A prospective study of cumulative job stress in relation to mental health

Instability of employment, rapid change of demands and intensification of work pressure are widely prevalent consequences of economic globalization and technological change [1]. Even in established sectors of industrial production, administration and services of advanced societies experiences of downsizing, mergers and outsourcing are increasingly shared by employees [2]. Surveys of working conditions in Europe indicate that stressful experience recently increased in the European workforce although variations between countries and sectors are observed [3]. Chronic stressful experience at work can adversely affect physical and mental health. This has been documented in a large number of epidemiological studies based mainly on two complementary theoretical concepts, the demand-control model [4‐6] and the effort-reward imbalance model [[7, 8]; see also [9‐12]]. The demand-control model posits that jobs characterised by high quantitative demands in combination with low decision latitude adversely affect health. The focus of the effort-reward imbalance model is put on contractual non-reciprocity where high efforts at work are not met by adequate rewards in terms of money, esteem, promotion prospects and job security.

Both models have been tested in the frame of prospective epidemiological studies, but in a majority of these investigations the measurement of exposure was restricted to baseline assessment. Thus, effects of cumulative job stress on incident disease have not been sufficiently explored so far. However, with regard to the demand-control model, there are a few important exceptions indicating, firstly, that recurrent job stress is indeed associated with elevated risk of ill health [13, 14], morbidity [15, 16], and mortality [17], and, secondly, that reduction of job stress over time results in improved health [18]. No comparable data on health effects of cumulative stress are available from the effort-reward imbalance model. This is particularly critical as this model documents close links with macroeconomic changes, suggesting that downsizing, mergers and outsourcing result in increases of effort-reward imbalance [19] and, thus, may indirectly affect health.

This study analyses the dynamics of stressful work experience over time, based on the effort-reward imbalance model, in relation to mental health, using longitudinal data of a large cohort. We test the hypothesis that the risks of poor mental health after one year are higher among employees who either continuously experience high job stress or who experience an increase in job stress from the first to the second measurement, compared to the remaining employees with either continuously low levels or decreasing levels of job stress over time. Both conditions, continuous exposure and incident exposure to job stress, are more likely to occur under conditions of downsizing and related macroeconomic constraints.

For two additional reasons, the effort-reward imbalance model is chosen to measure stressful experience at work. First, as mentioned, two of the three reward components specified in this model provide a direct link with the labour market dynamics that are becoming increasingly relevant in a globalized economy: promotion prospects including job security and level of salary or wage [20]. In addition the model is composed by an extrinsic component (perceived demands, perceived rewards) and an intrinsic component (coping with demands at work; overcommitment as a motivational risk factor), thus allowing for a differentiation of 'subjectively perceived' situational and personal characteristics of stressful experience at work. Secondly, the standard measurement of this model, a Likert-scaled self-administered questionnaire (see Methods) was shown to be highly sensitive to change of exposure over time [21]. This is an essential prerequisite of studying dynamics of stressful experience at work in a reliable way.

Socio-demographic, psychosocial and health-related characteristics of the sample are shown in Table 1, separately for the four enterprises. As specified above, the different combinations of sub-samples (all participants to T1, all participants to T2 and the paired sample, i.e. participants to both measures) do not differ in terms of sex, age, education, professional qualification as they are very similar in their mental health indicators.

While mean age of the sample is about 40 years and while about half of the sample is composed of women, we observe a considerable variation for those variables across the four enterprises. In particular, participants in enterprise 4 are significantly older and there are less women. In addition, their educational degree is also lower, and the rate of employees with poor health indicators is remarkably high compared to the other workplaces. These differences are partly explained by the fact that this latter enterprise is a telecommunication company which hires mainly (male) technicians and workers with low degree of qualification.

As mentioned, the dynamics of job stress are explored by defining four subgroups. The proportions of employees in each category of this summary variable are indicated in Table 2, together with socio-demographic characteristics. Besides education, a proxy measure of socioeconomic status, the groups did not differ with regard to these characteristics.

Relationships between the four job stress categories and mental health indicators show a steep gradient for depression, anxiety and somatisation, both in men and women (Figures 1 and 2). Less clear patterns were observed for chronic fatigue and psychotropic drug consumption (data not shown). Men and women reporting continuous or incident job stress during the observation period show higher proportions of mental health problems compared to those with low or decreasing job stress. Interestingly, the prevalence varies from about ten per cent in the group without stressful experience at work to about fifty per cent in the continuously stressed group, particularly among men.

As a final step, multivariate analysis is conducted to test the research hypothesis.

As indicated, logistic regression analysis is performed separately for men and women, with adjustment for age, education, threat from global economy, job dissatisfaction and work place instability. The dynamics of effort-reward imbalance at work over time are related to risks of poor mental health as suggested by our hypothesis. We do not observe a single effect in the group of employees who experienced a decrease in stressful work over time (group 2). In contrast, almost all odds ratios of mental health problems are significantly elevated in the two groups with newly emerging or continuously high job stress. In only 3 cases out of 20, odds ratios do not reach significance. There are interesting differences between gender. For men, recent onset of stressful experience at work is associated with a relatively highest risk of poor mental health, whereas for women relatively highest risks are observed in the group with continuously high job stress level for four indicators out of five. Odds ratios range between 1.4 and 7.1 for women and between 1.8 and 4.6 for men. When comparing the extrinsic versus the intrinsic component of the effort-reward imbalance model, it is obvious that associations of overcommitment with mental health are less consistent and generally weaker than those observed with the extrinsic component.

This study documents consistent associations of stressful experience at work over time with newly emerging mental health problems, using five mental health indicators. Employees with continuous job stress over a one year observation period and those with recently evolving job stress were at higher risk of developing poor mental health compared to the remaining groups, after exclusion of men and women with manifest mental problems at the study onset. One employee out of four in this large cohort belongs to one of the two groups with critical experience of job stress where, overall, a threefold elevated risk of incident poor mental health is observed. As the measurement of stressful experience at work is sensitive to change over time, it is unlikely that the observed systematic differences are spurious. Moreover, results remain true for men and women and after adjustment for relevant confounders, including the level of contextual job instability. While the gradient of mental health according to job stress is similar between men and women in bivariate analysis (Figure 1 and 2), an interesting gender difference results from multivariate analysis. Men are more reactive to a recent stress exposure, whereas women are more responsive to cumulative job stress (Table 3). This new finding needs further confirmation before being interpreted in a broader context.

Despite this evidence, this study suffers from several limitations. First, although employees with manifest mental health problems at T1 were excluded from multivariate analysis, we cannot rule out the possibility that a deterioration of mental health during the observation period has affected the measurement of job stress at T2. Secondly, even with a prospective protocol, two measurement waves only were conducted, and time of exposure was limited. However, structural and organizational changes occurred in all four enterprises under study even during this short observation period, thus reflecting the accelerated dynamics of work-related stress in current economic situation.

Thirdly, as the five indicators of mental health were not independent, some of the reported effects in multivariate analysis may have been overestimated. On the other hand, excluding employees with a respective mental health problem at study onset results in a conservative estimate, as this group may have suffered from previous job stress and that controlling for 'caseness' attenuates the associations under study. A further limitation concerns our decision of categorizing both the predicting and criterion variables, thus loosing information available from continuous data. Linear regressions performed with the normally distributed dependent variables did not yield different results. As this is not a clinical study, the decision of using scores in the upper quartile of the distribution of mental health indicators seems justified. Similarly, using the upper quartile of the ratio between effort and reward as a measure of stressful experience at work is in line with the evidence from several epidemiological studies although analysis of a log-transformed continuous ratio might reveal even stronger effects [25, 30, 31]. The same remains true with regard to a further test of main effects and interaction terms of the variables 'effort' and 'reward' which formed the basis for the ratio. Studies repeatedly revealed that the effect size of the ratio exceeds the effect sizes of the single variables [8, 30, 31]. Finally, we cannot exclude a 'healthy worker' effect of the final sample with full data (although this effect would point to a conservative estimate of the observed effect), and we cannot solve the methodological problem of reporting bias as both types of variables, the measure of stressful experience at work and the measures of mental health, were based on self-reported data.

Nevertheless, this study has remarkable strengths. It tests a theoretical model, effort-reward imbalance, that captures some of the core aspects of dynamics of stressful experience at work in a globalized, rapidly changing economy. Moreover, this model was shown to predict a variety of health problems in prospective observational studies in different occupational groups of several countries. Health outcomes include coronary heart disease [25, 32], cardiovascular mortality [9] mild-to-moderate psychiatric disorder (mainly affective disorder) [33], alcohol dependence [34], type 2-diabetes [35] and poor self-rated health or poor mental and physical functioning [30, 36].

This prospective evidence is supplemented by findings from cross-sectional studies testing associations of effort-reward imbalance at work with mental health, in particular depression [31, 37, 38]. Having controlled for different socio-demographic and workplace characteristics permits to better identify the evolution of the stress component in a one-year interval and its effect on the worker's mental health. Obviously, the evolution of the stress component in terms of effort-reward imbalance reflects some of the organizational changes in the enterprises.

The validity of reported results is further supported by two observations. First, in a cross-sectional analysis of the baseline data from the Somstress study, effort-reward imbalance at work was found to be associated not only with mental health indicators, but also with sickness absence, an indicator that is less vulnerable to reporting bias [22]. Secondly, the psychometric properties of the Belgian (French) version of the effort-reward imbalance questionnaire were tested in a comparative methodological data analysis of five different European countries. The values of internal consistency, discriminant validity of the scales and goodness of fit of the three model components (effort, reward, overcommitment) were well comparable across the five data sets and met psychometric criteria with satisfactory degree in all samples [20]. Finally, to our knowledge, this is the first report on psychosocial occupational health research with reference to the effort-reward imbalance model that explores the dynamics of stressful experience at work over time in a large sample of male and female employees from enterprises with different degrees of economic stability, using a variety of established mental health indicators.

In conclusion, continuous experience of stress at work over time and recently emerging job stress experience are both associated with elevated risk of mental health problems. Results underline the importance of studying health effects of a globalized, rapidly changing economy and of developing appropriate measures to reduce the respective burden of disease in working populations.