You are here

  • Home
  • Archive
  • Volume 69, Issue 8
  • Job strain and the risk of disability pension due to musculoskeletal disorders, depression or coronary heart disease: a prospective cohort study of 69 842 employees

Article Text

Article menu
Workplace
Original article
Job strain and the risk of disability pension due to musculoskeletal disorders, depression or coronary heart disease: a prospective cohort study of 69 842 employees
  1. Anne Mäntyniemi1,
  2. Tuula Oksanen2,3,
  3. Paula Salo2,
  4. Marianna Virtanen4,
  5. Noora Sjösten2,
  6. Jaana Pentti2,
  7. Mika Kivimäki4,5,6,
  8. Jussi Vahtera2,7
  1. 1Occupational Health Care Unit of Vihti, Federation of Municipalities for Social and Health Services, Nummela, Finland
  2. 2Finnish Institute of Occupational Health, Turku, Finland
  3. 3Department of Society, Human Development, and Health, Harvard School of Public Health, Boston, Massachusetts, USA
  4. 4Finnish Institute of Occupational Health, Helsinki, Finland
  5. 5Department of Epidemiology and Public Health, University College London, London, UK
  6. 6Department of Behavioral Sciences, University of Helsinki, Helsinki, Finland
  7. 7Department of Public Health, University of Turku and Turku University Hospital, Turku, Finland
  1. Correspondence to Anne Mäntyniemi, Occupational Health Care Unit of Vihti, Federation of Municipalities for Social and Health Services, Ritalantie 4 E 6, 03100 Nummela, Finland; anne.mantyniemi{at}fimnet.fi

Abstract

Objectives Observational studies suggest that high job strain is a risk factor for retirement on health grounds, but few studies have analysed specific diagnoses. We examined job strain's association with all-cause and cause-specific disability pensions.

Methods Survey responses to questions about job strain from 48 598 (response rate, 68%) public sector employees in Finland from 2000 to 2002 were used to determine work unit- and occupation-based scores. These job strain scores were assigned to all the 69 842 employees in the same work units or occupations. All participants were linked to the disability pension register of the Finnish Centre of Pensions with no loss to follow-up. Cox proportional hazard models were used to calculate HRs and their 95% CIs for disability pensions adjusted by demographic, work unit characteristics and baseline health in analyses stratified by sex and socioeconomic position.

Results During a mean follow-up of 4.6 years, 2572 participants (4%) were granted a disability pension. A one-unit increase in job strain was associated with a 1.3- to 2.4-fold risk of requiring a disability pension due to musculoskeletal diseases in men, women and manual workers, depending on the measure of job strain (work unit or occupation based). The risk of disability pension due to cardiovascular diseases was increased in men with high job strain but not in women nor in any socioeconomic group. No consistent pattern was found for disability pension due to depression.

Conclusion High job strain is a risk factor for disability pension due to musculoskeletal diseases.

  • Disability pension
  • job strain
  • work
  • musculoskeletal disorders
  • occupational health practice
  • stress
  • public health
  • sickness absence
  • ageing
  • psychology
  • epidemiology
  • sleep
  • retired
  • longitudinal studies
  • disability
  • cardiovascular
  • psychology
  • mortality studies
  • fitness for work
  • workload
  • healthcare workers
  • epidemiology
  • health and safety
  • mental health
  • workload
  • epidemiology
  • longitudinal studies

https://doi.org/10.1136/oemed-2011-100411

Statistics from Altmetric.com

    Request Permissions

    If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.

    What this paper adds

    • High job strain has previously been found to be associated with an increased risk of illnesses such as musculoskeletal diseases, depression and cardiovascular diseases as well as increased likelihood of all-cause disability pension. However, evidence of the association between job strain and subsequent cause-specific disability pensions is sparse.

    • Previous analyses from the Finnish Public Sector study suggested that high job strain is associated with an elevated risk of self-reported all-cause disability pension in those below the age of 55.

    • In this study, cohort-specific measures of job strain, based on occupation and work unit, were used in order to assess job strain more objectively and to minimise subjectivity bias.

    • A one-unit increase in the job strain score was associated with a 1.5–2.4 times higher risk of disability pension due to musculoskeletal disorders in men and women. Among manual workers—a group of employees with a high risk of musculoskeletal disorders—this risk was 1.3- to 1.4-fold.

    • No consistent associations were observed between job strain and the risk of disability retirement due to depression or coronary heart disease.

    Introduction

    Despite increasing life expectancy, people spend fewer years in paid work.1 This poses a substantial economic burden on societies and governments, for example, to ensure future pensions.1 ,2 In the OECD countries (excluding Mexico, Korea and Turkey), 4%–12% of the working-age population receive disability pension benefits annually.1 ,3 Depressive disorders and cardiovascular diseases are the two leading contributors to disease burden across Europe.4 ,5 Mental disorders, along with musculoskeletal diseases, also account for the most common disease groups inflicting ill-health retirement on most Western societies.1 ,6–9 For example, in the USA (in 2009), the UK (in 2004) and Finland (in 2009), 21%–31% of all new disability pensions were granted due to musculoskeletal diseases, 28%–38% due to mental illnesses and 7%–8% due to cardiovascular diseases.7–9

    Tight deadlines, working overtime, job insecurity and the use of information and communication technology are common sources of work-related stress.1 ,6 One of the leading models of work stress, the job strain (demand-control) model, proposes that stress arising from a combination of high job demands and low job control may be harmful to health.10 Recent systematic reviews and meta-analyses suggest that job strain, or its components, is associated with an increased risk of cardiovascular diseases and mental health problems11–13 and an increased rate of all-cause sickness absence,14 a key factor in the long process eventually leading to an early exit from the work force due to ill health.15 ,16 Previous evidence on the association between job strain and musculoskeletal disorders is, however, ambiguous.17 The limited literature available also suggests a link between job strain and subsequent all-cause disability pensioning.3 ,18 However, this evidence is open to bias as earlier studies have been based on self-reports on the outcome and/or exposure and lacked specific information on diseases leading to disability pensions.

    We have earlier found that job strain is associated with the risk of all-cause disability pensions.18 However, that study was limited to self-reports on disability pensions occurring before the age of 55. We have also found that a high level of work time control, self-assessed as well as coworker assessed, is associated with a lower risk of register-based disability pension in general and specifically due to diseases of the musculoskeletal system.19 Although work time control is a specific subdimension of the general concept of job control, it is not included in the job strain model. The mechanisms linking control over working times and health are likely to differ from those linking job strain and health.20 ,21

    The aim of the present study was to investigate the relationship between job strain and the subsequent all-cause and diagnosis-specific disability pensions in a large cohort of public sector employees. We took advantage of the fact that over 47 400 employees (response rate 68%) with detailed information on their occupational title and work unit had provided self-reported data on psychosocial exposures, including job demands and job control, the components of job strain. Our main interest was occupation- and work unit-based job strain as a risk factor for disability pension due to musculoskeletal disorders, depression or coronary heart disease (CHD).

    Methods

    Study sample and design

    The data were derived from the Finnish Public Sector Study, an ongoing prospective cohort study targeting employees in the municipal services of 10 Finnish towns and 21 public hospitals. The Ethics Committees of the Finnish Institute of Occupational Health and the Hospital District of Helsinki and Uusimaa approved the whole study. From 2000 to 2002, the eligible participants for the baseline survey to measure job strain included 71 705 employees who had a job contract with duration of at least 6 months in the target organisations. Of these, 48 598 employees responded to the survey (a 68% response rate). We calculated the mean job strain score from individual responses of all respondents (1) in the same work unit and (2) in the same occupation within the same organisation (town or hospital district) and linked this information, by using personal identification numbers (ie, unique numbers containing the birth date and a code indicating sex assigned to all citizens in Finland), to the whole eligible population irrespective of whether they responded to the questionnaire or not. This enabled us to construct two cohort-specific job strain scores (occupation- and work unit-based job strain) that were used as objective indicators of exposure to work stress for all eligible employees, an effective but rarely available strategy to reduce subjectivity bias in large-scale cohort studies. Work unit-based job strain score has previously been used by Laine et al.18 However, that study did not enable extrapolation of job strain scores to the non-respondents and was further restricted to the survey responses of the participants below age of 55 at follow-up. In the present study, all the participants were linked by their personal identification number to their records in the national pension and health registers from 1994 to 2005. We excluded 824 employees who were working in work units or occupations with less than three respondents and 1039 employees who were either on extended (over 90 days) sick leave, had retired or died by the beginning of the follow-up that started 1st of January of the year following the survey. Thus, the analytical sample comprised 69 842 employees, aged 44.3 years on average (range 17.0–64.0) at baseline. Compared with the non-respondents, there were more women (81% vs 66%) and lower non-manual workers (53% vs 42%) and less manual workers (18% vs 25%) among the respondents. There were no differences in relation to age, medical conditions or job contract.

    Exposure: job strain

    We measured individual-level job strain from the 48 018 complete survey responses to the items measuring job demands (three items) and job control (nine items), derived from the Job Content Questionnaire (10, for details, see Laine et al18). A five-point Likert-type response format, ranging from 1 (totally agree) to 5 (totally disagree), was used for all items. The mean of individual job control scores was subtracted from the mean of job demand scores to form the job strain score.22 A higher score indicated greater job strain. For each participant, we then calculated two cohort-specific job strain scores based on the two groups (s)he belonged to: the work unit and the occupational group.

    Job strain based on work unit

    We determined each participant's work unit from the employers' administrative records and identified 3699 functional work units (mean size 12.0 person-years, range 3–397), that is, the unit at the lowest organisational level that was typically at a single location (eg, a school or a hospital ward). Work unit-based job strain was calculated as the mean of all individual job strain scores from the same work unit, and each participant in the same work unit was given same work unit-based job strain score regardless of their survey response status. Intraclass correlation23 was 18% indicating significant variance in job strain between work units (18% of the total individual differences in job strain are at the work unit level).23

    Job strain based on occupational title

    In a similar way, we used the survey responses to construct a job exposure matrix (JEM) to assess occupation-based job strain. We used participants' occupational titles (based on the International Standard Classification of Occupations (ISCO-88)), obtained from the employers' administrative records, to determine a job axis for JEM. The exposure axis was calculated as a mean of all the individual job strain scores of the same occupational title in each workplace (town or hospital district). A total of 1259 occupations with at least three respondents were identified (mean group size 40.8, range 3–1178). Each participant in the same workplace with the same occupational title was given the same JEM job strain value (occupation-based job strain) regardless of whether or not they were respondents. Intraclass correlation was 14%, indicating significant variance of job strain between occupations.

    Outcome: disability pension

    By using personal identification numbers, we linked the participants to the register kept by the Finnish Centre for Pensions. This institute provides virtually complete retirement data as it coordinates all earnings-related pensions for permanent residents in Finland. All gainful employment is insured in some pension scheme and accrues a pension. Thus, the pension data were available for all participants irrespective of their employment status, workplace or sector (public or private) at the follow-up. We obtained information of the dates and the main diagnoses of all granted permanent or fixed-term (full time or partial) disability pensions. The main diagnoses for disability pensions were coded according to the International Classification of Diseases, 10th Revision (ICD-10). In addition to all-cause disability pensions, we analysed three common disease categories for disability pensions: musculoskeletal diseases (ICD-10 codes M00–M99), depression (F32–F34) and CHD (I20–I25).

    Baseline covariates

    Baseline covariates included demographic factors, information on baseline health and work unit characteristics. Demographic characteristics (sex, age, socioeconomic position (SEP; higher non-manual, lower non-manual and manual; based on the classification of occupations by Statistics Finland) and the type of job contract (permanent or fixed term)) were obtained from the employers' registers. The Social Insurance Institution of Finland maintains data on physician-defined eligibility for special reimbursement for the costs of medication for chronic diseases. Baseline physical illness (yes/no) was indicated based on either eligibility for special reimbursement for diabetes, asthma, hypertension, cardiac insufficiency or CHD medication at baseline or a cancer diagnosis within the 5 years preceding the beginning of the follow-up. Information on cancer was obtained from the Finnish Cancer Registry recording all cancer diagnoses in Finland. Baseline mental health problems (yes/no) were ascertained if the person had recorded long-term sick leave (≥90 days), hospitalisation due to mental disorder (ICD-10 codes F00–F99), reimbursement for psychotherapy, antidepressant use (purchases equal to at least 3 months of an efficient dose of antidepressant medication, Anatomical Therapeutic Chemical (ATC) code N06A) or special reimbursement for antipsychotic medication due to psychotic disorders within the survey year or the three preceding years. These data were derived from the Hospital Discharge Register24 and the registers maintained by the Social Insurance Institution of Finland. Work unit characteristics, the size of the work unit, the mean age of all employees at the work unit and the proportion (%) of fixed-term employees in the work unit, were drawn from the employers' registers.25

    Statistical analysis

    Follow-up for the disability pension began on the 1st of January following the survey year and ended when the participant was granted a (full time or partial, fixed term or permanent) disability pension (study end point) or other pension, if (s)he died, or on 31 December 2005, whichever came first. Descriptive statistics included mean job strain and the incidence of disability pension per 1000 person-years by baseline characteristics; differences were studied with analysis of variance and Cox proportional hazard models. To study the associations of the two job strain scores (continuous variables) with all-cause and diagnosis-specific disability pensions, we used Cox's proportional hazard models to calculate HRs and their 95% CIs per one-unit increase in the score. All analyses were stratified by sex and SEP. Adjustments for the risk of all-cause disability pension were made in four steps: model 1: age, sex and type of work contract, model 2: model 1 and additionally work unit characteristics, model 3: model 2 and additionally baseline physical and mental illnesses and model 4: model 3 and additionally the participant's SEP (when applicable). For the diagnosis-specific disability pension, we report results based on model 1 only. The measure of job strain could vary between the values −4 to +4. To ease interpretation and avoid negative values in table 1, we rescaled the scores by adding a constant +4 to the mean values of job strain. The original job strain scores were used in all statistical analyses.

    View this table:
    Table 1

    Baseline covariates (n=69 842) and their association with job strain and incidence of all-cause disability pension (n=2572)

    As a sensitivity analysis, we examined the associations between self-assessed job strain and the study endpoints using individual scores for job strain in a subset of 45 579 participants who had responded to the survey at baseline and had no missing data in any of the study variables. Another sensitivity analysis examined the risk of all-cause disability pension related to work unit- and occupation-based job strain scores among the non-respondents and respondents separately. Both sensitivity analyses were adjusted for age, sex (when appropriate) and type of job contract.

    All analyses were performed using the SAS statistical software, V.9.1.3 (SAS Institute Inc.). Minimally adjusted and multivariate-adjusted HRs were calculated with the TPHREG procedure.

    Results

    Of the 69 842 participants, 76% were women, 49% lower non-manual workers, 81% had a permanent job contract and 3% were over 60 years old (table 1). Almost half worked in work units with <21 employees. The means (SD) of occupation- and work unit-based job strain were 3.57 (0.45) and 3.58 (0.56), respectively. There were notable differences in job strain by baseline characteristics: women, manual workers and those in larger work units had higher levels of job strain in their work units and occupations. During the mean follow-up of 4.6 years (range 0.02–5.5), 2572 participants (4%) were granted a disability pension (incidence per 1000 person-years 8.0). An increased risk of requiring a disability pension was observed in older age groups, permanent and manual employees and among those with a physical or mental illnesses at baseline.

    Table 2 shows the association of occupation- and work unit-based job strain with all-cause disability pension by sex and SEP. After adjustments for age and type of job contract, one-unit increase in job strain was associated with a 1.5–2.1 times higher risk of requiring a disability pension in men and a 1.3–1.7 times higher risk in women depending on the measure of strain. Further adjustment for work unit characteristics and baseline medical conditions had little effect on these associations. Although additional adjustment for SEP attenuated these associations 45%–57% in men and 71%–73% in women, they remained significant in all cases with one exception: work unit-based job strain was no longer associated with disability risk in women. In the analyses stratified by SEP, the association between job strain and disability pension was statistically significant in manual employees only, showing a 24%–39% increase in the risk for disability pension. The association was of similar magnitude in upper non-manual employees, although it did not reach statistical significance. Further adjustments did not notably change the associations.

    View this table:
    Table 2

    Job strain and all-cause disability pension by sex and SEP

    Two-thirds of all 2572 disability pensions were granted due to the three main causes studied: musculoskeletal disorders (40%), depression (19%) and CHD (2%) (table 3, for all diagnoses, see appendix 1 (provided only in the online journal)). One-unit increase in job strain was associated with a 1.5–2.4 times higher risk of disability pension due to musculoskeletal diseases in men and women and a 1.3–1.4 times higher risk in manual workers, depending on the measure of strain. In higher grade non-manual workers, elevated risk was observed for the occupation-based job strain score only. In relation to receiving disability pension due to depression, no consistent pattern emerged: the work unit-based strain score was predictive of an increased risk of disability pension in men and higher grade non-manual employees, while the occupation-based score was associated with an increased risk in women but not in any other group. In relation to retirement due to CHD, the risk was over two times higher in men irrespective of the measure of exposure, while no significant risk was observed either in women or in any socioeconomic group.

    View this table:
    Table 3

    Job strain and cause-specific disability pension by sex and SEP

    As a sensitivity analysis, we examined the associations between self-assessed job strain and the risk of disability pension in a subcohort of 45 579 respondents to the baseline survey (table 4). Of these participants, 1569 (3%) retired during the follow-up. Self-assessed job strain was associated with all-cause disability pension and disability pension due to musculoskeletal diseases and depression in both sexes and all socioeconomic groups but did not show a consistent association with disability risk in relation to disability pension due to CHD. Another sensitivity analysis (data not shown) examined the association between work unit- and occupation-based job strain and all-cause disability pensions in the subgroup of non-respondents (n=22 139), a group with a 1.14 times higher (95% CI 1.05 to 1.24) risk of disability pension compared with the respondents. All the statistically significant results obtained in the main analyses were replicated. For example, one-unit increase in work unit- and occupation-based job strain was associated with a 1.60 (95% CI 1.25 to 2.05) and 1.85 (95% CI 1.43 to 2.40) times higher risk of disability pension in men (adjusted for age and type of job contract). In women, the corresponding HRs were 1.34 (95% CI 1.16 to 1.53) and 1.61 (95% CI 1.37 to 1.89), respectively.

    View this table:
    Table 4

    Sensitivity analysis. Associations of self-assessed job strain and all-cause and cause-specific disability pension

    Discussion

    In this prospective cohort study of nearly 70 000 public sector employees, greater job strain was associated with an increased risk of disability pension in men, women and manual workers. The increased risk was mainly observed in relation to disability pension due to musculoskeletal disorders. These findings were consistent across two different cohort-specific measures of job strain based on work unit and occupational title. Moreover, these results were replicated in a subsample of the non-respondents as well as by self-assessed job strain among employees participating in the survey.

    Earlier studies have reported an association between job strain and/or its components and all-cause disability pensions3 ,18 ,26 or all-cause early retirement,27 while other studies have found no association.28 Of the previous studies, only one measured job strain objectively18 and another outcome objectively.3 Our results are consistent with the earlier studies3 ,18 ,26 and support a link between higher job strain and the increased risk of disability pension in men, women and manual workers (sex combined).

    Many non-medical factors (such as the coverage of disability pension system and the amount of disability benefit) potentially influence the incidence of disability pensioning.1 In Finland, a disability pension may be granted if, due to an illness or injury, the employee cannot return to work within 300 reimbursed sickness absence days (Sundays excluded) during two consecutive years.29 A certificate by a physician with diagnoses according to the ICD-10 is needed. Depending on the reduction in work capacity, a disability pension may be partial (a 40%–59% reduction) or full-time (a 60%–100% reduction) and can be granted permanently or as a fixed-term rehabilitation subsidy.29 In 2010, a total of 46% of the newly granted disability pensions were fixed term.30 A partial disability pension is more often used by municipal than by private sector employees.31 The two-tier pension system consists of a national pension that guarantees minimum income and an employment-based earnings-related pension. The level depends on the wage during the whole working career and is approximately 60% of the wage after 40 years of working history.29

    Disability pensions granted due to musculoskeletal diseases, depression or CHD

    To the best of our knowledge, this is the first study to report an association between job strain and the risk of disability pension in specific illness categories. Previous literature suggests an association of job strain with musculoskeletal diseases,17 ,32 mental disorders12 ,13 and CHD,11 common disease groups inflicting retirement due to ill health on most Western societies.1 ,6–9 Our findings showed that high job strain was associated with the risk of disability pension in some of these disease categories, although the consistency of the results varied between the categories.

    The robust association between job strain and disability pensions granted due to diseases of the musculoskeletal system is important. In most developed countries, 0.5%–2% of gross national product is attributed to the costs of back pain in terms of work loss, sick leave and other indirect costs.33 Diseases of the musculoskeletal system account for 21% of all incapacity benefit claims filed in the UK8 and over 30% of new long-term sick leaves and disability pensions in Finland and Sweden,7 ,34 figures well in line with our data. Previous reviews suggest, although inconsistently, an association between job strain and musculoskeletal problems.17 ,32 ,35 ,36 Moreover, organisational downsizing (a natural experiment on job stress) may cause musculoskeletal morbidity and induce disability pensions due to musculoskeletal causes in employees who keep their jobs.37–39 These associations are not mediated only via job insecurity, the expected correlate of threatened job loss, but also through adverse changes in other psychosocial work characteristics, such as job strain.39 Importantly, we found that job strain was associated with an increased risk of disability retirement due to musculoskeletal causes especially in manual workers, the occupational group with the highest levels of job strain and highest risk of musculoskeletal problems and work disability. Indeed, almost half of disability pensions granted due to musculoskeletal causes were received by manual workers, a group which consists of only one-fifth of the whole study population. Upper non-manual workers' risk for disability pension due to musculoskeletal causes may partly be explained by chance since the number of cases was limited and the CI was wide.

    We also found an association between job strain and disability pensions granted due to depression in a sensitivity analysis based on self-assessment of the exposure. Although there was no consistent association with occupation or work unit-based job strain, all estimates were in the same direction. The association in the sensitivity analysis is probably explained by reversed causality as depression increases the risk of disability pension and the likelihood of reporting high job strain. In parallel to our findings, one study examining the effect of work time control reported a strong association between self-assessed work time control and retirement due to mental disorders, but no association was observed in relation to work time control assessed by coworkers.19 The association of job strain with depression may also depend on SEP as we found work unit-based job strain to be associated with disability pensions granted due to depression in higher non-manual workers but not in employees with lower SEPs.

    In relation to disability pensions granted due to CHD, the findings were also inconsistent. Both objective measures of job strain were associated with disability pensions granted due to CHD in men but not in women, while self-assessed job strain predicted disability pension in women but not in men. However, the latter association is open to reverse causality bias since subclinical CHD may influence the reporting of exposure and simultaneously affect the end point. Because the majority of the participants were women, who have a lower risk of CHD than men, a lack of statistical power may also explain the inconsistency of our results.

    Strengths and limitations

    The strengths of this study include its prospective study design, large study sample and objective outcome measured from reliable registers that cover the whole population and contain practically no loss to follow-up. Indeed, in these data, only 0.1% of the participants moved abroad, and some of these were probably lost to follow-up. We were able to minimise non-response bias by including all employees irrespective of their participation in the survey. Using work unit- and occupation-based scores for job strain and the register-based study end point helped us to reduce the subjectivity bias and recall bias inherent in self-reported data. Moreover, estimates of job strain were derived from the same sample, in which the risk of outcome was studied, thus leading to a more accurate measure of exposure than using general JEM. Also, work units importantly act as a social network and people in same work unit may influence each other and share the same thoughts about work. This social aspect has often been forgotten or underestimated in the previous studies.

    There are also several potential limitations in relation to our study. First, the two measures of job strain we used may overlap because the same occupations are concentrated within certain work units. However, there are also substantial differences. For example, the occupation-based score of job strain is sensitive to differences between class teacher versus special education teacher, whereas the work unit-based score taps the differences between class teachers working at different schools. Second, the relationship between job strain and disability pensions may be different for the subgroup of respondents and for the eligible population because of selection bias. In other words, the difference may be due to distortions that result from factors that influence study participation. The non-respondents did not differ from the respondents in relation to age or medical conditions, but their risk of disability retirement was higher than that of the respondents. Because of the higher disability pension risk, the non-respondents may still be unhealthier. However, selection bias is an improbable explanation to our findings since overestimation of job strain is unlikely among the respondents. In contrast, respondents may be healthier and thus may experience less job strain. Third, reversed causality could affect the results if unhealthy employees report more job strain because of decreased psychological or physical working capacity or by having more negative views and expectations than those held by their healthy coworkers. Because we measured the outcome objectively from a reliable national register, we could also examine the association between job strain and disability in the subgroup of non-respondents. The fact that this association was of a similar magnitude among the non-respondents as among the respondents provides strong evidence against reverse causality bias and/or selection bias. Moreover, job strain defined by means of occupation or work unit is unlikely to be affected by an individual's health status.

    It should be noted that the risk of disability pensioning is likely to result from long-term exposure. Thus, measuring job strain only once may have underestimated the true risk as the estimate for those at risk due to constant job strain is less accurate.40 Given the broad criterion for classification as ‘exposed’ by means of occupation and work unit, it is also possible that we might have misclassified some subjects that might not actually have been exposed in that particular category. However, the results for self-report of job strain converged with work unit- and occupation-based measures. Although we controlled a large number of covariates associated with the risk of disability pension, it is always possible that our associations were confounded by unmeasured third factors. We may also have underestimated the influence of time-varying baseline health. Measuring baseline health from registers precludes identification of individuals with a medical condition who have not been diagnosed or do not receive treatment for their condition. Last, because our study data were derived from public sector employees of a Western welfare society, we encourage research in other branches of industry and other countries to confirm the generalisability of our findings.

    Practical relevance

    Early exit from the labour market poses a substantial economic burden on societies. Reducing job strain is a potential means of preventing disability pensions. However, further research is needed to examine this issue in intervention studies.

    Conclusions

    The results of this study suggest that high job strain increases the risk of subsequent disability pension particularly due to musculoskeletal diseases. It is important for employers and occupational healthcare professionals to take into account these findings when tailoring interventions to sustain work life participation and to reduce rates of disability pensions.

    References

    1. ↵
      OECD. Transforming Disability Into Ability: Policies to Promote Work and Income Security for Disabled People. Paris: OECD Publication Service, 2003:8–86.
    2. ↵
      News from European Commission 7.7.2010: “EU Launches Public Debate on Future Pensions”. http://ec.europa.eu/social/main.jsp?langId=en&catId=89&newsId=839&furtherNews=yes# (accessed 29 Jan 2012).
    3. ↵
      1. Stattin M,
      2. Järvholm B
      . Occupation, work environment and disability pension: a prospective study of construction workers. Scand J Public Health 2005;33:84–90.
      OpenUrlAbstract/FREE Full Text
    4. ↵
      1. Mathers CD,
      2. Loncar D
      . Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med 2006;3:2011–30.
      OpenUrl
    5. ↵
      WHO. The Global Burden of Disease: 2004 Update. Geneva: World Health Organization, 2008.
    6. ↵
      OECD. Employment Outlook. OECD, 2008:1–371.
    7. ↵
      Official Statistics of Finland, Finnish Centre for Pensions and The Social Insurance Institution of Finland. 2009 Statistical Yearbook of Pensioners in Finland. Helsinki: Edita Prima Oy, 2010:104. http://www.etk.fi/fi/service/julkaisut/440/julkaisut?contentPath=fi%2Fjulkaisut%2Ftilastojulkaisut%2Ftilastovuosikirjat%2Ftilasto_suomen_elakkeensaajista_2009&pageOffset=0&firstTime=false (accessed 29 Jan 2012).
    8. ↵
      Department for Work and Pensions. Work and Pension Statistics 2004. Newcastle upon Tyne, 2005:117. http://research.dwp.gov.uk/asd/asd1/workandpens/2004/wps2004.pdf (accessed 29 Jan 2012).
    9. ↵
      Social Security Administration. Annual Statistical Report on the Social Security Disability Insurance Program 2009. Washington, DC: SSA Publication, 2010:35. http://www.socialsecurity.gov/policy/docs/statcomps/di_asr/2009/di_asr09.pdf (accessed 29 Jan 2012).
    10. ↵
      1. Karasek R,
      2. Theorell T
      . Healthy Work: Stress, Productivity, and The Reconstruction Of Working Life. New York, NY: Basic Books, 1990.
    11. ↵
      1. Kivimäki M,
      2. Virtanen M,
      3. Elovainio M,
      4. et al
      . Work stress in the etiology of coronary heart disease–a meta-analysis. Scand J Work Environ Health 2006;32:431–42.
      OpenUrlCrossRefPubMedWeb of Science
    12. ↵
      1. Bonde JP
      . Psychosocial factors at work and risk of depression: a systematic review of the epidemiological evidence. Occup Environ Med 2008;65:438–45.
      OpenUrlAbstract/FREE Full Text
    13. ↵
      1. Stansfeld S,
      2. Candy B
      . Psychosocial work environment and mental health–a meta-analytic review. Scand J Work Environ Health 2006;32:443–62.
      OpenUrlCrossRefPubMedWeb of Science
    14. ↵
      1. Virtanen M,
      2. Vahtera J,
      3. Pentti J,
      4. et al
      . Job strain and psychologic distress influence on sickness absence among Finnish employees. Am J Prev Med 2007;33:182–87.
      OpenUrlCrossRefPubMedWeb of Science
    15. ↵
      1. Kivimäki M,
      2. Forma P,
      3. Wikström J,
      4. et al
      . Sickness absence as a risk marker of future disability pension: the 10-town study. J Epidemiol Community Health 2004;58:710–11.
      OpenUrlFREE Full Text
    16. ↵
      1. Lund T,
      2. Kivimäki M,
      3. Labriola M,
      4. et al
      . Using administrative sickness absence data as a marker of future disability pension: the prospective DREAM study of Danish private sector employees. Occup Environ Med 2008;65:28–31.
      OpenUrlAbstract/FREE Full Text
    17. ↵
      1. Macfarlane GJ,
      2. Pallewatte N,
      3. Paudyal P,
      4. et al
      . Evaluation of work-related psychosocial factors and regional musculoskeletal pain: results from a EULAR Task Force. Ann Rheum Dis 2009;68:885–91.
      OpenUrlAbstract/FREE Full Text
    18. ↵
      1. Laine S,
      2. Gimeno D,
      3. Virtanen M,
      4. et al
      . Job strain as a predictor of disability pension: the Finnish Public Sector Study. J Epidemiol Community Health 2009;63:24–30.
      OpenUrlAbstract/FREE Full Text
    19. ↵
      1. Vahtera J,
      2. Laine S,
      3. Virtanen M,
      4. et al
      . Employee control over working times and risk of cause-specific disability pension: the Finnish Public Sector Study. Occup Environ Med 2010;67:479–85.
      OpenUrlAbstract/FREE Full Text
    20. ↵
      1. Ala-Mursula L,
      2. Vahtera J,
      3. Linna A,
      4. et al
      . Employee work time control moderates the effects of job strain and effort-reward imbalance on sickness absence: the 10-Town study. J Epidemiol Community Health 2005;59:851–7.
      OpenUrlAbstract/FREE Full Text
    21. ↵
      1. Ala-Mursula L,
      2. Vahtera J,
      3. Kouvonen A,
      4. et al
      . Long hours in paid and domestic work and subsequent sickness absence: does control over daily hours matter? Occup Environ Med 2006;63:608–16.
      OpenUrlAbstract/FREE Full Text
    22. ↵
      1. Courvoisier DS,
      2. Perneger TV
      . Validation of alternative formulations of job strain. J Occup Health 2010;52:5–13.
      OpenUrlCrossRefPubMedWeb of Science
    23. ↵
      1. Merlo J,
      2. Chaix B,
      3. Yang M,
      4. et al
      . A brief conceptual tutorial of multilevel analysis in social epidemiology: linking the statistical concept of clustering to the idea of contextual phenomenon. J Epidemiol Community Health 2005;59:443–9.
      OpenUrlAbstract/FREE Full Text
    24. ↵
      1. Leppälä JM,
      2. Virtamo J,
      3. Heinonen OP
      . Validation of stroke diagnosis in the National Hospital Discharge Register and the Register of Causes of Death in Finland. Eur J Epidemiol 1999;15:155–60.
      OpenUrlCrossRefPubMedWeb of Science
    25. ↵
      1. Virtanen M,
      2. Kivimäki M,
      3. Pentti J,
      4. et al
      . School neighborhood disadvantage as a predictor of long-term sick leave among teachers: prospective cohort study. Am J Epidemiol 2010;171:785–92.
      OpenUrlAbstract/FREE Full Text
    26. ↵
      1. Krause N,
      2. Lynch J,
      3. Kaplan GA,
      4. et al
      . Predictors of disability retirement. Scand J Work Environ Health 1997;23:403–13.
      OpenUrlCrossRefPubMedWeb of Science
    27. ↵
      1. Friis K,
      2. Ekholm O,
      3. Hundrup YA,
      4. et al
      . Influence of health, lifestyle, working conditions, and sociodemography on early retirement among nurses: the Danish Nurse Cohort Study. Scand J Public Health 2007;35:23–30.
      OpenUrlAbstract/FREE Full Text
    28. ↵
      1. Mein G,
      2. Martikainen P,
      3. Stansfeld SA,
      4. et al
      . Predictors of early retirement in British civil servants. Age Ageing 2000;29:529–36.
      OpenUrlAbstract/FREE Full Text
    29. ↵
    30. ↵
      Finnish Center for Pensions. 2011 Finnish Centre for Pensions Pocket Statistics. Tampere: Tammerprint oy, 2011. http://www.etk.fi/fi/service/julkaisut/440/julkaisut?contentPath=fi%2Fjulkaisut%2Ftilastojulkaisut%2Ftaskutilasto%2Fpocket_statistics_2011&pageOffset=0&firstTime=false (accessed 29 Jan 2012).
    31. ↵
      1. Gould R,
      2. Nyman H,
      3. Lampi J
      . Osatyökyvyttömyyseläkkeen toimivuus työeläkejärjestelmässä. Eläketurvakeskuksen keskustelualoitteita. Eläketurvakeskus. Function of part-time disability pension in earnings-related pensions. Helsinki: Finnish Centre for Pensions, 2008.
    32. ↵
      1. Linton SJ
      . Occupational psychological factors increase the risk for back pain: a systematic review. J Occup Rehabil 2001;11:53–66.
      OpenUrlCrossRefPubMedWeb of Science
    33. ↵
      1. Cats-Baril WL,
      2. Frymoyer JW
      . The Economics of Spinal Disorders. The Adult Spine: Principles and Practice. New York: Raven Press, 1991.
    34. ↵
      1. Kivimäki M,
      2. Ferrie JE,
      3. Hagberg J,
      4. et al
      . Diagnosis-specific sick leave as a risk marker for disability pension in a Swedish population. J Epidemiol Community Health 2007;61:915–20.
      OpenUrlAbstract/FREE Full Text
    35. ↵
      1. Hoogendoorn WE,
      2. van Poppel MN,
      3. Bongers PM,
      4. et al
      . Systematic review of psychosocial factors at work and private life as risk factors for back pain. Spine 2000;25:2114–25.
      OpenUrlCrossRefPubMedWeb of Science
    36. ↵
      1. Ariëns GA,
      2. van Mechelen W,
      3. Bongers PM,
      4. et al
      . Psychosocial risk factors for neck pain: a systematic review. Am J Ind Med 2001;39:180–93.
      OpenUrlCrossRefPubMedWeb of Science
    37. ↵
      1. Vahtera J,
      2. Kivimäki M,
      3. Forma P,
      4. et al
      . Organisational downsizing as a predictor of disability pension: the 10-town prospective cohort study. J Epidemiol Community Health 2005;59:238–42.
      OpenUrlAbstract/FREE Full Text
    38. ↵
      1. Kivimäki M,
      2. Vahtera J,
      3. Ferrie J,
      4. et al
      . Organisational downsizing and musculoskeletal problems in employees: a prospective study. Occup Environ Med 2001;58:811–17.
      OpenUrlAbstract/FREE Full Text
    39. ↵
      1. Kivimäki M,
      2. Vahtera J,
      3. Pentti J,
      4. et al
      . Factors underlying the effect of organisational downsizing on health of employees: longitudinal cohort study. BMJ 2000;320:971–5.
      OpenUrlAbstract/FREE Full Text
    40. ↵
      1. Kivimäki M,
      2. Head J,
      3. Ferrie JE,
      4. et al
      . Why is evidence on job strain and coronary heart disease mixed? An illustration of measurement challenges in the Whitehall II study. Psychosom Med 2006;68:398–401.
      OpenUrlAbstract/FREE Full Text

    Supplementary materials

    • Supplementary Data

      This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.

      Files in this Data Supplement:

    Footnotes

    • Funding The study was supported by grants from the Academy of Finland (projects 124271, 124322, 126602, 129262, and 132944), the Social Insurance Institution of Finland, the Finnish Association of Occupational Health Physicians, the Finnish Medical Foundation and the participating organisations.

    • Competing interests None.

    • Ethics approval Ethics approval was provided by The Ethics Committee of the Finnish Institute of Occupational Health and The Ethics Committee of the Hospital District of Helsinki and Uusimaa.

    • Provenance and peer review Not commissioned; externally peer reviewed.