Decision Authority

Int Arch Occup Environ Health (2013) 86:591605
DOI 10.1007/s00420-012-0791-3
ORIGINAL ARTICLE
Demands, skill discretion, decision authority and social climate

at work as determinants of major depression in a 3-year follow-up
study
Andres Fandino-Losada Yvonne Forsell
Ingvar Lundberg
Received: 13 July 2011 / Accepted: 18 June 2012 / Published online: 4 July 2012
Springer-Verlag 2012
Abstract
Purpose The psychosocial work environment may be a
determinant of the development and course of depressive
disorders, but the literature shows inconsistent findings.
Thus, the aim of this study is to determine longitudinal
effects of the job demandscontrolsupport model
(JDCSM) variables on the occurrence of major depression
among working men and women from the general
population.
Methods The sample comprised 4,710 working women
and men living in Stockholm, who answered the same
questionnaire twice, 3 years apart, who were not depressed
during the first wave and had the same job in both waves.
The questionnaire included JDCSM variables (demands,
skill discretion, decision authority and social climate) and
other co-variables (income, education, occupational group,
social support, help and small children at home, living with
an adult and depressive symptoms at time 1; and negative
life events at time 2). Multiple logistic regressions were run
to calculate odds ratios of having major depression at time
2, after adjustment for other JDCSM variables and covariables.
A. Fandino-Losada (&) Y. Forsell
Division of Public Health Epidemiology, Department of Public
Health Sciences, Karolinska Institutet, Norrbacka, Plan 7,
Karolinska Hospital, 171 73 Stockholm, Sweden
e-mail: andres.fandino@ki.se; fandino.andres@gmail.com
A. Fandino-Losada
School of Public Health/CISALVA Institute,
Universidad del Valle, Cali, Colombia
I. Lundberg
Department of Medical Sciences,
Occupational and Environmental Medicine,
Uppsala Universitet, Uppsala, Sweden
Results Among women, inadequate work social climate

was the only significant risk indicator for major depression.
Surprisingly, among men, high job demands and low skill
discretion appeared as protective factors against major
depression.
Conclusions The results showed a strong relationship
between inadequate social climate and major depression
among women, while there were no certain effects for the
remaining exposure variables. Among men, few cases of
major depression hampered well-founded conclusions
regarding our findings of low job demands and high skill
discretion as related to major depression.
Keywords Major depression Psychosocial factors
Work environment Demandscontrolsupport
Population survey Follow-up
Introduction
The psychosocial work environment can be an important
determinant of the development and course of depressive
disorders (for reviews, see Stansfeld and Candy 2006;
Bonde 2008; Netterstrm et al. 2008). There are different
models to describe this environment, but the job demands
control(support) model (JDCSM) has been the most
influential since its appearance in the late 1970s. The
JDCSM involves two logical and easily understandable
concepts: job demands and job control (or decision latitude). Job demands refer to task requirements and work
load. Job control refers to individuals ability to control
their work activities. Furthermore, Karasek emphasised
that both demands and control determine the psychological
strain at work, and job strain was defined as the combination of high demands and low control (Karasek 1979).
123
592
A third dimension was later added to the model: social

support at work. It is defined as the support received from
colleagues and supervisors at work, which hypothetically
buffers the effect of job strain situations on health outcomes (Johnson and Hall 1988; Johnson et al. 1989). Thus,
the last model is called the job demandscontrolsupport
model (JDCSM).
The literature shows inconsistent results regarding the
role of JDCSM variables as determinants of psychiatric
disorders, among other reasons because most studies have
been cross-sectional (van der Doef and Maes 1999; de
Lange et al. 2005). Longitudinal studies are necessary,
although not necessarily sufficient, to study causal relationships (Ahola and Hakanen 2007). In a pioneering
review, Stansfeld and Candy (2006) performed a metaanalysis of 11 longitudinal studies on psychosocial occupational exposures as determinants of psychological distress,
depression and anxiety.
Specifically for major depression, results from original
longitudinal studies are also inconsistent: some authors
have found that only job demands were determinants of
depressive disorders (Plaisier et al. 2007), and others have
found decision latitude (or its components) as the unique
risk factor (Griffin et al. 2002; Rugulies et al. 2006; Clays
et al. 2007; Joensuu et al. 2010); other studies indicate that
both job demands and control are independent determinants of the development of depressive disorders or
depressive symptoms (Paterniti et al. 2002). Additionally,
social support at work has appeared as a direct and independent determinant of depressive symptoms or depressive
disorders in several longitudinal studies (Niedhammer et al.
1998; Ylipaavalniemi et al. 2005; Griffin et al. 2007;
Stoetzer et al. 2009). Two recent systematic reviews have
tried to consolidate findings from longitudinal studies
(Bonde 2008; Netterstrm et al. 2008), but only Bonde
calculated averaged odds ratios by meta-analysis. Both
reviews agreed on that high demands, low job social support and high job strain appeared more consistently as
moderate risk factors for depression, while results regarding job control were contradictory. In any case, both groups
of authors acknowledged the heterogeneity of JDCSM
measures used by the original studies selected into each
review. Moreover, these systematic reviews are based on
small number of papers per JDCSM dimension, that is,
between 2 and 5 papers.
Although longitudinal studies imply progress in the
study of psychosocial determinants of depressive disorders,
a number of methodological problems remain: first, most
follow-up studies have been performed in specific occupations (e.g. nurses, civil servants, hospital employees,
forest industry employees) (Bourbonnais et al. 1999;
Stansfeld et al. 1999; Ylipaavalniemi et al. 2005; Joensuu
et al. 2010). Although studies on these groups could have
123
high internal validity and power, their conclusions may be

difficult to extrapolate to other working groups (Bourbonnais et al. 1999; Joensuu et al. 2010). Thus, research also
focuses on the psychosocial work situation in the general
population (Rugulies et al. 2006; Magnusson Hanson et al.
2009; Stoetzer et al. 2009; Wang et al. 2009).
Second, rather few studies have assessed major
depression by using more exact measures such as physicians diagnoses, hospital admission diagnoses or validated lay-administered diagnostic algorithms. In contrast,
most longitudinal studies have used scales of general
depressive symptoms as a proxy for diagnoses, with the
disadvantage of having lower sensitivity and specificity in
the assessment of the health outcome (e.g. de Lange et al.
2002; Paterniti et al. 2002; Clays et al. 2007; Andrea
et al. 2009).
Third, psychosocial work factors have been addressed by
using different conceptual approaches, although the concept
of work stress lies behind them in most studies (Stansfeld
and Candy 2006). This situation does not allow easy comparisons between different studies or generalisation of study
findings. Some studies have used shorter forms of the
JDCSM scales (Shields 2006; Wang et al. 2009), and others
have developed their own scales (Tokuyama et al. 2003;
Ylipaavalniemi et al. 2005; Rugulies et al. 2006; Joensuu
et al. 2010). Furthermore, a combined index was created in
a study (Wang 2005), which did not gauge the orthogonal
dimensions originally proposed by the JDCSM (Karasek
and Theorell 1990).
Finally, most longitudinal studies of the JDCSM and
depressive disorders have used a dichotomic approach to
operationalise the JDCSM scales, that is, a cut-off point
was selected and then the scale was dichotomised by
splitting scores into high and low categories (Stansfeld and
Candy 2006). In his seminal study, Karasek proposed a
median split methodology to create the dichotomic categories (Karasek 1979), which has been followed by most
researches (Wang 2004; Rugulies et al. 2006; Virtanen
et al. 2007). However, to identify doseresponse effects of
JDCSM variables, it is necessary to use finer exposure
categories. Accordingly, some cross-sectional studies have
addressed this issue by using scale tertiles (Shields 2006)
or scale quartiles (Michelsen and Bildt 2003), although the
longitudinal studies lack this methodological approach
(Blackmore et al. 2007; Bonde 2008; Netterstrm et al.
2008).
Hence, the aims of this study are as follows:
To determine the longitudinal effects of job demands,

skill discretion, decision authority and social climate
(support) at work on the occurrence of major depression (assessed by a diagnostic algorithm) among
working men and women from the general population.
To test the dose response of the job demandscontrol

support model (JDCSM) components on risk of major
depression over time.
We hypothesise that high job demands, inadequate skill

discretion, inadequate decision authority and inadequate
social climate at work are risk factors for the occurrence of
major depression among working men and women.
Methods
Population and sample
The study sample was drawn from the PART studya
longitudinal study of mental health, work and relations
ongoing in Stockholm County, Sweden. The study population at wave 1 included 19,744 individuals aged
2064 years registered in the county and randomly selected
from the county council register. The ethical committee at
the Karolinska Institutet, Stockholm, approved the study,
and informed consents were obtained from all participants.
Wave 1 (19982000)
(Time 1: T1). Subjects in the intended population sample
received a questionnaire by mail, and 10,443 individuals
responded (53 % of the intended sample). An extensive
non-response analysis was done by using available official
registers. Participation was related to female gender, higher
age, higher income and education, being born in the Nordic
countries, and having no psychiatric diagnosis in the hospital discharge register or in the early retirement register.
Associations between age, gender, income, country of
origin and inpatient hospital care due to psychiatric diagnosis were calculated for participants and non-participants
separately. The odds ratios (ORs) for these associations
were very similar among participants and non-participants
(Lundberg et al. 2005).
Wave 2 (20012003)
(Time 2: T2). All participants from the first wave received a
second, almost identical, questionnaire 3 years later. The
participation rate at the follow-up phase (T2) was 83 %
(n = 8,622). Attrition in the second phase was associated
with the same conditions as in the first phase, but associations
between putative risk factors and major depression, determined at T1, were the same in both the T2-participant and
T2-non-participant subsamples (Bergman et al. 2010).
Additionally, participants diagnosed with major depression
at time 1 had a higher probability of non-participation at time
2 (OR = 1.6; 95 % confidence interval (CI) = 1.12.4).
593
Study sample in the present study

For this study, only participants who were working at T1
(employed and self-employed) and who continued in the
same job at T2 were selected, that is, 4,710 subjects
(55.3 % from the T2 sample), and among them, 244 persons were excluded because they fulfilled the case criteria
for major depression at T1 according to the MDI algorithm
(see measures below), that is, 5 % of the remaining sample.
Additionally, one case was excluded because there was no
information about his/her depressive status at T2. Finally,
our study sample included 4,427 subjects (51 % from the
T2 sample), 2,415 women and 2,012 men.
Measures and instruments
Exposure variables
Inadequate social climate at work is a scale measuring
social support at work, extracted from the Swedish
DemandControlSupport Questionnaire (DCSQ) (Sanne
et al. 2005), which did not cover instrumental support at the
time of the first PART data collection (Hallstrom et al.
2003) and was oriented towards the atmosphere at the
worksite (Theorell and Karasek 1996; Landsbergis et al.
2000). The social climate scale has six questions: There is
a calm and pleasant atmosphere at my work place; There
is good sense of fellowship; My workmates support
me; If I have a bad day, my colleagues understand it;
I have a good relationship with my superiors; I have a
good relationship with my workmates. The four response
alternatives were Perfectly true, True, Not true and
Not true at all for all questions, using a Likert scale (14
points), adding up items into the scale score (624 points)
with higher scores indicating lower social climate. Validity
studies show that this scale explained 72 % of the variance
in a total index of social support originally containing 15
items (Ahlberg-Hulten et al. 1995) with an adequate
solution in principal component analyses and a satisfactory
internal consistency, that is, Cronbachs alpha = 0.83
(Sanne et al. 2005). In the present study, the scale showed a
Cronbachs alpha of 0.86 at T1.
Also the DCSQ, which is a translation/adaptation of
Karaseks demanddecision latitude questionnaire (Theorell
and Karasek 1996; Landsbergis et al. 2000), was used to
assess psychological demands, skill discretion and decision
authority (Karasek and Theorell 1990; Karasek 1979).
There were five questions on job psychological demands
and six on job control (four questions on skill discretion
and two questions on decision authority). The questions on
psychological demands concerned requirements to work
very fast, to work very hard, to do an excessive amount of
work, to have enough time to get the job done and to have
123
594
conflicting demands at work. Cronbachs alpha for the job

demands scale at T1 was 0.74. The four questions on skill
discretion concerned requirements to learn new things:
whether skills were required, whether creativity was
required and whether the work was monotonous or variable. Cronbachs alpha for the skill discretion scale at T1
was 0.66. The two questions on decision authority concerned freedom to decide what should be done and how
the job should be done. Cronbachs alpha for decision
authority scale at T1 was 0.77. The last two scales were
reversed for the analyses; thus, they expressed inadequate
job skill discretion and inadequate job decision authority. Cronbachs alpha for the job control scale at T1 was
0.73.
Outcome variable
Major depression was assessed by the Major Depression
Inventory (MDI) algorithm (Bech et al. 1997; Bech and
Wermuth 1998). Validity studies for the MDI have been
performed in population-based settings in Sweden (Forsell
2005), as well as in clinical settings in European countries
(Bech et al. 2001; Olsen et al. 2003; Cuijpers et al. 2007).
The sensitivity of MDI was 0.78 and the specificity was
0.78 in a study based on the first PART data collection
(Forsell 2005). The MDI scale includes DSM-IV criteria
for major depression; each symptom must have been
present more than 2 weeks, using a six-point intensity
Likert scale (from not at all to all the time). The MDI
can be used as a diagnostic instrument in which the items
are dichotomised to indicate the presence or absence of
each symptom following a diagnostic algorithm, but it can
also be used as a continuous scale of depressive symptoms
(ranging from 0 to 50 points). This scale had a Cronbachs
alpha of 0.92 at T1 and 0.92 at T2 in the study sample. The
outcome measure was defined as the point prevalence of
major depression assessed by the MDI algorithm at the
follow-up (T2).
2008; Virtanen et al. 2008; Kaikkonen et al. 2009;

Vanroelen et al. 2010). Small children were defined as those
younger than 7 years of age. Education was divided into
three levels depending on years of completed studies:
compulsory school (B9 years), uncompleted upper secondary school (1011 years) and completed upper secondary school or more (C12 years). Income was divided into
four levels: 149 K SEK or less (thousands of Swedish
Crowns), 150199 K SEK, 200299 K SEK, 300 K SEK or
more (1 & 9.33 SEK). Occupational groups were defined
according to the first level of classification in the Nordic
Standard Occupational Classification (NSOC) of 1989
(Statistiska centralbyran 1989) and still valid in 19982000
(T1), which originally consists of 12 main groups. These
were reclassified into 7 groups due to small numbers in
some NSOC groups: unskilled manual (blue collar) workers, skilled manual (blue collar) workers, lower non-manual
(white collar) workers, middle non-manual (white collar)
workers, higher non-manual (white collar) workers, selfemployed professionals and entrepreneurs. Availability of
attachment in the social network was assessed with three
questions from a Swedish modification of the Interview
Schedule for Social Interaction (ISSI) (Henderson et al.
1980; Unden and Orth-Gomer 1984); each question used a
four-point Likert scale (1 = completely true to 4 = not at
all), and they were added up in a single score; the higher the
score, the worse the social attachment. Cronbachs alpha for
the inadequate availability of the social attachment scale at
T1 was 0.75. Depressive symptoms at baseline were
assessed by the Major Depression InventoryMDI (Bech
et al. 1997; Bech and Wermuth 1998)as a continuous
score ranging from 0 to 50 points, with a Cronbachs alpha
of 0.92. A list of 22 potentially negative life events (e.g.
divorce, death of a child) was provided in the questionnaire
(Hallstrom et al. 2003), and each respondent stated whether
they occurred or not during the last 12 months before filling
the questionnaire at T2. The number of potentially negative
life events was categorised in the analyses as none, one,
two, and three or more events.
Confounders
Statistical analyses
In this study, the following variables were selected as
potential confounders in multivariable regression models,
because they have been associated with both depressive
disorders and psychosocial work factors in several studies:
age, living without another adult, little help with household
chores, small children at home, education level, income
level, occupational groups, inadequate availability of social
attachment (as a continuous score), depressive symptoms at
baseline (T1), also as a continuous score and the number of
potentially negative life events at T2 (Muntaner et al. 1998;
Griffin et al. 2002; Michelsen and Bildt 2003; Mitchell et al.
2003; Bonde 2008; Choi and Marks 2008; Netterstrm et al.
123
In this paper, the JDSCM scales on job demands, inadequate job skill discretion, inadequate job decision authority
and inadequate work social climate were split by cut-off
points defined by scale-specific quartiles obtained from the
current study sample, that is, subjects without major
depression at T1 according to the MDI algorithm and who
had the same job at both waves. Thus, sample distribution
quartiles, independently for each JDCSM variable, were
used for defining the exposure groups (i.e. exposure categories grouping scores of JDCSM scales). Due to the
limited number of scale points, the scales could not be
divided into exact medians or quartiles. The true percentages by intended quartiles and median are shown in
Tables 1, 2, 3 and 4. Median split exposure groups were
defined merging the two lowest (Q1 and Q2) and the two
highest (Q3 and Q4), respectively. The quartile and median
cut-off scores for defining the JDCSM exposure groups are
shown in Tables 1, 2 and 3 for the quartile splits and
Table 4 for the median splits. The JDCSM quartiles
resulted in percentages different to the theoretical 25 % in
each one, given the empirical distribution of scales scores;
for example, for job demands, a 12-point score was not
only the 25th percentile but also the 35th percentile. There
were not significant differences in cut-offs between women
and men for the JDCSM dimensions; thus, the same divisions were used for generating the exposure categories in
Tables 1 (womens results) and 2 (mens results).
Nearly eleven per cent (10.64 %) of the subjects records
had at least one variable with missing values; thus, the
regression imputation method (Raghunathan 2004; Donders
et al. 2006; Groves et al. 2009) was used to complete the
missing values in exposure variables and confounders before
running the regression models. For testing the robustness of
results from these models, all regressions were repeated
without imputed values. Initially, cross-tabulations and single-regression analyses were done to explore associations
between the outcome variable and exposures and confounders using the crude data (Tables 1, 2). Then, regression
analyses followed this sequence: firstly, exposure categories
(quartiles of JDCSM variables) were used in block enter
logistic regressions, starting with the simplest logistic
regression models (only one JDCSM variable) with occurrence of major depression at T2 as the outcome, using the
lowest quartile as the reference group for each variable.
These regression models were called Model 0, that is, the
unadjusted ones. Afterwards, multiple logistic regressions
were run to calculate the ORs of having the outcome (i.e.
major depression) at follow-up (T2) for each psychosocial
work environment variable quartiles compared with the
reference quartile, adding the confounding variables in each
step. In this manner, Model 1 adjusted for all JDCSM variables (i.e. job demands, skill discretion, decision authority
and job social climate) and age. Model 2 included Model 1
variables and additionally adjusted for living alone status,
inadequate social availability (scale score), having little help
with household chores, having small children at home and
number of negative life events at T2. Model 3 included
Model 2 variables and additionally adjusted for education
level, income level and occupational groups. Model 4
included Model 3 variables and additionally adjusted for
depressive symptoms score (MDI) at T1.
Model 4 could be considered an over-adjusted regression model, because it is controversial whether a high level
of baseline depressive symptoms (T1) was a confounder or
595
an intermediate variable (Thielen et al. 2011), but on the

other hand, directed acyclic graphs (DAGs) analysis
(Hernan et al. 2002) showed that it was necessary to control
for depressive symptoms at T1 as a confounder, due to
relationships between peoples mood and self-assessment
of their psychosocial work environment (de Lange et al.
2005; Stansfeld and Candy 2006). All described analyses
were done for mens and womens subsamples, separately.
Among men, a median split approach for JDCSM variables
was used instead of the quartile approach, due to the small
number of cases among them. Supplementary analyses
were done among women using a median split approach in
order to compare their results with the median approach
among men. Collinearity among co-variables in regression
models was checked using variance inflation factors (VIFs)
(Belsley et al. 1980); no collinearity was found. Finally,
trends over quartiles of significant JDCSM variables were
explored with the trend Walds test and orthogonal polynomial contrasts for linear, quadratic and cubic trends
(Szklo and Nieto 2007). Analyses were done by using
STATA 11.2 (StataCorp 2009).
Results
The point prevalence of major depression at the third year
follow-up (T2) was 3.77 % (95 % CI = 3.214.33 %),
being 5.63 % (95 % CI = 4.716.55 %) among women
and 1.54 % (95 % CI = 1.002.08 %) among men, based
on 136 cases among women and 31 cases among men.
Tables 1 and 2 describe the distribution of the JDCSM
variables and the selected confounders among cases and noncases for each gender (Tables 1 and 2, respectively). Among
women, the cases tended to have higher job demands, lower
job skill discretion and lower job social climate, to receive
less help with household chores, to have lower education and
income levels, to belong to the unskilled manual (blue collar)
workers, to the lower non-manual (white collar) workers or
to the entrepreneurs, to be younger, to have lower availability
of social attachment, to have higher levels of depressive
symptoms at T1 and to have experienced more negative life
events (one, two or three or more categories) at T2 when
compared with controls (Table 1). For men, the cases tended
to have lower job demands, lower job skill discretion, lower
job decision authority and lower job social climate, to live
without other adults, to receive less help with household
chores, to have lower income levels, to belong to the skilled
manual (blue collar) workers or to the lower non-manual
(white collar) workers, to have lower availability of social
attachment and to have higher levels of depressive symptoms
at T1 and to have experienced more negative life events (one,
two or three or more categories) at T2 when compared with
the controls (Table 2).
123
596
Table 1 Crude associations between study variables at time 1 and major depression at time 2 for women
Exposure variables
Depressed T2
N = 136
Non-depressed T2
N = 2,279
Crude ORs
Categorical variables T1 and T2
Na
Na
OR
Lowest (512 points)c
39
29.6
850
37.7
Low (13 points)

High (1415 points)
20
32
15.2
24.2
310
539
13.8
23.9
1.41
1.29
(0.812.45)
(0.802.09)
41
31.1
555
24.6
1.61*
(1.032.53)
Lowest (46 points)c
38
27.9
728
32.1
Low (7 points)
30
22.1
432
19.0
1.33
(0.812.18)
High (89 points)
34
25.0
739
32.6
0.88
(0.551.42)
Highest (1016 points)
34
25.0
370
16.3
1.76*
(1.092.84)
Lowest (2 points)c
39
28.7
693
30.5
Low (3 points)
18
13.2
483
21.2
0.66
(0.371.17)
High (4 points)
37
27.2
528
23.2
1.25
(0.781.98)
Highest (58 points)
42
30.9
572
25.1
1.30
(0.832.05)
(95 % CI)
Job demands T1b

b
Inadequate skill discretion T1
Inadequate decision authority T1b
Inadequate job social climate T1b

Lowest (69 points)c
22
17.2
760
35.4
Low (1011 points)

High (1213 points)
29
36
22.7
28.1
462
569
21.5
26.5
2.17**
2.19**
(1.233.82)
(1.273.76)
41
32.0
356
16.6
3.98***
(2.336.78)
Living without an adult T1d
37
27.2
630
27.7
0.98
(0.661.44)
Little help with house chores T1d
72
53.3
954
42.1
1.57*
(1.112.23)
Small children at home T1d
31
22.8
393
17.3
1.41
(0.932.13)
Negative life events at T2

Nonec
10
7.3
760
33.4
One
25
18.4
678
29.8
2.80**
(1.345.88)
Two
28
20.6
439
19.3
4.85***
(2.3310.07)
Three or more
73
53.7
400
17.5
13.87***
(7.0827.16)
Education level T1
9 years or less
24
17.7
303
13.3
1.84*
(1.113.04)
1011 years
62
45.6
813
35.7
1.77**
(1.212.59)
12 years or morec
50
36.8
1159
51.0
15
18
11.0
13.2
153
255
6.7
11.2
2.05*
1.48
(1.143.69)
(0.862.53)
(1.032.34)
Yearly income level T1

149 K SEK or less
150199 K SEK
200299 K SEK
39
28.7
526
23.1
1.55*
300 K SEK or morec
64
47.1
1339
58.9
18
14.1
243
11.2
1.87
(0.933.81)
6.2
111
5.1
1.84
(0.764.44)
Occupational groups T1
Unskilled manual workers
Skilled manual workers
Lower non-manual workers
39
30.5
459
21.1
2.16*
(1.173.99)
Middle non-manual workers
32
25.0
672
30.9
1.21
(0.652.27)
Higher non-manual workersc
15
11.7
382
17.6
Self-employed professionals
Entrepreneurs
123
4.7
183
8.4
0.83
(0.322.19)
10
7.8
121
5.6
2.10
(0.924.81)
597
Table 1 continued
Continuous variables T1
Mean
SD
Mean
SD
Age
40.43
10.54
42.58
10.95
0.98*
(0.971.00)
4.72
1.81
4.02
1.28
1.34***
(1.211.48)
10.82
6.76
5.50
5.26
1.14***
(1.111.17)
Inadequate social availability

Depressive symptoms score
OR
(95 % CI)
OR odds ratio, CI confidence interval, SD standard deviation
Significance level at p \ 0.10; * Significance level at p \ 0.05; ** Significance level at p \ 0.01; *** Significance level at p \ 0.001
N before imputation of missing values,
quartiles split,
reference category,
Table 3 describes the simple and adjusted logistic

regression results for study variables in relation to
occurrence of major depression at T2 among women. In
this manner, job demands (highest quartile, OR = 1.69,
95 % CI = 1.082.64) and lack of skill discretion at
work (highest quartile, OR = 1.7595 % CI = 1.092.83)
were significant predictors of major depression only in
the crude model (Model 0). When analysis was adjusted
for all JDCSM variables and age in Model 1, job
demands and lack of job skill discretion effects were no
longer significant in that and subsequent models. Inadequate social climate at work appeared as a significant
determinant of major depression at T2 in all adjusted
models; furthermore, the effect was significant for all
quartiles (when compared with the lowest quartile as the
reference category). Indeed, the highest quartile of
inadequate social climate remained as that with the
highest OR in Model 3 (OR = 3.35; 95 % CI = 1.88
5.99), but in Model 4, its OR was the second highest after
the low quartile (OR = 2.09 for the latter vs. OR = 2.06
for the former). The trend analysis rendered a significant
linear trend (X2(1) = 4.31; p = 0.038) over inadequate
social climate quartiles, but neither a quadratic nor a cubic
trend.
Analyses with non-imputed variables showed slightly
lower ORs (data not shown) compared with those in
Table 3, and some ORs in non-imputed analyses were not
statistically significant, although close to be significant
(p \ 0.10). Supplementary analyses for median splits of
JDCSM variables, using the same model definitions as
above, were done for women. Only inadequate social climate over the median was significant as a risk factor for
major depression occurrence in some adjusted models (1
and 2) with ORs ranging from 1.94 (95 % CI = 1.352.81)
to 1.52 (95 % CI = 1.032.23). In the fully adjusted model
(which included depressive symptoms at T1), inadequate
social climate was not significant (p = 0.20). Since the
median split did not show any significant association, we
kept the quartiles split of inadequate social climate. We
found that the three upper quartiles showed similar ORs in
Model 4, which were all significantly higher than the first
quartile (see Table 3).
the reference category is No
Table 4 describes the crude and adjusted logistic

regression results among men by using the median split
approach for the JDCSM. Due to the few cases among
men, no JDCSM variable was significant in the crude
model (Model 0), but when analysis was adjusted for all
JDCSM variables and age in Model 1, high job demands
and low social climate at work appeared as a significant
protective factor (OR = 0.41; 95 % CI = 0.190.88) and
as a risk factor for major depression (OR = 2.37; 95 %
CI = 1.125.02), respectively. Furthermore, high job
demands remained as a protective factor in subsequent
models, which included more confounders (OR Model
3 = 0.23; 95 % CI = 0.090.57; OR Model 4 = 0.24;
95 % CI = 0.100.60). In contrast, the social climate at
work effect was no longer significant in models including
more confounders (Models 2 to 4). On the other hand, low
skill discretion appeared as a protective factor against
major depression in Model 3 (OR = 0.30; 95 %
CI = 0.110.79) and Model 4 (OR = 0.32; 95 %
CI = 0.110.90). Additionally, analyses with non-imputed
variables showed significant and very similar ORs compared to those in Table 4 (data not shown).
Discussion
The literature shows inconsistent results for the longitudinal relationships between JDCSM variables and depressive
disorders, but two recent meta-analyses and a systematic review (Stansfeld and Candy 2006; Bonde 2008;
Netterstrm et al. 2008) showed coherent results on high
job demands and low occupational social support as risk
factors for subsequent major depressive disorders or mental
illness occurrence. In contrast, these reviews showed contradictory results regarding job control, or its components,
as risk or protective factors for major depression. The
findings about major depression in these reviews were
based on small number of papers for each JDCSM
dimension. In our study, the findings regarding inadequate
job social climate are in line with the current evidence and
show that, in particular, a good social climate seems protective against major depression among women, but our
123
598
Table 2 Crude associations between study variables at time 1 and major depression at time 2 for men
Exposure variables
Depressed T2
N = 31
Non-depressed T2
N = 1,981
Categorical variables T1 and T2
Na
16
53.3
4
3
13.3
10.0
Na
Crude ORs
%
OR
(95 % CI)
722
36.7
273
546
13.9
27.7
0.66
0.25*
(0.222.00)
(0.070.86)
23.3
428
21.7
0.74
(0.301.81)
Job demands T1b

Lowest (512 points)c
Low (13 points)
High (1415 points)
b
Inadequate skill discretion T1

Lowest (46 points)c
14
45.2
819
41.6
Low (7 points)
19.4
372
18.9
0.94
(0.362.47)
High (89 points)
9.7
566
28.7
0.31
(0.091.08)
25.8
213
10.8
2.20
(0.915.31)
Inadequate decision authority T1b

Lowest (2 points)c
22.6
782
39.5
Low (3 points)
29.0
466
23.5
2.16
High (4 points)
16.1
407
20.6
1.37
(0.434.35)
10
32.3
325
16.4
3.44*
(1.309.11)
Highest (58 points)
(0.805.83)
Inadequate job social climate T1b

Lowest (69 points)c
24.1
611
34.5
Low (1011 points)

High (1213 points)
5
6
17.2
20.7
429
470
24.2
26.6
1.02
1.11
(0.323.23)
(0.373.34)
11
37.9
260
14.7
3.69**
(1.429.63)
13
41.9
450
22.8
2.45*
(1.195.04)
21
70.0
873
44.3
2.93**
(1.346.44)
29.0
343
17.4
1.94
(0.894.26)

Living without an adult T1d
d
Little help with house chores T1

Small children at home T1d
Negative life events at T2
Nonec
3.3
753
38.0
One
20.0
575
29.1
7.86
(0.9465.45)
Two
20.0
360
18.2
12.55*
(1.51104.63)
17
56.7
291
14.7
43.99***
(5.83332.05)
Three or more
Education level T1
9 years or less
16.1
236
11.9
1.49
(0.534.22)
1011 years
13
41.9
827
41.8
1.11
(0.512.40)
12 years or morec
13
41.9
915
46.3
2
5
6.5
16.1
90
142
4.6
7.2
Yearly income level T1

149 K SEK or less
150199 K SEK
2.75
4.36**
(0.6012.61)
(1.4912.73)
(1.879.46)
200299 K SEK
13
41.9
383
19.4
4.21***
300 K SEK or morec
11
35.5
1363
68.9
Unskilled manual workers
10.0
197
10.2
2.17
(0.4310.86)
Skilled manual workers
16.7
203
10.5
3.51
(0.8314.85)
Lower non-manual workers
16.7
182
9.4
3.92
(0.93-16.57)
Middle non-manual workers
23.3
412
21.4
2.42
(0.629.44)
Higher non-manual workersc
10.0
428
22.2
Occupational groups T1
Self-employed professionals
13.3
264
13.7
2.16
(0.489.73)
Entrepreneurs
10.0
243
12.6
1.76
(0.358.79)
123
599
Table 2 continued
Continuous variables T1
Mean
Age
Inadequate social availability
Depressive symptoms score
SD
Mean
SD
OR
(95 % CI)
40.65
8.60
43.18
11.00
0.98
(0.951.01)
5.45
2.20
4.39
1.55
1.37***
(1.151.62)
12.03
5.43
4.18
4.45
1.23***
(1.171.29)
OR odds ratio, CI confidence interval, SD standard deviation
N before imputation of missing values,
quartiles split,
reference category,
the reference category is No
Table 3 Adjusted multiple logistic regression models for major depression at T2 among women
Outcome: major
depression at T2
Quartiles split of
Model 0
OR
Model 1
(95 % CI)
OR
Model 2
(95 % CI)
OR
Model 3
(95 % CI)
OR
Model 4
(95 % CI)
OR
(95 % CI)
Job Demands T1a

Lowest (512 points)b
Low (13 points)

High (1415 points)
1.41
1.33
(0.812.46)
(0.832.14)
1.36
1.21
(0.772.40)
(0.741.98)
1.23
1.01
(0.682.21)
(0.601.68)
1.21
0.99
(0.672.20)
(0.591.67)
1.23
0.94
(0.672.26)
(0.551.59)
1.69*
(1.082.64)
1.33
(0.832.13)
1.19
(0.731.94)
1.31
(0.792.15)
1.07
(0.641.79)
Inadequate skill discretion T1a

Lowest (46 points)b
Low (7 points)
1.29
(0.782.12)
1.29
(0.782.14)
1.28
(0.762.16)
1.18
(0.692.02)
1.13
(0.651.95)
High (89 points)
0.88
(0.551.42)
0.86
(0.531.42)
0.79
(0.471.32)
0.63
(0.371.08)
0.65
(0.371.13)
1.75*
(1.092.83)
1.63
(0.962.77)
1.57
(0.912.74)
1.12
(0.612.05)
1.12
(0.602.10)
Inadequate decision authority T1
Lowest (2 points)b
Low (3 points)
0.65
(0.371.15)
0.57
(0.321.01)
0.58
(0.321.05)
0.61
(0.331.11)
0.63
High (4 points)
1.23
(0.771.95)
1.00
(0.621.62)
1.08
(0.651.78)
1.05
(0.621.76)
1.02
(0.601.76)
Highest (58 points)
1.25
(0.801.97)
0.79
(0.481.31)
0.82
(0.491.38)
0.75
(0.441.30)
0.74
(0.421.30)
1
(0.341.17)
Inadequate job social climate T1a

Lowest (69 points)b
Low (1011 points)

High (1213 points)
2.32**
2.42***
(1.334.05)
(1.424.13)
2.35**
2.41**
(1.344.14)
(1.404.16)
1.97*
1.92*
(1.103.52)
(1.093.38)
2.08*
2.08*
(1.153.75)
(1.183.68)
2.09*
1.85*
(1.153.81)
(1.033.31)
4.19***
(2.477.11)
3.84***
(2.206.71)
2.55**
(1.424.56)
2.78***
(1.535.06)
2.06*
(1.103.83)
Model 0: Simple logistic regression, that is, unadjusted by each other JDCSM variable
Model 1: Adjusted for each other JDCSM variables (i.e. job demands, skill discretion, decision authority and job social climate) and age at T1
Model 2: Model 1 and additionally adjusting for living alone status, inadequate social availability (scale score), having little help with home
chores and having small children at home at T1, and negative life events at T2
Model 3: Model 2 and additionally adjusting for education level, income level and occupational groups at T1
Model 4: Model 3 and additionally adjusting for score of depressive symptoms (MDI) at T1
OR odds ratio, CI confidence interval
a
quartiles split,
reference category
findings regarding job demands and skill discretion among

men are opposite to recent reviews, that is, high job
demands appeared as a protective factor for major
depression occurrence and skill discretion appeared as a
factor that increased the occurrence of major depression.
The fact that high job demands among men appeared as

a protective factor against major depression in the adjusted
models was not expected, although the finding was consistent in analyses with and without imputation. We have
found one nested casecontrol study derived from the
123
600
Table 4 Adjusted multiple logistic regression models for major depression at T2 among men
Outcome: major
depression at T2
Median split of:
Model 0
OR
Model 1
(95 % CI)
OR
Model 2
(95 % CI)
OR
Model 3
(95 % CI)
OR
Model 4
(95 % CI)
OR
(95 % CI)
Job demands T1a

Low (513 points)b
High (1420 points)
0.49
1
(0.231.04)
0.41*
1
(0.190.88)
0.33**
(0.140.73)
0.23**
1
(0.090.57)
0.24**
(0.100.60)
Inadequate skill discretion T1a

Low (47 points)b
High (816 points)
0.84
Inadequate decision authority T1

Low (23 points)b
High (48 points)
1.60
(0.401.76)
0.58
1
(0.271.27)
0.50
(0.221.15)
0.30*
1
(0.110.79)
0.32*
(0.110.90)
1
(0.793.25)
1.44
1
(0.683.08)
1.20
(0.532.72)
0.91
1
(0.372.27)
0.82
(0.322.07)
Inadequate job social climate T1a

Low (611 points)b
High (1223 points)
2.03
(0.994.16)
2.37*
1
(1.125.02)
1.84
(0.814.18)
1.54
1
(0.633.78)
1.40
(0.563.48)
Model 0: Simple logistic regression, that is, unadjusted by each other JDCSM variable
Model 1: Adjusted for each other JDCSM variables (i.e. job demands, skill discretion, decision authority and job social climate) and age at T1
Model 2: Model 1 and additionally adjusting for living alone status, inadequate social availability (scale score), having little help with home
chores and having small children at home at T1, and negative life events at T2
Model 3: Model 2 and additionally adjusting for education level, income level and occupational groups at T1
Model 4: Model 3 and additionally adjusting for score of depressive symptoms (MDI) at T1
OR odds ratio, CI confidence interval
Median split,
reference category
Danish Work Environment Cohort Study (DWECS) that

showed a significant protective effect of high demands on
depressive disorders among men, but the authors did not
offer a plausible explanation for their findings (Wieclaw
et al. 2008). Additionally, we found no association between
high job demands and major depression among women.
This is in accordance with several previous studies where
high psychological demands were not related to depression
among women (Clumeck et al. 2009; Virtanen et al. 2010;
Thielen et al. 2011). We lack explanations for these findings, and further investigation is warranted. Furthermore,
we stratified the mens analyses by three major occupational groups (blue-collar workers, white-collar workers
and other workers) in order to test the robustness of this
protective effect among different occupational groups.
Thus, the OR of high job demands remained significantly
protective only in the blue-collar strata (data not shown);
however, the ORs for high job demands in the other two
strata were also below unity.
We found contradictory findings regarding job control
components (i.e. skill discretion and decision authority)
when the results of adjusted models were compared
between genders. Among women, inadequate skill discretion appeared as a risk factor for the occurrence of major
depression, but only in the crude model. The opposite
123
occurred among men; inadequate skill discretion appeared

as a protective factor in adjusted models (3 and 4).
Moreover, the literature reports inconsistent findings related to decision latitude or job control appearing as a
protective factor in some studies, as a risk factor in others
and without significant results in others, in samples of
either gender or samples of both genders included. In
systematic reviews, job control or its components showed
contradictory findings (Bonde 2008; Netterstrm et al.
2008). These reviews showed that decision latitude was
addressed only as a whole dimension in the cited studies,
that is, its components (skill discretion and decision
authority) were not individually scrutinised. Uniquely, two
recent longitudinal studies addressed decision authority as
a separate scale (Magnusson Hanson et al. 2009; Joensuu
et al. 2010) with contradictory findings between studies: in
the former, high decision authority appeared as a protective
factor against depression, and in the latter, it resulted as a
risk factor.
Additionally, only the Finnish Still Working Study
(Joensuu et al. 2010) addressed skill discretion as a separate scale, showing a protective effect on depression
occurrence. In contrast, the DWECS study showed nonsignificant results for possibilities for development, a
scale conceptually related to skill discretion (Rugulies et al.
2006). Adding more empirical evidence to this controversy, we found a protective effect of inadequate skill
discretion on the occurrence of major depression, but only
among men. Such findings have been obtained before
(Kawakami et al. 1995; Kondo et al. 2006). A possible
explanation for these peculiar results is that using skills has
become more of a demand when work intensity increases
(Joensuu et al. 2010). Furthermore, we stratified the mens
analyses by three major occupational groups (blue-collar
workers, white-collar workers and other workers) in order
to test the robustness of this protective effect among different occupational groups. Thus, the OR of low skill
discretion remained significantly protective only in the
blue-collar strata (data not shown); however, the ORs for
low skill discretion in the other two strata were also below
unity. Additionally, our findings strongly suggest that skill
discretion and decision authority should be addressed
separately in future studies about the psychosocial work
environment.
Finally, inadequate social climate at work appeared as a
predictor of major depression only among women, and it
remained significant over different adjusted models. Trend
analyses showed a significant linear trend over the quartiles
of inadequate social climate with the reference category
significantly lower than the upper three categories. Our
results confirm findings about inadequate social climate as
a risk factor for depressive disorders in other longitudinal
studies (Bonde 2008; Sinokki et al. 2009), but our genderspecific findings were opposite to those reported in the
literature, where inadequate social support at work was a
stronger risk factor among men than among women
(Griffin et al. 2002; Shields 2006; Netterstrm et al. 2008)
or only significant among men, when the outcome was use
of antidepressant medication (Thielen et al. 2011). The
limited number of men with major depression in our
sample (only 31 subjects) should have implied a low power
to detect significant relationships. Additionally, our measure of social climate did not cover instrumental social
support. It is possible that inadequate social climate could
have a stronger relation with major depression among
women than among men while inadequate instrumental
social support, which we have not measured, could be more
related to major depression among men.
Limitations of the study
First, our sample of working individuals was plausibly
comprised by individuals healthier in terms of their physical and mental health than those unemployed or not in the
labour market (i.e. the healthy worker effect), which is in
line with the patterns of participation in both waves of the
PART study (Lundberg et al. 2005; Bergman et al. 2010);
thus, the occurrence of major depression in our study was
601
lower compared to some other studies (for reviews, see

Stansfeld and Candy 2006; Bonde 2008; Netterstrm et al.
2008). Additionally, conclusions about the relationship
between JDCSM variables and major depression in this
study should be interpreted with caution because our
sample included only subjects who remained in the same
job over the two study waves. Subjects who were not
included in this sample may have changed jobs or have
become unemployed due to work stress problems (Haahr
et al. 2007) or depressive disorders (Lerner et al. 2004).
In contrast, persons who remained in the same job over
two waves could have done so because of interesting and
challenging work (Stansfeld et al. 2008). To assess these
alternative explanations, we did supplementary analyses:
there were no differences in the prevalence of major
depression at T2 between people who had the same job in
both waves and people who left their T1 job (3.77 vs.
4.54 %, Fishers exact test p = 0.12). Also, we made
regression analyses using the sample of people who did not
have the same job at T2, but no significant results were
found for any JDCSM variable neither among women nor
among men (data not shown). Finally, we have tested
independently each item of the job demands scale (some
items could be related to interesting and challenging work)
among persons with the same job, but non-significant
results were found for such items in models adjusted by
JDCSM and other co-variables; thus, no single job demands
item was significantly related to the study outcome.
Second, in this study, the measurements of exposures
and outcome cannot be considered completely independent
because they were both obtained through self-reports
(Rugulies et al. 2006; Bonde 2008). This dependence could
spuriously have increased the odds ratios due to the
triviality trap or bias due to common method (Kristensen 1996; Kasl 1998). Thus, we have tried to control the
common method bias in two ways: firstly, we removed
from the study sample individuals who met the criteria for
major depression, because individuals with major depression could have negatively assessed their work environment (de Lange et al. 2005). Secondly, we have measured
the work exposures and the major depression outcome
3 years apart. Moreover, the level of baseline depressive
symptoms was adjusted for in the regression analyses.
However, of course there might be non-measured factors, such as personality and social desirability, which
could have affected our results via the common method
bias mechanism. Two papers exploring this issue showed
that the associations between depressive symptoms and
psychosocial work factors were not biased by personality
factors (Paterniti et al. 2002; Jurado et al. 2005). The
depressive symptoms are part of several personality disorders; thus, controlling for the former would also partially
control for the latter.
123
602
Finally, we have no incidence estimates given that our

population was examined twice over a 3-year lag. Thereby,
we could have lost incident cases appearing and disappearing between the two data collection times. This fact
could reduce the study power, but it would hardly affect
our point prevalence estimates and the exposureoutcome
relationships.
Strengths of the study
In the literature, there are rather few longitudinal and
population-based studies about the influence of psychosocial work environment factors on major depression. The
advantage of population studies is that they allow including
a broader range of psychosocial work factors in comparison
with occupational sample studies where the extent of
demandscontrolsupport arrangements is limited to the
organisation boundaries. But population studies imply
logistic limitations, which make it more difficult to properly assess the outcome (i.e. major depression) and the
putative determinants (i.e. psychosocial work environment
variables).
Our study has several strengths compared to other population studies: first, the four psychosocial work dimensions
(demands, decision authority, skill discretion and social
climate) were assessed by specific state-of-the-art scales
originated from Karasek and Theorells work (Karasek
1979; Karasek and Theorell 1990; Johnson and Hall 1988;
Karasek et al. 1998) instead of using summary scales for
each or a composite index for all JDCSM factors (e.g. job
stress index or job strain quotient). Our approach also
allowed an individual assessment of each psychosocial
work environment dimension, yielding independent conclusions about demands, decision authority, skill discretion
and social climate at work. In the literature, only Joensuu
et al. study analysed the four Karasek JDCSM dimensions
individually, although it was a study about a specific population: private sector industrial employees (Joensuu et al.
2010). Our study sought to contribute to the understanding
of psychosocial work environment factors as determinants
of major depression in the general working population by
using the four JDCSM dimensions individually.
Second, we utilised a sample of subjects working in the
same job at both waves (T1 and T2); thus, it allowed
having an assessment of psychosocial work exposure
variables that should be more stable (over 3 years) and
more accurate compared to other follow-up studies, where
working conditions should have changed over the followup (Stansfeld and Candy 2006; Bonde 2008; Netterstrm
et al. 2008).
Finally, usually previous studies applied a dichotomised
split approach for psychosocial work variables, which
could bring statistical modelling problems when the
123
behaviour of the phenomenon is complex (Theorell and

Karasek 1996; Altman and Royston 2006). Then, analyses
using more categories are warranted. Conversely, a continuous scale approach could be very useful for determining doseresponse effects, and indeed, this method has
been used in some studies (Plaisier et al. 2007; Virtanen
et al. 2007; Magnusson Hanson et al. 2009), but it implies
that a linear relationship exists between the predictors and
the outcome variable (MacCallum et al. 2002). Thereby,
split analysis using a number of ordered categories based
on scale threshold criteria (e.g. quartiles) seems a good
alternative in the middle of dichotomic and continuous
approaches. Hence, we utilised quartile cut-off points for
the JDCSM variables (i.e. the 25th, 50th and 75th sample
percentiles) in the analyses of the womens subsample.
Conclusions
The results showed that a good work social climate was
protective against major depression among women. However, there were no clear effects of other JDCSM exposure
variables among women. Among men, high job demands
and low skill discretion appeared as protective factors
against major depression; both are novel and controversial
findings considering the recent literature reviews. Thus,
more studies are warranted to disentangle such relationships. Finally, decision authority and skill discretion
seemed to have opposite effects on the occurrence of major
depression; therefore, they should be addressed separately
in future studies about the psychosocial work environment.
Acknowledgments The authors thank the PART project grant
providers: the Swedish Medical Research Council, the Swedish
Council for Working Life and Social Research and the Stockholm
County Council (Grant number VR K2007-61X-20381-01-3). The
PhD student support was provided by the COLFUTURO Foundation
(Colombia), the ERACOL academic exchange programme (Erasmus
Mundus External CooperationEurope) and Universidad del Valle
(Colombia). We also wish to thank Professor Tores Theorell for
valuable suggestions during the development of the study, Peeter
Fredlund and Michael Lundberg for statistical advice and the
Reviewers for suggestions for improving the manuscript.
Conflict of interest
of interest.
The authors declare that they have no conflict
References
Ahlberg-Hulten GK, Theorell T, Sigala F (1995) Social support, job
strain and musculoskeletal pain among female health care
personnel. Scand J Work Environ Health 21(6):435439
Ahola K, Hakanen J (2007) Job strain, burnout, and depressive
symptoms: a prospective study among dentists. J Affect Disord
104(13):103110

Altman DG, Royston P (2006) The cost of dichotomising continuous
variables. BMJ 332(7549):1080
Andrea H, Bultmann U, Van Amelsvoort LGPM, Kant Y (2009) The
incidence of anxiety and depression among employeesthe role
of psychosocial work characteristics. Depress Anxiety
26(11):10401048
Bech P, Wermuth L (1998) Applicability and validity of the major
depression inventory in patients with Parkinsons disease. Nord J
Psychiatry 52(4):305310
Bech P, Stage KB, Nair NP, Larsen JK, Kragh-Sorensen P, Gjerris A
(1997) The major depression rating scale (MDS). Inter-rater
reliability and validity across different settings in randomized
moclobemide trials. Danish University Antidepressant Group.
J Affect Disord 42(1):3948
Bech P, Rasmussen NA, Olsen LR, Noerholm V, Abildgaard W
(2001) The sensitivity and specificity of the major depression
inventory, using the present state examination as the index of
diagnostic validity. J Affect Disord 66(23):159164
Belsley DA, Kuh E, Welsch RE (1980) Regression diagnostics:
identifying influential data and sources of collinearity. Wiley,
New York
Bergman P, Ahlberg G, Forsell Y, Lundberg I (2010) Non-participation in the second wave of the PART study on mental disorder
and its effects on risk estimates. Int J Soc Psychiatry
56(2):119132
Blackmore ER, Stansfeld SA, Weller I, Munce S, Zagorski BM,
Stewart DE (2007) Major depressive episodes and work stress:
results from a national population survey. Am J Public Health
97(11):20882093
Bonde JPE (2008) Psychosocial factors at work and risk of
depression: a systematic review of the epidemiological evidence.
Occup Environ Med 65(7):438445
Bourbonnais R, Comeau M, Vezina M (1999) Job strain and evolution
of mental health among nurses. J Occup Health Psychol
4(2):95107
Choi H, Marks NF (2008) Marital conflict, depressive symptoms, and
functional impairment. J Marriage Fam 70(2):377390
Clays E, De Bacquer D, Leynen F, Kornitzer M, Kittel F, De Backer
G (2007) Job stress and depression symptoms in middle-aged
workersprospective results from the Belstress study. Scand J
Work Environ Health 33(4):252259
Clumeck N, Kempenaers C, Godin I, Dramaix M, Kornitzer M,
Linkowski P, Kittel F (2009) Working conditions predict
incidence of long-term spells of sick leave due to depression:
results from the Belstress I prospective study. J Epidemiol
Community Health 63(4):286292
Cuijpers P, Dekker J, Noteboom A, Smits N, Peen J (2007) Sensitivity
and specificity of the major depression inventory in outpatients.
BMC Psychiatry 7:39
de Lange AH, Taris TW, Kompier MA, Houtman IL, Bongers PM
(2002) Effects of stable and changing demand-control histories
on worker health. Scand J Work Environ Health 28(2):94108
de Lange AH, Taris TW, Kompier MAJ, Houtman ILD, Bongers PM
(2005) Different mechanisms to explain the reversed effects of
mental health on work characteristics. Scand J Work Environ
Health 31(1):314
Donders AR, van der Heijden GJ, Stijnen T, Moons KG (2006)
Review: a gentle introduction to imputation of missing values.
J Clin Epidemiol 59(10):10871091
Forsell Y (2005) The major depression inventory versus schedules for
clinical assessment in neuropsychiatry in a population sample.
Soc Psychiatry Psychiatr Epidemiol 40(3):209213
Griffin JM, Fuhrer R, Stansfeld SA, Marmot M (2002) The
importance of low control at work and home on depression
and anxiety: do these effects vary by gender and social class?
Soc Sci Med 54(5):783
603
Griffin JM, Greiner BA, Stansfeld SA, Marmot M (2007) The effect
of self-reported and observed job conditions on depression and
anxiety symptoms: a comparison of theoretical models. J Occup
Health Psychol 12(4):334349
Groves RM, Fowler FJ, Couper MP, Lepkowski JM, Singer E (2009)
Survey methodology, 2nd edn. Wiley, Hoboken
Haahr J, Frost P, Andersen J (2007) Predictors of health related job
loss: a two-year follow-up study in a general working population. J Occup Rehabil 17(4):581592
Hallstrom T, Damstrom Thakker K, Forsell Y, Lundberg I, Tinghog P
(2003) The PART study. A population based study of mental
health in the Stockholm County: study design. Phase l
(19982000). Report, Samhallsmedicin, Sweden
Henderson S, Duncan-Jones P, Byrne DG, Scott R (1980) Measuring
social relationships. The interview schedule for social interaction. Psychol Med 10(4):723734
Hernan MA, Hernandez-Daz S, Werler MM, Mitchell AA (2002)
Causal knowledge as a prerequisite for confounding evaluation:
an application to birth defects epidemiology. Am J Epidemiol
155(2):176184
Joensuu M, Vaananen A, Koskinen A, Kivimaki M, Virtanen M,
Vahtera J (2010) Psychosocial work environment and hospital
admissions due to mental disorders: a 15-year prospective study
of industrial employees. J Affect Disord 124(12):118125
Johnson JV, Hall EM (1988) Job strain, work place social support,
and cardiovascular disease: a cross-sectional study of a random
sample of the Swedish working population. Am J Public Health
78(10):13361342
Johnson JV, Hall EM, Theorell T (1989) Combined effects of job
strain and social isolation on cardiovascular disease morbidity
and mortality in a random sample of the Swedish male working
population. Scand J Work Environ Health 15(4):271279
Jurado D, Gurpegui M, Moreno O, Fernandez MC, Luna JD, Galvez
R (2005) Association of personality and work conditions with
depressive symptoms. Eur Psychiatry 20(3):213222
Kaikkonen R, Rahkonen O, Lallukka T, Lahelma E (2009) Physical
and psychosocial working conditions as explanations for occupational class inequalities in self-rated health. Eur J Public
Health 19(5):458463
Karasek RA (1979) Job demands, job decision latitude, and mental
strainimplications for job redesign. Adm Sci Q 24(2):285
308
Karasek R, Theorell T (1990) Healthy work: stress productivity and
the reconstruction of working life. Basic Books, New York
Karasek R, Brisson C, Kawakami N, Houtman I, Bongers P, Amick B
(1998) The job content questionnaire (JCQ): an instrument for
internationally comparative assessments of psychosocial job
characteristics. J Occup Health Psychol 3(4):322355
Kasl SV (1998) Measuring job stressors and studying the health
impact of the work environment: an epidemiologic commentary.
J Occup Health Psychol 3(4):390401
Kawakami N, Roberts RE, Lee ES, Araki S (1995) Changes in rates
of depressive symptoms in a Japanese working population: lifetable analysis from a 4-year follow-up study. Psychol Med
25(06):11811190
Kondo K, Kobayashi Y, Hirokawa K, Tsutsumi A, Kobayashi F,
Haratani T, Araki S, Kawakami N (2006) Job strain and sick
leave among Japanese employees: a longitudinal study. Int Arch
Occup Environ Health 79(3):213219
Kristensen TS (1996) Job stress and cardiovascular disease: a
theoretic critical review. J Occup Health Psychol 1(3):246260
Landsbergis P, Theorell T, Schwartz J, Greiner BA, Krause N (2000)
Measurement of psychosocial workplace exposure variables.
Occup Med 15(1):163188
Lerner D, Adler DA, Chang H, Lapitsky L, Hood MY, Perissinotto C,
Reed J, McLaughlin TJ, Berndt ER, Rogers WH (2004)
123
604
Unemployment, job retention, and productivity loss among
employees with depression. Psychiatr Serv 55(12):13711378
Lundberg I, Damstrom Thakker K, Hallstrom T, Forsell Y (2005)
Determinants of non-participation, and the effects of nonparticipation on potential cause-effect relationships, in the PART
study on mental disorders. Soc Psychiatry Psychiatr Epidemiol
40(6):475483
MacCallum RC, Zhang S, Preacher KJ, Rucker DD (2002) On the
practice of dichotomization of quantitative variables. Psychol
Methods 7(1):1940
Magnusson Hanson L, Theorell T, Bech P, Rugulies R, Burr H, Hyde
M, Oxenstierna G, Westerlund H (2009) Psychosocial working
conditions and depressive symptoms among Swedish employees.
Int Arch Occup Environ Health 82(8):951960
Michelsen H, Bildt C (2003) Psychosocial conditions on and off the
job and psychological ill health: depressive symptoms, impaired
psychological wellbeing, heavy consumption of alcohol. Occup
Environ Med 60(7):489496
Mitchell PB, Parker GB, Gladstone GL, Wilhelm K, Austin M-PV
(2003) Severity of stressful life events in first and subsequent
episodes of depression: the relevance of depressive subtype.
Muntaner C, Eaton WW, Diala C, Kessler RC, Sorlie PD (1998)
Social class, assets, organizational control and the prevalence of
common groups of psychiatric disorders. Soc Sci Med
47(12):20432053
Netterstrm B, Conrad N, Bech P, Fink P, Olsen O, Rugulies R,
Stansfeld S (2008) The relation between work-related psychosocial factors and the development of depression. Epidemiol Rev
30(1):118132
Niedhammer I, Goldberg M, Leclerc A, Bugel I, David S (1998)
Psychosocial factors at work and subsequent depressive symptoms in the Gazel cohort. Scand J Work Environ Health
24(3):197205
Olsen LR, Jensen DV, Noerholm V, Martiny K, Bech P (2003) The
internal and external validity of the major depression inventory
in measuring severity of depressive states. Psychol Med
33(2):351356
Paterniti S, Niedhammer I, Lang T, Consoli SM (2002) Psychosocial
factors at work, personality traits and depressive symptoms:
longitudinal results from the GAZEL Study. Br J Psychiatry
181(2):111117
Plaisier I, de Bruijn JGM, de Graaf R, Have MT, Beekman ATF,
Penninx BWJH (2007) The contribution of working conditions
and social support to the onset of depressive and anxiety
disorders among male and female employees. Soc Sci Med
64(2):401410
Raghunathan TE (2004) What do we do with missing data? Some
options for analysis of incomplete data. Annu Rev Public Health
25:99117
Rugulies R, Bultmann U, Aust B, Burr H (2006) Psychosocial work
environment and incidence of severe depressive symptoms:
prospective findings from a 5-year follow-up of the Danish work
environment cohort study. Am J Epidemiol 163(10):877887
Sanne B, Torp S, Mykletun A, Dahl AA (2005) The Swedish demandcontrol-support questionnaire (DCSQ): factor structure, item
analyses, and internal consistency in a large population. Scand J
Public Health 33(3):166174
Shields M (2006) Stress and depression in the employed population.
Health Rep 17(4):1129
Sinokki M, Hinkka K, Ahola K, Koskinen S, Kivimaki M, Honkonen
T, Puukka P, Klaukka T, Lonnqvist J, Virtanen M (2009) The
association of social support at work and in private life with
mental health and antidepressant use: the health 2000 study.
123

Stansfeld S, Candy B (2006) Psychosocial work environment and
mental healtha meta-analytic review. Scand J Work Environ
Health 32(6):443462
Stansfeld SA, Fuhrer R, Shipley MJ, Marmot MG (1999) Work
characteristics predict psychiatric disorder: prospective results
from the Whitehall II study. Occup Environ Med 56(5):302307
Stansfeld SA, Clark C, Caldwell T, Rodgers B, Power C (2008)
Psychosocial work characteristics and anxiety and depressive
disorders in midlife: the effects of prior psychological distress.
Occup Environ Med 65(9):634642
StataCorp (2009) Stata. Statistics/data analysis, 11.2 edn. College
Station
Statistiska centralbyran (1989) Yrkesklassificeringar i FoB 85 enligt
Nordisk yrkesklassificering (NYK) och Socioekonomisk indelning (SEI): alfabetisk version = Occupations in population and
housing census 1985 (FoB 85) according to Nordic standard
occupational classification (NYK) and Swedish socio-economic
classification (SEI): alphabetical version. Meddelanden i samordningsfragor, 1989:5. SCB, Stockholm
Stoetzer U, Ahlberg G, Johansson G, Bergman P, Hallsten L, Forsell
Y, Lundberg I (2009) Problematic interpersonal relationships at
work and depression: a Swedish prospective cohort study.
J Occup Health 51(2):144151
Szklo M, Nieto FJ (2007) Epidemiology: beyond the basics, 2nd edn.
Jones and Bartlett Publishers, Sudbury
Theorell T, Karasek RA (1996) Current issues relating to psychosocial job strain and cardiovascular disease research. J Occup
Health Psychol 1(1):926
Thielen K, Nygaard E, Rugulies R, Diderichsen F (2011) Job stress and
the use of antidepressant medicine: a 3.5-year follow-up study
among Danish employees. Occup Environ Med 68(3):205210
Tokuyama M, Nakao K, Seto M, Watanabe A, Takeda M (2003)
Predictors of first-onset major depressive episodes among whitecollar workers. Psychiatry Clin Neurosci 57:523531
Unden A-L, Orth-Gomer K (1984) Socialt Stod Och Halsa Delrapport
2: Utveckling av en enkatmetod for att mata socialt stod i
befolkningsstudier. (Social supportReport No 2: development
of a method for assessment of social support in population
surveys.). Stressforskningsrapporter 178:72
van der Doef M, Maes S (1999) The job demand-control (-support)
model and psychological well-being: a review of 20 years of
empirical research. Work Stress 13(2):87114
Vanroelen C, Levecque K, Louckx F (2010) Differential exposure and
differential vulnerability as counteracting forces linking the
psychosocial work environment to socioeconomic health differences. J Epidemiol Community Health 64(10):866873
Virtanen M, Honkonen T, Kivimaki M, Ahola K, Vahtera J, Aromaa
A, Lonnqvist J (2007) Work stress, mental health and antidepressant medication findings from the health 2000 study. J Affect
Disord 98(3):189197
Virtanen M, Koskinen S, Kivimaki M, Honkonen T, Vahtera J, Ahola
K, Lonnqvist J (2008) Contribution of non-work and workrelated risk factors to the association between income and mental
disorders in a working population: the health 2000 study. Occup
Environ Med 65(3):171178
Virtanen M, Batty GD, Pentti J, Vahtera J, Oksanen T, Tuisku K, Salo
P, Terho K, Ahola K, Elovainio M, Kivimaki M (2010) Patient
overcrowding in hospital wards as a predictor of diagnosisspecific mental disorders among staff: a 2-year prospective
cohort study. J Clin Psychiatry 71(10):13081312
Wang JL (2004) Perceived work stress and major depressive episodes
in a population of employed Canadians over 18 years old. J Nerv
Ment Dis 192(2):160163
Wang J (2005) Work stress as a risk factor for major depressive
episode(s). Psychol Med 35(6):865871

Wang J, Schmitz N, Dewa C, Stansfeld S (2009) Changes in
perceived job strain and the risk of major depression: results
from a population-based longitudinal study. Am J Epidemiol
169(9):10851091
Wieclaw J, Agerbo E, Mortensen PB, Burr H, Tuchsen F, Bonde JP
(2008) Psychosocial working conditions and the risk of
605
depression and anxiety disorders in the Danish workforce.
BMC Public Health 8:280
Ylipaavalniemi J, Kivimaki M, Elovainio M, Virtanen M, KeltikangasJarvinen L, Vahtera J (2005) Psychosocial work characteristics
and incidence of newly diagnosed depression: a prospective cohort
study of three different models. Soc Sci Med 61(1):111
123
Copyright of International Archives of Occupational & Environmental Health is the property

of Springer Science & Business Media B.V. and its content may not be copied or emailed to
multiple sites or posted to a listserv without the copyright holder's express written permission.
However, users may print, download, or email articles for individual use.

Decision Authority

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Decision Authority

Uploaded by

Copyright:

Available Formats

Int Arch Occup Environ Health (2013) 86:591605

Demands, skill discretion, decision authority and social climate

Results Among women, inadequate work social climate

A third dimension was later added to the model: social

Int Arch Occup Environ Health (2013) 86:591605

high internal validity and power, their conclusions may be

To determine the longitudinal effects of job demands,

Int Arch Occup Environ Health (2013) 86:591605

To test the dose response of the job demandscontrol

We hypothesise that high job demands, inadequate skill

Study sample in the present study

conflicting demands at work. Cronbachs alpha for the job

Int Arch Occup Environ Health (2013) 86:591605

2008; Virtanen et al. 2008; Kaikkonen et al. 2009;

Int Arch Occup Environ Health (2013) 86:591605

an intermediate variable (Thielen et al. 2011), but on the

Int Arch Occup Environ Health (2013) 86:591605

Categorical variables T1 and T2

Lowest (512 points)c

Low (13 points)

Lowest (46 points)c

High (89 points)

Highest (1016 points)

Highest (58 points)

Job demands T1b

Highest (1620 points)

Inadequate skill discretion T1

Inadequate decision authority T1b

Inadequate job social climate T1b

Low (1011 points)

Highest (1423 points)

Living without an adult T1d

Little help with house chores T1d

Small children at home T1d

Negative life events at T2

Yearly income level T1

300 K SEK or morec

Middle non-manual workers

Higher non-manual workersc

Int Arch Occup Environ Health (2013) 86:591605

Inadequate social availability

OR odds ratio, CI confidence interval, SD standard deviation

N before imputation of missing values,

Table 3 describes the simple and adjusted logistic

the reference category is No

Table 4 describes the crude and adjusted logistic

Int Arch Occup Environ Health (2013) 86:591605

Categorical variables T1 and T2

Job demands T1b

Inadequate skill discretion T1

High (89 points)

Highest (1016 points)

Inadequate decision authority T1b

Highest (58 points)

Inadequate job social climate T1b

Low (1011 points)

Highest (1423 points)

Little help with house chores T1

Yearly income level T1

300 K SEK or morec

Unskilled manual workers

Skilled manual workers

Lower non-manual workers

Middle non-manual workers