Article Text
PDF
Abstract
Background In this article, the link between (1) psychosocial working conditions (job demands, job autonomy, task variation, social support), (2) self-reported health (persistent fatigue, musculoskeletal complaints, emotional well-being) and (3) socioeconomic position (skill levels, occupational status) is explored. The two theoretical pathways linking the psychosocial work environment to socioeconomic differences in health are explored: differential exposure and differential vulnerability. Previously, the focus has often been on social inequalities in exposure to the stressors. The pathway of differential vulnerability in different socioeconomic positions is often neglected.
Methods In a representative cross-sectional sample of 11 099 Flemish (Belgian) wage earners, 16–65 years of age (47.5% women), logit modelling is applied.
Results Higher exposure to psychosocial occupational stressors is associated with a higher prevalence of adverse health outcomes. Lower skill levels and subordinate occupational positions show a higher prevalence of musculoskeletal complaints, but not of persistent fatigue or emotional well-being. High demands, job strain and iso-strain are more common in higher-skilled, supervisory and managerial positions, but have the strongest health-damaging effects in lower socioeconomic positions. Low control is more prevalent in lower-skilled and subordinate positions, while having stronger adverse health effects in higher socioeconomic positions—the same holds for social support, although it has no clear socioeconomic distribution.
Conclusion Differential exposure and differential vulnerability constitute two counteracting forces in constituting the association between the psychosocial work environment and socioeconomic differences in self-reported health complaints among wage earners.
- Social class
- psychosocial factors
- effect modifiers
- employees
- health status me
- employed CG
Statistics from Altmetric.com
Today's work poses primarily mental and emotional instead of physical demands, having specific consequences for health and well-being.1 Probably the most important contribution to the study of these “immaterial working conditions” comes from occupational stress models such as the demand–control–support model (DCS model).2 The DCS model consists of three dimensions. (1) Job demands refer to the work load and the pace of work. (2) Control has two subdimensions: job autonomy, indicating the extent of decision-making autonomy of a worker, and task variation, aiming at the possibility of applying and developing skills and creativity in a job.3 (3) Social support concerns the instrumental and emotional support provided by colleagues and superiors. Moreover, specific combination hypotheses exist. It is important to differentiate between additive and interactive effects.4 The additive effects of job strain (high demands and low control) and iso-strain (high demands, low control, low support) are generally conceived as the most acute situations of job stress.3 The interactive buffer hypotheses assume a moderating effect on the impact of high demands by high control and support, conceived as resources.3 Strong associations of the separate DCS dimensions and the additive combinations with fatigue, musculoskeletal complaints and emotional well-being are reported previously.5–10 This is not the case for the buffer hypotheses6 7; therefore, they will not be tested in this article.
Less often addressed is the distribution and moderation of the DCS dimensions and their health effects by socioeconomic position (SEP). In this study, skill levels and occupational status represent SEP. The first is a combination between educational attainment and skills or knowledge acquired at the job. The latter is conceived as a classification of formal authority positions, ranging from no authority over other workers whatsoever to strategic managerial authority. In this article, the relationships between these indicators of SEP and the dimensions of the DCS model are central. Two pathways are linking them together while (re)producing socioeconomic health differences: differential exposure (mediation) and differential vulnerability (effect modification).11 12 Differential exposure concerns the unequal social distribution of stressors and resources, while differential vulnerability concerns the modification of their effects in different social positions.12 The first pathway presupposes the health relatedness of the dimensions of the DCS model and their unequal socioeconomic distribution in favour of higher SEP. It has been shown that low autonomy and task variation, as well as job strain and iso-strain, are more common in lower SEP, while a reverse pattern can be seen for (psychological) demands.13–15 Social support often shows no clear socioeconomic distribution.14 16 As a result, the contribution of differential exposure to socioeconomic health differences caused by the DCS dimensions is not straightforward. The available evidence on differential vulnerability shows a strengthening impact of low SEP on the effects of high demands, low control and job strain.17–20 As a result, this latter pathway may be expected to reinforce socioeconomic health inequalities.
Methods
Subjects
The data are derived from the 2004 Flemish Quality of Labour Monitor, a survey representative for all wage earners living in the Flemish region of Belgium. A random sample of 20 000 wage earners received a postal questionnaire. Usable questionnaires were returned by 12 095 respondents (60.6% response rate). Of these, 996 were excluded because they stopped working as wage earners between the moment of sampling and the completion of their questionnaires. The remaining 11 099 cases, upon which our current analyses are based, are largely representative for the population with regard to gender, economic sectors and occupational categories.21
Measures
The psychosocial stressors representing the dimensions of the DCS model are job autonomy (11 items), task variation (6 items), quantitative demands of work (11 items) and support from superiors (9 items). The item scales are taken from the Dutch VBBA quality of labour questionnaire, which can be conceived as a benchmark measuring instrument for the quality of work in the Dutch language community.22 Although they are not completely in line with the full recommended version of the Job Content Questionnaire,3 autonomy, task variation, quantitative demands and superior support can be considered as proxies for the corresponding Job Content Questionnaire scales.3 The VBBA questionnaire has been tested frequently, and the measurement scales are found to be one-dimensional, reliable and valid in previous samples23 as well as in the sample considered here.6 The four psychosocial indicators are included in the analyses as tertiles. For the combination hypotheses, job autonomy and task variation are combined into one indicator—job control. A dummy variable for the job strain category (high demands and low control) is created based on the combination of median cut-off demands and control. The iso-strain variable is created in a similar way, combining high demands, low control and low support into one category.
Skill levels
The skills dimension is a constructed indicator. Respondents with no or lower secondary education and those who combine higher secondary education with un/semi-schooled manual, non-manual routine, educational or healthcare occupations are classified as lower skilled (45.6% of the sample). Schooled manual workers, professionals and managerial employees with higher secondary educational attainment, as well as manual, non-manual routine, educational and healthcare workers with higher non-university educational attainment are categorised into the semi-skilled category (35%). Finally, professionals and managerial employees with non-university higher education and all respondents with a university degree are classified as experts (19.3%).
Occupational status
The properties of strategic policy-making and surveillance authority over employees are central to our concept of occupational status. These properties are conceived as prerogatives of the employer.24 In this regard, Clement and Myles25 suggest to make a distinction between “higher managers”, who combine strategic decision-making and surveillance authorities, and “supervisors”, who have no strategic decision-making authority but only delegated surveillance authority. The category of “workers” is lacking both these properties. The indicator resulting from this distinction is constructed by cross-classifying the occupational qualification of higher manager with the property of having authority over other workers. The resulting indicator is composed of higher managers (4.6%), supervisors (17.7%) and workers (77.7%).
Controls
Age is included as a variable into three categories: 16–29 years (20.5%), 30–49 years (61.9%) and 50 years or older (17.6%). Additionally, gender is also included—47.5% of the sample are women.
The dependent variables are three self-reported health outcomes: persistent fatigue, musculoskeletal complaints and emotional well-being—each of them is measured with a reference period of 2 weeks. Persistent fatigue is represented by a single question in the survey, to be answered with “no” or “yes” (prevalence 32.2%). The indicator for “musculoskeletal complaints” is based on a combination of four dichotomous variables: “backache”, “pain in the neck and shoulders”, “muscular pains in the limbs” and “tingling or numb feeling in the limbs”. The responses on these four indicators are summed together and subsequently dichotomised into “0–2 complaints” and “3–4 complaints”—there is a 21.5% prevalence of three complaints or more. The indicator for “problems with emotional well-being” is based on the question: “To what extent were you bothered with emotional problems, such as anxiety, depressive feelings, feelings of irritation or dejection during the last 2 weeks?”. The original five answering categories are dichotomised into “not at all or a little” versus “moderately, rather or very much” (prevalence 25.4%).
Statistical analyses
The socioeconomic distribution of the DCS components is reported by means of descriptive statistics and χ2 significance tests. Next, for each of the outcomes, a number of effect-coded logit models are performed. The analyses in the first step report on the effects of skill levels and social class relations. The associations with the psychosocial stressors are then reported in three models. In the first model, the effects of all psychosocial stressors, adjusted for gender, age and mutual effects are estimated (models 1). The combination variables are only gender- and age-adjusted. Skill levels (models 2) and occupational status (models 3) are added separately. For models 2 and 3, the main effects as well as the interaction effects of the stressors with skill or occupational status are specified. Based on the combination of the main and interactive effects, within-group associations for skill and occupational status categories are reported. All individual parameter effects are transformed into ORs using the method of Alba.26 Moreover, for each categorical difference, 95% CIs are calculated from the asymptotic standard errors of the parameter effects. The CI of a specific category of an independent is to be compared with the reference category of the same independent. The general significance of the parameter effects is assessed by means of Wald statistics.
Results
Social distribution of the psychosocial stressors
The social distribution of the psychosocial stressors shows that low autonomy, task variation, job strain and iso-strain are occurring more frequently in lower-skilled employees, while high demands are more prevalent in semi-skilled and expert-level employees (table 1). The latter category also has a significantly lower percentage of low superior support. Low autonomy, task variation, job strain and iso-strain are also most prevalent in workers without authority. Supervisors and higher managers more often report high demands. Low social support is least prevalent in managerial employees.
Descriptive distribution of the highest-exposure tertiles of the psychosocial stressors over the categories of gender, age, skill levels and authority positions (absolute numbers and percentages)
Health effects of skill levels and occupational status
Skill is significantly associated with the three outcomes (table 2). There is a small negative association with persistent fatigue and emotional well-being, showing a slightly lower prevalence in the lower and semi-skilled categories. For musculoskeletal complaints, the lower-skilled employees have a 2.5 times higher prevalence compared to expert-level employees. These patterns are not fundamentally changed when controlling for DCS dimensions. Occupational status only shows significantly higher odds for musculoskeletal complaints in workers and supervisors. The associations with persistent fatigue and emotional well-being are not significant.
Categorical differences for the main effects of skill levels and occupational status (ORs)
Associations of the psychosocial dimensions with persistent fatigue
In table 3, the effects of the psychosocial stressors on persistent fatigue are shown. In model 1, the effect of job autonomy is insignificant. The other stressors show a clear association, with the “stressed” categories showing a higher prevalence of persistent fatigue. Quantitative demands show the steepest differences of the single dimensions. The gender- and age-adjusted job strain and iso-strain combinations present the most elevated odds in general.
Effects on persistent fatigue: main effects and within-group effects for skill levels and occupational status (ORs)
In the model additionally controlling for skill (model 2), the associations of the stressors diminish; however, they are not fundamentally altered. Task variation, quantitative demands, job strain and iso-strain interact with skill, which leads to significantly different between-group associations. The association with persistent fatigue of low task variation is stronger in the category of expert-level employees. The opposite holds for quantitative demands, job strain and iso-strain, showing stronger associations in lower and semi-skilled employees.
In the model incorporating occupational status (model 3), the effect of task variation becomes insignificant. The effects of the other independent variables are only reduced. No significant interaction effects are found.
Associations of the psychosocial dimensions with musculoskeletal complaints
In relation with musculoskeletal complaints, the effects in model 1 are all significant (table 4). Again, the most adverse tertiles and the categories of job strain and iso-strain are showing the highest prevalence of musculoskeletal complaints, with the most elevated odds for quantitative demands and the combination effects.
Effects on musculoskeletal complaints: main effects and within-group effects for skill levels and occupational status (ORs)
When skill is added (model 2), all main associations remain significant, although they are further reduced. Significant interaction effects with skill are seen for task variation and quantitative demands. Although at the limit of being significant, low task variation is more strongly associated with musculoskeletal complaints in expert-level employees compared to the other skill levels. Quantitative demands are again showing stronger associations in the lower skill levels.
In the model with occupational status (model 3), only the effects of task variation, quantitative demands, job strain and iso-strain remain significant. There are no significant interactive effects.
Associations of the psychosocial dimensions with emotional well-being
The effects on emotional well-being in model 1 are all significant, except for job autonomy (table 5). The most adverse categories are related to lower emotional well-being, with the steepest differences for superior support, quantitative demands and the combined indicators.
Effects emotional well-being: main effects and within-group effects for skill levels and occupational status (ORs)
Incorporating skill (model 2) does not fundamentally alter these results. The interactive effect of skill and superior support is significant: low social support from superiors is more strongly related with the outcome in the higher-skilled employees.
In the model with occupational status (model 3), the associations of the remaining significant stressors again become smaller. No significant interactions are found.
Discussion
First of all, modest skill-related and occupational status-related health differences are reported. Musculoskeletal complaints are more prevalent in the lower-skilled positions and in workers in a subordinate position. Expert-level workers show a slightly higher prevalence of persistent fatigue and emotional well-being. Similar associations between SEP indicators and these self-reported health outcomes are previously shown among wage earners.27–31 The non-significant or even inverse associations of SEP indicators with fatigue may be related to the underlying distribution of (psychological) demands, which are, on average, less high in lower-skilled workers. High psychological demands often show particularly strong associations with fatigue-related outcomes.10 32 The dimensions of the psychosocial work environment are all significantly related with the outcomes. The control dimension is generally rather moderately associated. For persistent fatigue and musculoskeletal complaints, quantitative demands are most strongly associated. For emotional well-being, steep associations exist with superior support. Comparable results are seen in previous studies.10 33 Additional gender-specific analyses based on our data (results not shown) reveal no strong deviations from the picture drawn above.
A small number of significant effect modifications by skill levels of the association between DCS indicators and the health outcomes are reported. The lack of effect modifications by occupational status may be due to the small number of managerial employees in the sample. The effect modifications are not all going into the same direction: task variation and support have a stronger impact in the higher-skilled positions, while demands, job strain and iso-strain have a higher impact in the lower-skilled positions. The latter results are in line with those in previous studies.17 18 20 The direction of the effect modification of superior support and task variation contradicts with previous findings on cardiovascular outcomes.17 18 In one study, incorporating general self-reported outcomes, the socioeconomic effect modification of job strain is not significant; however, this can be due to the rather small sample size.20 In addition, the findings in our study supplement these previously reported results by showing that effect modification in the separate dimensions of the DC(S) model is worth investigating. The effect associated with job strain (or iso-strain) may obscure some counteracting tendencies for the separate components.
The interpretation of this differential vulnerability may be related to the social stress theory.34 35 While control is considered as a resource to deal with demands,3 skills and formal authority positions in the labour process can also be conceived as general social resources.20 These general social resources are tied to important inequality-generating processes, determining a person's SEP in the labour market or society in general. Skills and knowledge are governed by the process of “credentialism”—that is, the mechanism distributing the resource of valuable knowledge in society.36 Credentialism runs through a number of “labelling systems”, such as the educational system, seniority rules, on the job training, etc, offering “certificates of knowledge” that determine someone's social position. A similar process is constituted by the “social class relations” underlying authority positions. Based on the ownership or strategic control over the means of production, social positions in the production process with legitimised authority over the work of others are established.25 These positions are related to the distribution of material and psychosocial demands and resources in the work environment. As a result, privileged positions according to these two SEP dimensions may, on the one hand, help to better deal with elevated job demands and the acute situations of job strain and iso-strain. On the other hand, the combination of high “general resources” (skills, knowledge, authority positions) with low “situational resources” (control) may lead to situations of role conflict,35 leading to a stronger impact of, in this case, low task variation in the higher-skilled employees. Finally, the stronger association of low superior support with emotional well-being in higher-skilled employees can be related to the more important value of social relations in positions requiring complex (interpersonal) problem-solving and decision-making.34 However, it seems obvious that these often subtle effects of resources and role strains are more likely to become manifested in general psychological and somatic complaints than in “harder” clinical outcomes, such as cardiovascular problems.
In interpreting the results, some limitations should be kept in mind. First, in cross-sectional data, the causality going from the DCS components to the health outcomes can only be assumed on the basis of previous empirical research.37 Another possible limitation concerns the 39.4% non-response rate. However, a comparison of the survey respondents with the general wage-earning population and a comparison of early and late responders suggest no important biases. Third, our data contain no formal controls for common method variance or trait effects.38 39 Nevertheless, we know from a previous study on these data that the amount of common variance between all psychosocial stressors and the dependent variables is very limited.6 A final limitation concerns the one-item measurement of persistent fatigue and emotional well-being. The general prevalence of these outcomes is, however, in line with findings in studies using multi-item measures.10 27
In sum, our results add to existing knowledge on the link between the psychosocial work environment and socioeconomic health differences. Given the counteracting socioeconomic distribution of psychosocial stressors—with high demands being more prevalent in higher SEP, and, reversely, low control being more prevalent in lower SEP—the finding of small significant effects of differential vulnerability further complicates the association. It shows that two contradictory forces determine the association between the DCS dimensions and socioeconomic differences in health complaints. On the one hand, high demands of work, while being less prevalent, have stronger effects when present in lower SEP. On the other hand, low control is less prevalent in higher SEP, while having stronger consequences in these positions. Consequently, the DCS dimensions are not univocally increasing socioeconomic health inequalities. However, because of the unequal distribution of other forms of work-related adversity,12 these findings are not necessarily contradicting the general assumption of the work environment as a major source of socioeconomic health differences.
What is already known on this subject
The general conditions of the work environment are generally held responsible for contributing to an important extent to socioeconomic inequalities in self-reported health complaints in the working population.
The role of the psychosocial work environment in constituting these inequalities is not straightforward since the socioeconomic distribution of exposure to high job demands is opposite to that of low control.
What this study adds
In this article, we focus on the two pathways linking the psychosocial work environment to socioeconomic differences in self-reported health: differential exposure and differential vulnerability, in different socioeconomic positions.
Both pathways constitute two sometimes contradictory forces in determining the association between the psychosocial stressors and the socioeconomic differences in health complaints: a double effect of differential exposure and a counteracting (double) effect of differential vulnerability.
Acknowledgments
For this study, the data from the “Flemish Quality of Labour Monitor, 2004”, originated and owned by the Socio-Economic Council of Flanders, are used. The content of this article is at the full responsibility of the authors. The authors would like to thank the owners of the database for giving permission to use their data for the purpose of scientific research.
References
Footnotes
Funding Research Council of the Vrije Universiteit Brussel, Pleinlaan 2, 1050 Brussels, Belgium.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.