Background
Job stress can lead to absenteeism [
1] and presenteeism [
2], and has been estimated to contribute to 40 % of all job turnover [
3]. Evidence favours causal links between job stress and increased risk of down-stream illness [
4‐
7]. The World Health Organisation cites workplace health promotion (WHP) as beneficial to job stress prevention, stating that health-promoting workplaces should address health at a systemic- (policies, practices, systems) as well as individual-level [
8]. However, findings in favour of effective systems-level intervention to prevent job stress remain inconclusive [
9,
10]. Evidence for stress prevention largely stems from individual-level stress management interventions [
9‐
12].
Comprehensive WHP, a term given to interventions targeting both individual- and system-levels [
13], has proven popular among employers, with associated decreases in absenteeism, presenteeism [
14] and financial returns on investment [
15,
16]. Nevertheless, publications citing research on comprehensive forms of WHP with job stress outcomes are uncommon [
17] with the majority focusing on employee participation [
18].
Conceptually, WHP appears associated with job stress in two key ways. First, investment in the ‘social capital’ of the organisation [
19‐
21] may contribute to workers’ perceptions of support [
19] from their organisation because the employer shows care for their health and well-being [
22,
23]. The presence of WHP may also serve to reduce the stigma associated with reporting health-related issues or to enhance general health awareness among employees [
23]. These emotional and cognitive effects have been linked to improved job satisfaction [
18,
24] and mental health [
25]. However, we were unable to identify any published articles that separated WHP availability from WHP participation when assessing whole-of-workforce job stress.
Second, exposure to job stress can provoke short-term behavioural responses such as inappropriate nutrition [
26], smoking [
27], physical inactivity [
28] and alcohol consumption [
29]. Extended exposure to situations stimulating stress responses can also lead to chronic arousal or strain [
30]. Participation in workplace activities targeting known health risks or enhancing work-related coping strategies aims to reduce job stress. Meta-analytic research supports this link between participation in WHP programs, reduced job stress and improved mental health [
9,
10,
18,
31]. Comprehensive strategies have also been shown to be more effective than approaches tackling only organisational or individual-level factors [
32]. However, gaps remain in understanding time-related effects [
17] and intervention effectiveness when WHP is scaled-up in size to intervene with whole working populations [
33].
Effort-reward imbalance (ERI) concepts appear suited to assessing both pathways and a strong evidence-base is available across a broad range of occupations supporting the association between self-reported measures of ERI and enduring health outcomes such as cardiovascular disease [
34,
35] and diabetes [
36]. Effort-reward imbalance theory asserts that work is a form of mutual exchange, or reciprocity, where job-related efforts are traded for rewards (i.e., job security, career advancement, self-esteem) as a type of ‘social contract’. The theory proposes that insufficient reward for work effort can negatively impact upon the capacity of an individual to regulate their emotions, thoughts and behaviours, which in turn can lead to job strain [
37]. Workplace health promotion may be viewed as an organisational benefit signaling regard for an employee’s welfare, thereby increasing perceived organisational support and enhancing self-esteem [
19]. Research has highlighted that ERI measures explain unique variance in relation to the macro- or contractual factors contributing to mental health outcomes, and that the effort dimension can be likened to job-related demands [
38,
39].
This project evolved from a collaboration between university researchers and government (public sector) that had the goal of evaluating the long-term effectiveness of a comprehensive multi-component WHP initiative, named
Healthy@Work. Baseline workforce survey data had indicated that ERI was a key correlate of high psychological distress among employees, and that mental health varied by sex when compared with working population norms [
40]. We hypothesized that i) higher availability of WHP would be positively associated with perceived reward, particularly through improved self-esteem (given job security and career progression were unlikely to be impacted by WHP), and ii) higher participation in WHP would be negatively associated with perceived effort.
Methods
Setting and study population
Tasmania is an Australian region with a population of around half a million people. The Tasmanian Government employed approximately 28,000 public sector workers across 14 separate organizations (government departments), which are highly diverse in their functions (e.g., health, education, fire services), locations (e.g., urban, rural, remote) and occupations. Participants were drawn from this working population and gave informed consent for study involvement, including publication. Ethics approval for the study was obtained from the Human Research Ethics Committee (Tasmania) Network (ID: H0010501).
Intervention
Overview
Between 2009 and 2012, the Tasmanian Government invested approximately $2 million in a whole-of-workforce WHP intervention called Healthy@Work. A small, centralised Healthy@Work team was responsible for the associated structural changes including strategy, model development, principles and implementation cycle, and was tasked with oversight of this new government policy focus on WHP. Implementation was mandatory and was delegated to the senior executive of each government department. It was internally audited each year until its conclusion in mid-2012. Our research team commenced a partnership with the Tasmanian Government in 2010 to conduct a naturalistic evaluation of the intervention.
Department-based activities
Departments were responsible for establishing in-house WHP vision, strategies and action plans, and for reporting on progress. Grant funding was available to departments as an incentive for WHP including for example, development of a workplace health promotion resource toolkit, funding equipment or recreation spaces, development of a computer-based system to interrupt sitting time and prompt healthy activity, and individual assessment, activity or education programs. The number of departments with an established WHP program increased from 6 in 2009 to 13 in 2012. The mean number of initiatives per department increased from 13 in 2009 to 48 in 2012 (Additional file
1: Figure S1).
Exposures
Healthy@Work strategies targeted i) individuals via mental health and well-being, health education, health assessments, physical activity, and injury management, and ii) organizational change through initiatives such as increasing physical space for health-activities, making healthy food options available, funding onsite gymnasiums, giving access to stairs, promoting health via information bulletins and implementing health-promoting policies. Primary job stress prevention strategies (e.g., job control) were not included in Healthy@Work.
For analysis of individual exposures we first calculated a score indicating the ‘
availability’ of Healthy@Work strategies [
41]. This score was obtained from questions asking respondents to provide a ‘yes’ or ‘no’ answer to a specified list of Healthy@Work amenities and programs (Additional file
2: Figure S2). The availability timeframe was
‘the previous 12 months’ in 2010 as a baseline reference period and
‘the previous 3 years’ in 2013 to capture the period over which Healthy@Work was implemented. A ‘total availability’ score was derived from per-person counts of positive responses to question items and a minimum score of 1 was needed to calculate participation. This approach was taken so that respondents could distinguish between work situations were WHP interventions were available (including health-promoting environments), and where they actually participated in activities. Where participants provided a ‘yes’ answer to activities, they were asked for the number of times they had participated and we used this information to calculate a total participation score per person (Additional file
2: Figure S2 describes calculations).
Outcomes
Our overall measure of job stress outcome was Effort-reward imbalance (ERI). We applied the 17 item ERI questionnaire, which is a validated self-report survey with 6 items measuring Effort and 11 items dedicated to Reward [
42]. A ratio is typically calculated for every person by first adding all scores for each of the effort (
e) and reward (
r) scales, then applying the formula
e/(r x
c) where
c equals the proportion 6/11. Scores ≥1 are argued to indicate job strain conditions. The procedure for calculating the Reward component and its subscales of self-esteem, job security and career advancement has been described elsewhere [
42]. Continuous scale scores were used to maximize the data available for analysis.
Participants and sample size
We collected data via repeated, cross-sectional postal survey (2010 and 2013), selecting a 40 % random population sample from the total pool of workers, stratified according to employment condition, employment category and department (Additional file
3: Figure S3). In 2013 a portion of workers were re-selected by chance and survey respondents from this group were referred to as the ‘cohort’ (men = 161; women = 423). Survey responses were merged with de-identified administrative data and this process enabled propensity weighting to adjust for possible non-response bias (described below).
Statistical analysis and methods
The repeated cross-sectional surveys were analyzed together in two-stages: 1) assessing whether mean ERI or its subcomponent scores changed over time and estimating associations between these scores and the availability of, or participation in Healthy@Work programs in 2010 and 2013; and 2) assessing whether there were changes in availability or participation over time. Survey responses were anticipated to be more similar within government departments, and for those who were in the cohort of repeat respondents. Mixed-effects linear regression modelling with random intercepts for department and participants was used to allow for correlated responses. Models were stratified by sex due to known differences in employee reporting of psychological distress. In stage 1, linear mixed-models were constructed with the outcome ERI (or its subcomponents) and a dummy variable for
‘survey year’ in the fixed effect section of each model along with covariates for confounders [
43]. This process allowed us to determine whether ERI scores or their components changed by survey year. We then added covariates for total availability or participation. We tested for interaction between survey year and Healthy@Work exposure variables in each model to assess whether the effect of exposure changed between surveys. Confounders were identified via regression modelling techniques described by Hosmer, Lemeshow and Sturdivant [
44] and were defined as those variables that were associated with the outcome and which also produced more than 10 % change in an estimated coefficient of the model.
Poisson regression with random effects as above was used to assess whether mean availability of, or participation in Healthy@Work strategies had changed over time. Model diagnostics from linear mixed effects models showed that residuals were skewed and an inverse transformation was applied to the ERI values. We then back-transformed the ERI results to present mean estimates on the original scale of measurement. Further we applied propensity weighting as described by Little and Rubin [
45] to deal with potential non-response bias; the propensity model included age, sex, government department, employment category, employment condition, and tenure using the human resources administrative database as the reference population.
Models showing relationships between the exposure and outcome were corroborated by replicating the analysis with the repeat-respondent cohort. All analyses were conducted using STATA 12.1 (StataCorp LP, Texas, USA).
Discussion
Our first hypothesis, that higher availability of WHP would be positively associated with perceived reward through improved self-esteem was supported among women but not men. Our second hypothesis, that higher participation in WHP would be negatively associated with perceived effort was supported for men but not women. However, the magnitudes of effect for these additive associations were modest and were not reflected as statistical differences in perceived effort or reward (including self-esteem) at a working-population level over time. We found a high corroboration between results for the repeat-responder cohort and the broader respondent group, which was randomly sampled and weighted to minimize non-response bias. Therefore these results seem generalizable to the source population of public sector workers under study.
To show effects at a population level, additive relationships rely on increased dosage of exposure (e.g., higher volumes of availability or higher participation levels). In 2013 self-reported WHP availability increased by 14 % and participation approximately doubled over time. Systematic differences in occupational exposures between sexes, linked to disparities in perceptions and/or reporting and variations in exposure between or within jobs [
46] may also have contributed to our results. Recently published results from this project have indicated that where activities were available, participation was less likely among employees with cardio-metabolic conditions, those who smoked and workers with variable work schedules. Participation was more likely among administrative staff and those who participated in leisure time physical activity [
47].
For women, we infer WHP availability contributed to perceptions of organisational support thereby enhancing self-esteem [
20]. The ERI self-esteem construct was derived from items capturing perceptions of i) respect from supervisors and colleagues, ii) adequacy of support in difficult situations, and iii) effects of job interruptions [
42]. Research using this concept of social exchange for other forms of non-monetary employee benefits, such as manager trustworthiness and procedural justice has supported their relationship with job satisfaction and employee turnover [
48]. However, the increases in WHP availability may have been of insufficient dose, or may have needed supplement from other non-monetary benefits to show changes in self-esteem at a population-level. A study of Chinese physicians, by Li and colleagues [
49] has found differences in the way men and women perceive the reward sub-factor of ERI when facing similar work environments.
The time period for Healthy@Work implementation coincided with the global economic downturn, which had major financial ramifications for the Tasmanian Government. During the implementation period, government directives also focused on long-term reduction of operating costs, including labor costs via vacancy control and productivity management. For men, who were higher wage-earners and more likely to be working in full-time or management positions, it is possible that the adverse events reported here may have contributed to perceived or real threats of job-loss and work intensification at population-level [
50]. Higher WHP participation may have enhanced work-related coping or personal well-being but it was only one side of the effort-reward equation. We interpret that men did not perceive WHP availability as a reward. It is possible that men in this workforce were more sensitive to job security than socio-emotional relationship issues [
51]. We note that sex-based differences in occupational exposures been noted previously [
46] and the distinctions identified by employment band for participating men require further investigation. However, attention to areas such as self-esteem, job security and promotion prospects through stress management programs or primary stress prevention interventions may have been more suited to addressing increased job stress among men.
Limitations
Repeated-cross sectional designs offer advantages in cost and allow for changes in working population characteristics but they do not allow causal inferences. Neither do they control for baseline differences in exposure to interventions or between individuals, or influences on results due to inter-departmental migration, [
52,
53]. Other research has shown that for large population samples repeated cross-sectional designs can be superior to cohort designs [
53]. Linear mixed-modelling analysis also provides robust estimates in the face of modest associations [
54]. Further, even though response rates were arguably low, they were typical for organizational surveys [
55] and have been addressed here through weighting procedures. We acknowledge it is possible that people with greater stress may have chosen not to respond to the surveys [
56,
57]. Our study did not measure societal trends and commonly changing features of public sector workforces may have influenced the observed changes in effort and reward over time. Furthermore, our self-reported measures of exposure may have been too crude. The focus of Healthy@Work was on comprehensive intervention and departmental programs were necessarily different, catering for working circumstances, and employee needs and preferences. This meant that activities also had different levels of content, intensity and levels of delivery and were implemented across a large and diverse working population. Furthermore, participation questions did not ask about the specific dose of activity (i.e., whether the number of times represented a full dose or its portion). Other authors have acknowledged the innate challenges in measuring customized WHP programs [
58], and in this setting although the study’s design did not capture specific details of availability or participation, it was able to deal with the heterogeneity of comprehensive WHP. We do not know whether the changes in wording of the response period for our exposures affected the results but this approach enabled us to capture the full period of WHP. The self-reported increases in WHP availability appeared to reflect increases in departmental data obtained from the employers’ audit processes (Additional file
1: Figure S1) but further investigation is needed to assess self-reported versus actual overlap. More detail on specific types of interventions from organizations would have been an advantage. Identification of further exposure effects may require differently timed data collection. Recall bias can also be an issue in self-reported data [
59]. Broader conclusions about the generalizability of this study would benefit from follow-up research in other public or private sector workplaces.
Competing interests
The authors have no financial or competing interests in this research.
Authors’ contributions
LJ conceptualised the article, participated in data collection, data management and cleaning and wrote the manuscript. AM and AV helped with interpretation of the results and revised the manuscript. PO advised on data analysis and interpretation and revised the manuscript. KS helped with conceptualisation of the paper, interpretation of the results and revised the manuscript. All authors read and approved the final manuscript.