Work

Cohort Profile: the Work Environment and Health in Denmark study

Year: 2025

Type: article

Abstract: The Work Environment and Health in Denmark (WEHD) study was established to monitor employees’ work environment and health and to analyse the association between working conditions and health. The nationwide scope and comprehensive linkage to national registries are unique, allowing long-term follow-up with minimal attrition. Survey data of employees in Denmark were collected biennially from 2012 to 2018 with continuous follow-up in national registries, yielding three cohorts, with responders 18–64 years old and 53% women: i) N = 88 076 responding to at least one survey and followed up in registries; ii) N = 15 852 responding in 2012 and followed up in subsequent waves (2014, 2016, 2018, responders to all four waves: N=5933, 37.4%) and registries; iii) N = 30 070 responding in either 2012 or 2016 and invited to survey follow-up 2 years later (2014 and 2018, respectively, responders: N = 22 445, 74.7%) and followed up in registries. The surveys collected data on physical, chemical, biological, and psychosocial working conditions, and self-reported health. The registries provided sociodemographic, labour market, and health data for responders and responders’ cohabiting partners, children, and parents. Data are stored at Statistics Denmark. For collaboration, please contact the National Research Centre for the Working Environment, Copenhagen, Denmark ([email protected], Dr Jeppe Karl Sørensen). Conditions at the workplace, including physical, chemical, biological, and psychosocial exposures, are important determinants of population health [1–3]. Considering the ethical and practical constraints of conducting randomized controlled trials on potentially hazardous work conditions, and the limitations of cross-sectional and case-control studies regarding causal interpretations, cohort studies are essential for elucidating the relationships between working conditions and health outcomes. Examples for occupational cohort studies with specific job groups and industries are the British Whitehall II study (civil servants) [4], the French GAZEL study (employees with civil servant-like status in an utility firm) [5], the Finnish Public Sector Study (public employees, predominately hospital workers) [6], and the US-American Nurses’ Health Study [7]. Examples of occupational cohort studies with a focus on a broader range of occupations are the Danish Work Environment Cohort Study (DWECS) [8], the Swedish Longitudinal Occupational Survey of Health (SLOSH) [9], and the French CONSTANCES Cohort [10]. The Work Environment and Health in Denmark (WEHD, Danish: Arbejdsmiljø og helbred i Danmark, AH) study was established in 2012. WEHD has a unique nationwide scope with both random and workplace-stratified samples of employees and a comprehensive linkage to national registries. Its main objectives are: To monitor work environment and health of employees in Denmark over time. To investigate the short- and long-term effects of physical, chemical, biological, and psychosocial working conditions on health and labour market participation. Data collection of WEHD was carried out by the Danish National Research Centre for the Working Environment (Danish: Det Nationale Forskningscenter for Arbejdsmiljø). Data are stored on a secured research server of Statistics Denmark and are managed by the Danish Working Environment Authority (Danish: Arbejdstilsynet). Data collection, basic data management, and the monitoring of working conditions were funded by a grant from the Danish government. Analyses on the association between working conditions and health, and analyses for scientific publications have been funded by different grants, e.g. from the Danish Working Environment Research Fund on risk factors for mental disorders (AMFF-10–2019-03) and the influence of age on the association between work environment and health (AMFF-21–2019-09). WEHD combines self-administered questionnaire data concerning working conditions and health, with national register data concerning sociodemographic characteristics, labour market participation, and health status. Survey data and register data are merged using the participants’ unique civil registration number that is assigned to all Danish residents. WEHD was registered and approved by the Danish Data Protection Agency (Datatilsynet, journal number 2015–57-0074), and informed consent was collected from participants. According to Danish law, studies based on questionnaire and register data only do not need ethical approval. The study design includes both cross-sectional and longitudinal components. We collected survey data in cross-sectional samples in 2012, 2014, 2016, and 2018. Responders from the 2012 survey were invited to the surveys in 2014, 2016, and 2018. Responders from the 2016 survey were invited to the 2018 survey. All responders are followed up continuously in Danish national population, labour market, and health registries. WEHD consists therefore of three cohorts: Cohort I: Individuals who responded at least once to any WEHD survey (N=88 076). Cohort II: Individuals who responded to the WEHD survey in 2012 and were invited to the waves in 2014, 2016, and 2018 (N=15 852). Of these, N=14 085, N=10 961, and N=5933 responded in at least two, three, or all four waves, respectively. Cohort III: Individuals who responded to the WEHD survey for the first time in either 2012 or 2016 (N=30 070) and were invited to another survey 2 years later (2014 and 2018, respectively). Of these, N=22 445 responded. In spring 2012, NFA received from Statistics Denmark home addresses of Danish residents that fulfilled the following criteria: (i) aged 18 to 64 years; (ii) subject to taxation in Denmark; (iii) had been employed for at least 35 hours between 30 November 2011 and 31 January 2012; (iv) monthly income of at least 3000 Danish kroners (€404, £337, $523 as of 1 January 2012). This yielded a source population of 2 267 485 individuals [11], which was divided into two parts: The ‘stratified workplace population’, consisting of 198 789 employees from 1060 small (10–99 full-time employees, FTEs), medium-size (100–499 FTEs), and large workplaces (≥500 FTEs). Home addresses were available for 169 762 individuals. From those, a random population sample of 15 811 individuals were drawn, of which 15 767 were alive and residing in Denmark in September 2012, constituting the final stratified workplace sample. The ‘random population’ consisting of the remaining 2 086 696 employees from the source population. Addresses were available for 1 750 266 individuals. From those, a random sample of 35 111 individuals were drawn, of which 35 039 were alive and residing in Denmark in September 2012, constituting the final random population sample. NFA sent a link to an online questionnaire to the potential participants by postal mail in April 2012. Non-responders were contacted again by letter, then text message, then received a printed questionnaire with a stamped return envelope. In 2014, 2016, and 2018, we recruited new random population samples, following the same procedure as in 2012. In 2016, we further recruited a new stratified workplace sample. In addition, eligible responders from the 2012 random population sample were invited to participate in the 2014, 2016, and 2018 waves. Eligible responders from the 2016 random population sample were invited to participate in the 2018 wave. Table 1 presents the number of invited individuals, responders, and response rates across the four waves and sampling methods. In 2012 and 2016, 50 806 and 49 968 individuals, respectively, were invited, from the workplace sample and the random population sample. In 2014 and 2018, 37 737 and 37 331 individuals, respectively, were invited from the random population sample only. Response rates ranged from 48.2% in 2016 to 56.3% in 2018. Number of initially invited individuals, responders, and response rate for the 2012, 2014, 2016, and 2018 Work Environment and Health in Denmark (WEHD) waves Number of initially invited individuals, responders, and response rate for the 2012, 2014, 2016, and 2018 Work Environment and Health in Denmark (WEHD) waves Table 2 compares the sociodemographic characteristics of responders (n = 88 076) and non-responders (n = 81 846) combined for the four WEHD waves. If individuals were by chance invited to more than one wave, we only considered the first response. Sociodemographic characteristics of unique responders and non-responders in Work Environment and Health in Denmark (WEHD) (Cohort I) Responders and non-responders combined from the WEHD cross-sectional waves in 2012, 2014, 2016, and 2018. If an individual was by chance invited to more than one wave, only the data from the first wave are included. Data on education were retrieved from a registry in 2011, i.e. the year preceding the WEHD 2012 wave. Sociodemographic characteristics of unique responders and non-responders in Work Environment and Health in Denmark (WEHD) (Cohort I) Responders and non-responders combined from the WEHD cross-sectional waves in 2012, 2014, 2016, and 2018. If an individual was by chance invited to more than one wave, only the data from the first wave are included. Data on education were retrieved from a registry in 2011, i.e. the year preceding the WEHD 2012 wave. Responders were more often women, of higher age, and had a higher education, compared to non-responders. A more detailed non-response analysis of the 2012 wave is published elsewhere [11]. Figure 1 presents the intersection prevalence for the different possible combinations of WEHD responders across the four waves. Responders of possible combination of the four Work Environment and Health in Denmark (WEHD) waves. In Cohort II, of the initial 17 662 responders from the 2012 random population sample, 1810 were not eligible for follow-up because of self-reported non-employment, migration, dead, or address-protection, yielding a sample of 15 852 responders that were followed up in surveys in 2014, 2016, and 2018. Number of responders were 11 471 in 2014, 10 307 in 2016, and 9201 in 2018, yielding attrition rates of 27.6%, 35.0%, and 42.0%, respectively. There were 14 085, 10 961, and 5933 individuals who responded to at least two, three, or all four waves, respectively, thus attrition rate considering all four waves was 62.6%. In Cohort III, of the initial 34 066 responders from the random population samples in 2012 and 2016 (N = 17 662 and N = 16 404, respectively), 30 070 were eligible for follow-up and re-invited in 2014 (N = 15 852) and 2018 (N = 14 218), respectively. Of these, 22 455 responded (N = 11 781 and N = 10 674 in 2014 and 2018, respectively), yielding an attrition rate of 25.3%. Table 3 shows sociodemographic characteristics for Cohort II and Cohort III at baseline and follow-up. In Cohort II, those who participated in all four waves were older, more educated, more likely to report excellent or very good self-rated health, and less likely to report a probable depressive disorder at baseline. In Cohort III, those who responded to the follow-up (2014 or 2018) were older, slightly more educated, reported slightly better self-rated health, and were less likely to report a probable depressive disorder at baseline. Sociodemographic characteristics, self-rated health, and depressive symptoms Cohort II and Cohort III at baseline and at follow-up Data on education were retrieved from a registry the year preceding the first Work Environment and Health in Denmark (WEHD) wave. Probable depressive disorder defined by scoring ≥21 points on the Major Depression Inventory in the first WEHD wave. Sociodemographic characteristics, self-rated health, and depressive symptoms Cohort II and Cohort III at baseline and at follow-up Data on education were retrieved from a registry the year preceding the first Work Environment and Health in Denmark (WEHD) wave. Probable depressive disorder defined by scoring ≥21 points on the Major Depression Inventory in the first WEHD wave. Table 4 shows the variables measured in 2012, 2014, 2016, and 2018. The four surveys were, with a few exceptions, identical. The exceptions included, for example, that from 2014 onwards, items on job satisfaction, sickness absence, and working from home were added and that the wording of three items were slightly modified. See the Supplementary Material for a detailed documentation. Measurements in the Work Environment and Health in Denmark (WEHD) cohorts in the surveys and in Danish national population and health registries DISCO codes: Danish version of the ‘International Standard Classification of Occupations’; ICD codes: ‘International Statistical Classification of Diseases and Related Health Problems’; ATC codes: ‘Anatomical Therapeutic Chemical Classification System’. Measurements in the Work Environment and Health in Denmark (WEHD) cohorts in the surveys and in Danish national population and health registries DISCO codes: Danish version of the ‘International Standard Classification of Occupations’; ICD codes: ‘International Statistical Classification of Diseases and Related Health Problems’; ATC codes: ‘Anatomical Therapeutic Chemical Classification System’. In the surveys, we measured general working conditions, physical, chemical, biological, and psychosocial working conditions, accidents at work, health-related behaviours, personal attitudes, and physical and mental health. A detailed overview of all items is provided in the Supplementary Material. Table 4 also shows the registries that were linked to the survey and that allow retrieving information for example on sociodemographic characteristics (sex, age, cohabitation, education, job title, occupational grade, industry), hospital admission for somatic and psychiatric disorders, purchase of prescription medicine, mortality and cause of mortality, sickness absence, disability pension, and labour market participation (in work, on leave, unemployed, retired). The registries cover the time period at baseline (when the survey was filled in), the time period before baseline (in principal back until inception of the registry), and the time period after baseline (in principal forward until the registry ceases to exist). The registries include but are not limited to the Danish Civil Registration System [12], the Danish Register of Causes of Death [13], The Danish Labour Market Account [14], The Danish National Patient Register [15], The Danish Psychiatric Central Research Register [16], The Danish National Prescription Registry [17], and the Danish Register for Evaluation of Marginalisation [18]. By linking data from participants’ cohabiting partners, children, and parents, we further included measures on participants’ childhood adversity (e.g. death of a parent, family disruption, childhood poverty), and on adverse life events both before baseline and during follow-up (e.g. diagnosed mental disorder or severe somatic illness of a partner, child, or parent, divorce/separation from a partner). WEHD does not provide such information for all participants due to limitations in the registries. For example, participants born before the establishment of the registry or those for grow up outside of Denmark. Previous publications have assessed childhood adversities for approximately 40% of the participants. For details on the operationalization and availability of childhood adversity and adverse life events, see Sørensen et al. [19]. Until today, around 40 peer-reviewed research articles have been published with WEHD data. See the Supplementary Material for a bibliography. Psychosocial work environment factors (e.g. workplace bullying, emotional demands, effort-reward imbalance) were the most examined exposures, followed by physical and ergonomic exposures. The most frequently examined outcome was labour market participation (e.g. sickness absence, working life expectancy, labour market exit), followed by mental health outcomes (e.g. depressive and anxiety symptoms and disorders), and various physical disorders (e.g. musculoskeletal pain, diabetes, cardiovascular disease). Examples are studies that examined associations between effort-reward imbalance at work and risk of type 2 diabetes [20], between occupational lifting and risk of long-term sickness absence [21], and between self-reported work-related stress and labour market costs [14]. WEHD data have been included in multi-cohort studies with cohorts from other Nordic countries. These studies reported among other things that exposure to workplace violence was associated with an increased risk of suicide [22] and that higher levels of psychosocial resources at work were associated with a decreased risk of cardiovascular disease [23]. Furthermore, studies used WEHD data to develop job exposure matrices for psychosocial and physical working conditions and applied these matrices in population register-based studies [24–27]. Supplementary Table S1 gives an example of two studies that examined the association between psychosocial work environment and mental health in two different WEHD cohorts. The study by Sørensen et al. linked survey data on exposure to positive leadership behaviours to register data on treatment for a mental disorder (purchase of antidepressants or anxiolytics or hospital diagnosis of depressive or anxiety disorder) for individuals from Cohort I that fulfilled additional eligibility criteria (e.g. had a leader, no missing values on key covariates, no history of treatment for depressive or anxiety disorders, N = 59 743). The hazard ratio for onset of mental disorders was 1.36 (95% CI: 1.03 to 1.81) among workers who reported a low level of positive leadership behaviours compared to workers who reported a high level [19]. Estimates were adjusted for a wide range of covariates, including childhood adversities and adverse life events (see Supplementary Table S1 for details). The study by Rugulies et al. analysed the association between onset of sexual harassment in WEHD 2014 (among individuals who had been free of sexual harassment in WEHD 2012) and changes in depressive symptoms (measured by the MDI-scale) from 2012 to 2016 (Cohort II, N = 6647, 2018 data were not available, when the analysis was conducted) [28]. Responders who reported onset of sexual harassment from workplace personnel in 2014 had a higher mean depressive symptoms in 2016 than responders who did not report onset of sexual harassment in 2014. Estimates were adjusted for several covariates, including depressive symptoms in 2012 (see Supplementary Table S1 for details). The main strengths of the WEHD cohort are the large nationwide data that comprised all types of job groups and was based on both random and company-stratified sample of employees in Denmark, and the linkage of comprehensive survey data, including detailed assessments of working conditions, with numerous population, labour market, and health registries. The repeated measurements of working conditions allow analysing the association between onset and changes in exposure to certain working conditions and subsequent incidence of diseases and disorders. The linkage of data from participants’ cohabiting partners, children, and parents allows estimating participants’ childhood adversity and adulthood life events, which are important potential confounders when analysing the association between working conditions and health [19]. A weaknesses of the WEHD cohort is possible selection bias. Baseline response rate was about 50% and analyses indicate that responders were more often women, of higher age, and had higher education than responders. Whereas the extent to which selection bias is a concern in exposure-outcome analyses is debated [29], selection is considered a concern when reporting the prevalence of specific working conditions, diseases and disorders. WEHD therefore includes statistical weights to adjust prevalence rates to the source population. Reporting bias is another concern. Working conditions are assessed by self-report and unmeasured confounders (e.g. personal dispositions, mood, subclinical health conditions) may have influenced both the reporting of working conditions and the likelihood of subsequent onset of diseases and disorders. One mitigation strategy for this bias is to use onset of exposure in individuals who reported absence of exposure in an earlier wave [28]. Another strategy is to use WEHD data to build job exposure matrices and apply these matrices in national registries [24]. However, job exposure matrices have their own limitations, e.g. exposure misclassification, and they are applicable only to certain working conditions [30]. All data are stored at a secured research server located at Statistics Denmark and can be accessed by authorized researchers with an affiliation to a Danish research institution. Researchers interested in collaboration should contact the National Research Centre for the Working Environment, Copenhagen, Denmark (email: [email protected], contact persons: Dr Jeppe Karl Sørensen and Prof. Reiner Rugulies). WEHD was registered and approved by the Danish Data Protection Agency (Datatilsynet, journal number According to Danish law, studies that are based on questionnaire and register data only do not need from the Danish National on Health Research The study with the ethical of the We all individuals who participated in the survey and all research personnel who with and the data. and the design of the cohort and to the of and the data analysis for the cohort with by and All to the of the data. and the first and all other the All approved the final version for and all to be for all of the is the for the Supplementary data is available at of to this employed at the Danish National Research Centre for the Working Environment and to NFA as a is employed at a among other analyses data on had no in the data analyses and in the of the and in the to the All other no of This work was by a grant from the Danish to the Danish National Research Centre for the Working Environment and the Danish Working Environment Authority for Data and Data grant Data analysis for this cohort and the of this was by a grant from the Danish Working Environment Research Fund See I of the were not used in or analysing data, or the

Source: International Journal of Epidemiology

DOI: 10.1093/ije/dyaf085