-
Two Alternative Job Stress Models andthe Risk of Coronary Heart
Disease
Hans Bosma, PhD, Richard Peter, PhD, Johannes Siegrist, PhD,and
Michael Marmot FFPHM
Introduction
....ts-..h
. . .d.... ,: ; ::68 Amerian J of X P
There is a cumulating body of evidenceon the association between
psychosocialhazards at work and the physical health ofworkers. The
job stress model most fre-quently used is that of job strain.'3
Thismodel postulates that job strain results fromthe joint effects
of high job demands andlow job control. Recent
publicationsincreasingly underscore the special impor-tance of low
job control for a range of out-comes, including cardiovascular
disease andsickness absence.7 Another recently devel-oped job
stress model is that of effort-reward imbalance.) In this model,
there is anexplicit emphasis on individual attributes-
.^.: that is, coping characteristics of high intrin-sic effort,
as defined by the concept of "needfor control."9 Extrinsic efforts,
as defined bya high workload, are also specified. On theother side,
this model also takes three differ-ent sources of rewards into
account: money,esteem, and occupational status control(promotion
prospects, job security). Thefocus is on a negative trade-off
betweenexperienced "costs" and "gains" at workrather than on
specific job task characteris-tics, as in the job strain model.
In a prospective study among Germanblue-collar men, poor
promotion prospectsand job insecurity (low rewards) in menhaving a
high workload and a high need forcontrol (high efforts) predicted
new cardio-
.... vascular events.'0 Thus, worksite-specificstressful
experience has been linked withstressful experience related to
broadersocioeconomic influences of the labor mar-ket and income
distribution. While the twojob stress models clearly differ, there
is alsosome overlap with respect to the dimen-sions of extrinsic
effort (job demands) andesteem reward (social support at
work)."
Given both the promising results andthe different analytical
perspectives of the
two job stress models, more information isgreatly needed on how
they compare intheir association with coronary heart dis-ease,
especially if the potentially overlap-ping dimensions (i.e.,
extrinsic effort andesteem reward) are excluded from the
com-parative analysis. The longitudinal phase ofthe Whitehall II
study of British civil ser-vants12 allowed us to examine whether
cru-cial components of effort-reward imbalanceand job strain are
independently associatedwith new reports of coronary heart
disease.
Subjects and MethodsStudy Population
The Whitehall IL study was set up toinvestigate the degree and
causes of thesocial gradient in morbidity; to study
workcharacteristics, social support, and addi-tional factors
related to the gradient in mor-tality; and, importantly, to include
women.In the study, a new cohort of civil servantswas established
between 1985 and 1988(phase 1). All male and female civil ser-vants
between 35 and 55 years of age in 20London-based civil service
departmentswere sent an introductory letter and screen-ing
questionnaire and were offered ascreening examination for
cardiovascular
Hans Bosma and Michael Marmot are with theInternational Centre
for Health and Society,Department of Epidemiology and Public
Health,University College London Medical School, Lon-don, England.
Richard Peter and Johannes Siegristare with the Institute of
Medical Sociology, Uni-versity of Dusseldorf, Dusseldorf,
Germany.
Requests for reprints should be sent to HansBosma, PhD,
Department of Epidemiology andPublic Health, University College
London MedicalSchool, 1-19 Torrington P1, London, WC1E
6BT,England.
This paper was accepted March 14, 1997.
January 1998, Vol. 88, No. 1
.. .. ...
...,.: .: .:
..
*;111...s* . a........
*I;:..c...i..ci...
..... a....,
01:.
.:.
*:.
-
Job Stress and Coronary Heart Disease
diseases. The response rate was 73% andprobably would have been
higher, hadabout 4% of the civil servants on the listsprovided by
the civil service not movedbefore the study and thus become
ineligiblefor inclusion. In total, 10 308 civil servantswere
examined: 6895 men (67%) and 3413women (33%). The participants
wereapproached again in 1989/1990 (phase 2:postal questionnaire)
and in 1991/1993(phase 3: postal questionnaire and
screeningexamination). The participation rates atthese two phases
were 79% and 83%,respectively; 7372 subjects (72%) partici-pated in
all three phases and 9302 subjects(90%) participated in either
phase 2 orphase 3. The length of follow-up was 5.3years on average,
with a range of 3.7 to 7.6years. Full details of the screening
examina-tions are reported elsewhere. 12,13
Coronary Heart Disease
Three indicators of coronary heart dis-ease were analyzed:
angina pectoris,doctor-diagnosed ischemia, or either ofthese
outcomes. Angina pectoris was meas-ured by the Rose questionnaire
and definedas pain located over the sternum or in boththe left
chest and left arm that is precipi-tated by exertion, causes the
person to stop,and goes away in 10 minutes or less.'4
Doc-tor-diagnosed ischemia depended onwhether the subject reported
that a generalpractitioner or hospital doctor ever sus-pected or
confirmed a heart attack or anginapectoris.
The outcomes were assessed at allthree phases. Excluding 230 men
and 144women with any coronary heart disease atphase 1, there were
239 men and 174women reporting any new coronary heartdisease
outcome at phases 2 and 3.
Effort-Reward Imbalance
As there was no original measurementof effort-reward imbalance
at phase 1,proxy measures (available from theauthors) had to be
constructed for the cru-cial components of the model. In the
origi-nal measurement, adverse personal charac-teristics of high
need for control (highintrinsic effort) were assessed using a
well-tested scale that contains 29 items.9 Theseitems measure
characteristics such as com-petitiveness, latent hostility, a high
need forapproval, and an excessive work commit-ment. On the basis
of the highest face valid-ity compared with the original "need
forcontrol" items, 10 items were selected fromthe Framingham type A
questionnaire'5 andthe Cook-Medley hostility questionnaire'6
for assessment at phase 1. The dimensional-ity of these 10 items
was confirmed withfactor analysis. The three
identifiedscales-competitiveness (3 items), work-related
overcommitment (4 items), and hos-tility (3 items)-were considered
validproxy measures of the original "need forcontrol" scale.
On the reward side, information wasavailable on core aspects of
occupationalstatus control, especially on perceived poorpromotion
prospects (1 item) and-on objec-tively determined restricted
occupationalmobility ("poor promotion prospects" and"blocked
career"). A blocked career wasdetermined by a lower than expected
currentemployment grade level. The expectationwas based on the mean
current grade levelfor the different grade levels in which
sub-jects started their careers. Job insecurity, afurther important
aspect of occupational sta-tus control, was not assessed at phase
1because most civil servants were not (yet)threatened by
unemployment (see "Discus-sion'). Information on financial rewards
wasnot available. Thus, the effort-reward imbal-ance in our study
reflected a mismatchbetween personal characteristics (highefforts)
and occupational career characteris-tics (low rewards).
Each effort scale was constructed bysummning the scores on the
individulal itemsand dichotomizing the resulting score(O = bottom
two tertiles and 1= highest ter-tile). The highest tertile
indicated highefforts. Poor promotion prospects were
alsodichotomized using the most adverse tertileto indicate low
rewards. Blocked career wasalready a dichotomous variable
(lowrewards). The indicators were used to createthe effort-reward
imbalance indicator, whichhad three categories: 1= neither high
effortsnor low rewards; 2 = either high efforts orlow rewards; and
3 = both high efforts andlow rewards. This indicator reflects the
theo-retically postulated cumulative effect of highefforts (adverse
personal characteristics) andlow rewards (adverse occupational
career).
Because the hostility questionnairewas included in the
questionnaire at a laterstage of the first phase, 3790 subjects
(37%)had no hostility score. However, these sub-jects did not
differ from the other group intheir risk of newly reported coronary
heartdisease. Furthermore, using an imputationmethod did not result
in substantially differ-ent odds ratios (ORs) for the
effort-rewardimbalance indicator.
Job Strain
At all three phases, the subjects weregiven a questionnaire
asking them about the
central components of the job strain model:high job demands, low
job control, and lowwork support. The items, which were basedon the
job strain questionnaire,217, 8 can befound elsewhere.5
Furthermore, jobs wereexternally assessed at phase 1: in 18 of
20departments, 140 well-informed personnelmanagers provided
information on workpace ("How often does the job involveworking
very fast?") and on the level ofcontrol ("How often does the job
permitcomplete discretion and independence indetermining how, and
when, the work is tobe done?"). Detailed information wasobtained on
individual jobs because 5766different jobs were rated and 8838
subjectsoccupied them. Both self-reported andexternally assessed
work characteristicswere grouped into tertiles.
Job strain was modeled by assigningsubjects who simultaneously
scored abovethe median of job demands and below themedian ofjob
control to the group with jobstrain. All others were assigned to
the "nojob strain" category. In an alternative strat-egy,
interaction terms of job demands andjob control were introduced
into a logisticregression model with the main compo-nents. This was
done separately for the cat-egorical and continuous work
scales.
Statistical Analysis
Logistic regression analyses were usedto determine whether
effort-reward imbal-ance and job strain at phase 1 were relatedto
new reports of coronary heart diseaseduring follow-up (phases 2 and
3). In allanalyses, coronary heart disease cases atphase 1 were
excluded and age and lengthof follow-up were controlled for. In a
sub-sequent analysis, effort-reward imbalanceand job strain were
controlled for eachother. This model was successively adjustedfor
employment grade level, negative affec-tivity, and classical
coronary risk factors,including smoking (never, stopped, orsmoked I
to 10, 11 to 20, or 21 or more cig-arettes per day), cholesterol
(mmol/L),hypertension according to diastolic(> 95 mm Hg) or
systolic (> 160 mm Hg)blood pressure or drug treatment for
hyper-tension, and body mass index (kg/M2),which were all assessed
at phase 1. Nega-tive affectivity, which is the disposition
torespond negatively to questionnaires andwhich may inflate
correlations betweenself-reported work characteristics and
self-reported disease,'9-22 was measured withthe negative affect
subscale of the affectbalance scale.23 24 Ordinal variables, such
asemployment grade level, were representedby dummy indicators in
the analyses.
American Journal of Public Health 69January 1998, Vol. 88, No.
I
-
Bosma et al.
Results
Table 1 shows that women reportedhigh effort and low reward
conditions moreoften than men (48% and 41%, respec-tively).
Moreover, men and women in loweremployment grades reported
effort-rewardimbalance more often than subjects inhigher grades.
This was mainly owing to ahigher prevalence of hostility, a
blockedcareer, and poor promotion prospects in thelower grades;
work-related overcommitmentand competitiveness were more prevalent
inthe higher grades. High job strain was alsosomewhat more
prevalent among womenthan among men (self-reported job strain:17%
and 15%, respectively).
The lowest prevalence ofjob strain wasfound in the highest
employment grades.
This was primarily owing to the strong asso-ciation between high
job control and highemployment grade level. In the highestgrade
level, 60% reported high job controlcompared with 5% in the lowest
grade. Jobdemands were also highest in the highergrades, reported
by 45% compared with10% in the lowest grade level. Similarresults
were found with the external assess-ments.
Table 2 shows that effort-reward imbal-ance is associated with
elevated risks of sub-sequent coronary heart disease. The risk
ofcoronary heart disease for men and womenwho have both high
efforts and low rewardsis about three times as high as that for
sub-jects with low efforts and high rewards(ORs varied from 2.59 to
3.63). Job strainwas not consistently related to new coronary
heart disease reports; only the associationbetween self-reported
job strain and anycoronary heart disease outcome in men
wasstatistically significant (OR = 1.45). Theinteraction terms of
job demands and jobcontrol were not statistically significant
andnot consistently in the expected direction;this is primarily
because (strong) adverseeffects of high job demands were absent.Nor
did low work support affect the coro-nary heart disease outcomes.
These negativefindings did not change when additionalinformation
from phase 2 was used. Lowjob control was, however,
consistentlyrelated to new coronary heart diseasereports. Because
low job control, as meas-ured on two occasions (phases 1 and 2),
wasearlier found to have cumulative effects onnew coronary heart
disease at phase 3,5 sub-sequent analyses will use the mean
ofphases 1 and 2 job control to predict newcoronary heart disease
outcomes at phase 3.(Coronary heart disease cases at phases 1and 2
were excluded in these analyses.)
The interaction terms between sex andeffort-reward imbalance or
job control werenot statistically significant, so all
furtheranalyses were based on the total sample andsex was
controlled for in each logisticregression model. Table 3 shows that
botheffort-reward imbalance and low job con-trol have strong and
significant associationswith the coronary heart disease outcomes
inthe total sample. Odds ratios of any coro-nary heart disease
outcome were 3.14, 2.04,and 1.57 for effort-reward imbalance
(highefforts and low rewards), self-reported lowjob control, and
extemally assessed low jobcontrol, respectively. The
associationbetween job control and doctor-diagnosedischemia was
somewhat smaller.
70 American Joumnal of Public Health
TABLE 1-Number and Percentage of Men and Women Reporting
Effort-Reward Imbalance and Job Strain at Phase 1, by Employment
GradeLevel"
Effort-Reward Job Strain: Job Strain: ExternalNumber Imbalance
(%) Self-Report (%) Assessment(%)
Men 6895 41.2 14.7 11.9Grade level
High 2647 33.3 10.5 7.5Intermediate 3607 44.4 17.6 13.4Low 641
55.6 15.8 20.3
Women 3413 48.1 17.2 18.8Grade level
High 381 36.9 12.4 9.2Intermediate 1336 37.5 20.4 15.7Low 1696
60.8 15.7 23.4
aEmployment grades were grouped into three levels: unified
grades 1-7 (administrators inWhitehall I), executive officers, and
clerical and office support grades. Professionalgrades were
classified with the equivalent administrative or executive
grade.
TABLE 2-Odds Ratios (ORs)8 and 95% Confidence Intervals (Cis) of
New Coronary Heart Disease Reports, by Effort-RewardImbalance and
Job Strain at Phase I
Men WomenAny Coronary Any Coronary
Angina Diagnosed Heart Disease Angina Diagnosed Heart
DiseasePectoris Ischemia Outcome Pectoris lschemia Outcome
OR (95% Cl) OR (95% Cl) OR (95% Cl) OR (95% Cl) OR (95% Cl) OR
(95% Cl)Effort-reward imbalanceLow efforts and high rewards 1.00
1.00 1.00 1.00 1.00 1.00High efforts or low rewards 2.13 (0.97,
4.70) 2.13 (0.75, 6.03) 2.12 (1.05, 4.27) 2.08 (0.63, 6.84) 1.45
(0.18, 11.6) 2.41 (0.74, 7.91)High efforts and low rewards 2.59
(1.17, 5.73) 3.63 (1.30,10.2) 2.98 (1.48, 5.99) 3.14 (0.96,10.3)
3.10 (0.40, 23.8) 3.59 (1.10, 11.7)
Number (events) 3751 (129) 3910 (97) 3724 (178) 1589 (110) 1687
(30) 1588 (125)Job strain (self-reported) 1.40 (0.93, 2.10) 1.16
(0.70,1.94) 1.45 (1.03, 2.06) 1.01 (0.65,1.58) 1.89 (0.90, 3.99)
1.14 (0.76,1.72)Number (events) 4817 (168) 5027 (118) 4784 (227)
2116 (141) 2247 (35) 2116 (160)Job strain (external assessment)
0.91 (0.53,1.57) 1.18 (0.65, 2.14) 1.03 (0.66,1.61) 1.27(0.81,1.98)
0.97 (0.40, 2.39) 1.22 (0.80,1.86)Number (events) 4169 (149) 4351
(106) 4143 (204) 1839 (122) 1958 (32) 1841 (140)aAdjusted for age
and length of period between phases 1 and 3; coronary heart disease
cases at phase 1 were excluded; new coronary heartdisease reports
at phases 2 or 3 were the outcome.
January 1998, Vol. 88, No. I
-
Job Stress and Coronary Heart Disease
Subjects with low job control reportedeffort-reward imbalance
conditions moreoften than subjects with high job control(51% and
36%, respectively). Despite thisassociation, both characteristics
were par-tially independently associated with newreports of any
coronary heart disease out-come (Table 4). The odds ratios only
mar-ginally decreased when job control andeffort-reward imbalance
were simultane-ously adjusted for (model 2). Additionaladjustments
for employment grade level,negative affectivity, and coronary risk
fac-tors only marginally affected the odds ratiosfor the work
characteristics. When gradewas controlled for, the odds ratios for
self-reported low job control increased strik-ingly (from 2.04 to
2.44). This was causednot by any unexpected direction of
theunderlying associations but by the combi-nation of a relatively
small number ofevents (115) with strong associationsbetween low job
control, low grade, andeffort-reward imbalance. In the
fullyadjusted model, subjects with high effort-low reward
conditions had more than twicethe risk of any new coronary heart
diseaseoutcome compared with their counterpartswithout such
conditions (OR = 2.15). Lowjob control had independent effects on
newcoronary heart disease reports because theodds ratios in the
fully adjusted model were2.38 and 1.56 for self-reported low job
con-
trol and extemally assessed low job control,respectively.
DiscussionThe findings in the Whitehall II study
further support the predictive validity ofcomponents of two
alternative job stressmodels-the effort-reward imbalancemodel and
the job strain model-for coro-nary heart disease morbidity. In the
lattermodel, only low job control was related tonew reports of
coronary heart disease.Hence, subjects experiencing high effortand
low reward conditions and subjectswith low job control had higher
risks ofnew coronary heart disease than their coun-terparts in less
adverse psychosocial workenvironments.
Effort-Reward Imbalance
Subjects experiencing a mismatchbetween their personal
characteristics andcharacteristics of their occupational careerhad
strongly elevated risks of subsequentcoronary heart disease. More
specifically,competitive, hostile, and overcommittedsubjects
experiencing poor promotionprospects and blocked careers had the
high-est risks. The association between thiseffort-reward imbalance
indicator and the
coronary heart disease outcomes was pres-ent after adjustment
for employment gradelevel, negative affectivity, and coronary
riskfactors and was not significantly different inmen and women.
These findings corroboratethe results found in male German
blue-collar and middle-management popu-lations.8 In a previous
paper based on theWhitehall II study, Ferrie and colleaguesfound
adverse health effects from anticipa-tion ofjob loss orjob
change.25 Their resultsmay be interpreted as providing further
evi-dence for the importance of "status control"(job insecurity,
poor promotion prospects) inthe effort-reward imbalance model.
Low Job Control
Low job control also increased therisks of coronary heart
disease. However,neither high job demands nor low socialsupport nor
the interactions between workcharacteristics (job strain) were
related tothe coronary heart disease outcomes. It ispossible that
specific characteristics of oursample of white-collar workers
contributedto the negative findings for high jobdemands and high
job strain. High jobdemands were more common in the
higheremployment grade levels and were posi-tively associated with
high job control,resulting in a relatively small number ofhigh
strain jobs. Our finding corresponds tothat in the review by
Schnall and col-leagues, in which they concluded that 17 of25
studies that examined main effects foundsignificant associations
between job controland cardiovascular outcome, whereas only8 of 23
studies found significant associa-tions with job demands.3
The importance of (job) control is fur-ther elaborated upon by
several investiga-tors.2-28 More details on the associationbetween
self-reported and externallyassessed low job control and coronary
heartdisease can be found in another Whitehall Hpaper.5 Other
reports based on the White-hall II study have shown the wider
predic-tive validity of low job control as it relatesto sickness
absence,7 fibrinogen,29 and psy-chiatric disorder.24 These results
underscorethe view expressed by Johnson and col-leagues that in the
job strain model, it iscontrol over the work process rather
thanhigh job demands or job strain that increas-ingly emerges as
the main critical compo-nent of a healthy work environment.6
Alternative Job Stress Models
To our knowledge, this is the firstreport that compares
components of theeffort-reward imbalance model and the job
American Journal of Public Health 71
TABLE 3-Odds Ratios (ORs)a and 95% Confidence Intervals (Cis) of
NewCoronary Heart Disease Reports by Effort-Reward Imbalance
atPhase 1, Self-Reported Job Control (Mean Phases I and 2)
andExternally Assessed Job Control at Phase I In the Total
Sample
Angina Diagnosed Any Coronary HeartPectoris Ischemia Disease
Outcome
OR (95% CI) OR (95% Cl) OR (95% Cl)Effort-reward imbalancebLow
efforts and high rewards 1.00 1.00 1.00High efforts or low rewards
2.06 (1.07, 3.98) 2.00 (0.79, 5.06) 2.17 (1.19, 3.95)High efforts
and low rewards 2.78 (1.44, 5.37) 3.55 (1.42,8.90) 3.14 (1.72,
5.71)
Number (events) 5340 (239) 5597 (127) 5312 (303)Self-reported
job contror
High job control 1.00 1.00 1.00Intermediate job control 1.36
(0.83, 2.23) 1.39 (0.79, 2.45) 1.61 (1.04, 2.48)Low job control
2.09 (1.29,3.37) 1 49 (0.81, 2.74) 2.04 (1.32, 3.16)
Number (events) 6565 (132) 6982 (73) 6489 (163)Extemally
assessed job controlb
High job control 1.00 1.00 1.00Intermediate job control 1.28
(0.91, 1.81) 1.08 (0.68,1.71) 1.26 (0.92,1.71)Low job control 1.47
(1.77, 2.02) 1.38 (0.74, 2.09) 1.57 (1.17, 2.08)
Number (events) 6003 (271) 6303 (137) 5979(343)
aAdjusted for age, sex, and length of period between phases 1
and 3.bCoronary heart disease cases at phase 1 were excluded; new
coronary heart diseasereports at phase 2 or phase 3 were the
outcome.
cMean phases 1 and 2 job control; coronary heart disease cases
at phases 1 and 2 wereexcluded; new coronary heart disease reports
at phase 3 were the outcome.
January 1998, Vol. 88, No. I
-
Bosma et al.
strain model. Effort-reward imbalance andlow job control were
independently relatedto the coronary heart disease outcomes.When
these were controlled for one anotherand for other potential
confounders, theodds ratio for effort-reward imbalance was2.15
whereas those for low job control were2.38 and 1.56 for
self-reported and exter-nally assessed job control,
respectively.This suggests that the further refinement ofjob stress
theories may benefit from inte-grating theories on control-related
personalattributes and theories on actual controlover environmental
factors, such as dailytasks (job control) and occupational
career(status control). The cumulative adversehealth impact of low
job control and effort-reward imbalance indicates that both
jobstress factors provide supplementary infor-mation on relevant
stressors in the psy-chosocial work environment.
Conceptual Overlap between JobStress Models
Possible conceptual overlap betweenthe effort-reward imbalance
model andother models needs to be explored in furtherstudies.
First, job demands and work sup-port from the job strain model
closely
resemble extrinsic efforts and esteemrewards from the
effort-reward imbalancemodel. However, the effort-reward imbal-ance
model attaches much importance tothe perception and appraisal of
adversework conditions, whereas the job strainmodel focuses
attention primarily on the"objective" psychosocial work
environment(we used both self-reported and objectivemeasures ofjob
control).2
Second, there might be overlap withhostility, which on its own
has been shownto be strongly related to future coronaryheart
disease.30 Although the odds ratios ofour imbalance indicator did
not change sub-stantially when the total 38-item hostilityscale was
controlled for, the theoretical andempirical contribution of
hostility to effort-reward imbalance is worth some
furtherelaboration. Third, it has also been sug-gested that the
adverse personal characteris-tics (e.g., competitiveness or
hostility) inthe effort-reward imbalance model could bethe result
of low job control and high jobdemands. If this were the case,
controllingeffort-reward imbalance for job controlshould have
resulted in decreased oddsratios of the former. However, the
oddsratio for high efforts and low rewardshardly decreased (from
2.68 to 2.54). More-
over, in a 6.5-year prospective study onblue-collar workers, an
impressive degreeof stability over time was reported for the"need
for control" measures.9'10 Fourth,proponents of the job strain
model haveoften referred to the need to expand the jobstrain model
by including information onjob insecurity (macro-level decision
lati-tude).2 In further studies, the links betweenlow job control
(job strain model) and lowoccupational status control
(effort-rewardimbalance model) should be elaboratedupon and tested
more vigorously.
Further Considerations
Despite our promising findings, somefurther considerations need
to be taken intoaccount. First, given the known variability
inangina reporting,31 which has been repli-cated in our data set,
using new reports ofangina pectoris as an indicator of
incidentcoronary pathology may be problematic.One might speculate
that the underlyingcondition has not altered but that the ten-dency
to report it has changed. A new reportof a doctor diagnosis is
likely to be a betterindicator of new disease, although other
fac-tors may also influence both recall of diag-nosis or access to
medical care. Preliminary
72 American Joumal of Public Health
TABLE 4-Odds Ratios (ORs) and 95% Confidence Intervals (Cis) of
Any New Coronary Heart Disease Outcome by Effort-Reward Imbalance
at Phase 1, Self-Reported Job Control (Mean Phases I and 2), and
Externally Assessed JobControl at Phase 1 in the Total Sample,
Adjusted for Age, Sex, and Length of Period between Phases I and
3(Model 1), Additionally Adjusted for Other Work Characteristic
(Model 2), and Model 2 Adjusted for EmploymentGrade Level, Negative
Affectivity, and Coronary Risk Factors.
Model 2 Separately Adjusted forModel la Model 2b Employment
Negative Coronary Fully
OR (95% CI) OR (95% CI) Grade Level Affectivity Risk Factors
AdjustedEffort-reward imbalancecLow efforts and high rewards 1.00
1.00 1.00 1.00 1.00 1.00High efforts or low rewards 1.93 (1.05,
3.55) 1.88 (1.02, 3.45) 1.87 (1.01, 3.44) 1.70 (0.92, 3.13) 1.93
(1.05, 3.57) 1.77 (0.95, 3.28)High efforts and low rewards 2.68
(1.46,4.91) 2.54 (1.38, 4.67) 2.52 (1.36, 4.65) 2.12 (1.15, 3.91)
2.56 (1.39, 4.72) 2.15 (1.15, 4.01)
N = 4393 (251 events)Self-reported job controld
High job control 1.00 1.00 1.00 1.00 1.00 1.00Intermediate job
control 2.05 (1.22, 3.44) 2.02 (1.20, 3.39) 2.08 (1.22, 3.53) 1.96
(1.17, 3.37) 2.05 (1.22, 3.44) 2.08 (1.22, 3.55)Low job control
2.15 (1.26, 3.67) 2.04 (1.19, 3.49) 2.44 (1.37, 4.38) 1.94 (1.13,
3.34) 1.97 (1.15, 3.40) 2.38 (1.32, 4.29)
N = 4702 (115 events)Externally assessed job controPc
High job control 1.00 1.00 1.00 1.00 1.00 1.00Intermediate job
control 1.53 (1.07, 2.14) 1.52 (1.06, 2.17) 1.52 (1.06, 2.19) 1.51
(1.06, 2.17) 1.52 (1.06, 2.18) 1.53 (1.06, 2.20)Low job control
1.72 (1.23, 2.02) 1.63 (1.16, 2.28) 1.61 (1.11, 2.33) 1.62 (1.16,
2.27) 1.53 (1.09, 2.15) 1.56 (1.08, 2.27)
N = 4393 (251 events)
Note. Subjects with missing values on any of the variables were
excluded from the separate analyses.aAdjusted for age, sex, and
length of period between phases 1 and 3.bModel 1 additionally
adjusted for other work characteristics. Effort-reward imbalance
was adjusted for extemally assessed job control(findings were
similar when mean self-reported job control was adjusted for); mean
self-reported job control was adjusted for effort-rewardimbalance;
externally assessed job control was adjusted for effort-reward
imbalance.
cCoronary heart disease cases at phase 1 were excluded; new
coronary heart disease reports at phase 2 or phase 3 were the
outcome.dMean phases I and 2 job control; coronary heart disease
cases at phases 1 and 2 were excluded; new coronary heart disease
reports atphase 3 were the outcome.
January 1998, Vol. 88, No. I
-
Job Stress and Coronary Heart Disease
results show that 87% of subjects reporting amyocardial
infarction at phase 3 had docu-mented coronary heart disease.
Furthermore,the classic coronary risk factors were relatedto both
doctor-diagnosed ischemia andangina pectoris, suggesting that the
end-points do reflect coronary heart disease, notjust reporting
bias. Despite the likely differ-ent levels of sensitivity and
specificity ofboth endpoints, effort-reward imbalance andlow job
control show consistent associationswith both angina and
doctor-diagnosedischemia. This supports an etiologicalhypothesis.
Future analyses will examine theeffects of effort-reward imbalance
and lowjob control on fatal and non-fatal myocardialinfarction and
biological mechanisms.
Second, small numbers did not permitextensive analyses of the
associationsbetween employment grade level, job con-trol,
effort-reward imbalance, and coronaryheart disease. Because the
interactionbetween grade and any of the job stress fac-tors was
never statistically significant, wewould not expect different
effects of workcharacteristics among grades. The absenceof
"confounding by" or "interaction with"employment grade level
suggests that theassociation between work and coronaryheart disease
does not depend on employ-ment grade level. A related issue is
theextent to which work contributes to theinverse association
between employmentgrade level and coronary heart disease
inWhitehall II. This is further investigated inanother paper, which
concludes that muchof the inverse social gradient in coronaryheart
disease incidence in Whitehall II canbe attributed to differences
in the psychoso-cial work environment.32
Third, some further methodologicalissues need to be addressed.
Although thethree-category effort-reward imbalanceindicator
reflects a mismatch betweenefforts and rewards, there was no
significant(multiplicative) interaction term betweenour proxy
measures of efforts and rewardsin the logistic regression analysis.
Thisshould be further elaborated upon usingoriginal effort and
reward measures. Thehigh number of subjects whose effort-reward
imbalance scores were not avail-able, mainly owing to a delayed
inclusionof the hostility scale in the phase 1 ques-tionnaire,
probably did not bias the resultsbecause these subjects did not
differ fromthe other group in their risk of newlyreported coronary
heart disease. Moreover.using an imputation method did not resultin
substantially different odds ratios for theeffort-reward imbalance
indicator. Giventhat individuals with effort-reward imbal-ance, low
job control, angina pectoris, and
doctor-diagnosed ischemia at phase 1 hadsomewhat lower
participation rates atphases 2 and 3, it is likely that the impact
ofeffort-reward imbalance and low job con-trol on newly reported
coronary heart dis-ease is somewhat underestimated in
theseanalyses. Controlling for whether individu-als had left the
civil service did not affectthe results. Theoretically, infonnation
biasmay have caused overestimated odds ratiosin our analyses
because a complaining atti-tude regarding work and health
(negativeaffectivity) may have resulted in negativereports about
job control, effort-rewardimbalance, and coronary heart
disease.33-36However, because baseline cases wereexcluded in a
longitudinal setting and nega-tive affect balance was controlled
for, it isunlikely that negative affectivity biased
theresults.1922
ConclusionLow job control and high cost/low
gain conditions influence the developmentof heart disease among
men and womenworking in British government offices. Thefinding that
competitive, overcommitted,and hostile subjects with a less
successfuloccupational career and low job controlhave higher risks
of coronary heart diseaseunderscores the advantage of a job
stressmodel combining personal and environ-mental factors. To our
knowledge, this isthe first report showing independent effectson
coronary heart disease of components oftwo alternative job stress
models: the effort-reward imbalance model and the job strainmodel
(ob control). FZ
AcknowledgmentsThe work presented in this paper was supported
bygrants from the Medical Research Council, BritishHeart
Foundation, National Heart Lung and BloodInstitute (2 RO1 HL36310),
Agency for HealthCare Policy Research (5 ROI HS06516), Healthand
Safety Executive, the Institute for Work andHealth, Toronto,
Ontario, and the John D. andCatherine T. MacArthur Foundation
Research Net-work on Successful Midlife Development.Professor
Marmot is supported by a MedicalResearch Council research
professorship. DrBosma is supported by grants from the EU BIO-MED
network "Socio-Economic Variations inCardiovascular Disease in
Europe: The Impact ofthe Work Environment (Heart at Work)."
We thank all participating civil servicedepartments and their
welfare and personnel offi-cers; the Civil Service Occupational
Health Ser-vice and their directors, Dr Elizabeth McCloy, DrGeorge
Somre, and Dr Adrian Semmence; and allparticipating civil
servants.
References1. Karasek RA. Job demands, job decision lati-
tude, and mental strain: implications for jobredesign. Admin Sci
Q. 1979;24:285-308.
2. Karasek RA, Theorell T. Healthy Work:Stress, Productivity,
and the Reconstructionof Working Life. New York, NY: BasicBooks;
1990.
3. Schnall PL, Landsbergis PA, Baker D. Jobstrain and
cardiovascular disease. Ann RevPublic Health. 1994;15:381-41 1.
4. Alterman T, Shekelle RB, Vernon SW, BurauKD. Decision
latitude, psychologic demand,job strain, and coronary heart disease
in theWestem Electric Study. Am J Epidemiol.1994; 139:620-627.
5. Bosma H, Marmot MG, Hemingway H,Nicholson AC, Brunner E,
Stansfeld SA. Lowjob control and the risk of coronary heart
dis-ease in the Whitehall II (prospective cohort)study.
BMJ.1997;314:558-564.
6. Johnson JV, Stewart W, Hall EM, Fredlund P,Theorell T.
Long-term psychosocial workenvironment and cardiovascular
mortalityamong Swedish men. Am J Public Health.1996;86:324-331.
7. North FM, Syme SL, Feeney A, Shipley M,Marmot M. Psychosocial
work environmentand sickness absence among British civil ser-vants:
the Whitehall II study. Am J PublicHealth. 1996;86:332-340.
8. Siegrist J. Adverse health effects of higheffort/low reward
conditions. J Occup HealthPsychol. 1996;1 :27-41.
9. Matschinger H, Siegrist J, Siegrist K,Dittmann K. Type A as a
coping career:towards a conceptual and methodologicalredefinition.
In: Schmidt H, Dembroski TM,Blimchen G, eds. Biological and
Psychologi-cal Factors in Cardiovascular Disease. Hei-delberg,
Germany: Springer; 1986:104-126.
10. Siegrist J, Peter R, Junge A, Cremer P, SeidelD. Low status
control, high effort at work andischemic heart disease: prospective
evidencefrom blue-collar men. Soc Sci Med. 1990;3 1:1127-1134.
11. Kasl SV. The influence of the work environ-ment on
cardiovascular health: a historical,conceptual, and methodological
perspective. JOccup Health Psychol. 1996; 1:42-56.
12. Marmot MG, Davey Smith G, Stansfeld S, etal. Health
inequalities among British civil ser-vants: the Whitehall II study.
Lancet.1991;337: 1387-1393.
13. Beksinska M, Yea L, Brunner E. Whitehall IIStudy. Operations
Manual for ScreeningExamination 1991-93 (Phase s3). London,England:
University College London, Depart-ment of Epidemiology and Public
Health;1995.
14. Rose G, Reid DD, Hamilton PS, McCartneyP, Jarrett RJ.
Myocardial ischaemia risk fac-tors and death from coronary heart
disease.Lancet. 1977;i:105-109.
15. Haynes SG, Levine S, Scotch N,.Feinleib M,Kannel WB. The
relationship of psychosocialfactors to coronary heart disease in
the Fram-ingham study, I: methods and risk factors. AmJEpidemiol.
1978;107:362-383.
16. Cook WW, Medley DM. Proposed hostilityand pharisaic-virtue
scales for the MMPI. JAppl Psychol. 1954;38:414-418.
January 1998, Vol. 88, No. I American Journal of Public Health
73
-
Bosma et al.
17. Karasek RA, Baker D, Marxer F, Ahlbom A,Theorell T. Job
decision latitude, jobdemands, and cardiovascular disease:
aprospective study of Swedish men. Am J Pub-lic Health. 1981 ;71
:694-705.
18. Quinn RP, Staines GL. The 1977 Quality ofEmployment Survey.
Ann Arbor, Mich: Sur-vey Research Center, Institute for
SocialResearch, University of Michigan; 1979.
19. Brief AP, Burke MJ, George JM, RobinsonBS, Webster J. Should
negative affectivityremain an unmeasured variable in the study
ofjob stress? JAppl Psychol. 1988;73: 193-198.
20. Chen PY, Spector PK. Negative affectivity asthe underlying
cause of correlations betweenstressors and strains. J Appl Psychol.
1991;76:398-407.
21. McCrea G. Controlling neuroticism in themeasurement of
stress. Stress Med. 1990;6:237-241.
22. Spector PE. A consideration of the validity andmeaning of
self-report measures of job condi-tions. Int Rev Ind Org Psychol.
1992; 123-151.
23. Bradburn NM. The Structure ofPsychologicalWellbeing.
Chicago, Ill: Aldine; 1969.
24. Stansfeld S, North FM, White I, Marnot MG.Work
characteristics and psychiatric disorder
in civil servants in London. JEpidemiol Com-munity Health.
1995;49:48-53.
25. Ferrie JE, Shipley MJ, Marmot MG, StansfeldS, Davey Smith G.
Health effects of anticipa-tion of job change and non-employment:
lon-gitudinal data from the Whitehall II study.BMJ.
1995:311:1264-1269.
26. deJonge J. Job Autonomy, Well-Being, andHealth. Maastricht,
The Netherlands: Datawyseb.v. Universitaire Pers Maastricht; 1996.
Mas-ter's thesis.
27. Sauter SL, Hurrell JJ, Cooper CL, eds. JobControl and Worker
Health. New York, NY:John Wiley & Sons; 1989.
28. Steptoe A, Appels A, eds. Stress, PersonalControl and
Health. New York, NY: JohnWiley & Sons, 1989.
29. Brunner E, Davey Smith G, Marmot M, Can-ner R, Beksinska M,
O'Brien J. Childhoodsocial circumstances and psychosocial
andbehavioural factors as determinants of plasmafibrinogen. Lancet.
1996;347: 1008-1013.
30. Miller TQ, Smith TW, Turner CW, GuijarroML, Hallet AJ. A
meta-analytic review ofresearch on hostility and physical
health.Psychol Bull. 1996;1 19:322-348.
31. Rose G. Variability of angina. Some implica-tions for
epidemiology. Br J Prev Soc Med.1968;22:12-15.
32. Marmot M, Bosma H, Hemingway H, Brun-ner E, Stansfield S.
Contribution of job con-trol and other risk factors to social
variationsin coronary heart disease incidence.
Lancet.1997;350:235-239.
33. Frese M. Stress at work and psychosomaticcomplaints: a
causal interpretation. J ApplPsychol. 1985;70:314-328.
34. Frese M, Zapf D. Methodological issues in thestudy of work
stress: objective vs subjectivemeasurement of work stress and the
question oflongitudinal studies. In: Cooper CL, Payne R,eds.
Causes, Coping and Consequences ofStress at Work New York, NY: John
Wiley &Sons, 1988:375-411.
35. Kasl SV. The challenge of studying the dis-ease effects of
stressful work conditions. Am JPublic Health. 1981;71:682-684.
36. Kasl SV. Methodologies in stress and health:past
difficulties, present dilemmas, future direc-tions. In: Kasl SV,
Cooper CL, eds. Stress andHealth: Issues in Research Methodology.
NewYork, NY: John Wiley & Sons, 1987:307-318.
74 American Joumal of Public Health January 1998, Vol. 88, No.
I
