Professional Documents
Culture Documents
Psychosocial Work Characteristics and Self Rated Healt in Four Post Comunits Countries
Psychosocial Work Characteristics and Self Rated Healt in Four Post Comunits Countries
Countries
Author(s): H. Pikhart, M. Bobak, J. Siegrist, A. Pajak, S. Rywik, J. Kyshegyi, A. Gostautas,
Z. Skodova and M. Marmot
Source: Journal of Epidemiology and Community Health (1979-), Vol. 55, No. 9 (Sep., 2001),
pp. 624-630
Published by: BMJ
Stable URL: http://www.jstor.org/stable/25569516
Accessed: 07-07-2016 04:39 UTC
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
http://about.jstor.org/terms
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted
digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about
JSTOR, please contact support@jstor.org.
BMJ is collaborating with JSTOR to digitize, preserve and extend access to Journal of Epidemiology and
Community Health (1979-)
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
624 J Epidemiol Community Health 2001 ;55:624-630
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
Work environment and health 625
shown to predict physical and mental health in valid answers. The effort/reward ratio was esti
a number of prospective and cross sectional mated as a continuous measure (the ratio of the
epidemiological investigations,22 23 and recent respective scores) in order to improve the
comparative studies revealed their independent statistical power of this construct. The continu
effects on health.24 25 Based on this evidence,
ous ratio was logarithmically transformed in
the present study includes both models to testorder to place inverse imbalance of the same
the associations of psychosocial work environ magnitude (for example 0.5 and 2) in the same
ment and self rated health. distance from 1 (when effort and reward are
equal). Job strain was defined as the combina
tion of low/high job demand and job control.
Methods
POPULATIONS AND SAMPLES
ADDITIONAL VARIABLES
This was a cross sectional study in five popula Information was collected on a range of other
tion samples in four countries of Central and factors known to influence self rated health.
Eastern Europe. Four samples were based on Subjects were classified into four categories of
populations participating in the WHO attained education: primary or less, vocational
MONICA Project26: six districts of the Czech (apprenticeship), secondary (A level equival
Republic; Warsaw and Tarnobrzeg, Poland; ent), and university degree. An indicator of
and Kaunas, Lithuania. The remaining sample material deprivation was assessed by three
consisted of a baseline survey for a prevention questions about how often the subject's house
programme in the town of Kalocsa, Hungary. hold had difficulties to buy enough food or
All samples were chosen randomly from popu clothes and to pay bills for housing, heating and
lation registers, all surveys were conducted electricity. The possible answers were "never or
between 1995 and 1996. Data were collected
almost never", "sometimes", "often" and
by postal questionnaires (all MONICA sam "always". These responses were coded as 0, 1,
ples) and by an interview (Hungary). Com 2 or 3, and a deprivation score was calculated
pleted questionnaires were received from 6642
as the sum. Occupation position was classified
subjects, and response rates were 73% in into three crude categories: "managerial/
Lithuania, 75% in the Czech Republic, 72% in supervisor", "other employee", and "self em
Warsaw, 76% in Tarnobrzeg, and 94% in Hun ployed". Study subjects were categorised by
gary. The analyses of work related factors was marital status as "married" and "unmarried"
restricted to 3941 working subjects.
people (more detailed information was not
available). We have also constructed a score of
SELF REPORTED HEALTH
"perceived general control" calculated from
Self rated health was assessed by the question nine questions, adapted from the Whitehall II
"How would you rate your health in the last 12 Study and by the MacArthur Study on
months?", with five possible answers: "very Successful Midlife, as described elsewhere.18 28
good", "good", "average", "bad" and "very
bad". For the present analyses, these responses STATISTICAL ANALYSIS
were dichotomised into two categories, with Data were first cross tabulated by centre
participants reporting "bad" or "very bad" (country), and descriptive measures were cal
health classified as "poor health". The question culated. Associations between self rated health
"Have you ever had heart trouble suspected or and work characteristics were estimated by
confirmed by a doctor" was used as a proxy logistic regression. The associations were
measure for history of coronary heart disease. similar across populations; data were therefore
pooled, and the overall results are reported.
PSYCHOSOCIAL FACTORS AT WORK Because all psychosocial factors at work were
Six characteristics of the psychosocial environ measured on a continuous scale, the odds
ment at work were measured: decision author ratios are reported for an increase by 1 stand
ity (four questions), job demand (one ques ard deviation. Thus, the effects of different
tion), job variety (four questions), social variables on self rated health can be directly
support at work (for questions), and effort and compared in quantitative terms. The associ
reward at work (6 and 11 questions, respec ation between self rated health and psychoso
tively) . The questions are shown in the appen cial work characteristics was analysed in
dix. In addition, decision authority and job several steps. Firstly, the odds ratios were
variety were combined into a single variable of adjusted for age, gender, and population. In a
job control.27 Internal consistency of the scores second step, odds ratios were further adjusted
defined above was assessed by Cronbach's a; it for history of coronary heart disease, type of
ranged from 0.74 to 0.81. employment, education, deprivation and mari
For decision authority, job demand, job vari tal status. Thirdly, perceived general control
ety and social support at work, responses were was added to the model. Finally, all psychoso
given at a 4 point scale. The answers were cial work characteristics and covariates were
coded to values 0 to 3, and average scores were entered into the model. Throughout the
calculated. For decision authority, job varietyanalyses, we have tested for interactions (in
and social support at work, the average score multiplicative models) among different psy
was calculated if at least three questions had chosocial factors at work and between psycho
valid data. Responses to questions on effort social and socioeconomic variables. All analy
and reward were coded to values 1 or 2, and ses were performed using STATA statistical
average scores were calculated if a minimum of software (Stata Corporation, College Station,
five and nine questions, respectively, contained USA).
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
626 Pikhart, Bobak, Siegrist, et al
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
Work environment and health 627
Table 3 Odds ratios (OR and 95% CI) of poor self rated health by work related
psychosocial work characteristics
KEY POINTS
Psychosocial work characteristics were
Adjustment 1 Adjustment 2 Adjustment 3 related to self rated health in population
Job demand 1.11 (0.94, 1.11) 1.09 (0.92, 1.29) 1.03 (0.86, 1.23)a samples of working men and women in
Decision authority 0.85 (0.73, 1.00) 0.91 (0.77, 1.07) 0.98 (0.82, 1.18)a four post-communist countries.
Job variety 0.82 (0.70,0.97) 0.85 (0.73, 1.00) 0.83 (0.69, 1.01)a The continuous measure of the imbal
Job control* 0.81 (0.69,0.95) 0.86 (0.73, 1.02) 0.86 (0.72, 1.02)b
Social support 0.87 (0.74, 1.02) 0.91 (0.77, 1.07) 1.04 (0.87, 1.25)a ance of effort and reward at work seemed
Log (effort/reward) 1.60 (1.38, 1.86) 1.51 (1.30, 1.76) 1.51 (1.29, 1.78)a to have the largest effect.
*Job control combines decision authority and job variety. Adjustment 1: age, sex, population, his Among job strain model variables, low job
tory of cardiovascular disease, type of employment, education, deprivation, marital status. Adjust variety was the most strongly associated
ment 2: age, sex, population, history of cardiovascular disease, type of employment, education, with poor health.
deprivation, marital status, perceived control. Adjustment 3: a: age, sex, population, history of
cardiovascular disease, type of employment, education, deprivation, marital status, perceived con
Education, material deprivation and per
trol, all remaining work characteristics except job control, b: age, sex, population, history of ceived control over life were also strongly
cardiovascular disease, type of employment, education, deprivation, marital status, perceived con associated with self rated health in these
trol, job demand, social support, effort-reward imbalance.
populations.
working subjects fulfilled this definition, we
examined whether the binary measured imbal
ance could be replaced by a continuous meas predictor of the outcome. Job variety and job
ure. The age-sex adjusted odds ratios for the control (combining job variety and decision
binary measure of effort/reward imbalance was authority) were marginally significantly associ
2.65 (95% CI 1.78, 3.95). When subjects were ated with poor self rated health.
grouped into quintiles of the effort-reward In additional analyses we explored whether
ratio, the odds ratios (95% CI) for the 2nd to the effects of work related psychosocial factors
5th quintiles, compared with the first quintile, on self rated health varied by socioeconomic
were 0.93 (0.49, 1.74), 2.53 (1.50, 4.27), 3.84 circumstances. To do so, the odds ratios for the
(2.32, 6.36), 4.06 (2.43, 6.79). This, as well as psychosocial factors at work were estimated
further analyses (not shown), suggest a linear within each stratum of education, material
relation. Another indicator of how well the data deprivation, and marital status (table 4). We
explain the dependent variable is the change in found only two statistically significant interac
the log likelihood after including an independ tions: the effect of high work demand on poor
ent variable into a model (and is equivalent to health was stronger among the better educated.
the %2 test at 1 degree of freedom). The %2 was Decision authority and social support were
26.0 after including the continuous measure of related to reduced risk of poor health in
effort reward-imbalance, compared with 7.4 married, but not in unmarried, subjects.
after including the binary measure. This clearly
confirms that the binary measure (which would Discussion
compare the top 10% with the remaining 90% This study found consistent associations of
of subjects) is inferior to the continuous meas education, perceived general control and se
ure in this study. lected psychosocial work characteristics with
Table 3 provides the results of the main self reported health in five population samples
analyses. The odds ratios, adjusted for age, from Central and Eastern Europe. In particu
gender, population, history of cardiovascular lar, there was a strong relation between the
disease, type of employment, education, depri continuous measure of effort-reward imbal
vation, and marital status, were statistically sig ance at work and self rated health. The
nificant for decision authority, job variety, and associations persisted after adjusting for socio
effort-reward imbalance. After further adjust demographic and socioeconomic conditions
ment for "perceived control over life", the and, overall, the effect of these characteristics
effort-reward imbalance and, with borderline on self reported health did not seem to be
significance, job variety, remained associated mediated by socioeconomic status. The fact
with poor health. With all job characteristics in that the associations were similar in the five
one model (adjustment 3), effort-reward im population samples and in both genders
balance at work remained the most powerful supports the robustness of findings. Whereas
Table 4 Odds ratios of psychosocial work characteristics according to different levels of
associations of distinct sociodemographic, so
socioeconomic!sociodemographic factors cioeconomic and psychosocial (for example,
perceived general control) conditions with self
Log Decision Job Work Job Job reported health have been reported
(EIR) authority demand support variety control*
previously,8-1018 no investigation to date, to our
Education
knowledge, explored the contribution of ad
primary 1.73 1.01 0.82 0.69 0.86 0.92
vocational 1.57 0.95 1.13 0.94 0.78 0.85 verse psychosocial work environments, in
secondary 1.68 0.77 1.35 0.85 0.79 0.73 explaining poor self rated health. According to
university 2.42 0.82 2.38 0.95 1.11 0.92 the theoretical assumptions, the risk of experi
test for interaction NS NS p<0.01 NS NS NS
Deprivation encing poor subjective health was higher the
low (0-4.9) 1.73 0.76 1.04 0.82 0.73 0.73 higher the imbalance between efforts and
high (5-10) 1.63 0.89 1.25 0.88 0.81 0.80 rewards.15 With respect to the demand
test for interaction NS NS NS NS NS NS
Marital status support-control model of job strain, no signifi
married 1.78 0.72 1.18 0.74 0.70 0.67 cant interaction terms were observed, in
unmarried 1.85 1.34 1.15 1.32 0.97 1.14
contrast with the theory14; only job variety and
test for interaction NS p<0.01 NS p<0.05 NS p<0.05
job control were consistently associated with
NS: p>0.10. *Job control is the combination of decision authority and job variety. self rated health.
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
628 Pikhart, Bobak, Siegrist, et al
The finding of an adverse effect on health Western societies but applies equally to Central
produced by effort-reward imbalance is con and Eastern European societies that are
sistent with an increasing body of evidence currently subject to rapid socioeconomic trans
derived from prospective and cross sectional formation. It could be speculated that effort/
studies (for overview see Siegrist22 29). How reward imbalance may be particularly stressful
ever, firm conclusions concerning its possible in a society in transformation, when many
causal effect cannot be drawn, because of the share the view that a privileged minority accu
limitations of the cross sectional design. Firstly, mulates fortune without any obvious efforts
some people may have taken worse jobs (with and disregarding the fundamental principle of
less favourable psychosocial conditions) be distributive justice.
cause they were less healthy (reverse causation, In our analyses, job strain and its compo
selection bias). This possibility could only be nents (job demand and decision authority/job
excluded in a prospective study, but previous control) did not predict poor health that well.
studies of social variation in health found little
This may partly be attributable to measure
evidence for selection bias.30 Secondly, the per ment error. The original questionnaire con
ception of self rated health is subjective and can tains three questions on job demand, but we
be influenced by other factors, including social only had one. Similarly, we used only four
and working circumstances (reporting bias). questions on decision authority, while the
People suffering from poor health may respond original questionnaire has eight questions. This
differently to questions measuring effort and
might have led to imprecise measurement of
reward at work, although we excluded the these variables (random misclassification),
impaired or chronically ill people and those on
which would bias the odds ratios towards unity.
long term sick leave. Both biases would result
in overestimation of the effects of work related The larger number of questions on effort and
reward, by contrast, could produce a relatively
factors. However, controlling for perceived
more precise measurement that would lead to
control (which contained three questions on
health locus of control and would therefore apparently larger effects of the effort/reward
ratio. On the other hand, when we combined
reflect differential reporting) did not remove
the effects of psychosocial factors at work. This
decision latitude and job variety into job
control,27 the effect of this combined variable
does not indicate a presence of a major bias.
was not stronger than those of the individual
In addition to the cross sectional design, sev
eral further limitations of this study need to be
components. In general, however, although the
considered. Firstly, by dichotomising the out effects of job demand were inconsistent in sev
come (self rated health), we may lose some eral studies on coronary heart disease,32 33 we
information. However, there is good reason to remain cautious in judging the merits of the
concentrate on a high risk group as previous two models with respect to self rated health.
studies documented adverse effects of subjec The results on the possible interactions
tive health on measures of morbidity and mor between variables are interesting for several
tality.1 3 5 Moreover, others have shown that self reasons. Firstly, there were no interactions
rated health is a continuous measure, and between job control, job demand and social
support at work. Although such interactions
results on the dichotomised measure agree well
with continuous answers.31 Secondly, the ma have been observed in several studies,1419 20 34
jority of information obtained from this study they were not found in other studies,24 32~37 and
was collected by postal questionnaire, a re it is increasingly recognised that job control is
search method susceptible to several sources of the more important dimension of the model.38
bias and error. However, in one population Secondly, it has been often speculated that job
(Hungary), questions were answered in a related "stress" would be more harmful in
standardised personal interview. The fact that lower socioeconomic groups. Our data do not
neither the distribution of answers nor the support this view. Low job variety and low
observed odds ratios of the predicting variables social support at work were related to increased
differed significantly according to data collec risk of poor health in married subject, who are
tion method may reduce the methodological not disadvantaged or isolated. Similarly, the
concern to some extent. The third limitation effects of job demand was more pronounced
concerns the sample selection, and sampling among subjects with better education. Both is
and overrepresentation of urban populations, the opposite than the anecdotal accounts.
with higher than average education and, prob An important methodological finding of this
ably, more favourable working conditions. study relates to the effort-reward imbalance.
Thus the findings may not be directly general Instead of the previously used binary measure,
isable to the whole population, although it is we applied a continuous indicator, the logarith
unlikely that the association between self rated mically transformed ratio of the effort and
health and work related factors would be reward scores. This continuous measure is
biased. interesting for two reasons. Firstly, and most
This study is the first large scale investigation important, it is statistically more efficient, as it
on associations of self reported health with makes use of all original values in the data,
adverse psychosocial work characteristics in rather than reducing the responses to two cat
working populations of post-communist socie egories. Secondly, this measure can be used in
ties. It confirmed that the relation between a populations with low prevalence of exposure to
stressful psychosocial work environment in effort-reward imbalance, defined by the binary
terms of effort-reward imbalance and poor indicator. Other studies should validate the
subjective health is not restricted to modern continuous measure in different populations.
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
Work environment and health 629
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms
630 Pikhart, Bobak, Siegrist, et al
32 Marmot M.G, Bosma H, Hemingway H, et al. Contribution 35 Schnall PL, Landsbergis PA, Baker D. Job strain and cardio
of job control and other risk factors to social variations in vascular disease. Annu Rev Public Health 1994;15:381-411.
coronary heart disease incidence. Lancet 1997;350: 36 Pieper C, LaCroix AZ, Karasek RA. The relation of psycho
235-9. social dimensions of work with coronary heart disease risk
33 Bobak M, Hertzman C, Skodova Z, et al. Association factors: a meta analysis of five United States databases. Am
between psychosocial factors at work and nonfatal myocar J Epidemiol 1989;129:483-94.
dial infarction in a population-based case-control study in 37 Johnson JV, Stewart W, Hall EM, et al. Long-term psychoso
Czech men. Epidemiology 1998;9:43-7. cial work environment and cardiovascular mortality among
34 Hallqvist J, Diderichsen F, Theorell T, et al. Is the effect of Swedish men. Am J Public Health 1996;86:324-31.
job strain on myocardial infarction risk due to interaction 38 Hemingway H, Marmot M. Evidence based cardiology: psy
between high psychological demands and low decision lati chosocial factors in the aetiology and prognosis of coronary
tude? Results from Stockholm Heart Epidemiology Pro heart disease. Systematic review of prospective cohort
gram (SHEEP). Soc SciMed 1998;46:1405-15. studies. BMJ 1999;318:1460-7.
www.jech.com
www.jech. com
This content downloaded from 139.184.14.159 on Thu, 07 Jul 2016 04:39:32 UTC
All use subject to http://about.jstor.org/terms