UBC Faculty Research and Publications

A comparison between the effort-reward imbalance and demand control models Ostry, Aleck S; Kelly, Shona; Demers, Paul A; Mustard, Cameron; Hertzman, Clyde Feb 27, 2003

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


52383-12889_2002_Article_54.pdf [ 492.56kB ]
JSON: 52383-1.0223259.json
JSON-LD: 52383-1.0223259-ld.json
RDF/XML (Pretty): 52383-1.0223259-rdf.xml
RDF/JSON: 52383-1.0223259-rdf.json
Turtle: 52383-1.0223259-turtle.txt
N-Triples: 52383-1.0223259-rdf-ntriples.txt
Original Record: 52383-1.0223259-source.json
Full Text

Full Text

ralssBioMed CentBMC Public HealthOpen AcceResearch articleA comparison between the effort-reward imbalance and demand control modelsAleck S Ostry*1, Shona Kelly1, Paul A Demers1, Cameron Mustard2 and Clyde Hertzman1Address: 1Department of Health Care and Epidemiology, University of British Columbia, Canada and 2Institute of Work and Health (Toronto), CanadaEmail: Aleck S Ostry* - ostry@interchange.ubc.ca; Shona Kelly - skelly@interchange.ubc.ca; Paul A Demers - pdemers@interchange.ubc.ca; Cameron Mustard - cmustard@iwh.on.ca; Clyde Hertzman - hertzman@interchange.ubc.ca* Corresponding author    AbstractBackground: To compare the predictive validity of the demand/control and reward/imbalancemodels, alone and in combination with each other, for self-reported health status and the self-reported presence of any chronic disease condition.Methods: Self-reports for psychosocial work conditions were obtained in a sample of sawmillworkers using the demand/control and effort/reward imbalance models. The relative predictivevalidity of task-level control was compared with effort/reward imbalance. As well, the predictivevalidity of a model developed by combining task-level control with effort/reward imbalance wasdetermined. Logistic regression was utilized for all models.Results: The demand/control and effort/reward imbalance models independently predicted poorself-reported health status. The effort-reward imbalance model predicted the presence of a chronicdisease while the demand/control model did not. A model combining effort-reward imbalance andtask-level control was a better predictor of self-reported health status and any chronic conditionthan either model alone. Effort reward imbalance modeled with intrinsic effort had marginallybetter predictive validity than when modeled with extrinsic effort only.Conclusions: Future work should explore the combined effects of these two models ofpsychosocial stress at work on health more thoroughly.BackgroundA strong body of evidence indicates that exposure to ad-verse psychosocial work conditions is a major hazard forthe health of workers in modern economies. Much of thisevidence, accumulated over the past two decades, is basedon the demand/control model [1] in which task-levelwork conditions characterized by low control and highvascular disease as well as high rates of sickness absence[2,3].One of the criticisms of this model is its reliance on "ob-jective" measures of the work environment only [4]. Ac-cording to many critics, workers will respond differentlyto the same constellation of control and demand condi-Published: 27 February 2003BMC Public Health 2003, 3:10Received: 29 October 2002Accepted: 27 February 2003This article is available from: http://www.biomedcentral.com/1471-2458/3/10© 2003 Ostry et al; licensee BioMed Central Ltd. This is an Open Access article: verbatim copying and redistribution of this article are permitted in all media for any purpose, provided this notice is preserved along with the article's original URL.Page 1 of 9(page number not for citation purposes)demand have been shown to predict high rates of cardio- tions leading to varied biological outcomes so that ameasure of individual worker differences, specifically inBMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10coping style, must therefore be included in any job strainmodel.In the early 1990s, the effort-reward imbalance model wasdeveloped [5]. This model postulates that jobs character-ized by a perceived imbalance between high effort andlow rewards are stressful and will lead to negative healthoutcomes, particularly in persons with limited copingabilities. This model is meant to tap the attribute of an in-dividual's "need for control"; a personality characteristicrelated to flexibility in coping. According to the model, aperson with high need for control will respond in an in-flexible way to work situations of high effort and low re-ward; and will therefore be more stressed and diseaseprone than a person in the same situation who has lessneed for control.Using well designed epidemiological studies both modelshave succesfully predicted "hard" disease outcomes (par-ticularly CHD) [6,7]. In the first comparative study withboth models, Bosma has demonstrated independent pre-dictive effects for new coronary heart disease of a compo-nent of the demand/control model (low control) as wellas effort/reward imbalance in a cohort of English white-collar workers [8].The models overlap to some extent as "extrinsic demands"in the effort/reward imbalance model is similar to "psy-chological and physical demands" in the demand/controlmodel. However, the models also differ. The effort/rewardimbalance model includes a measure of coping ability(need for control) which has no counterpart in the de-mand/control model. On the other hand, effort/rewardimbalance excludes any measurement of task-level con-trol. This is important, for as Bosma notes, "recent publi-cations increasingly underscore the special importance oflow job control for a range of health outcomes, includingcardiovascular disease and sickness absence" [[8] p68].The purpose of this investigation is to compare the predic-tive validity of the demand/control and effort reward/im-balance models for self-reported health status and the self-reported presence of any chronic disease condition in asample of former and current sawmill manufacturingworkers. As well, because task-level control is the only el-ement which is absent from the effort/reward imbalancemodel, and because this variable has been consistentlypredictive of a range of health outcomes for the past 2 dec-ades, the predictive validity of the effort/reward imbal-ance model in combination with task-level control is alsotested.Methodsinally gathered to study the impact of chlorophenol anti-sapstain chemicals on British Columbia (BC) sawmillworkers [9,10].Selection of sawmills and workers for the original studyFourteen medium to large sized sawmills, located mainlyin Southwest BC, participated in the original cohort study.Study sawmills were selected on the basis of a long-termhistory of chlorophenol use and availability of intact per-sonnel records. A total of 26,221 workers were enrolled inthe cohort, representing approximately 20 percent of allBC sawmill workers. (This increased to approximately29,000 as workers hired in study mills between 1986 and1996 were added to the cohort.) To be eligible, a workerhad to be employed at a study mill for at least one year be-tween January 1, 1950 and December 31, 1996.The investigation of this sample of 3,000 workers wasoriginally designed to study the impact of a recession andmajor restructuring of sawmills which began in 1980. Ac-cordingly, the year 1979 was chosen as the pre-recession/restructuring "baseline" year. All workers enrolled in thecohort during 1979 were included in this baseline sub-co-hort. A sample of 3,000 workers was randomly selectedfrom the 9,806 workers working in a study sawmill in1979.Locating intervieweesIn order to locate interviewees the 1979 sub-cohort waslinked to the British Columbia Linked Health Database(BCLHDB). This database includes provincial health min-istry files on physician services, hospital discharges, drugprescriptions for the elderly, long term care services,deaths, and births for the years 1985 through to 1996.These data are useful for finding individuals because theyinclude patient postal codes at time of contact with a phy-sician. Ethical approval limited our access to the first 3-digits of the 6-digit postal codes. This allowed us to iden-tify the community where cohort members lived, so thatwe could then find them through public informationsources.The 9,806 workers employed at a study mill in 1979 werelinked probabilistically to the BCLHDB. Linkage efficien-cy was 94.7% so that 3-digit postal codes were obtainedfor 9,282 workers; including 2,920 (97.3%) of the 3000randomly sampled workers. Searches of union pensionplans, electronic telephone databases, and telephonebooks (by hand) were undertaken to obtain full addressesfor the 3,000 workers. For the unlinked workers in thesample, address searches were undertaken using namesonly.Page 2 of 9(page number not for citation purposes)This investigation is based on a sample of 3,000 male saw-mill workers drawn randomly from a cohort that was orig-BMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10Administering the interviewsFace-to-face interviews were conducted between Novem-ber 1997 and March 1999. Subjects living in remote re-gions of the province were interviewed by telephone. Ashort version of the questionnaire (requiring about 20minutes compared to one hour) was administered by tel-ephone when a respondent was only willing to conduct abrief interview or when proxy interviews were conductedfor deceased or incapacitated interviewees. However, be-cause work-related variables were incompletely deter-mined with the short version of the questionnaire, onlythe long version of the questionnaire was used in thisinvestigation.The instrumentThe instrument was developed after a thorough review ofthe literature on technological change, restructuring, un-employment, and health and work. Two focus groupswere conducted with experienced sawmill workers to fi-nalize the questionnaire; it was then pilot tested on 29 re-tired sawmill workers.Socio-demographic information and health behaviours(smoking and alcohol consumption) were measured. In-come was measured for the year preceding interview aswas the number of dependants supported by each subject.Income per dependant over the year prior to interview wascategorized into quartiles. Education status was catego-rized as completed elementary, secondary, apprenticetraining, community college, and university.Because of downsizing/restructuring over the follow-upperiod, many workers moved out of the sawmill sectorand in and out of employment. Variables measuring non-psychosocial work conditions, which may confound asso-ciations between current psychosocial work conditionsand self-reported health outcomes, such as history of un-employment, sector of employment and occupational cat-egory at time of interview were developed as follows. Dataployed outside the sawmill sector). Data on unemploy-ment history was also obtained and categorized asfollows: no unemployment, 1 episode less than 1 year induration, 1 episode greater than 1 year in duration, 2 epi-sodes or more less than 1 year in duration, 2 episodes ofmore greater than 1 year in length.All job titles obtained in the interviews were re-coded us-ing the Canadian Standard Occupational Classification[12] and then translated into the Pineo16 OccupationalStatus Scale [13]. This 16 category scale was collapsed into4 basic categories; professional/managerial, trades, semi-skilled, and unskilled to measure current occupationalcategory.Task-level work characteristics were measured using ashortened version [10] of Karasek's demand/control in-strument [14,15]. Scores for control and psychological de-mand were divided into high and low categories at themedian. Jobs which were high in psychological demandand low in control were categorized as high strain.Esteem reward, status control, extrinsic effort, and needfor control were measured in the "full" effort/reward im-balance model [5]. The two effort scales were constructedby summing across questions and dichotomizing the scalescore with zero for the two bottom tertiles and one for thehighest tertile representing high extrinsic effort and highneed for control. Two reward scales were constructed bysumming across questions and dichotomizing the scalescore with zero for the two top tertiles and one for the bot-tom tertile representing low esteem reward and low statuscontrol. This "full" model was thus based on four varia-bles, intrinsic demand, extrinsic demand, esteem reward,and status control. These four dichotomous variables wereused to create the effort/reward imbalance indicators con-sisting of three categories: 1=neither high effort nor lowreward; 2=either high effort or low reward; and 3=bothhigh effort and low reward used in the "full" effort/rewardTable 1: Interview status of 3,000 randomly sampled workersInterview Status Number PercentLong questionnaire 1885 62.9Short questionnaire 270 9.1Questionnaire Respondents sub-total 2155 72.0Refusals 126 4.2Deceased 18 0.6Needs translator 8 0.3Not located 693 22.8Total 3000 100.0Page 3 of 9(page number not for citation purposes)on current work sector was obtained and dichotomized(currently employed in the sawmill sector vs currently em-imbalance model[5,8,16].BMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10A "partial" effort/reward imbalance model was also devel-oped by eliminating one of the four components, intrinsiceffort, from the full model. This "partial" model was thusbased on three variables of the variables (extrinsic de-mand, esteem reward, and status control) utilized in thefull model. Because the demand/control model containsno information on the personal characteristics of theworker, using a partial effort/reward imbalance model,i.e., with the psychological measure (intrinsic demand)removed, allows for a comparison of the "objective" ele-ments of both models.The combined effect of task-level control and the effort-re-ward imbalance models (full and partial) was tested bycategorizing effort/reward imbalance into three categories(none, medium, and high) and task-level control into thetwo categories high and low. In this way workers were cat-egorized into, at one extreme, a reference category of noeffort/reward imbalance and high task-level control and,at the other extreme, a category of high effort/reward im-balance and low task-level control with 4 categories repre-senting the possible combinations of effort/rewardimbalance and task-level control between the twoextremes.Two outcome variables were used in this investigation;self-reported health status, and the presence of one self-re-ported chronic condition. Self-reported health was report-ed on a 5 point scale for current job and dichotomizedinto "good" (good or excellent) and "poor" (fair, poor,bad) health status for use in logistic regression analysis.Self-reported health is dichotomized, with a cut point be-tween good and fair in most studies of work stress of thistype [17]. And, any worker who reported one or more ofthe following conditions at time of interview was consid-diabetes, CHD, hearing loss; or any other non-specifiedchronic condition.AnalysesLogistic regression was used to determine the associationbetween self-reported health status or self-reported chron-ic disease status and with various exposure variables. Inthe first model, the association between both outcomevariables and sociodemographic variables was examined(Table 3). In the second model, sociodemographic varia-bles and current smoking status were included as adjust-ment variables in a model examining the associationbetween the outcome variables and three non-psychoso-cial work variables, current sector, current occupationalcategory, and unemployment history (Table 4).In the third model, after controlling for socio-demograph-ic variables, smoking, and non-psychosocial work condi-tions, associations with demand/control and effort/reward imbalance (full and partial) were tested (Table 5).In the fourth model associations were tested after combin-ing effort/reward imbalance (full and partial) with task-level control (Table 6, 7).ResultsTable 1 shows that the overall response rate was 72 per-cent, the refusal rate was 4.2 percent, and 19% of respond-ents were not located. The proportion of workers notlocated was highest in isolated "mill towns" with a young-er more transient workforce than at other sawmills in thecohort. Refusal rates did not vary by age category but the"not found" rate was higher in younger age groups andworkers with lowest duration of work in a study sawmill.Some respondents did not speak English well enough forTable 2: Age and labour force status of long questionnaire respondents at time of interviewCurrent Status Number PercentAge Status65 Years or Over 464 24.664 Years or Less 1421 75.4Total 1885 100.0Labour force StatusSawmill Sector 600 42.3Non-Sawmill Sector 570 40.1Early Retired 131 9.2Unemployed 69 4.8Disabled 40 2.8Other 11 0.8Total 1885 100Page 4 of 9(page number not for citation purposes)ered to have a chronic disease: asthma, back problems (ex-cluding arthritis), chronic bronchitis or emphysema,the interview. Cantonese and Punjabi speaking translatorswere hired to administer interviews to subjects who spokeBMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10these languages. Eight people, with other languages re- Table 2 shows that 1,421 (75.4%) of the 1,885 long ques-Table 3: Age adjusted associations of socio-demographic and smoking factors with poor self-reported health status (SRHS) and self-reported chronic condition (Chronic).Variable Chronic SRHSAge35–39 1.00 1.0040–44 0.99 1.69*45–49 1.37 1.77*50–54 1.88** 1.72*55–59 2.32** 2.06**60–64 4.25** 2.50**BirthplaceCanada 1.00 1.00Outside Canada 0.66** 1.32EducationElementary 1.00 1.00Secondary 0.99 0.75Apprentice 1.34 0.76Community College 1.05 0.75University 0.71 0.60*Marital statusMarried 1.00 1.00Unmarried 1.08 1.05Income/dependant<$12,999 1.00 1.00$13,000–18,749 1.31 0.87$18,750–28,229 1.34 0.75>$28,00 1.20 0.65*Current Smoking statusNo 1.00 1.00Yes 1.11 1.55***p = 0.05–0.01; **p = <0.01Table 4: Socio-demographic and smoking adjusted associations of non-psychosocial work condition variables with poor self-reported health status and self-reported chronic condition.Variable Chronic SRHSSectorNon-Sawmill 1.00 1.00Sawmill 1.09 1.57**Current Occupational CategoryManager 1.00 1.00Trades 1.30 1.08Unskilled 0.89 1.26Unemployment History1 episode< 1 year 1.00 1.001 episode > 1 year 1.19 1.122 episodes or more <1 year 1.01 0.882 episodes or more >1 year 1.21 1.22*p = 0.05–0.01; **p = < 0.01Page 5 of 9(page number not for citation purposes)quired translators but due to expense and logistics theseinterviews were not conducted.tionnaire respondents were aged 64 or less. Of these, 1170(82.3%) were employed at time of interview, 131 (9.2%)BMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10were retired, 69 (4.9%) were unemployed, 40 (2.8%)were disabled, and 11 (0.8%) were in at home looking af-ter children, performing voluntary work, or attending ed-ucational institutions at time of interview. Of the 1170workers employed at time of interview, 600 (51.3%) wereemployed in a sawmill and 570 (48.7%) were employedoutside the sawmill sector. The analyses described in thispaper were based on the 1170 respondents who answeredthe long questionnaire and who were aged 64 or underand employed at time of interview.Table 3 presents age-adjusted associations between self-re-ported health status and chronic conditions and sociode-associated in a step-wise gradient with increasing age. Forthe remaining variables, no statistically significant associ-ations were observed with either outcome variable.Self-reported health status declined step-wise with in-creasing age. Increasing education was positively associat-ed, in gradient fashion, with increasing self-reportedhealth status. This was statistically significant for universi-ty education (OR = 0.60). Current smoking was associatedwith worse self-reported health status (OR = 1.55).Table 4 shows that, after adjustment, none of the non-psy-chosocial work variables demonstrated a statistically sig-Table 5: Associations between the demand/control and effort/reward imbalance models (full and partial) and poorself-reported health status and any chronic condition.Self-reported Health Status Any Chronic conditionControlHigh 1.00 1.00Low 1.60** (1.12–2.28) 1.09 (0.81–1.45)Psychological demandLow 1.00 1.00High 1.65** (1.21–2.26) 1.13 (0.86–1.48)Physical demandLow 1.00 1.00High 1.01 (0.68–1.50) 1.25 (0.86–1.81)Job StrainLow 1.00 1.00High 2.07* (1.18–3.66) 1.31 (0.77–2.22)Partial Effort/reward Imbalance ModelNone 1.00** 1.00*Medium 1.60* (1.06–2.45) 1.12 (0.0.84–1.44)High 3.13** (1.96–4.85) 1.59** (1.12–2.24)Full Effort/reward Imbalance ModelNone 1.00** 1.00**Medium 1.87** (1.26–3.34) 1.22 (0.91–1.85)High 3.35** (2.10–5.61) 1.70** (1.17–2.37)Numbers in parentheses are 95% confidence intervals. *p = 0.05–0.01;**p = 0.01–0.0001Table 6: Associations between effort/reward imbalance (full and partial) model combined with task-level control and poor self-reported health statusFull Model Partial ModelNO Imbalance with HIGH control 1** 1**NO Imbalance with LOW control 0.75 (0.37–1.48) 0.71 (0.41–1.25)MEDIUM Imbalance with HIGH control 1.40 (0.77–2.69) 1.16 (0.66–2.02)MEDIUM Imbalance with LOW control 1.74 (0.97–3.10 1.54 (1.07–2.55)HIGH Imbalance with HIGH control 2.20** (1.25–3.99) 2.09** (1.25–3.50)HIGH Imbalance with LOW control 3.50** (2.04–6.08 3.23** (2.01–5.18)Numbers in parentheses are 95% confidence intervals. *p = 0.05–0.01; **p < 0.01Page 6 of 9(page number not for citation purposes)mographic variables, smoking, and self-reported healthstatus in 1979. The presence of a chronic condition wasnificant relationship with the presence of a chroniccondition. In the case of self-reported health status, cur-BMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10rent employment in the sawmill sector was associatedwith greater odds (statistically significant) for reportingpoor health (OR = 1.57) and decreasing occupational sta-tus was associated with a (non-significant) gradient forworse self-reported health.Table 5 shows that low control (OR = 1.60; CI = 1.12–2.28) and high psychological demand (OR = 1.65; CI=1.21–2.26) predicted poor self-reported health. Effort/re-ward imbalance and job strain both predicted poor healthstatus. The risk of reporting poor health status for subjectswith high job strain was approximately twice as high asthose with low strain (OR = 2.07; CI= 1.18–3.66). And,for both the full and partial effort/reward imbalance mod-els the risk of reporting poor health status for subjectswith both high effort and low reward was approximately3 times higher than those with low effort and high reward(full model; OR = 3.35; CI= 2.10–5.51) and (partial mod-el; OR = 3.13; CI= 1.96–4.85).Effort/reward imbalance (full and partial models) was theonly variable which predicted the presence of a chroniccondition. The risk of reporting a chronic condition forsubjects with both high effort and low reward was 59 per-cent greater than those with low effort and high reward inthe case of the partial model and 70 percent greater withthe full model. The risk of reporting a chronic conditionfor subjects with high job strain was approximately 30%greater than those with low strain jobs.For self-reported health status, combining effort/rewardimbalance with task-level control produced odds ratioswhich increased in a regular gradient moving from the ref-erence category (no effort/reward imbalance with hightask-level control) to the "worst" category (effort/rewardimbalance with low task-level control) (Table 6). For thefull effort/reward imbalance model, the odds ratio for thislatter category was 3.50 (CI = 2.04–6.08). For the partialmodel the odds ratio for this latter category was 3.23 (CI=worst" category (effort/reward imbalance with low task-level control) was 1.98 (CI= 1.23–3.18) and for the partialmodel the odds ratio was 1.80 (CI= 1.14–1.80) (Table 7).Finally, in the case of self-reported health status, the Mod-el Chi Square for the full effort/reward imbalance modelcombined with task-level control was 71.95 compared to65.1 for effort/reward imbalance alone. Results were sim-ilar in size and trend for the partial effort/reward imbal-ance model combined with task-level control. In the caseof "any chronic condition" the Model Chi Square for thefull effort/reward imbalance model combined with task-level control was 46.67 compared to 44.97 for effort/re-ward imbalance alone. Results were similar in size andtrend for the partial effort/reward imbalance model com-bined with task-level controlDiscussionIn this investigation, effort/reward imbalance and jobstrain independently predicted self-reported health statusand both the full and partial effort/reward imbalancemodels predicted the presence of a chronic condition. Aswell, both the full and partial effort/reward imbalancemodels in conjunction with task-level control predictedself-reported health status and the presence of a chroniccondition.The odds ratio for self-reported health status with thecombination high effort/reward imbalance (full model)and low task-level control was 3.50 and the odds ratio inthe case of the partial effort/reward imbalance model incombination with low task-level control was 3.23. Theseodds ratios were slightly higher than those obtained usingthe effort/reward imbalance model alone and approxi-mately 50 percent greater than odds rations obtained us-ing the demand control model (i.e., the job strainvariable) alone. Similarl results were obtained for the out-come "any chronic condition".Table 7: Associations between effort/reward imbalance (full and partial) model combined with task-level control and poor self-report of any chronic conditionFull Model Partial ModelNO Imbalance with HIGH control 1* 1NO Imbalance with LOW control 1.04 *0.63–1.72) 0.97 (0.67–1.58)MEDIUM Imbalance with HIGH control 1.13 (0.70–1.83) 1.0 (0.65–1.56)MEDIUM Imbalance with LOW control 1.38 (0.83–2.27) 1.22 (0.77–1.93)HIGH Imbalance with HIGH control 1.57 (0.95–2.49) 1.38 (0.86–2.20)HIGH Imbalance with LOW control 1.98** (1.23–3.18) 1.80* (1.14–1.80)Numbers in parentheses are 95% confidence intervals. *p = 0.05–0.01; **p < 0.01Page 7 of 9(page number not for citation purposes)2.01–5.18). In the case of chronic conditions, for the fulleffort/reward imbalance model, the odds ratio for theIs the combined model (effort/reward imbalance withtask-level control) a better predictor of the two outcomeBMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10variables than effort/reward imbalance or task-level con-trol alone? As noted, odds ratios were slightly higher forthe combined models compared to effort/reward imbal-ance and task-level control on their own. The full effort/reward imbalance task-level control model explained11.7% and 41.1% more variance in self-reported healthstatus than the effort/reward imbalance model and task-level control alone. Results for the partial effort/rewardimbalance task-level control model were similar. And,similar, but less pronounced trends were observed for theoutcome "any chronic condition". These results both con-firm those obtained from the Whitehall study [8,18] andextends it.The predictive ability of the full compared to the partial(i.e., without intrinsic effort) effort/reward imbalancemodel was only marginally greater for both health out-comes. While intrinsic effort is a major theoretical compo-nent of the effort/reward imbalance model, its did not, inthis study, contribute markedly to enhanced predictiveability for the model. The use of the partial effort/rewardimbalance model should be further explored with otherdata, with "hard" outcomes, to determine empiricallywhether or not intrinsic effort adds substantially to thepredictive ability of the model.Major strengths of this study are that 1) both effort/rewardimbalance and demand/control measures were obtainedfor each individual, 2) the sample was large (1,170 work-ers), 3) the workers, were at the time of the survey, em-ployed across many sectors including manufacturing andthe service, transportation, and construction sectors, and4) the occupational status of workers ranged from un-skilled to skilled professionals. Thus, in spite of the origi-nal sampling frame (which meant that all those sampledhad been employed in the sawmill industry approximate-ly 20 years prior to the survey), the sample represented afairly heterogeneous group of middle-aged, male, formerresource manufacturing workers.The most serious, limitation of this study arises becauseboth explanatory and outcome variables are based on self-reports at time of interview. It is possible that "soft" de-pendent variables such as self-reported health status mayderive from the same conception of self as explanatoryvariables like psychosocial work conditions. In this case,there is a problem of common methods' variance inwhich the independent and dependent variables are hard-ly distinguishable [19] resulting in the possibility of con-tamination between measures [20].Bias arising from common methods' variance may be agreater problem for the full effort/reward imbalance mod-ure included in the full effort-reward imbalance model isessentially a measure of coping ability which because ofits subjective nature may be more vulnerable to commonmethods' variance than the other variables in the effort/re-ward and demand/control models. Use of the partialeffort/reward imbalance model may mitigate anycommon method's variance related to the intrinsic effortvariable.As well, some researchers have argued that any associa-tions observed between self-reports of psychosocial workconditions and health outcomes may be confounded bythe subjective "state" or personality of the worker [21–23]. According to this perspective, the major factor re-sponsible for this confounding is "negative affectivity"and that the impact of this confounding is so great thatself-reports of job work conditions are essentially a meas-ure of negative affectivity [21].However Bosma has demonstrated, using the demand/control model and data from the Whitehall study, "an ab-sence of consistently stronger effects of job control in par-ticipants with reported negative personal characteristics[which] also indicates that a neurotic tendency to com-plain cannot explain the job control-CHD association"[[24] p406]. These and other recent findings have demon-strated that it may not be useful to measure and controlfor negative affectivity in studies using self-reports of psy-chosocial work conditions [25].ConclusionsIn summary, task-level control and effort-reward imbal-ance were independently associated with self-reportedhealth status and effort reward imbalance was associatedwith self-reported presence of any chronic condition.Modeled in combination, effort/reward imbalance andtask-level control was more predictive for both outcomesthan the effort/reward imbalance and demand/controlmodels alone. The predictive power of the full effort/re-ward imbalance model was only marginally greater thanfor the partial model using both health outcome meas-ures. Future work should explore the combined effects ofthese two models of psychosocial stress at work on healthmore thoroughly.Competing interestsNone declared.Authors' contributionAO conducted the research, organized data gathering, andwrote the paper. SK carried out data cleaning tasks. PDprovided guidance in study design. CM provided writinghelp and conceptual design. CH provided conceptualPage 8 of 9(page number not for citation purposes)el relative to both the partial effort/reward balance modeland the demand/control model. The intrinsic effort meas-help.Publish with BioMed Central   and  every scientist can read your work free of charge"BioMed Central will be the most significant development for disseminating the results of biomedical research in our lifetime."Sir Paul Nurse, Cancer Research UKYour research papers will be:available free of charge to the entire biomedical communitypeer reviewed and published immediately upon acceptancecited in PubMed and archived on PubMed Central BMC Public Health 2003, 3 http://www.biomedcentral.com/1471-2458/3/10Note1In this study effort and reward variables were constructedfrom questionnaire items which were similar to but werenot the same as items used by Johannes Siegrist in his ef-fort/reward imbalance instrument.AcknowledgmentsI'd like to thank Dr. Johannes Siegrist and Dr. Micky Kerr for their help in reviewing this paper. Acknowledgments to the Canadian Institute of Health Research and Micheal Smith Foundation for Health Research for Dr. Os-try's salary support. As well, acknowledgments to the Canadian Population Health Initiative, the Institute of Work and Health, the Canadian Institute of Advanced Research, the Center for Health Services and Policy Research, and Forest Renewal British Columbia for their financial and intellectual sup-port for this project.References1. Karasek R Job demands, job decision latitude, and mentalstrain: Implication for job redesign. Admin Sci Quart 1979,24:285-3082. Landsbergis PA, Schnall P, Schwartz J, Warren K and Pickering TGJob strain, hypertension, and cardiovascular disease: Recom-mendations for further research. In Organisational risk factors forjob stress (Edited by: Sauter SL, Murphy LR) Washington: American Psycho-logical Association 1995, 97-1123. North F, Syme L, Feeney A, Shipley M and Marmot M Psychosocialwork environment and sickness absence among British civilservants: The Whitehall II study. Am J Public Health 1996, 86:332-3404. Frese M and Zapf D Methodological issues in the study of workstress: Objective vs subjective measurement of work stressand the question of longitudinal studies. In: Causes, coping, andconsequences of stress at work (Edited by: Cooper C, Payne R) NewYork:John Wiley & Sons Ltd 1988, 375-4115. Siegrist J, Peter R, Junge A, Cremer P and Seidel D Low status con-trol, high effort at work and ischemic heart disease: Prospec-tive evidence from blue-collar men. Soc Sci & Med 1990,31(10):1127-11346. Karasek R and Theorell T Healthy work. New York: Basic Books 1990, 7. Siegrist J Adverse health effect of high-effort/low-rewardconditions. Occup Health 1996, 1:27-418. Bosma H, Peter R, Siegrist J and Marmot M Two alternative jobstress models and the risk of coronary heart disease. Am J Pub-lic Health 1998, 88(1):68-749. Hertzman C, Teschke K, Ostry A, Dimich-Ward H, Kelly S andSpinelli J Mortality and cancer incidence in sawmill workersexposed to chlorophenols. Am J Public Health 1996, 87:71-7910. Ostry A, Marion SA, Green L, Demers PA, Hershler R and Kelly STechnological change in relation to changes in psychosocialconditions of work in BC sawmills (1950–1996). Scand J WorkEnviron Health 2000, 26(3):273-27711. Ostry A, Marion SA, Green L, Demers PA, Hershler R and Kelly SComparison of expert-rater methods for assessing psycho-social job strain. Scand J Work Environ Health 2001, 27(1):1-612. Statistics Canada. Standard occupational classifications, Ottawa 1980, 13. Pineo P Revisions of the Pinio-Porter-McRoberts socioeconomic classifica-tion of occupations for the 1981 census, QSEP Research Report No. 125,Program for Quantitative Studies in Economics and Population, McMasterUniversity, Hamilton, Ontario February 198514. Karasek R, Gordon G, Pietrokovsky C, Frese M, Pieper C andSchwartz J Job content instrument: questionnaire and users' guide. Low-ell(MA): University of Massachusetts 1985, 15. Johnson J, Hall E and Theorell T Combined effects of job strainand social isolation on cardiovascular disease morbidity andmortality in a random sample of the Swedish male workingpopulation. Scand J Work Environ Health 1989, 15:271-27916. Stansfeld S, Bosma H, Hemingway H and Marmot M Psychosocialwork characteristics and social support as predictors of SF-36 health functioning: the Whitehall II study Psychoom Med17. Kelly S Self-reported Health: Stitching together a picturefrom the fabric of life. P.h.D. Thesis, University of BritishColumbia. 2003, 18. de Jonge J, Bosma H, Peter R and Siegrist J Job strain, effort-re-ward imbalance and employee well-being: a large scalecross-sectional study. Soc Sci & Med 2000, 50:1317-132719. Kasl S Measuring job stressors, and studying the health im-pact of the work environment: an epidemiologiccommentary. J Occup Health Psychol 1998, 3(4):390-40120. Kristensen T Job stress and cardiovascular disease: a theoreticcritical review. J Occup Health Psychol 1996, 1(3):246-26021. Chen PY, Spector PE and Jex SM Effects of manipulated job stres-sors and job attitude on perceived job conditions:A simula-tion. In: Organizational risk factors for job stress (Edited by: Sauter SL,Murphy LR) Washington D.C.: American Psychological Association 1995, 22. Spector P, Dwyer D and Jex S Relation of job stressors to affec-tive, health, and performance outcomes: a comparison ofmultiple data sources. J Appl Psychol 1988, 73:11-1923. Spector P and Jex S Relations of job characteristics from multi-ple data sources with employee affect, absence, turnover in-tentions, and health. J Appl Psychol 1991, 76:46-5324. Bosma H, Stansfeld S and Marmot M Job control, personal char-acteristics, and heart disease. J Occup Health Psychol 1998,3(4):402-40925. Spector P, Zapf D, Chen P and Frese M Why negative affectivityshould not be controlled in job stress research: don't throwout the baby with the bath water. J Org Behaviour 2000, 21:79-95Pre-publication historyThe pre-publication history for this paper can be accessedhere:http://www.biomedcentral.com/1471-2458/3/10/prepubyours — you keep the copyrightSubmit your manuscript here:http://www.biomedcentral.com/info/publishing_adv.aspBioMedcentralPage 9 of 9(page number not for citation purposes)1998, 60:247-255


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items