Workplace Psychosocial Resources and Risk of Sleep Disturbances Among Employees

Key Points Question What are the associations of clustering of and changes in workplace psychosocial resources (ie, leadership quality, procedural justice, culture of collaboration, and coworker social support) with sleep disturbances? Findings In this cohort study, including 219 982 participant-observations nested within 114 971 participants, clustering of favorable workplace psychosocial resources was associated with a statistically significant lower risk of sleep disturbances. Improvements in leadership quality and procedural justice (ie, vertical resources) and in culture of collaboration and coworker social support (ie, horizontal resources) were associated with a lower risk of persistent sleep disturbances in a dose-response fashion. Meaning These findings suggest that multilevel workplace interventions are essential to promote short- and long-term sleep quality among workers.


Introduction
Sleep is essential to physiological restitution and recovery. 1,2Impaired sleep and prolonged activations of the physiological stress response may have far-reaching effects on the hormonal and immune system and metabolism, eventually contributing to a higher risk of cardiometabolic diseases. 3As the workplace is an important foundation for social and professional networks, investigating the preventive potential of favorable psychosocial work environment on reducing sleep disturbances is essential. 4rge-scale studies on workplace psychosocial resources, such as leadership quality, organizational justice, culture of collaboration, and coworker support, have only recently started to emerge.While higher workplace psychosocial resources have been associated with reduced risks of stress-related cardiometabolic diseases, 5,6 previous findings on sleep disturbances are mixed.3][14] In these studies, each workplace resources was addressed individually, while it is more likely that workplace psychosocial resources interact and thereby affect sleep and health through multiple mechanisms. 4Thus, it is important to investigate how different combinations of workplace psychosocial resources may be associated with sleep disturbances.
0][11] Importantly, such longitudinal associations may result from long-term effects of workplace psychosocial resources, but we cannot exclude the possibility of individual differences in perceptions of workplace characteristics and reporting sleep disturbances as an explanation for our results.Future studies emulating a target trial and investigating the extent to which changes in clustering of workplace resources are associated with changes in the risk of sleep disturbances are warranted to reduce the likelihood of this limitation and to evaluate the benefits of interventions with an aim to improve workplace resources. 15We are not aware of such studies having been conducted.
To address these limitations, we assessed the clustering of and changes in 4 major workplace psychosocial resources (ie, leadership quality, procedural justice, culture of collaboration, and coworker social support) to examine their concurrent and longitudinal associations with sleep disturbances.We studied 114 971 men and women from Denmark, Finland, and Sweden, with a up to 6 years of follow-up.

Study Population
This cohort study used data from the Work Environment and Health in Denmark (WEHD) study, Finnish Public Sector Study (FPS), and Swedish Longitudinal Occupational Survey of Health (SLOSH).
Ethical approvals were obtained from the Regional Ethical Review Board in Stockholm (for use of SLOSH data) 16 and the Ethics Committee of the Hospital District of Helsinki and Uusimaa (for use of FPS data). 17Use of WEHD data was approved by and registered with the Danish Data Protection Agency.Participants were informed about the purpose of the SLOSH, WEHD, and FPS, and by returning the questionnaire, the participants consented to participate.For the concurrent analysis, we included participants who had information on workplace resources, sleep disturbances, and key confounders in 1 or more waves.For the longitudinal analysis, we further included only individuals who participated in all follow-up waves.For the analysis of changes in resources and changes in sleep, only participants of at least 2 consecutive waves were included.The flowchart of participants in each specific analysis is presented in Figure 1.

Workplace Psychosocial Resources
We applied the framework of hierarchical dimensions of workplace psychosocial resources related to group-, leader-, and organizational-level variables, 4 including culture of collaboration (horizontal), support from colleagues (horizontal), leadership quality (vertical), and procedural justice (vertical) (eTable 1 in Supplement 1).Horizontal resources are resources from the same hierarchical level. 19rtical resources are resources from different hierarchical levels. 19All resources were measured Culture of collaboration was dichotomized (low vs high) and was defined as the workplace collaborative efforts to achieve the best available results or to develop or apply new ideas.Coworker support was assessed as perceived colleagues' support and was dichotomized (low vs high).The leadership variable included multiple dimensions, such as listening and appreciative, and was grouped into 4 levels (high, intermediately high, intermediately low, and low).The variable for procedural justice (ie, fairness in the principles and processes leading to decision-making) was also categorized into 4 groups (high, intermediately high, intermediately low, and low).Detailed descriptions of the variables can be found in eTable 1 in Supplement 1. 6 Clustering of workplace psychosocial resources has been detected previously from these same cohorts using a latent class model. 6We used the same predefined in Supplement 1). 6Changes of resources were defined as moving from 1 class to another across 2 consecutive waves, implying improvement or decline of either or both vertical and horizontal dimensions (eTable 2 in Supplement 1).

Assessment of Sleep Disturbances
Sleep disturbances were defined as self-reported difficulties in initiating or maintaining sleep, nonrestorative sleep, or daytime tiredness during the past 4 weeks (for WEHD and FPS) or 3 months (for SLOSH) according to the Karolinska Sleep Questionnaire (3 or 4 items; for WEHD and SLOSH) or Jenkins Sleep Problem Scale (4 items; for FPS) (eTable 3 in Supplement 1).Sleep disturbances were identified if respondents reported any of the symptoms at least often or all the time (in WEHD), 2 to 4 nights per week (in FPS), or 3 to 4 nights per week (in SLOSH), and the same criteria were used for indicating insomnia. 10,20An alternative definition was applied to indicate the overall severity of the sleep disturbances (identified when the mean value was greater than or equal to the corresponding value of the aforementioned threshold).

Covariates
Potential confounders were identified using directed acyclic graphs based on previous evidence, 21 and

Statistical Analysis
Two different analytical approaches were used to determine the concurrent and longitudinal associations of clusters of workplace psychosocial resources with sleep disturbances and associations of changes in workplace psychosocial resources with changes in sleep disturbances (eFigure 1 in Supplement 1).In terms of missing values, we used complete-case analysis.
In the analysis for assessing the concurrent association between resources and sleep disturbances (eFigure 1 in Supplement 1), all available observations were included.To bridge the knowledge gap of varying follow-up intervals, we performed longitudinal analysis between workplace resources and sleep measured 2 and 6 years after the exposure measurement among individuals who participated in all follow-up waves (eFigure 1 in Supplement 1).
Next, we applied an emulated trial design to investigate the concurrent association between changes of resources and the risk of maintaining or developing sleep disturbances (eFigure 1 in Supplement 1). 15To address the issue of reverse causation, we applied a stricter restriction criteria to changes of resources occurring prior to changes of sleep disturbances.eTable 2 in Supplement 1 shows restrictions in each analysis.
We used multiple logistic regression models for the analyses.Some individuals were included in more than 1 baseline for the analysis of the concurrent association and the longitudinal analysis with the 2-year follow-up interval.Generalized estimating equations were used to account for the intraindividual correlation.All potential confounders were adjusted in all analyses, except that night work was adjusted for in a sensitivity analysis restricted to individuals with this information.All analyses were repeated for the alternative definition of sleep disturbances.
We used SAS statistical software version 9.4 (Proc Genmod; SAS Institute) to perform cohortspecific analyses.Given the small numbers of cohorts, 22 fixed-effect meta-analyses were then performed to combine the cohort-specific estimates using R package meta version 4.16-2 (R Project for Statistical Computing).We used I 2 statistics to quantify the cross-cohort heterogeneity.P values were 2-sided, and statistical significance was set at P = .05.Data were analyzed from November 2020 to June 2022.

Results
A

Concurrent and Longitudinal Associations Between Resources and Sleep
Using the largest available sample size (114 971 participants), we found relatively stable trajectories in the prevalence of disturbed sleep according to resource classes and cohorts (eFigure 2 in Supplement 1).In the confounder-adjusted model, compared with the general low resource class, having a general high level of resources was associated with an overall lower prevalence of sleep disturbances (odds ratio [OR], 0.38; 95% CI, 0.37-0.40).Although there was substantial crosscohort heterogeneity (I 2 > 80%), all cohort-specific estimates pointed to the same direction (Figure 2A).2B).
There was a dose-response pattern for vertical resources when horizontal resources were high (Figure 2B).However, observing the contrast between intermediate vertical and low horizontal vs intermediate vertical and high horizontal, the role of horizontal resources was less clear in the presence of an intermediate level of vertical resources in the concurrent analysis (OR, 0.82; 95% CI, 0.75-0.89),and it lost statistical significance after 2 years (OR, 0.90; 95% CI, 0.81-1.00)and 6 years (OR, 0.94; 95% CI, 0.81-1.10).Nevertheless, when vertical resources were low, perceiving high horizontal resource was associated with a lower risk of sleep disturbances from the concurrent analysis to 6 years of follow-up (Figure 2B).

Association Between Changes in Resources and Changes in Sleep Disturbances
Among 51 259 employees participating 2 consecutive waves, 27 167 (53%) experienced a change in resources across 2 consecutive follow-ups (Figure 3A).Changes in vertical resources accounted for most of the change, including 8233 employees (16%) who experienced improvement and 8130 employees (16%) who experienced decline.
In the analyses of improvement (Figure 3B), 72% to 75% of employees reported persisting sleep problems between baselines and follow-up.
Similar findings were observed when changes in resources occurred before changes in sleep disturbances.However, these temporal associations were weaker, less precisely estimated, and did not achieve statistical significance at conventional levels (eFigure 3 in Supplement 1).

Sensitivity Analyses
Additional adjustment for night shifts did not materially change the results (eFigure 3 and eFigure 4 in Supplement 1).Results were similar when applying the alternative definition of sleep disturbances (eFigure 5 in Supplement 1).

Discussions
The findings of this cohort study, including data on 219 982 participant-observations from 114 971 participants in 3 large longitudinal cohorts, suggest that better psychosocial resources at work were associated with lower risk of sleep disturbances, both concurrently and longitudinally, with vertical resources of greater importance in the long term.Supporting modifiability of sleep disturbance risk by targeting workplace psychosocial resources, we observed that improvements in workplace psychosocial resources were associated with a reduced risk of persisting sleep disturbances and deterioration of resources was associated with the development of sleep disturbances.
Our findings are generally in line with several previous smaller-scale studies, 1,2,7,[9][10][11] and add to this literature by examining clustering of vertical and horizontal psychosocial resources at work with an emulated trial design in observational data.3][14] Moreover, our findings on changes in resources confirm findings from a single-cohort study, suggesting an association between changes in relational justice and changes in sleep. 10cording to the effort-recovery model, improved psychosocial resources from work could contribute to a better switch-off from work, in turn leading to better recovery process from work, including improved sleep. 23[26] Our findings suggest that vertical resources might be more crucial than horizontal resources in associations with long-term sleep disturbances.In the presence of high horizontal resources, vertical resources showed a long-term dose-response association, while the same association was not observed for horizontal resources, given an intermediate level of vertical resources was present.This finding is plausible, considering the greater power of leaders to affect a positive working environment, 4 such that vertical resources may promote horizontal resources in the long term. 27A study by Framke et al 28 suggested workplace horizontal social capital may be particularly relevant for sleep disturbances; however, their study was based on a smaller occupation-specific sample.Future investigations of the interactive cross-dimensional mechanisms of work resources may facilitate the design of cost-effective strategies for the prevention of sleep disturbances.
The concept of workplace psychosocial resources overlaps a number of occupational and organizational theories, including the demand-control model, the job-demand resource model, the effort-reward imbalance model, and the concept of organizational justice. 29,30A 2017 study by Nielsen et al 4 proposed that workplace resources at the individual (eg, psychological capital), group (eg, team climate), leader (eg, leadership quality), and organizational levels (eg, perceived organizational support and justice) may be associated with better job performance and well-being among employees.Following this framework, we have shown that vertical (organization or leader) and horizontal (the work group) clusters of resources were associated with health-related outcomes. 5,6Our findings further highlight the potential benefits of multilevel workplace interventions in promoting sleep of employees.

Limitations and Strengths
This study has some limitations.We applied a novel clustering of workplace resources based on predefined information on 4 important workplace resources.Thus, caution should be taken when generalizing our findings to other workplace resources, such as job control, rewards, and autonomy.
We included more than 100 000 participants with 6 years of follow-up measured up to 3 times and reported results from both concurrent and longitudinal analyses.Although we applied an emulated trial design, reverse causation cannot be fully ruled out due to lack of statistical power to draw solid conclusions on temporality.Furthermore, it is likely that sleep, stress, and workplace resources are affecting each other in a dynamic manner, which may create feedback loops over time.
Future studies with larger sample sizes are needed to disentangle these intertwined mechanisms.
Our population-based sample included employees from different occupational groups, industries, and employment types.Although point estimates varied substantially across these cohorts, most cohort-specific results pointed to the same direction, implying the validity and generalizability across countries and different settings.The data sets were only followed until 2018, which is before the COVID-19 pandemic began.Therefore, future studies are needed to understand whether, for example, the COVID-19-related changes in work-from-home opportunities would modify the observed associations.Other potential modifiers, including age group, gender, employment sector, educational level, and occupational grade, should be explored in future studies to detect at-risk groups that would benefit most from interventions.
With a large sample size and repeated follow-ups, we assessed the general association between resources and sleep as an extension to prior studies, allowing for direct comparisons with previous findings addressing different follow-up lengths and observing short-to long-term transitions.Five combinations of vertical and horizontal dimensions of resources enabled us to weigh the importance of 1 dimension at a time, with or without the presence of the other.A further step was taken to analyze the associations with changes in resources, mimicking a trial, which can be informative for intervention development when data from an actual trial are not available.

Conclusions
In this cohort study including 3 Scandinavian cohorts, nearly half of the employees experienced a low level of 1 or more vertical or horizontal dimensions of workplace psychosocial resources, which included procedural justice, leadership quality, coworker support, and culture of collaboration.

Figure 2 .
Figure 2. Concurrent and Longitudinal Associations Between Workplace Psychosocial Resources and Sleep Disturbances

Figure 3 .
Figure 3. Association Between Changes in Resources and Changes of Sleep Disturbances

Favorable 2 . 3 . 2 . 2 .
resources at work were associated with a lower risk of sleep disturbances, and improvements in resources were associated with lower risk of sleep disturbances.Our findings justify future intervention studies to examine the extent to which improvements in workplace psychosocial resources could facilitate remission or recovery from sleep disturbances and prevent development, deterioration, or prolongation of sleep disturbances among employees.SUPPLEMENT 1. eAppendix 1.The Choice of Measurements, Dimensions, and Cutoffs of Workplace Psychosocial Resources eAppendix Comparisons with Other Possible Categorizations eAppendix Measurement of Covariates eTable 1. Items Measuring Workplace Resources eTable Changes of Resources and Restrictions eTable 3. Measurement Invariances for Sleep Measurement Across Waves, Using the Confirmatory Factor Analysis eFigure 1. Study Design eFigure Trajectory of Sleep Disturbances eFigure 3. Results From Analyses of Changes, Before and After the Adjustment of Night Shifts eFigure 4. Additional Adjustment for Night Shifts for Analyses on Concurrent and Longitudinal Associations eFigure 5. Results From Alternative Definition of Sleep Disturbances SUPPLEMENT 2. Data Sharing Statement

JAMA Network Open | Public Health Workplace
17ychosocial Resources and Risk of Sleep Disturbances Among EmployeesThe WEHD is a nationwide population-based biennial occupational survey, with baseline in 2012.It consists of individuals from the general working population and supplementary surveys on selected workplaces in Denmark.18Data for this study were obtained from the 2012, 2014, and 2018 waves (wave 2016 was dropped due to sleep measurement inconsistencies).The FPS is based on a dynamic public sector cohort in Finland followed up with repeated questionnaire surveys at 2-or 4-year intervals since 2000.17Datafor this study were obtained from The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.JAMA Network Open.2023;6(5):e2312514.doi:10.1001/jamanetworkopen.2023.12514(Reprinted) May 9, 2023 2/12 Downloaded From: https://jamanetwork.com/ by a Copenhagen University Library User on 09/12/2023

Downloaded From: https://jamanetwork.com/ by a Copenhagen University Library User on 09/12/2023
5-class categorization in this study: general low, intermediate vertical and low horizontal, low vertical and high horizontal, intermediate vertical and high horizontal, and general high (Table 1), because it showed better robustness across data waves, cohorts, and employment sectors compared with other categorizations (eAppendix 2

Table 1 .
Categorization of Workplace Psychosocial Resources Based on Previous Latent Class AnalysisDiseases and Related Health Problems, Tenth Revision (ICD-10) codes from the national patient registries (eAppendix 3 in Supplement 1).Self-reported night shift work (yes and no) was also a potential confounder; however, night shift data were not available in FPS wave 2010 (wave 2).
included age, sex, country of birth (categorized as Scandinavian countries, other European countries, and other continents), marital status (married or cohabiting, single, separated or divorced, and widowed), highest educational level achieved (Յ9, 10-12, and Ն13 years), type of employment contract (permanent, and nonpermanent), preexisting comorbidities, and preexisting mental disorders.Information on these variables were extracted from national registers in Denmark, Finland, and Sweden, except that marital status and employment contract in SLOSH and marital status in FPS were self-reported.Information on country of birth was not available in FPS.Preexisting comorbidities according to the Charlson Comorbidity Index and preexisting mental disorders were detected using the International Classification of Diseases, Eighth Revision (ICD-8), International Classification of Diseases, Ninth Revision (ICD-9), and International Statistical Classification of a Vertical resources include procedural justice and leadership quality.b Horizontal resources include colleagues' support and culture of collaboration.JAMA Network Open | Public Health Workplace Psychosocial Resources and Risk of Sleep Disturbances Among Employees JAMA Network Open.2023;6(5):e2312514.doi:10.1001/jamanetworkopen.2023.12514(Reprinted) May 9, 2023 4/12 Downloaded From: https://jamanetwork.com/ by a Copenhagen University Library User on 09/12/2023

Table 2 .
Demographic, Clinical, and Work-Related Characteristics of Participant-Observations Based on the Largest Available Sample of Participant-Observations FPS indicates Finnish Public Sector Study; NA, not applicable; OR, odds ratio; SLOSH, Swedish Longitudinal Occupational Survey of Health; and WEHD, Work Environment and Health in Denmark.