Job Strain, Long Work Hours, and Suicidal Ideation in US Workers: A Longitudinal Study

Abstract

PurposeTo investigate whether chronic psychosocial work stressors (low job control, high job demands, job strain, low supervisor and coworker support, job insecurity, and long work hours) are longitudinally associated with suicidal ideation in a working population. Methods
Five-hundred seventy-eight workers (aged 34–69) were chosen for this analysis from those who participated in both project 1 (2004–2006 at baseline) and project 4 (2004–2009 at follow-up) of the Midlife Development in the United States II study. The median time interval between the two projects was 26 months (range 2–62 months). ResultsAbout 11% of the workers reported suicidal ideation at follow-up, while 3% of them reported moderate/severe suicidal ideation at follow-up. After controlling for age, marital status, race, family history of suicide, and suicidal ideation at baseline, low skill discretion and job strain (a combination of low job control and high job demands) were associated with total suicidal ideation. After excluding those with suicidal ideation at baseline from analysis and further controlling for other work stressors, job strain was strongly associated with moderate/severe suicidal ideation: ORs, 4.29 (1.30–14.15) for quartile-based job strain and 3.77 (1.21–11.70) for median-based job strain. Long work hours (> 40 h/week vs. ≤ 40 h/week) also increased the likelihood for moderate/severe suicidal ideation: OR 4.06 (1.08–15.19). Conclusions
Job strain and long work hours were longitudinally associated with moderate/severe suicidal ideation. Increasing job control and ensuring optimal level of work demands, including 40 h or less of work per week may be an important strategy for the prevention of suicide in working populations.

Full text

Vol.:(0123456789)
1 3
International Archives of Occupational and Environmental Health
https://doi.org/10.1007/s00420-018-1330-7
ORIGINAL ARTICLE
Job strain, long work hours, andsuicidal ideation inUS workers:
alongitudinal study
BongKyooChoi1,2,3
Received: 30 October 2017 / Accepted: 26 June 2018
© Springer-Verlag GmbH Germany, part of Springer Nature 2018
Abstract
Purpose To investigate whether chronic psychosocial work stressors (low job control, high job demands, job strain, low
supervisor and coworker support, job insecurity, and long work hours) are longitudinally associated with suicidal ideation
in a working population.
Methods Five-hundred seventy-eight workers (aged 34–69) were chosen for this analysis from those who participated in
both project 1 (2004–2006at baseline) and project 4 (2004–2009at follow-up) of the Midlife Development in the United
States II study. The median time interval between the two projects was 26months (range 2–62months).
Results About 11% of the workers reported suicidal ideation at follow-up, while 3% of them reported moderate/severe
suicidal ideation at follow-up. After controlling for age, marital status, race, family history of suicide, and suicidal ideation
at baseline, low skill discretion and job strain (a combination of low job control and high job demands) were associated
with total suicidal ideation. After excluding those with suicidal ideation at baseline from analysis and further controlling
for other work stressors, job strain was strongly associated with moderate/severe suicidal ideation: ORs, 4.29 (1.30–14.15)
for quartile-based job strain and 3.77 (1.21–11.70) for median-based job strain. Long work hours (> 40h/week vs. ≤ 40h/
week) also increased the likelihood for moderate/severe suicidal ideation: OR 4.06 (1.08–15.19).
Conclusions Job strain and long work hours were longitudinally associated with moderate/severe suicidal ideation. Increas-
ing job control and ensuring optimal level of work demands, including 40h or less of work per week may be an important
strategy for the prevention of suicide in working populations.
Keywords MIDUS II study· Job strain· Skill discretion· Work demands· Work hours
Introduction
Suicide is a serious public health concern across the globe,
including the United States (US) (Rockett etal. 2012;
World Health Organization 2014). About 800,000 people
died due to suicide globally in 2012, which is equivalent to
one suicide death every 40s (World Health Organization
2014, 2017). In the US, the population suicide mortality
rate increased by 24% from 1999 (10.5 per 100,000) to 2014
(13.0 per 100,000) (Curtin etal. 2016), particularly among
middle-aged adults (Centers for Disease Control and Pre-
vention, 2013; Phillips and Nugent 2014). And the suicide
mortality rate in US working populations has been also on
the rise (Tiesman etal. 2015).
For the primary prevention of suicide in working popula-
tions, it is essential to identify the important work-related
risk factors for suicidal ideation that is a strong predictor
for suicide attempt within 1 or 2years (Nock etal. 2008;
Simon etal. 2013; Ten Have etal. 2009). Several chronic
psychosocial work stressors such as low job control, high
psychological job demands, job strain (as a combination
of low job control and high job demands), low social sup-
port at work, job insecurity, and long work hours have been
implied as important risk factors for suicide mortality case
studies (Amagasa etal. 2005; Lee etal. 2016) or for suicidal
* BongKyoo Choi
b.choi@uci.edu
1 Center forOccupational andEnvironmental Health,
Department ofMedicine, University ofCalifornia Irvine,
100 Theory, Suite 100, Irvine, CA92617, USA
2 Enviromental Health Sciences Graduate Program, University
ofCalifornia Irvine, Irvine, USA
3 Program inPublic Health, University ofCalifornia Irvine,
Irvine, USA

International Archives of Occupational and Environmental Health
1 3
ideation in many cross-sectional studies (Loerbroks etal.
2016; Milner etal. 2016, 2017).
However, only a handful of longitudinal studies (Baumert
etal. 2014; Ostry etal. 2007; Tsutsumi etal. 2007) have been
undertaken to examine chronic psychosocial work stress-
ors in relation to suicidal attempt or mortality in working
populations. With regard to suicide ideation, only one study
(Kim etal. 2017) has examined and reported job insecurity
as a significant risk factor. However, in the study, neither
other chronic psychosocial work stressors nor a family his-
tory of suicide (Wang etal. 2017) was considered in analy-
sis. Thus, there is still a possibility of residual confounding
by unmeasured occupational and individual risk factors for
suicidal ideation. In addition, no studies have examined the
longitudinal associations of job strain and its components,
supervisor and coworker support, and long work hours with
suicidal ideation in working populations.
The purpose of this study is to investigate whether chronic
psychosocial work stressors (job strain and its components,
supervisor and coworker support, job insecurity, and work
hours) are longitudinally associated with suicidal ideation
in a middle-aged US working population particularly after
controlling for family history of suicide.
Methods
MIDUS II study data
Psychosocial work stressors were assessed in Project 1
(2004–2006) of the National Survey of Midlife Development
in the United States (MIDUS) II study. Suicidal ideation was
measured in both Project 1 and Project 4 (2004–2009) of the
MIDUS II study (Fig.1).
Project 1 of the MIDUS II study was conducted in
2004–2006 as a follow-up study of the MIDUS I study
(1994–1995) that had been originally designed to investigate
the roles of behavioral, psychological, and social factors in
understanding age-related differences in physical and mental
health (Ryff etal. 2007). Initially, 7108 persons (males, 48%
and females, 52%) participated in the baseline phone inter-
view of the MIDUS I study. 6329 participants additionally
completed a self-administered questionnaire. All of the
participants were non-institutionalized, English-speaking
adults, aged 25–74 in the US. They were drawn from four
subsamples: (a) a main national random-digit-dial (RDD)
sample (N = 3034); (b) oversamples from five metropolitan
areas (N = 658); (c) siblings of individuals from the RDD
sample (N = 869); and (d) a national RDD sample of twin
pairs (N = 1764). The response rates of the four subsamples
ranged from 60 to 70%. The socio-demographic character-
istics of the main RDD subsample were comparable to those
of a US population representative sample, the October 1995
Current Population Survey (http://www.censu s.gov/cps).
However, the main RDD subsample relatively underrepre-
sented those who were black, young (e.g., aged 25–34), or
had less formal education (i.e., 12 or less than 12years of
formal education) (Choi etal. 2010a; Ryff etal. 2007).
In total, 4963 persons (males 47% and females 53%) par-
ticipated in the follow-up phone interview of the MIDUS I
study. The longitudinal retention rates among the four sub-
samples ranged from 65 to 78% (on average 70%). There
were no significant (p < 0.01) differences in age and gender
between the follow-up participants and non-participants.
However, during the follow-up, less-educated persons and
non-whites were relatively more likely to have dropped out
of the study (Choi etal. 2010a).
Among the phone interview participants of the MIDUS
II study, 4032 persons additionally completed a self-
administered questionnaire (i.e., Project 1 of the MIDUS
II study), including the items about psychosocial work
stressors: (a) the main national RDD sample (N = 1805);
(b) the city oversamples (N = 386); (c) the sibling sam-
ple (N = 637); and (d) the twin national RDD sample
(N = 1204). All the main and twin RDD participants of the
Project 1 of the MIDUS II study, along with a small sub-
group of the sibling sample, were eligible for participation
in the Project 4 of the MIDUS II study if their existing
health information indicated an ability to travel to clinics
for comprehensive bioindicator and physical examinations
(N = 2680) (Dienberg Love etal. 2010). Among the eligi-
ble participants (N = 2680), 1054 persons (response rate,
39.2%, and 43.1% after adjusting for those who could not
be located or contacted) were additionally recruited to a
Fig. 1 Description of the
MIDUS II study as the basis
of this study. *Participated
in a phone interview and a
self-administered questionnaire
survey

International Archives of Occupational and Environmental Health
1 3
subsequent project, called the Biomarker project (Project
4) of the MIDUS II study in 2004–2009 (Fig.1). For the
recruitment for the Project 4, a study invitation letter was
sent to all eligible participants and then follow-up calls
were made by the research staff. The participants of the
Project 4 were given $200 as a stipend for their 2-day visit
to the medical clinic. The purpose of Project 4 was to add
a subsample of MIDUS participants including informa-
tion on comprehensive biological assessments (Ryff etal.
2013). There were no significant (p > 0.01) differences
between the participants and non-participants of Project
4 in terms of gender, marital status, household income,
working conditions (e.g., job strain, work hours, and job
insecurity), and suicidal ideation at T1 (see below) in the
MIDUS II study. But, the participants of Project 4 were
younger (mean ages, 55.3 vs. 56.6years), more whites
(93 vs. 90%) and less less-educated persons (24% for 12
or less years of formal education vs. 35%), compared to
the non-participants of Project 4. And the participants of
Project 4 were mostly from the main RDD and twin RDD
subsamples.
Study subjects: 582 workers
For the current study, among those (N = 1054) who par-
ticipated in both Project 1 (called hereafter T1) and Pro-
ject 4 (called hereafter T2) of the MIDUS II study, study
subjects were first restricted to those (N = 582) who were
aged less than 70years at T1, had a full-time or part-time
job in 2003 and were working (at least 1h per week at a
main job) at T1, and had valid information on the expo-
sure and outcome variables at both T1 and T2 for analy-
sis. The median time interval between the two projects was
26months (range 2–62months) in the study subjects of the
current study. Then, study subjects were further restricted
to those (N = 552) who did not reported suicidal ideation at
T1. Suicidal ideation at T1 was only approximately assessed
with the following two items for screening major depression
in the World Health Organization Composite International
Diagnostic Interview Short-Form (CIDI-SF) (Kessler etal.
1998): “During the past 12months, was there ever time
when you felt sad, blue, or depressed for 2weeks or more in
a row? (Yes/No)”; and “Did you think a lot about death—
either your own, someone else’s, or death in general—dur-
ing those 2weeks? (Yes/No)”. Suicidal ideation cases at T1
were defined as those (N = 30) who agreed to the above two
items. Several previous studies (Crandall etal. 2010; Ten
Have etal. 2009) based on the suicidal process perspective
from early stage (death ideation) to later stages (e.g., suicidal
ideation and suicide attempt) (Neeleman etal. 2004) have
used the above more inclusive CIDI-SF items as a proxy
measure of suicidal ideation or for suicidality research.
Suicidal ideation atT2
Suicidal ideation at T2 was measured with one question of
the anhedonia depression scale in the Mood and Anxiety
Symptoms Questionnaire (MASQ) (Watson etal. 1995):
“During the past week, how much you have felt or experi-
enced thought about death or suicide?” Total suicidal idea-
tion cases in the current study were defined as those who
responded to the above question with one of the following
four response options (a little bit, moderately, quite a bit, or
extremely) vs. not at all (i.e., non-suicidal ideation cases).
In addition, in consideration of a strong dose–response
relationship between the degree of suicidal ideation and
future suicidal attempt (Walker etal. 2010), moderate or
severe suicide ideation cases were also defined with those
who responded to the above question with one of the fol-
lowing three response options (moderately, quite a bit, or
extremely). A similar single questionnaire item [item 9 of
the Patient Health Questionnaire (Kroenke etal. 2001)] has
been commonly used for research on work and suicidal idea-
tion (Loerbroks etal. 2016). Its validity was demonstrated
against cross-sectional clinical assessment (Walker etal.
2010) and subsequent suicide attempt or death (Simon etal.
2013).
Psychosocial working conditions atT1
Several psychosocial working conditions (job control, psy-
chological job demands, supervisor and coworker support at
work, job insecurity, working hours, and serious on-going
stress at work) were measured with a self-administered
questionnaire at T1. The items for job control (five items:
two skill discretion items about variety of work and learn-
ing opportunities on the job and three decision authority
items about on-the-job decision-making opportunities),
psychological job demands (three items; time pressure and
workload), and immediate supervisor (two items) and cow-
orker support (two items) were similar to the ones of the Job
Content Questionnaire (JCQ) (Karasek etal. 1985). Cron-
bach alphas of job control and psychological job demands
were 0.81 and 0.68, respectively (Choi etal. 2010b). More
detailed information about the items is available elsewhere
(Choi etal. 2010a). The items had a five-point Likert type of
response set: all of the time to never, and were summed up
for scaling-scoring. The scores of the aforementioned scales
were dichotomized at their medians in those who partici-
pated in Project 1 of MIDUS II study for analyses. There was
one additional response option for the immediate supervisor
and coworker support items: does not apply. Thus, those who
responded with the option were categorized into no immedi-
ate supervisor and no coworker groups, respectively.
Job strain, a combination of low job control and high
job demands based on the Karasek’s demand-control model

International Archives of Occupational and Environmental Health
1 3
(Karasek 1979), was operationalized in the following two
ways: (a) median-based dichotomous job strain (vs. non-
job strain) using the medians of job control and psycho-
logical job demands, and (b) quartile-based dichotomous job
strain, using the quartiles of job control and psychological
job demands (the 3 bottom right corner cells for job strain
vs. the other 13 cells for non-job strain) (Karasek etal. 2007)
for avoiding potential misclassification of job strain around
the medians of job control and psychological job demands.
Job insecurity was measured with one item (“If you
wanted to stay in your present job, what are the chances that
you could keep it for the next 2years?”). Those with the
response options (fair or poor vs. good, very good, or excel-
lent) were considered to be the high-job-insecurity group.
Work hours per week at a main job and other paid jobs were
added up for analysis.
Covariates
Several potential confounders (Table1) were considered in
analysis: data sources, socio-demographic measures (age,
sex, marital status, race, annual household income, and edu-
cation), number of chronic diseases, alcohol consumption,
and family history of suicide. Race was categorized into
three groups: White, Black, and the remaining group. In
the last remaining group (N = 26), there were Native Ameri-
cans/Alaska native islanders/Eskimo (N = 9), Asians (N = 1),
and others (N = 16). The number of chronic diseases (those
who have experienced or been treated for any of the follow-
ing during the past 12months: arthritis, sciatica, recurring
stomach trouble or diarrhea, persistent foot troubles, trouble
with varicose veins, multiple sclerosis, stroke, and hernia;
or those who have ever had heart problems or ever had can-
cer) was counted. Alcohol consumption was categorized into
heavy, moderate, and non-drinking. Moderate drinking was
defined as up to two drinks per day for men and one drink
per day for women during the past month and heavy drinking
was defined as more than moderate drinking. Family history
of suicide was measured with one question: “Has any one
in your immediate family (father, mother, and siblings), or
maternal/paternal family (grandparents and uncle/aunt) had
suicide?” (Yes/No).
Statistical analyses
Descriptive statistics of total and moderate/severe suicidal
ideation are presented in Table1. The univariate associations
of each of the study variables with total and moderate/severe
suicidal ideation were examined by χ2 or Fisher’s exact test.
When a psychosocial working condition was at least margin-
ally (p < 0.20) associated with suicidal ideation in the uni-
variate analysis, it was further investigated through a series
of multivariate logistic regression models after controlling
for socio-demographic variables (Model 1); controlling for
socio-demographic variables, health behaviors and condi-
tions, and suicidal ideation at T1 (Model 2); and if neces-
sary, controlling for socio-demographic variables, health
behaviors and conditions, suicidal ideation at T1, and other
psychosocial work stressors (Model 3). In addition, the
above multivariate analyses were replicated without those
who reported suicidal ideation at T1. Statistical significance
testing was based on a two-sided test.
Results
Total andmoderate/severe suicidal ideation atT2
Sixty-six (11.3%) of the 582 workers in this study reported
suicidal ideation at T2 and most of them (47 workers)
thought of suicide a little bit. Nineteen (3.3%) of the 582
workers reported that they thought of suicide moderately (14
workers), quite a bit (4 workers), or extremely (1 worker).
Among the 552 workers without suicidal ideation at T1, 52
workers (9.4%) newly reported suicidal ideation at T2. Most
of the 52 workers (38 workers) thought of suicide a little bit,
while 14 workers (2.5%) thought of suicide moderately (13
workers) or quite a bit (1 worker).
Univariate associations betweenpsychosocial
working conditions andtotal suicidal ideation atT2
Low skill discretion and job strain (quartile based) were pos-
itively and significantly (p < 0.05) associated with suicidal
ideation in univariate analyses (Table1). High psychological
job demands and job strain (median based) were marginally
(p < 0.20) associated with suicidal ideation. Low decision
authority, low job control, no coworkers, job insecurity, and
long work hours were all positively associated with suicidal
ideation; however, their associations with suicidal ideation
did not reach the marginal statistical significance (p = 0.20).
Immediate supervisor support was not associated with sui-
cidal ideation.
Age, marital status, race, and family suicide history were
at least marginally associated with suicidal ideation. Suicide
ideation was greatest in the Native Americans/Alaska native
islanders/Eskimo (four out of nine, 44.4%). Suicide ideation
was relatively greater in the younger (< 40years old), the
windowed or never married, non-white and non-black peo-
ple (i.e., native Americans/Alaska native islanders/Eskimo/
Asians/ Others), and those with family history of suicide.
Data source, sex, education, household income, number of
chronic disease, alcohol consumption, and the time interval
between Time 1 and Time 2 were not significantly associated
with suicidal ideation in univariate analyses. As expected,

International Archives of Occupational and Environmental Health
1 3
Table 1 Suicidal ideation in relation to study variables in 582 US workers
Major category Minor category Subcategory Frequency (%) Total suicidal
ideation (%)
Moderate or severe
suicidal ideation (%)
Data source Subsamples Main RDD 57.9 13.4 3.3
City or Siblings 2.7 6.3 0.0
Twin 39.3 8.7 3.3
Socio-demographic vari-
ables
Sex Men 50.0 12.4 3.1
Women 50.0 10.3 3.4
Age (years) <40 11.2 18.5‡6.2†
40–49 35.7 12.5‡2.4†
50–59 36.1 11.9‡4.8†
60–69 17.0 3.0‡0.0†
Marital status Married 73.5 10.3†3.0*
Separated 1.9 9.1†0.0*
Divorced 11.9 7.2†4.3*
Windowed 2.1 33.3†16.7*
Never married 10.7 19.4†1.6*
Race White 92.6 10.2‡2.8*
Black 2.9 11.8‡5.9*
Native Americans/Alaska
native islanders/Eskimo/
Asian/Others
4.5 34.6‡11.5*
Education High school or less 20.7 15.0 4.2
Some college 26.2 10.5 2.0
University or more 53.2 10.4 3.6
Annual household income
($)
<60,000 31.1 9.4 3.3
60,000–99,999 33.2 13.5 3.6
≥100,000 35.7 11.1 2.9
Psychosocial working condi-
tions
Skill discretion Low 35.7 15.4†5.3†
High 64.3 9.1†2.1†
Decision authority Low 41.2 12.9 4.6*
High 58.8 10.2 2.3*
Job control Low 45.4 12.9 4.9†
High 54.6 10.1 1.9†
Psychological job demands Low 43.1 9.2* 1.6†
High 56.9 13.0* 4.5†
Job strain (median based) No 74.4 10.2* 2.1‡
Yes 25.6 14.8* 6.7‡
Job strain (quartile based) No 84.7 10.1†2.2‡
Yes 15.3 18.0†9.0‡
Supervisor support Low 42.4 10.5 3.2
High 41.6 12.4 3.3
No immediate supervisors 16.0 10.8 3.2
Coworker support Low 46.7 11.8 3.3
High 44.5 9.7 3.5
No coworkers 8.8 17.6 2.0
Job insecurity No 95.5 11.2 3.1
Yes 4.5 15.4 7.7
Hours ofwork per week ≤ 40 49.6 11.5 2.1*
41 or more 50.4 11.3 4.4*
Hours ofwork per week ≤ 40 49.6 11.5 2.1*

International Archives of Occupational and Environmental Health
1 3
suicidal ideation at T1 was strongly associated with suicidal
ideation at T2.
Multivariate associations betweenpsychosocial
working conditions andtotal suicidal ideation atT2
After controlling for age, marital status, race, and family
suicide history in multivariate logistic regression analysis
(Model 1), the odds ratios (ORs) for suicidal ideation were
significantly higher in the workers with low skill discretion
and job strain (quartile based) (Table2): 1.84 (95% CI: 1.07,
3.17, p = 0.027) and 2.03 (1.08, 3.85, p = 0.029), respec-
tively. On the other hand, high psychological job demand
and job strain (median based) were not significantly associ-
ated with suicidal death after controlling for age, marital
status, race, and family suicide history. The ORs of low
skill discretion and job strain (quartile based) only slightly
changed after additionally controlled for suicidal ideation
at T1 (Model 2): 1.95 (1.11–3.43) and 1.95 (1.00–3.80),
respectively. When the above multivariate analyses were
replicated in those without suicidal ideation at T1 (N = 552),
the results were very similar (Table3).
Univariate associations betweenpsychosocial work
stressors andmoderate/severe suicidal ideation
atT2
Generally, the distribution patterns of moderate or severe
suicide ideation by the study variables were similar to those
of total suicide ideation (Table1). However, in univariate
analyses, job strain (both quartile and median based) and its
components (skill discretion, decision authority, job con-
trol and psychological job demands), job insecurity, and
work hours per week were all more strongly associated with
Table 1 (continued)
Major category Minor category Subcategory Frequency (%) Total suicidal
ideation (%)
Moderate or severe
suicidal ideation (%)
41–48 18.4 9.3 2.8*
49–56 21.2 11.4 4.9*
≥ 57 10.8 14.3 6.3*
Health behaviors and condi-
tions
Number of chronic diseases 0 26.8 10.3 3.2*
1–2 44.3 10.5 1.9*
≥ 3 28.9 13.7 5.4*
Family history of suicide No 90.0 10.7* 3.6
Yes 10.0 17.2* 0.0
Alcohol consumption No 29.7 12.1 5.2*
Moderate 67.4 10.5 2.3*
Heavy 2.9 23.5 5.9*
Suicidal ideation at T1 No 94.8 9.4‡2.5‡
Yes 5.2 46.7‡16.7‡
*p < 0.20, †p < 0.05, and ‡ p < 0.01 at χ2 test or Fisher’s exact test
Table 2 Odds ratios and their
95% confidence intervals
of psychosocial working
conditions at T1 for total
suicidal ideation at T2 in US
workers (N = 582)
Model 1 was controlled for age, marital status, race, and family suicide history. Model 2 was controlled for
age, marital status, race, family suicide history, and suicidal ideation at T1
Variables Model1 P value Model2 P value
Low skill discretion 1.84 (1.07–3.17) 0.027 1.95 (1.11–3.43) 0.020
High psychological job demands 1.45 (0.83–2.53) 0.197 1.25 (0.70–2.23) 0.454
Job strain (median based) 1.46 (0.82–2.58) 0.196 1.40 (0.78–2.55) 0.263
Job strain (quartile based) 2.03 (1.08–3.85) 0.029 1.95 (1.00–3.80) 0.050
Table 3 Odds ratios and their 95% confidence intervals of psychoso-
cial working conditions at T1 for total suicidal ideation at T2 in US
workers who did not report suicidal ideation at T1 (N = 552)
Model 1 was controlled for data source, age, marital status, race, and
family history of suicide
Variables Model1 P value
Low skill discretion 1.87 (1.03–3.39) 0.040
High psychological job demands 1.24 (0.67–2.27) 0.498
Job strain (median based) 1.44 (0.76–2.73) 0.260
Job strain (quartile based) 1.96 (0.96–4.01) 0.065

International Archives of Occupational and Environmental Health
1 3
moderate/severe suicidal ideation than total suicidal idea-
tion. In particular, there was a linear association (p = 0.043)
between work hours per week and moderate/severe suicidal
ideation. However, supervisor and coworker support at work
were not associated with moderate/severe suicidal ideation.
Also, there was no moderate or severe suicidal ideation case
at T2 among those with family history of suicide (N = 58).
Multivariate associations betweenpsychosocial
work stressors andmoderate/severe suicidal
ideation atT2
Table4 shows the results of the multivariate associations
between psychosocial work stressors with moderate/
severe suicidal ideation after controlling for age, mari-
tal status, race, family history, and suicidal ideation at
T1. The ORs for moderate/severe suicidal ideation were
4.18 (1.57–11.16) for job strain (quartile-based) and 3.14
(1.22–8.10) for job (median based). These results were
similar after additionally controlling for job insecurity and
long work hours (Model 3, Table4). Low skill discretion,
low decision authority, low job control, and psychologi-
cal job demands were at least at the marginal significance
level associated with moderate/severe suicidal ideation.
Job insecurity was positively, but not significantly associ-
ated with moderate/severe suicidal ideation. Work hours
per week were associated with moderate/severe suicidal
ideation. In particular, there was a significant linear
increase in moderate/severe suicidal ideation as hours
of work per week increased in all multivariate models in
Table4. The above multivariate associations were simi-
lar, but stronger in those without suicidal ideation at T1
(Table5). Low skill discretion, low decision authority,
low job control were significantly associated with moder-
ate/severe suicidal ideation. The ORs for moderate/severe
suicidal ideation after controlling for age, marital status,
race, family history of suicide, and other work stressors
(Model 3 in Table5) were 4.29 (1.30–14.15) for job strain
(quartile-based) and 3.77 (1.21–11.70) for job (median
based) in the 552 workers without suicidal ideation at T1.
Long work hours per week (> 40h vs.≤40h per week)
also increased the risk for moderate/severe suicidal idea-
tion: OR, 4.06 (1.08–15.40). In particular, the OR for mod-
erate/severe suicidal ideation was significantly higher in
those who reported 49–56h of work per week (vs.≤40h
per week): 6.34 (1.48–27.26).
Discussion
This is the first longitudinal study that clearly demon-
strated significantly positive associations of job strain
and long work hours with moderate/severe suicidal idea-
tion in a working population after controlling for other
chronic work stressors and also family history of suicide.
Job control and its components were strongly associated
with moderate/severe suicidal ideation. While both high
psychological job demands and job insecurity increased
the likelihood of total or moderate/severe suicidal idea-
tion, their associations were not statistically significant.
This study indicates that improving psychosocial working
conditions may be an important strategy for the prevention
of suicide in working populations.
Table 4 Odds ratios and their 95% confidence intervals of psychosocial working conditions at T1 for moderate/severe suicidal ideation at T2 in
US workers (N = 582)
Model 1 was controlled for age, marital status, race, and family suicide history. Model 2 was controlled for age, marital status, race, family sui-
cide history, and suicidal ideation at T1. Model 3 for job strain or its components was controlled for the covariates in Model 2 plus job insecurity
and work hours per week. Model 3 for job insecurity was controlled for the covariates in Model 2 plus job strain (quartile based) and work hours
per week. Model 3 for work hours per week was controlled for the covariates in Model 2 plus job strain (quartile based), and job insecurity
*P for linear trend = 0. 019. †P for linear trend = 0. 013. ‡P for linear trend = 0. 028
Variables Model1 P value Model2 P value Model3 P value
Low skill discretion 2.38 (0.93–6.10) 0.070 2.49 (0.94–6.58) 0.065 2.54 (0.94–6.92) 0.067
Low decision authority 2.15 (0.83–5.51) 0.113 2.10 (0.81–5.45) 0.126 2.38 (0.88–6.40) 0.086
Low job control 2.61 (0.97–7.04) 0.059 2.68 (0.98–7.35) 0.055 2.94 (1.03–8.35) 0.044
High psychological job demands 2.99 (0.97–9.17) 0.056 2.51(0.81–7.83) 0.113 1.99 (0.61–6.43) 0.253
Job strain (median based) 3.46 (1.36–8.77) 0.009 3.14 (1.22–8.10) 0.018 2.99 (1.13–7.95) 0.028
Job strain (quartile based) 4.56 (1.75–11.87) 0.002 4.18 (1.57–11.16) 0.004 4.08 (1.47–11.34) 0.007
Job insecurity 2.79 (0.60–13.06) 0.192 3.05 (0.63–14.81) 0.167 2.36 (0.41–13.72) 0.339
Hours of work per week(≥ 41 vs. ≤40) 2.30 (0.85–6.20) 0.100 2.43 (0.88–6.66) 0.085 2.60 (0.91–7.37) 0.073
Hours of work per week (41–48 vs. ≤40) 1.40* (0.34–5.79) 0.641 1.50† (0.36–6.32) 0.581 1.82‡ (0.42–7.93) 0.425
(49–56 vs. ≤40) 2.63* (0.82–8.44) 0.104 2.61† (0.79–8.60) 0.115 2.90‡ (0.85–9.87) 0.088
(≥ 57 vs. ≤40) 3.21* (0.86–12.06) 0.084 3.74† (0.97–14.39) 0.055 3.10‡ (0.75–12.81) 0.118

International Archives of Occupational and Environmental Health
1 3
Comparison withtheprevious studies
None of the previous longitudinal studies on chronic psy-
chosocial work stressors and suicidal mortality or suicidal
ideation have considered a family history of suicide in their
analysis. The current study newly showed that chronic psy-
chosocial work stressors can increase the risk for suicidal
ideation in working populations, independently of a fam-
ily history of suicide. A family history of suicide was only
marginally associated with total suicidal ideation, but not
moderate/severe suicidal ideation in the current study.
Low job control, its components (low skill discretion
and low decision authority), and long work hours increased
the risk for moderate/severe suicidal ideation in the current
study. This is consistent with the previous longitudinal study
in a Japanese working population (Tsutsumi etal. 2007) in
which job control was inversely associated with suicidal
mortality, although there was no separate analysis for the
components of job control in the Japanese study. Also it is
similar to the previous longitudinal study in a German work-
ing population (Baumert etal. 2014) in which a very crude
composite measure of several work stressors (i.e., overtime,
shift work, night shifts, taskwork, and assembly-line work)
was positively associated with suicidal mortality. However,
the current study has a merit over the German study in dem-
onstrating clearly long work hours per week as a risk factor
for suicidal ideation.
On the other hand, as opposed to the previous longitu-
dinal studies (Baumert etal. 2014; Ostry etal. 2007; Tsut-
sumi etal. 2007), both job strain and high psychological
job demands (albeit not statistically significant) were posi-
tively associated with suicidal ideation in the current study.
In the previous longitudinal studies (Baumert etal. 2014;
Ostry etal. 2007; Tsutsumi etal. 2007), job strain and high
psychological job demands unexpectedly appeared to be pro-
tective against suicidal mortality or suicidal attempt. There
is no clear answer for the discrepancy between the current
study and the previous studies. And they are not completely
comparable to each other due to the difference in the main
outcome of interest (suicidal ideation vs. suicidal mortal-
ity or suicide attempt). Nonetheless, two methodological
differences between the studies deserve some discussion
here for future studies. In the previous longitudinal stud-
ies (Ostry etal. 2007; Tsutsumi etal. 2007), psychological
job demands were assessed with the standard JCQ or JCQ-
like five-item scale. And study subjects were mostly manual
workers (sawmills workers and a rural community sample).
But in the current study, psychological job demands were
measured with time pressure-focused three items without
the two items, “work fast” and “work hard”, of the standard
JCQ demand scale. In addition, study subjects were mostly
non-manual workers (Choi etal. 2010a), although they had
a wide range of occupations. The construct validity of the
standard JCQ job demand scale has been reported to be weak
particularly in physically demanding occupations (Choi etal.
2008, 2012). Also, an alternative three-item JCQ demand
scale without the “work fast” and “work hard” items was
stronger in terms of predictive validity for mental health
indicators and cardiovascular disease risk factors than the
standard five-item JCQ demand scale (Choi etal. 2008;
Garcia-Rojas etal. 2015).
The current study demonstrated that quartile-based
dichotomous job strain is a better predictor for total or mod-
erate/severe suicidal ideation than the typical median-based
dichotomous job strain. This is consistent with the previous
cross-sectional study (Choi etal. 2010b) on job strain and
leisure-time physical activity; however, the current study
newly demonstrated that the differential operationalization
Table 5 Odds ratios and their
95% confidence intervals
of psychosocial working
conditions at T1 for moderate/
severe suicidal ideation at T2 in
US workers who did not report
suicidal ideation at T1 (N = 552)
Model 1 was controlled for age, marital status, race, and family suicide history. Model 2 for job strain or
its components was controlled for the covariates in Model 1 plus job insecurity and work hours per week.
Model 2 for job insecurity was controlled for the covariates in Model 1 plus job strain (quartile based) and
work hours per week. Model 2 for work hours per week was controlled for the covariates in Model 1 plus
job strain (quartile based), and job insecurity
Variables Model1 P value Model2 P value
Low skill discretion 3.04 (0.99–9.31) 0.052 3.48 (1.11–10.94) 0.033
Low decision authority 2.46 (0.80–7.59) 0.117 3.41 (1.05–11.05) 0.041
Low job control 3.33 (1.01–10.99) 0.048 4.09 (1.20–13.98) 0.025
High psychological job demands 2.09 (0.64–6.84) 0.222 1.56 (0.46–5.32) 0.481
Job strain (median based) 3.62 (1.20–10.62) 0.022 3.77 (1.21–11.70) 0.022
Job strain (quartile based) 4.03 (1.26–12.92) 0.019 4.29 (1.30–14.15) 0.017
Job insecurity 1.59 (0.19–13.20) 0.668 1.72 (0.19–15.82) 0.630
Hours of work per week (≥ 41 vs. ≤40) 3.91 (1.06–14.41) 0.040 4.06 (1.08–15.19) 0.038
Hours of work per week (41–56 vs. ≤40) 1.88 (0.30–11.75) 0.499 1.90 (0.29–12.40) 0.502
(49–56 vs. ≤40) 5.92 (1.40–25.16) 0.016 6.34 (1.48–27.26) 0.013
(≥ 57 vs. ≤40) 4.08 (0.77–21.62) 0.098 3.89 (0.73–20.87) 0.113

International Archives of Occupational and Environmental Health
1 3
of job strain could also make a difference in longitudinal
research (Choi etal. 2015; Karasek etal. 2007).
Job insecurity was positively associated with moderate/
severe suicidal ideation in the current study as in the pre-
vious longitudinal study (Kim etal. 2017). However, the
associations were not statistically significant in the current
study. But, the non-significant associations in the current
study should be cautiously interpreted due to the following
two reasons. First, only 4.5% of the study subjects in the
current study reported job insecurity. It should be reminded
that the majority of the study subjects in the MIDUS II
study (2004–2009) were recruited during non-economic
downturns (e.g., before the Great Recession, 2008–2012).
Second, the sample size of the current study was relatively
small. Thus, job insecurity should be further tested in future
studies in a larger sample of working populations.
Supervisor support and coworker support were not
associated with suicidal ideation in the current study. In a
case–control study in a German working population (Sch-
neider etal. 2011), dissatisfaction with supervisors or col-
leagues was not associated with suicidal mortality, while
psychic strain due to contact with clients was associated with
suicidal mortality. In a cross-sectional study in an Australian
working population (Milner etal. 2016), supervisor support
was only marginally associated with suicidal ideation, while
bullying or harassment at work was significantly associated
with suicidal ideation. In one previous longitudinal study
(Ostry e al. 2007), coworker support was inversely associ-
ated with suicide attempt, but not suicidal mortality. How-
ever, the measure for coworker support in the previous study
(e.g., “The worker could interact with co-workers while
they worked”) was different from those (general emotional
and informational support from immediate supervisor and
coworkers) in the current study. Experiences of bullying or
harassment at work, and a lack of social interactions with
coworkers may be better predictors for suicidal behaviors
than general low social support at work from immediate
supervisor and coworkers.
Implications fortheprevention ofsuicide inworking
populations
This study indicates that job strain and long work hours may
be important occupational risk factors for suicidal ideation
in working populations. However, contemporary suicide
prevention programs at the workplace are mostly focused
on training and education for detecting those at high risk
of suicide and connecting them with mental health services
(Milner etal. 2015). Those programs are necessary as the
secondary or tertiary prevention approaches, but insufficient
for addressing the sources of suicide ideation in working
populations. The current study suggests that creating and
maintaining healthy work organization should be another
important strategy for the prevention of suicide in work-
ing populations. Several workplace intervention studies tar-
geting low job control (Bond and Bunce 2001) including
monotonous tasks (Orpen 1979), high workload (Evans etal.
1999), and long work hours (Schiller etal. 2017) have dem-
onstrated that changing adverse psychosocial working condi-
tions are possible and it is also beneficial for workers’ mental
health and job satisfaction. The World Health Organization
(World Health Organization 2006) has also recommended
promoting a healthy workforce by eliminating or reducing
job-related stressors for the prevention of suicide in working
populations.
Limitations
This study has four limitations. First, the information on
psychosocial work stressors was available only at baseline.
Repeated assessments of psychosocial work stressors were
more strongly associated with exhaustion and coronary
heart disease than one-time assessment (Kivimaki etal.
2006). Thus, it is likely that the associations between psy-
chosocial work stressors and suicidal ideation in the cur-
rent study were underestimated to some extent. Second, no
identical measure for suicidal ideation was administered
twice (at T1 and at T2) in the MIDUS II project. Thus, the
two different measures for suicidal ideation at T1 and at T2
were used in the current study. However, it is unlikely that
the main findings of the current study were significantly
affected by the different measures for suicidal ideation due
to the following reasons. The more inclusive (death idea-
tion), proxy measure of suicidal ideation at T1 (Crandall
etal. 2010; Ten Have etal. 2009) in the current study
would have captured most, if not all, suicidal ideation
cases at T1. In addition, there was a very high correlation
(0.76) between death ideation and suicidal ideation based
on a single item each (Milos etal. 2004). Furthermore, the
results in those with and without suicidal ideation at T1 in
the current study were very similar to each other. On the
other hand, the wording of the measure for suicidal idea-
tion at T2 was suicide specific, although it included death
(i.e., though about death or suicide). Despite some possi-
ble overestimation of the proportion of suicidal ideation at
T2 in the current study, it is unlikely that the overestima-
tion was differential according to exposures at T1, includ-
ing job strain and long works hours. Third, the findings of
the current study should be carefully interpreted because
of the underrepresentation of the following groups in the
MIDUS study: younger workers, racial minority workers,
and workers with less formal education. A more represent-
ative sample of US workers will be needed for testing and
confirming the findings of the current study. Nonetheless,
in the current study, workers with the racial background
of Native Americans, Alaska Native Islanders, or Eskimo

International Archives of Occupational and Environmental Health
1 3
were at a higher risk for suicidal ideation than black work-
ers, which is generally in line with the existing literature
(Olson and Wahab 2006; Tiesman etal. 2015). Fourth, the
sample size of the current study is relatively small. Some
of chronic psychosocial work stressors (job insecurity) in
the current study remain to be further tested and confirmed
in relation to suicidal ideation in future larger samples
of working populations. However, it should be balanced
against the fact that the current study is the first study that
examined the longitudinal associations between multiple
chronic psychosocial work stressors and suicidal ideation
in a working population.
Compliance with ethical standards
Conflict of interest The author declares that he has no conflict of inter-
est.
Ethical approval For this type of study formal consent is not required.
References
Amagasa T, Nakayama T, Takahashi Y (2005) Karojisatsu in Japan:
characteristics of 22 cases of work-related suicide. J Occup
Health 47(2):157–164
Baumert J, Schneider B, Lukaschek K etal (2014) Adverse condi-
tions at the workplace are associated with increased suicide risk.
J Psychiatr Res 57:90–95
Bond FW, Bunce D (2001) Job control mediates change in a work
reorganization intervention for stress reduction. J Occup Health
Psychol 6(4):290–302
Centers for Disease Control and Prevention (CDC) (2013) Suicide
among adults aged 35–64 years—United States, 1999–2010.
Mortal Morb Wkly Rep 62(17):321–325
Choi B, Kawakami N, Chang S etal (2008) A cross-national study
on the multidimensional characteristics of the five-item psy-
chological demands scale of the job content questionnaire. Int
J Behav Med 15(2):120–132
Choi B, Schnall PL, Yang H etal (2010a) Sedentary work, low physi-
cal job demand, and obesity in US workers. Am J Ind Med
53(11):1088–1101
Choi B, Schnall PL, Yang H etal (2010b) Psychosocial working con-
ditions and active leisure-time physical activity in middle-aged
us workers. Int J Occup Med Environ Health 23(3):239–253
Choi B, Kurowski A, Bond M etal (2012) Occupation-differential
construct validity of the job content questionnaire (JCQ) psy-
chological job demands scale with physical job demands items:
a mixed methods research. Ergonomics 55(4):425–439
Choi B, Schnall P, Landsbergis P etal (2015) Recommendations for
individual participant data meta-analyses on work stressors and
health outcomes: comments on IPD-Work Consortium papers.
Scand J Work Environ Health 41(3):299–311
Crandall M, Sridharan L, Schermer C (2010) Depression and
thoughts of death among disadvantaged mothers: risk factors
and impact on maternal and child health. Arch Suicide Res
14(3):248–260
Curtin SC, Warner M, Hedegaard H (2016) Increase in suicide in
the United States, 1999–2014. US Department of Health and
Human Services, Centers for Disease Control and Prevention,
National Center for Health Statistics. https ://www.cdc.gov/nchs/
produ cts/datab riefs /db241 .htm
Dienberg Love G, Seeman TE, Weinstein M etal (2010) Bioindica-
tors in the MIDUS national study: protocol, measures, sample,
and comparative context. J Aging Health 22(8):1059–1080
Evans GW, Johansson G, Rydstedt L (1999) Hassles on the job:
a study of a job intervention with urban bus drivers. J Organ
Behav 20(2):199–208
Garcia-Rojas IJ, Choi B, Krause N (2015) Psychosocial job factors
and biological cardiovascular risk factors in Mexican workers.
Am J Ind Med 58(3):331–351
Karasek RA (1979) Job demands, job decision latitude, and mental
strain: implications for job redesign. Adm Sci Q 24(2):285–308
Karasek RA, Gordon G, Pietrokovsky C etal (1985) Job Content
Questionnaire and User’s Guide. Los Angeles (CA). University
of Southern California/University of Massachusetts, Lowell
(MA)
Karasek R, Choi B, Ostergren PO etal (2007) Testing two meth-
ods to create comparable scale scores between the Job Content
Questionnaire (JCQ) and JCQ-like questionnaires in the Euro-
pean JACE Study. Int J Behav Med 14(4):189–201
Kessler RC, Andrews G, Mroczek D etal (1998) The world health
organization composite international diagnostic interview short
form (CIDI-SF). Int J Meth Psych Res 7(4):171–185
Kim MS, Hong YC, Yook JH etal (2017) Effects of perceived job
insecurity on depression, suicide ideation, and decline in self-
rated health in Korea: a population-based panel study. Int Arch
Occup Environ Health 90(7):663–671
Kivimäki M, Head J, Ferrie JE etal (2006) Why is evidence on
job strain and coronary heart disease mixed? An illustration of
measurement challenges in the Whitehall II study. Psychosom
Med 68(3):398–401
Kroenke K, Spitzer RL, Williams JB (2001) The PHQ-9: valid-
ity of a brief depression severity measure. J Gen Intern Med
16(9):606–613
Lee J, Kim I, Roh S (2016) Descriptive study of claims for occu-
pational mental disorders or suicide. Ann Occup Environ Med
28:61
Loerbroks A, Cho SI, Dollard MF etal (2016) Associations between
work stress and suicidal ideation: Individual-participant data from
six cross-sectional studies. J Psychosom Res 90:62–69
Milner A, Page K, Spencer-Thomas S etal (2015) Workplace suicide
prevention: a systematic review of published and unpublished
activities. Health Promot Int 30(1):29–37
Milner A, Page K, Witt K etal (2016) Psychosocial working conditions
and suicide ideation: evidence from a cross-sectional survey of
working Australians. J Occup Environ Med 58(6):584–587
Milner A, Witt K, LaMontagne AD etal (2017) Psychosocial job
stressors and suicidality: a meta-analysis and systematic review.
Occup Environ Med. https ://doi.org/10.1136/oemed -2017-10453
1. (Epub ahead of print)
Milos G, Spindler A, Hepp U etal (2004) Suicide attempts and suicidal
ideation: links with psychiatric comorbidity in eating disorder
subjects. Gen Hosp Psychiatry 26(2):129–135
Neeleman J, de Graaf R, Vollebergh W (2004) The suicidal process;
prospective comparison between early and later stages. J Affect
Disord 82(1):43–52
Nock MK, Borges G, Bromet EJ etal (2008) Cross-national prevalence
and risk factors for suicidal ideation, plans and attempts. Br J
Psychiatry 192(2):98–105
Olson LM, Wahab S (2006) American Indians and suicide: a neglected
area of research. Trauma Violence Abuse 7(1):19–33
Orpen C (1979) The effects of job enrichment on employee satisfaction,
motivation, involvement, and performance: a field experiment.
Hum Relat 32(3):189–217

International Archives of Occupational and Environmental Health
1 3
Ostry A, Maggi S, Tansey J etal (2007) The impact of psychoso-
cial work conditions on attempted and completed suicide among
western Canadian sawmill workers. Scand J Soc Med 35:265–271
Phillips JA, Nugent CN (2014) Suicide and the Great Recession of
2007–2009: the role of economic factors in the 50 US states. Soc
Sci Med 116:22–31
Rockett IR, Regier MD, Kapusta ND etal (2012) Leading causes of
unintentional and intentional injury mortality: United States,
2000–2009. Am J Public Health 102(11):e84–e92
Ryff C, Almeida DM, Ayanian JS etal (2007) Midlife Development
in the United States (MIDUS II), 2004–2006 [computer file].
ICPSR04652-v1. Ann Arbor (MI): Inter-university Consortium
for Political and Social Research [distributor], 2007-03-22. https
://doi.org/10.3886/ICPSR 04652
Ryff C, Seeman T, Weinstein M (2013) National Survey of Midlife
Development in the United States (MIDUS II): Biomarker Biblio-
graphic Citation. Project, ICPSR29282-v6. Ann Arbor. https ://doi.
org/10.3886/ICPSR 29282 .v6 (distributor, 2013-12–20)
Schiller H, Lekander M, Rajaleid K etal (2017) The impact of reduced
work time on sleep and perceived stress—a group randomized
intervention study using diary data. Scand J Work Environ Health
43(2):109–116
Schneider B, Grebner K, Schnabel A etal (2011) Impact of employ-
ment status and work-related factors on risk of completed sui-
cide. A case-control psychological autopsy study. Psychiatry Res
190(2–3):265–270
Simon GE, Rutter CM, Peterson D etal (2013) Does response on the
PHQ-9 Depression Questionnaire predict subsequent suicide
attempt or suicide death? Psychiatr Serv 64(12):1195–1202
ten Have M, de Graaf R, van Dorsselaer S etal (2009) Incidence and
course of suicidal ideation and suicide attempts in the general
population. Can J Psychiatry 54(12):824–833
Tiesman HM, Konda S, Hartley D etal (2015) Suicide in U.S. Work-
places, 2003–2010: a comparison with non-workplace suicides.
Am J Prev Med 48(6):674–682
Tsutsumi A, Kayaba K, Ojima T etal (2007) Low control at work and
the risk of suicide in Japanese men: a prospective cohort study.
Psychother Psychosom 76(3):177–185
Walker J, Hansen CH, Hodges L etal (2010) Screening for suicidality
in cancer patients using Item 9 of the nine-item patient health
questionnaire; does the item score predict who requires further
assessment? Gen Hosp Psychiatry 32(2):218–220
Wang YG, Chen S, Xu ZM etal (2017) Family history of suicide and
high motor impulsivity distinguish suicide attempters from suicide
ideators among college students. J Psychiatr Res 90:21–25
Watson D, Clark LA, Weber K etal (1995) Testing a tripartite model:
II. Exploring the symptom structure of anxiety and depres-
sion in student, adult, and patient samples. J Abnorm Psychol
104(1):15–25
World Health Organization (WHO) (2006) Preventing Suicide: a
resource at work. World Health Organization. http://apps.who.
int/iris/bitst ream/10665 /43502 /1/92415 94381 _eng.pdf
World Health Organization (WHO) (2014) Preventing Suicide: a global
imperative. WHO. http://apps.who.int/iris/bitst ream/10665 /13105
6/8/97892 41564 878_eng.pdf?ua=1&ua=1
World Health Organization (WHO) (2017) Suicide. http://www.who.
int/media centr e/facts heets /fs398 /en