencia en sectores focalizados.
Article
Full-text available
The international evidence base shows that the mass communication industry is characterized by elevated levels of employee burnout while there is scant research addressing parallel situations in Africa and more specifically Ethiopia. Accordingly, the objective, of the present study was to determine the prevalence and magnitude of burnout among journalists working for Amhara TV, Oromia Broadcasting Network and Tigray TV using the Maslachburnout inventory and a sample of 123 journalists. The aim was also to determine the contribution of job control, job demands, organizational support and biographical factors to burnout using hierarchical regression methods. A two-step hierarchical regression procedure was employed to determine the best fit to the data. Results indicated that biographical predictors failed to qualify as significant explanatory factors, but the model explained an improved 32 % of the variance in burnout. ANOVA tests showed no significant scores in burnout were attributable to gender, job level, age or work experience differences except for region. Implications are discussed.
Article
Full-text available
Background: Little documentation exists on relationships between long-term residential care facilities (LTRCFs), staff working conditions and residents' quality of care (QoC). Supporting evidence is weak because most studies examining this employ cross-sectional designs. Methods: Systematic searches of twelve bibliographic databases sought experimental and longitudinal studies, published up to May 2021, focusing on LTRCF nursing staff's working conditions and the QoC they provided to older adults. Results: Of the 3577 articles identified, 159 were read entirely, and 11 were retained for inclusion. Higher nursing staff hours worked per resident per day (HPRD) were associated with significant reductions in pressure sores and urinary tract infections. Overall staff qualification levels and numbers of RNs had significant positive influences on QoC. Conclusions: To the best of our knowledge, this systematic review is the first to combine cohort studies with a quasi-experimental study to explore associations between LTRCF nursing staff's working conditions and older adult residents' QoC. Human factors (including HPRD, staff turnover, skill mix, staff ratios) and the specific working contribution of RNs had overwhelmingly significant influences on QoC. It seems essential that LTRCF supervisory and decision-making bodies should promote optimal working conditions for nursing staff because these have such a direct impact on residents' QoC.
Article
Full-text available
objective and subjective measures of stress at work discuss the consequences of using longitudinal designs problems of longitudinal studies / different models of the time course of cause and effect / different kinds of stressors and of dysfunctioning and the time course of the models / person and environmental parameters and their mediating and moderating effect on the time course and the models problem of small correlations in stress research (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Full-text available
Karasek (1979) drew attention to the possibilities that job characteristics may be non-linearly associated with employee well-being, and that they may combine interactively in relation to well-being. This paper examines those issues, and finds that both linear and non-linear components are present in relationships between job features and well-being. However, there is no evidence for a synergistic interaction between decision latitude and job demands. Those job features are differentially predictive of two aspects of well-being: job-related depression-enthusiasm and anxiety-contentment.
Book
Clinicians and those in health sciences are frequently called upon to measure subjective states such as attitudes, feelings, quality of life, educational achievement and aptitude, and learning style in their patients. This fifth edition of Health Measurement Scales enables these groups to both develop scales to measure non-tangible health outcomes, and better evaluate and differentiate between existing tools. Health Measurement Scales is the ultimate guide to developing and validating measurement scales that are to be used in the health sciences. The book covers how the individual items are developed; various biases that can affect responses (e.g. social desirability, yea-saying, framing); various response options; how to select the best items in the set; how to combine them into a scale; and finally how to determine the reliability and validity of the scale. It concludes with a discussion of ethical issues that may be encountered, and guidelines for reporting the results of the scale development process. Appendices include a comprehensive guide to finding existing scales, and a brief introduction to exploratory and confirmatory factor analysis, making this book a must-read for any practitioner dealing with this kind of data.
Article
A study was designed to test the association between job demands and job decision latitude and coronary heart disease (CHD) risk and psychological strain among working men and women in the FRG. The data are from the 1984 FRG national health survey. The sample used for this research was composed of 795 persons: 476 working men and 319 working women. Following the work of Karasek, the results indicate that psychological strain was related to high job demands and low job decision latitude. However, an association between CHD risk and high job demands and low job decision latitude was not supported by the findings. Correlational analyses revealed moderate correlations between psychological strain and CHD risk. Multiple regression analyses indicated that low decision latitude may be a stressor that grows more severe in the presence of high job demands in terms of negative health outcomes. Significant differences between men and women were also found.
Book
This book makes clear to researchers what item-bias methods can (and cannot) do, how they work and how they should be interpreted. Advice is provided on the most useful methods for particular test situations. The authors explain the logic of each method - from item-response theory to nonparametric, categorical methods - in terms of how differential item functioning (DIF) is defined by the method and how well the method can be expected to work. A summary of findings on the behaviour of indices in empirical studies is included. The book concludes with a set of principles for deciding when DIF should be interpreted as evidence of bias.
Article
The test-item bias literature is summarized, emphasizing the conceptual basis for bias detection methods and the technical issues involved in choosing among methods. It describes both judgmental and statistical methods for identifying biased items, and discusses the reconciliation of these two types of evidence. (Author/BW)
Article
The possible physiological significance of individual control over environmental/occupational stressors has not been discussed systematically in relation to cardiovascular disease. In this paper we review associations between stressors and control in the work situation and endocrine, metabolic, and cardiovasclar activity, using the categories “anabolic” and “catabolic” to organize the findings in relation to the development of ischemic cardiovascular disease. We propose a comprehensive hypothesis, related to existing stress models which incorporates potential physiological regenerative or ameliorating—as well as pathologic—effects of stressor exposure. Regenerative processes are hypothesized to occur in circumstances where equilibrium or a match exists between environmental demands and possibilities for individual control over those demands, whereas pathologic consequences would occur when demands exceed control. This theory is further proposed to account for cardiovascular pathophysiologic differences found between occupational groups. Finally, empirical testing of the validity of this theory is discussed.
Article
health risk / employment status / occupational stressors / work environment (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
The Stress Profile is a psychosocial instrument for measuring stress in life in general and at work at the levels of the individual, the group and the organization. It has been tested and standardized on more than 4000 men and women. The present article outlines the design and the developmental stages of the Profile. It also describes its use at individual and company levels. The impact of the scientific development of behavioural medicine has greatly improved and extended the application of behavioural methods. The Stress Profile is based on this methodological and scientific development. It is a questionnaire consisting of 224 questions, 20 of which concern background variables and ten criteria. The remaining questions map a number of areas derived from current stress research. The data are computer processed and a graphic profile is produced. This presentation includes illustrations of results from Swedish companies, where the Stress Profile has been used.