Suominen, Sakari MD, PhD; Vahtera, Jussi MD, PhD; Korkeila, Katariina MD, PhD; Helenius, Hans MSc; Kivimäki, Mika PhD; Koskenvuo, Markku MD, PhD
* Outline any observed associations between background variables in this population of Finnish employees and adverse life events, high job strain, and episodes of sickness absence lasting over 8 days during a 9-year follow-up period.
* Describe associations between adverse life events and job strain, analyzed separately, and sickness absence in male and female employees.
* Relate whether and how adverse life events and job strain interacted to influence sickness absence.
For working-aged populations, work and private life are important sources of psychosocial strain. Work-related strain has been shown to affect several public health domains such as perceived health,1,2 mental health,3 cardiovascular morbidity,4 sickness absence,5–7 and mortality.8–10 By specifying the backgrounds of these health effects, the leading job strain model has identified health risks resulting from a combination of high job demands with low control over job content.11 Observational studies have also shown corresponding health problems after adverse life events such as death12 or severe illness of a close person, divorce, and financial difficulties.13
Although the evidence reviewed suggests that stressors originating in work as well as in other spheres of life affect health, less is known about their independent effects or their relative importance on men's and women's health. Accumulation of life stressors may contribute to health-related outcomes.14,15
However, the effects of job strain and life events may vary between men and women. For example, some studies report a greater health risk for women,13,16 others report a greater risk for men,17 and still other studies suggest that most life event categories influence the risk of health problems equally among men and women.18 In addition, much of the evidence on the job strain-health association is based on male-dominated samples11 or is derived from special populations.19
To examine the role of job strain and life events on health, we monitored subsequent sickness absence in a large population sample. For working populations, sickness absence data provide an objective daily-based measure of health, covering a full-range of illnesses.5–7,19 Both job strain and severe life events have been shown to predict increased rates of sickness absence.5–7,13,19 Earlier research on public sector employees during severe economic decline suggests that both are independent risk factors,19 but with general populations in less extreme economic circumstances this has, according to our knowledge, not been studied previously. The aim of this study was to explore the extent to which job strain and life events are predictive of sickness absences among men and women.
Materials and Methods
The study is part of the Perceived Health and Life Control study, which focused on a random sample of 5000 non-institutionalized Finns at age 15 to 64. In order to allow additional regional comparative studies, the city of Turku in the Southwest of Finland, with, at that time, 160,000 inhabitants, was by purpose 4-fold over-represented in the sample. A total of 3421 (68.4%) responded to a mailed questionnaire survey on job strain, life events, and other factors in 1989. Of these, 3068 were identifiable according to a unique identification number on the questionnaire. A non-response and dropout analysis20 showed no considerable bias in terms of the respondents' demographic characteristics. In 2000, the survey data were, with appropriate authority consent, complemented by linking the unique social security numbers with health register data on sick absences, disability pensioning, and mortality (Statistics Finland) from 1987 to 1998. Also, register data on occupation from 1990 was linked to the survey material. The data linkage was successful for a total of 2368 individuals. After linkage, the data were anonymized before delivery. According to additional dropout analyses, the complete number of original respondents did not, in relation to sex, age group, and educational level, in any way differ from the individuals to whom register data could be linked.
For this study, we selected the 1806 respondents having reported being in gainful employment in 1989. The concurrent Joint Ethical Committee of the University of Turku and Turku University Central Hospital approved the study.
The measurement of job strain was based on the work stress model introduced by Karasek21 and further developed by Karasek and Theorell.11 The model proposes that employees who do not have enough job control to meet their job demands are in a job strain situation, which, if prolonged, increases the risk of stress-related diseases. Thus, adverse stress reactions are hypothesized to occur when a worker's decision latitude is low in a task and the psychological demands of the job are high. Job control was measured as a mean of four items measuring possibilities to take part in the planning of work, to apply skills and abilities, the extent to which work was monotonous and uninteresting (reverse score), and to influence the pace of the work (Cronbach's alpha 0.77; range 1–4; mean 3.20 and SD 0.65). Job demands were measured as a mean of two items, which were “Do you agree with the claim that your work makes too high demands on you?” and “Do you agree with the claim that your work in detail is arranged in a compulsory way?” (Cronbach's alpha 0.52; range 1–4; mean 1.52 and SD 0.59). Job strain situations were cross-tabulated from the median splits of job control and job demands, and categorized as low strain (low demands with high control), active work (high demands with high control), passive work (low demands with low control), and high strain (high demands with low control).11 The ability of the individual items to describe the named work characteristics was tested by factor analysis with polychoric correlation coefficients. The construction of the concepts job control and job demands used in this study corresponded to results on primarily loadings of the items.
The occurrence of adverse life events (0 or ≥1) outside work was measured with eight items focusing on major type of events during the previous 1 to 2 years. The items comprised questions on financial difficulties, divorce, separation from another close relationship, serious illness of a family member, death of spouse, relative, or a close friend, and long-term unemployment (>6 months).
The outcome variable was the total number of spells of sickness absences (>8 days) during the follow-up. Sickness absences proceeding the initial survey date were excluded. Individuals were followed up until the end of 1998. We used the participants' social security numbers, a unique number assigned to each Finnish citizen, to link the data to the electronic records kept on sickness absence by the Social Insurance Institution of Finland. Each sickness absence eligible for reimbursement has to be based on a medical certificate and is forwarded to the institution and recorded, including the dates when each period of absence starts and ends. In Finland, employers/employees are reimbursed by the Social Insurance Institution for loss of salary/costs of sickness absence exceeding 8 days.
Background and Health Behavioral Variables
Age was categorized into three groups (15–34, 35–49, and 50–64 years). Initial socioeconomic status was based on register data from Statistics Finland on occupation in 1990 comprising the classes unskilled workers, skilled blue-collar workers, lower professionals, middle professionals, and upper professionals.
The frequency of high alcohol consumption was inquired by the question “How often do you get drunk?” with five response alternatives: a couple of times a week, once a week, a couple of times per month, less than a couple of times per month, never. For the statistical analyses, the first three categories were combined to form the high consumption group, the fourth category forming the moderate-consumption group, and the fifth category the low consumption group. The frequency of physical exercise was inquired by 1 of 11 questions on leisure time activities. The general question was “How many times during the last month have you done any of the following tasks?” with four response alternatives: not at all, once, two to three times, more often. For statistical analyses the categories once and two to three times were combined leaving three classes (0, 1–3, 4+).
Additional loss of observations was observed in the multi-variate analysis because only 1300 individuals had complete data (data not shown). The greatest loss (N = 323) was caused by missing values on the items on job strain followed by register data on occupation in 1990 (N = 259). The number of missing values for the rest of the variables studied, including the response variable, was considerably lower compared to the two former ones ranging from 0 to 57. The persons thus excluded belonged more frequently to the youngest or oldest age groups and had lower socioeconomic status as compared to the rest. No difference in relation to sex distribution was detected. The rate of sickness absence was statistically significantly lower (P < 0.001) (rate ratio [RR] 0.62, 95% confidence interval [CI] = 0.51–0.75) among the respondents excluded (N = 506) compared to among the remaining subjects (data not shown).
The records were checked for inconsistencies. Overlapping, consecutive, or duplicate spells of sickness absences were combined. For each employee, the number of spells of sickness absences was computed and the follow-up period was measured in person-years. The participants were followed until 1998, or if earlier, until they died or retired. Sickness absence is a rare event and constitutes count data. We applied in the analysis generalized linear models with negative binomial distribution assumption. These models are analogical with more often applied Poisson regression models with one exception. In the negative binomial model there are separate parameters for mean and variance, which leads to the model where so called “over-dispersion” is not a problem as it is in the Poisson model.
The associations of job strain and life events with the subsequent rates of sick leaves per person-years were studied separately and the strengths of the associations were expressed as RRs and their 95% CI. Adjustments were made for background variables (age, socioeconomic status) and additionally for frequency of high alcohol consumption and frequency of physical exercise. Job strain and life events were later included in the same multi-variate model adjusted for demographic and health behavioral variables. The potential synergistic effects of job strain and life events on sickness absence was tested by including the corresponding interaction term in regression models that already included the main effects. All the analyses were performed separately for men and women. The statistical computation was performed with SAS 8.2 software (SAS Institute, Inc., Cary, NC).
Altogether, 1789 spells of sickness absences and 11,462 person-years were recorded with a mean follow-up of 8.9 years. Table 1 shows the association between the background variables, life events, and job strain. Life events were more common among those with lower socioeconomic status and moderate consumption of alcohol. High job strain was more common among women, employees in the intermediate age group, those of a lower level of socioeconomic status, and among physically active participants.
Table 1 also shows the association between the background variables and sickness absence. The rate of sickness absence was 1.5-fold in women compared to in men, 2.0-fold in the oldest age group compared to in the youngest age group, and 2.0-fold among those of the lowest level of socioeconomic status compared to among those of the highest level. Lack of physical exercise was associated with a 1.4 times higher rate of sickness absence. For alcohol consumption, the association was U-shaped. Respondents reporting moderate alcohol consumption had a 16% to 42% (not significant) lower rate of sickness absence than those with high or low consumption had, respectively.
Table 2 shows the associations of life events and job strain analyzed separately with sickness absence. Among men, life events as such were not associated with sickness absence, whereas high job strain predicted a 1.7-fold increase in absence rate adjusted for background variables. Among women, on the contrary, exposure to life events was associated with a 1.4-fold increase in absence rate, whereas job strain showed no association. Additional adjustments for health risk behaviors, as indicated by alcohol consumption and physical exercise, did not substantially attenuate these associations.
We also tested the independent effects of life events and job strain by concomitantly including both job stain and life events into the same multivariate model. As expected from the previously described results, this did not alter the principal results.
Finally, we tested the interaction between life events and job strain on sickness absence. Both in men and women, the interaction was not significant (for men P = 0.36, for women P = 0.70).
In this random sample-based data of 1806 Finns, high job strain was an independent risk factor of sickness absence among men but not among women, whereas for women (in contrast to men) the occurrence of adverse life events outside work independently predicted sickness absence. However, no synergistic effect related to job strain and life events was detected because an interaction term formed of these variables was not significant and did not alter the principal effects of these variables.
According to our knowledge, no previous investigations are available that initially represent a general population with concomitant information on job strain as well as life events and subsequent sickness absence. Our data were of a sufficient size for this analysis. According to the Finnish social security legislation valid throughout the follow-up of this study, the sickness absence has to be medically certified and last for more than 8 days, Sundays and the first day of the period excluded, before sick absences are reimbursed (and recorded) by the Social Insurance Institution. Because the data on sickness absences were gained from the Social Insurance Institution, all observations actually refer to relatively long spells of sickness absences certified by a physician. This means that the outcome variable of the study represents at least relatively serious and long-lasting health problems.22 Although shorter spells of sickness absences may represent a concrete additional problem for production continuity they could not be included in the study, due to characteristics of the Finnish health register data. However, we do not consider this to have caused any remarkable bias because we have been able to focus on the more severe health consequences.
Our findings suggest that job strain and adverse life events outside work play an independent role of their own on health depending on sex. Of these factors, the predominant factor for men seems to be job strain and for women, adverse life events. For men, the results related to job strain were in accordance with the assumptions of Karasek's theory,11 in other words, individuals with low job control in combination with high job demands are in the highest risk of health impairment.
The finding that occurrence of life events is a stronger predictor of sickness absence among women compared to among men is supported by previous findings.13,16 In a longitudinal study of the association between serious events and health among 6095 male and 21,217 female Finnish public sector employees, a greater risk for women was observed16 in terms of number of sick days and length of recovery periods. One possible explanation is that women, on average, have a greater responsibility in family-related roles.16 However, there is also a possibility that our measure of life events was too general to detect sex differences in the health effects of events, as the nature and frequency/occurrence of stress exposure and subsequent health outcome may vary by sex.16,18
One potential limitation of the study was the self-reported and retrospective nature of the life event measurements from a checklist of events, with no objective possibility to measure variation in the severity of the event. The simultaneous reporting of life events and job strain makes the results prone to co-linearity bias (eg, more depressed participants reporting both more negative events and more job strain). This does not, however, jeopardize the outcome measurements, because sickness absence due to depressive disorder is of great public health concern.
Our outcome variable, records of sickness absence based on medical certificates, is a by-product of the clinical care of working-aged populations and can be considered a valid measure of health if the concept of health is understood in terms of social, physical, and mental functioning. Sickness absence records are independent archival data that minimize the possibility of co-linearity bias,6–8 the risk of selective recall bias, and other problems characterizing research with self-reported data. In the Whitehall II study, the number of medically certified sick leaves was a better predictor of all-cause mortality than established self-reported health measures and available objective measures of specific physical illnesses and medical conditions.22 In addition, such sick leaves are a strong predictor of specific causes of death, such as cardiovascular disease, cancer, alcohol-related causes, and suicide,23 and a risk marker for future disability retirement.24
The follow-up covered the whole time period from which register data on sickness absences were available (mean follow-up, 8.9 years), since job strain and/or adverse life events do not necessarily have any immediate impact on the risk of sickness absence. However, the principal findings could be replicated when the follow-up was limited to 3 years.
Although all the participants included were currently employed, a limitation may have been the measurement of long-term (>6 months) unemployment as one of the adverse life events. Being unemployed for a long time may affect job strain in two directions: on the other hand, it may attenuate the total strain as well as inhibit sickness absences, but on the other hand, employment after a long-term unemployment may be stressful in itself. Because of these two potential, mutually opposite directions, the long-term unemployment was, however, accepted to be included as one of the life events studied. Moreover, the percentages of men and women reporting long-term unemployment, was 1.3 and 1.1, respectively.
Missing data on other study variables reduced the number of participants to 1300 in the multi-variate analysis with the greatest loss caused by missing values on job strain (N = 323) and occupation in 1990 (N = 259). The number of missing values for the rest of the variables studied, including the response variable, did not exceed 57. The persons excluded from multi-variate analyses had a lower rate of sickness absence, were younger, and had lower socioeconomic status as compared to the rest. This suggests that these dropouts constituted a group with a more labile labor market position at baseline and also potentially explains the loss of data on job strain and occupation. Because this study aimed to focus on persons with a stabilized labor market situation, we consider that sample attrition is an unlikely source of considerable bias on the results.
Health behavior may play a role in several pathways25 mediating health-related outcomes of social or psychological stressors. In this study, health behavior was approximated by alcohol consumption and physical exercise, which were included in the fully adjusted model as covariates. In our study, health risk behaviors unlikely explained the linkage between psychological stressors and ill health, because adjusting for health behavior did not substantially attenuate the rate ratios observed.
The association between social or psychological stressors and health-related outcomes also seems to be far more complex than what studies have revealed so far. It is probable that several causative agents and webs of interactions are needed before harmful outcomes will occur, for instance, gene-environment interaction.26 The mechanisms by which the influence of life events as well as sustained psychosocial stress are mediated is gradually beginning to be understood, albeit far from completely. The neuroendocrinological pathways (mediated by the hypothalamic-pituitary-adrenal axis), as well as pathways involving cardiovascular reactivity, inflammatory, immunological, and platelet activities, seem to play a role here.25
These data suggest that occurrence of adverse life events is the predominant risk factor of sickness absence in comparison to job strain among women whereas the opposite is true in the case of men. No synergism between these variables was detected.
When the need of sickness absence is evaluated, eg, within occupational health care, not only the evaluation of health and job strain, but also strain originating in private life should be kept in mind in the case of women.
Drs Koskenvuo and Vahtera were supported by the Academy of Finland (projects 105195 and 117604) and the Finnish Work Environment Fund. The Academy of Finland financed the linkage with register data.
1. Stansfeld SA, Bosma H, Hemingway H, et al. Psychosocial work characteristics and social support as predictors of SF-36 health functioning: the Whitehall II study. Psychosom Med. 1998;60:247–255.
2. Amick BC III, Kawachi I, Coakley EH, et al. Relationship of job strain and iso-strain to health status in a cohort of women in the United States. Scand J Work Environ Health. 1998;24:54–61.
3. Stansfeld SA, Fuhrer R, Shipley MJ, et al. Work characteristics predict psychiatric disorder: prospective results from the Whitehall II study. Occup Environ Med. 1999;56:302–307.
4. Bosma H, Peter R, Siegrist J, et al. Two alternative job stress models and the risk of coronary heart disease. Am J Public Health. 1998;88:68–74.
5. North F, Syme SL, Feeney A, et al. Psychosocial work environment and sickness absence among British civil servants: the Whitehall II study. Am J Public Health. 1996;86:332–340.
6. Niedhammer I, Bugel I, Goldberg M. Psychosocial factors at work and sickness absence in the Gazel cohort: a prospective study. Occup Environ Med. 1998;55:735–741.
7. Vahtera J, Kivimäki M, Pentti J, et al. Effect of change in the psychosocial work environment on sickness absence: a 7-year follow-up of initially healthy employees. J Epidemiol Community Health. 2000;54:484–493.
8. Johnson JV, Stewart W, Hall EM, et al. Long-term psychosocial work environment and cardiovascular mortality among Swedish men. Am J Public Health. 1996;86:324–331.
9. Amick BC III, McDonough P, Chang H, et al. Relationships between all-cause mortality and cumulative working life course psychosocial and physical exposures in the United States labor market from 1968 to 1992. Psychosom Med. 2002;64:370–381.
10. Kivimäki M, Riihimäki H, Vahtera J, et al. Work stress and risk of cardiovascular mortality: prospective cohort study of industrial employees. BMJ. 2002;325:857–860.
11. Karasek R, Theorell T. Healthy Work: Stress, Productivity, and the Reconstruction of Working Life. New York: Basic Books; 1990.
12. Kaprio J, Koskenvuo M, Rita H. Mortality after bereavement: a prospective study of 95,647 widowed persons. Am J Public Health. 1987;77:283–287.
13. Maciejewski PK, Prigerson HG, Mazure CM. Sex differences in event-related risk for major depression. Psychol Med. 2001;31:593–604.
14. McEwen BS. Protective and damaging effects of stress mediators. N Engl J Med. 1998;338:171–179.
15. Korkeila K, Korkeila J, Vahtera J, et al. Childhood adversities, adult risk factors and depressiveness: a population study. Soc Psychiatry Psychiatr Epidemiol. 2005;40:700–706.
16. Vahtera J, Kivimäki M, Väänänen A, et al. Sex differences in health effects of family death or illness: are women more vulnerable than men? Psychosom Med. 2006;68:283–291.
17. Bruce ML, Kim KM. Differences in the effects of divorce on major depression in men and women. Am J Psychiatry. 1992;149:914–917.
18. Kendler KS, Thornton LM, Prescott CA. Gender differences in the rates of exposure to stressful life events and sensitivity to their depressogenic effects. Am J Psychiatry. 2001;158:587–593.
19. Kivimäki M, Vahtera J, Thomson L, Griffiths A, Cox T, Pentti J. Psychosocial factors predicting employee sickness absence during economic decline. J Appl Psychol. 1997;82:858–872.
20. Suominen S, Gould R, Ahvenainen J, Vahtera J, Uutela A, Koskenvuo M. Sense of coherence (SOC) and disability pensions. A nation-wide, register-based prospective population study of 2196 adult Finns. J Epidem Comm Health. 2005;59:455–459.
21. Karasek RA. Job demands, job decision latitude and mental strain: implications for job redesign. Admin Sci Q 1979;24:285–308.
22. Kivimäki M, Head J, Ferrie JE, Shipley MJ, Vahtera J, Marmot MG. Sickness absence as a global measure of health: evidence from mortality in the Whitehall II prospective cohort study. BMJ. 2003;327:364–369.
23. Vahtera J, Pentti J, Kivimäki M. Sickness absences as a predictor of mortality among male and female employees. J Epidem Comm Health. 2004;58:321–326.
24. Kivimäki M, Forma P, Wikström J, et al. Sickness absence as a risk marker of future disability pension: the 10-Town study. J Epidem Comm Health. 2004;58:710–711.
25. Strike PC, Steptoe A. Psychosocial factors in the development of coronary artery disease. Proc Cardiovasc Dis. 2004;46:337–347.
26. Caspi A, Sugden K, Moffitt TE, et al. Influence of life stress on depression: moderation by a polymorphism in the 5-HTT gene. Science. 2003;301:386–389.