Emotional demands and all-cause and diagnosis-specific long-term sickness absence: a prospective cohort study in Sweden

Abstract Background High emotional demands at work require sustained emotional effort and are associated with adverse health outcomes. We tested whether individuals in occupations with high emotional demands, compared with low demands, had a higher future risk of all-cause long-term sickness absence (LTSA). We further explored whether the risk of LTSA associated with high emotional demands differed by LTSA diagnoses. Methods We conducted a prospective, nationwide cohort study on the association between emotional demands and LTSA (>30 days) in the workforce in Sweden (n = 3 905 685) during a 7-year follow-up. Using Cox regression, we analyzed sex-stratified risks of all-cause and diagnosis-specific LTSA due to common mental disorders (CMD), musculoskeletal disorders (MSD) and all other diagnoses. Multivariable adjusted models included age, birth country, education, living area, family situation and physical work demands. Results Working in emotionally demanding occupations was associated with a higher risk of all-cause LTSA in women [hazard ratio (HR) = 1.92, 95% confidence interval (CI): 1.88–1.96] and men (HR = 1.23, 95% CI: 1.21–1.25). In women, the higher risk was similar for LTSA due to CMD, MSD and all other diagnoses (HR of 1.82, 1.92 and 1.93, respectively). In men, risk of LTSA due to CMD was pronounced (HR = 2.01, 95% CI: 1.92–2.11), whereas risk of LTSA due to MSD and all other diagnoses was only slightly elevated (HR of 1.13, both outcomes). Conclusions Workers in occupations with high emotional demands had a higher risk of all-cause LTSA. In women, risk of all-cause and diagnosis-specific LTSA were similar. In men, the risk was more pronounced for LTSA due to CMD.

Introduction D ealing with sick or dying patients, taking care of clients' emotional needs, responding to clients' sorrows and worries or handling aggressive customers entails high emotional demands at work. 1,2 Emotional demands at work require a sustained emotional effort 3,4 and have been linked to a higher risk of long-term sickness absence (LTSA). [5][6][7][8] Most previous studies on emotional demands at work and LTSA have been limited by using self-reported data on emotional demands and by analyzing smaller study populations. [6][7][8][9] The use of self-reported exposure measurement may be problematic as the experience of emotional demands may be influenced by the affective state of the respondent. 2,10 Small study samples may be vulnerable to selection bias and may have low statistical power. Studies of emotional demands and LTSA in the working populations of Denmark, Norway, Netherlands and Europe were conducted based on 3188; 6758; 31 884 and 32 708 respondents, respectively. All four studies had non-response rates about 40%. [6][7][8][9] Furthermore, all but one 9 of the previous studies on emotional demands at work and LTSA have not included the LTSA diagnoses and were restricted to examine all-cause LTSA. This is a limitation, as diagnosis-specific LTSA may help to understand through which pathways-e.g. psychological, psychobiological or behavioural-emotional demands may affect workers' health. The only study that included LTSA diagnoses 9 reported a higher risk of LTSA due to a mental diagnosis, after adjustment for other psychosocial work factors.
A nationwide, complete cohort study from Denmark reported a higher risk of LTSA among employees in jobs with a high level of emotional demands compared with employees in jobs with a low level of emotional demands. 11 The study measured emotional demands at work with a job exposure matrix (JEM), i.e. exposure data were not on the individual level but aggregated on the job group level. The study was limited, though, by analyzing all-cause LTSA only, as diagnoses are not available in Danish sickness absence registers. In Sweden, sickness absence registers include diagnoses, however, a Swedish JEM on emotional demands has not yet been developed, to our knowledge. Therefore, in the present study, we used the Danish JEM on emotional demands at work 11 and applied this JEM to Swedish register data on working populations and their sickness absence. This allowed us analysing not only allcause LTSA but also diagnosis-specific LTSA, in terms of LTSA due to common mental disorders (CMD) and musculoskeletal disorders (MSD), respectively. CMD and MSD are the two leading categories of LTSA diagnoses in high-income countries, 12 including Sweden. 13 The aim of this study was to test the hypothesis that women and men in occupations with high emotional demands at work, compared with low demands, had a higher risk of future all-cause LTSA. In addition, we explored whether this differed by LTSA diagnoses in terms of CMD, MSD and all other diagnoses except CMD and MSD, respectively.

Methods
We conducted a nationwide 7-year prospective cohort study on the association between emotional demands at work and the future risk of LTSA in the workforce in Sweden. We examined the risk of allcause LTSA and risk of LTSA due to CMD, due to MSD and due to all diagnoses except CMD and MSD.

Exposure to emotional demands at work
We measured emotional demands at work in 2009 with a JEM based on information from the Danish Work Environment Cohort Study (DWECS). The construction of the Danish emotional demands at work JEM is described in detail elsewhere. 11 Briefly, DWECS is a survey about working conditions and health in Denmark, conducted in a nationwide sample of the Danish workforce from 1990 to 2010. 14,15 DWECS includes three items on emotional demands at work, each rated on a five-point scale: (i) Does your work put you in emotionally disturbing situations? (ii) Is your work emotionally demanding? (iii) Do you get emotionally involved in your work?
Using data from the 2000 and 2005 waves of DWECS, we estimated the JEM as the predicted level of emotional demands at work given job group [coded according to DISCO-88, 16 the Danish version of the International Standard Classification of Occupations (ISCO)-88 system 17 ] sex, age and year of data collection (2000,2005). Next, we translated Swedish occupational codes, Swedish Standard for Occupational Classification (SSYK-96), 18 the Swedish version of ISCO-88 codes, into the corresponding DISCO-88 codes. We then assigned the predicted levels of emotional demands at work to each individual in the study population at baseline in 2009, according to their occupational group, sex and age. Finally, we categorized individuals into groups with low, medium-low, medium-high and high levels of emotional demands at work based on a quartile split of the distribution in the study population.

The public sickness absence system in Sweden
All individuals living in Sweden, aged 16 years and above, are covered by public sickness absence insurance in case of reduced work capacity due to disease or injury, provided that the individual has an income from work, parental leave benefits or unemployment benefits. The first day of a sickness absence spell is a waiting day and the following 13 days are covered by the employer. Sickness absence days from day 15 onwards are covered by the Social Insurance Agency, which also covers benefits from day 2 onwards for unemployed individuals. 13 After the 7th day, a medical certificate is required, issued by the physician responsible for assessing the patient's work capacity. The certificate includes the diagnosis leading to the work incapacity, coded according to the International Classification of Diseases, 10th revision (ICD-10).

LTSA diagnoses
We defined LTSA as having had a sickness absence spell of >30 days between 1 January 2010 and 31 December 2016. First, we coded allcause LTSA as LTSA due to all sickness absence diagnoses. Next, using the ICD-10 codes on the sickness absence certificates, we constructed three categories of LTSA: (i) CMD [depressive disorders (F32, F33), anxiety disorders (F40, F41), reaction to severe stress and adjustment disorders (F43) and problems related to life management difficulty, including burnout (Z73)]; (ii) MSD (M00-M99) and (iii) all other diagnoses including those with missing information on diagnosis.

Covariates
All covariates were measured at baseline in December 2009. As covariates, we included age (categorized as 18-24, 25-34, 35-44, 45-54, 55-60 years); country of birth [Sweden, other Nordic country, other EU27, rest of world (including missing)]; type of living area [large city (Stockholm, Gothenburg, Malmö with surrounding municipalities), medium-sized town (municipalities with >90 000 inhabitants within 30 km of municipal centre), small town or rural (municipalities with <90 000 inhabitants within 30 km of municipal centre)]; family situation (married/cohabiting with children <18 years living at home, married/cohabiting with no children <18 years living at home, single with children <18 years living at home, single with no children <18 years living at home); educational level [elementary or equivalent ( 9 years or missing), high school or equivalent (10-12 years), university/college or equivalent (>12 years)] and, finally, physical demands at work (measured with a JEM, obtained in a similar way from DWECS as we obtained the JEM for emotional demands 11 and categorized into quartiles based on the distribution in the study population).

Statistical analysis
All analyses were sex-stratified using SAS 9.4. We analyzed the association between emotional demands at work at baseline and the risk of LTSA during the 7-year follow-up by estimating hazard ratios (HR) and 95% confidence intervals (95% CI) using Cox proportional hazards regression models with calendar time as the underlying time axis.
We followed each individual from 1 January 2010 until the first episode of LTSA or censoring due to disability pension, early old-age pension, emigration, death or end of follow-up (31 December 2016), whichever came first. First, we analyzed the association between emotional work demands and all-cause LTSA. Next, we analyzed the association between emotional demands and diagnosis-specific LTSA according to the three diagnostic groups: CMD, MSD and all other diagnoses. In addition to the reasons for censoring mentioned above, for each of the three diagnosis-specific analyses, we also censored due to LTSA due to all other diagnoses than the diagnostic group analyzed, whichever came first.
We fitted three models. Estimates were unadjusted in model 1, adjusted for age in model 2, and further adjusted for birth country, type of living area, family situation, educational level and physical demands at work in model 3.
In sensitivity analyses, we repeated the main analyses while changing the definition of LTSA from a sickness absence spell of more than 30 days to a sickness absence spell of more than 14 days. In post-hoc analyses, we explored the reason for the changes in the estimates from model 2 to model 3. This project was approved by the Regional Ethical Review Board in Stockholm.

Role of the funding source
The funders of the study had no role in study design, data collection, data analysis, data interpretation, writing of the report or decision to submit for publication. E.F., J.P., K.A. and K.F. had full access to all data. E.F., K.A., R.R. and K.F. had final responsibility for the decision to submit for publication.

Results
Characteristics of the study population Table 1 shows the baseline characteristics of the study population in the year 2009, separately for the 1 841 250 women and 2 064 435 men. About 86% in both sexes were born in Sweden. Most had at least some high school education and were either single without children or married or cohabitant with children. In total, about three quarters of the population were living in big or medium-sized cities.  Table 2 shows person-years, LTSA cases, LTSA cases per 1000 person-years, the unadjusted (model 1), the minimal (age-) adjusted (model 2) and the most-adjusted (model 3) estimates for the associations between emotional demands and all-cause and diagnosisspecific LTSA in women and men. Notable is that the number of person-years in employees in occupations with high emotional demands is three times higher among women than among men.

Emotional demands and all-cause LTSA
Among women, high emotional demands were associated with a higher risk of all-cause LTSA in all three models, with HRs of 1.34 (unadjusted, model 1), 1.30 (age-adjusted, model 2) and 1.92 (mostadjusted, model 3), respectively. Among men, high emotional demands were associated with a lower risk of LTSA in model 1 (HR ¼ 0.80) and model 2 (HR ¼ 0.64) and with a higher risk of LTSA in model 3 (HR ¼ 1.23). Figure 1 depicts the most-adjusted estimates (HR and 95% CI for model 3) for both all-cause and diagnosis-specific LTSA.
In the age-adjusted model (model 2), emotional demands were also associated with a higher risk of LTSA due to CMD among both women and men; however, associations were weaker than in the most-adjusted model (table 2).

Musculoskeletal diagnoses
In the most-adjusted model, among women, compared with working in occupations with low emotional demands, working in occupations with medium-low (HR ¼ 1.14, 95% CI: In the age-adjusted model (model 2), among women, emotional demands were associated with a higher risk of LTSA due to MSD (albeit with a weaker estimate than in the most-adjusted model), whereas among men, high emotional demands were associated with a lower risk of LTSA (table 2).

All other diagnoses except CMD and MSD
In the most-adjusted model, among women, compared with working in occupations with low emotional demands, working in occupations with medium-low (adjusted HR ¼ 1.25, 95% CI: 1.22-1.28), medium-high (adjusted HR ¼ 1.34, 95% CI: 1.31-1.38) and high emotional demands (adjusted HR ¼ 1.93, 95% CI: 1.88-1.98) was associated with a higher risk of LTSA due to all other diagnoses, except CMD and MSD. Among men, compared with working in occupations with low emotional demands, working in occupations with medium-low (adjusted HR ¼ 0.89, 95% CI: 0.88-0.91) and medium-high emotional demands (adjusted HR ¼ 0.79, 95% CI: 0.77-0.80) was associated with a lower risk of LTSA due to all other diagnoses, whereas working in occupations with high emotional demands was associated with a higher risk of LTSA due to all diagnoses (adjusted HR ¼ 1.13, 95% CI: 1.11-1.16).
In the age-adjusted model (model 2), among women, emotional demands were associated with a higher risk of LTSA due to all other causes (albeit with a weaker estimate than in the most-adjusted model), whereas among men, high emotional demands were associated with a lower risk of LTSA (table 2).

Sensitivity analyses
When repeating the main analyses while defining LTSA as sickness absence spells with a duration of more than 14 days instead of sickness absence spells of more than 30 days, results were similar (results shown in Supplementary table S1 and Supplementary figure S1).

Post-hoc analyses
Post-hoc analyses revealed that the main reason for the changes in the estimates from model 2 to model 3 was the adjustment for physical demands at work (Supplementary table S2). After adding physical demands as a covariate, the association between high emotional demands and risk of LTSA became stronger in all analyses. In men, in some analyses, the estimate for emotional demands even changed direction, from being associated with a lower risk of LTSA before adjustment to a higher risk of LTSA after adjustment.

Discussion
In this prospective, nationwide and complete cohort study in Sweden, we found that working in emotionally demanding jobs Emotional demands and all-cause and diagnosis-specific LTSA 439 was associated with a higher risk of future all-cause LTSA in both women and men. The risk of LTSA did not differ by diagnosis in women. In men, exposure to emotional demands was associated with a high risk of LTSA due to CMD, whereas risk of LTSA due to MSD and due to all other diagnoses was only slightly elevated.
The findings on all-cause LTSA are in agreement with previous studies, [5][6][7]11 reporting that emotional demands at work are a risk factor for all-cause LTSA. To our knowledge, previously only one study has examined the risk of diagnosis-specific LTSA in relation to exposure to emotional demands at work. 9 Van Hoffen et al. 9 linked survey data of 31 884 non-sick-listed employees in the Netherlands to diagnosis-specific register-based sickness absence data. Adverse psychosocial working conditions were associated with a higher risk of LTSA due to mental diagnoses, and after adjusting for other psychosocial work factors the association between emotional demands and LTSA due to mental diagnoses was the strongest. In this study, we did not account for other psychosocial factors at work. Other JEM-based studies of emotional demands and depressive disorder 19 and all-cause LTSA 11 have accounted for other work-related psychosocial factors and found that estimates for emotional demands were robust for this adjustment.
In our study, women had more than three times more personyears in occupations with high emotional demands than men. Some of the associations between emotional demands and LTSA differed between women and men. In women, the risk was higher in all diagnostic groups with risk estimates of approximately equal magnitude, whereas in men, the risk was higher for LTSA due to CMD and only slightly elevated for the other diagnostic groups. We do not know the reasons for these different patterns between women and men across LTSA diagnoses. One explanation could be that the psychological, biological and behavioural mechanism through which emotional demands affect workers' health may be partly different for women and men. This should be examined in further studies.
Post-hoc analyses showed that after adjustment for high physical demands at work, the estimates for the association between high emotional demands and risk of LTSA became stronger and in some analyses (among men) even changed direction from a lower to a higher risk of LTSA. Thus, high physical demands seem to confound (mask) the association between high emotional demands and risk of LTSA. We recommend to routinely adjust for high physical demands in future studies on emotional demands results highlight the need that employers, public health decision makers and physicians understand the role of emotional demands at work as a potential contributor to sickness absence and that organizational policies are developed to help workplaces to handle this stressor. 20 Such policies in relation to emotional demands have, however, not yet been tested in intervention studies. It may be debatable whether reduction of exposure to emotional demands at work is possible, as emotional demands may be considered part of the job in several occupations. Identifying other modifiable work factors that may modify possible harmful effects of work-related emotional demands may therefore be important. Previous research suggests that increasing possibilities for development and reducing work-related role conflicts, 11 but not increasing leadership quality or influence or reducing physical demands, 5,11 may modify harmful effects of emotional demands at work. Intervention studies, including randomized controlled trials, are needed on these and other factors to evaluate the effectiveness of interventions.

Strengths and limitations
The strengths of this study include: First, the nationwide and complete cohort of all 3.9 million women and men aged 18-60 years in paid work in Sweden without LTSA or disability pension at baseline, thereby circumventing possible selection bias. Second, the use of non-self-reported measurement of emotional demands at work and of administrative data on covariates and LTSA not affected by recall or reporting bias. Third, there were no drop-outs during the 7-year follow-up. Fourth, the availability of information on LTSAdiagnoses, provided by the patient's treating physician.
The study also has some limitations. First, JEM's may led to exposure misclassification as they disregard heterogeneity in exposure between workers with the same job code. Therefore, interpretations should be made on the level of occupations (i.e. risk of LTSA in job groups with specific average exposure levels of emotional demands) and caution is needed when interpreting results on the individual level. 21 Second, this is an observational study and therefore bias by residual confounding, including by other psychosocial work factors, cannot be ruled out. However, previous studies have shown that emotional demands remained as a risk factor for all-cause LTSA, LTSA due to mental diagnoses, and depressive disorder, also after adjusting for other psychosocial factors. 9,11,19 Third, we applied a JEM developed based on Danish data in a cohort of the workforce in Sweden, as a Swedish JEM on emotional demands was not available. Fourth, generalizability may be a limitation. Assuming that the Danish JEM for emotional demands at work can reasonably be applied in a Swedish context, we consider our results generalizable to the workforce in Sweden. Since administrative systems of LTSA differ from one country to another, however, our results may not be generalizable to countries with other types of welfare systems.

Conclusion
Workers in occupations with high emotional demands had a higher risk of all-cause LTSA. In women, risk of all-cause and diagnosisspecific LTSA were similar. In men, the risk was more pronounced for LTSA due to CMD. Post-hoc analyses suggested that the association between emotional demands and risk of LTSA were partially masked when analyses were not adjusted for physical demands at work.

Supplementary data
Supplementary data are available at EURPUB online.