There are psychosocial aspects of work that may be either protective or hazardous for mental health.1 One of the dominant models linking working conditions and health is the job strain model.2 In this model the combination of high job demands and low decision latitude (constituting job strain) is a risk factor for psychological distress. Low social support at work was later added to the model as an additional work stressor.3 Decision latitude includes decision authority (or control over the working environment) and skill discretion that includes variety of work and opportunity for the use of skills. A further novel risk factor for mental health is job insecurity.4 Changes in the labour market over the last two decades have led to demands for a more “flexible” workforce with the disadvantage that jobs have less security of tenure that may relate to increased risk of psychological distress.5 6

Self-reported work stressors have been consistently associated with increased risk of psychological distress.7^–15 A meta-analysis of longitudinal studies between 1994 and 2005 found strong relationships linking job strain with psychological distress.16 Modest associations were also found linking psychological distress with low social support at work, high demands, low decision authority and low decision latitude.16 However, studies of psychosocial work characteristics and health suggesting a causal role of working conditions on mental health have been criticised for an undue reliance on self-report outcome measures and for not addressing the possibility of response bias confounding the associations.17 One way of making measurement of work characteristics less subjective is to construct a job exposure matrix assigning mean work characteristics scores based on occupations to individuals. Most studies have used self-report measures of psychological distress or depressive symptoms and only a few have used standardised structured interview measures for psychiatric disorder such as the Composite International Diagnostic Interview10 18 or Clinical Interview Schedule. Studies using interview measures of psychiatric diagnoses have found that job strain10 and high job demands, low skill discretion, job insecurity and low work social support18 were associated with increased risk of major depressive episode. The small number of studies examining psychiatric diagnoses has focused on major depressive rather than anxiety disorders.

There has been increasing recognition of the reciprocal nature of the association between work characteristics and mental health, in that mental health may influence working conditions and perceptions of the workplace as well as working conditions affecting mental health.19 Two types of mechanism might explain this. First, psychological distress in adolescence might lead to selection into jobs with poorer work characteristics. Having a mental illness may lead to the choice of an undemanding but more manageable job, possibly associated with less positive benefits. Secondly, prior psychological distress may negatively influence current perceptions of the working environment. Depressed mood is associated with people having a negative view of their surroundings. This perceptual explanation encompasses not only psychological distress but also dispositional traits such as emotional instability or negative affectivity that may be linked to negative perceptions of the environment.19 Thus in both explanations prior psychological distress may confound the association between current working conditions and psychological distress. Furthermore, prior psychological distress may influence the ability to maintain or keep a job and thus may affect perceptions of job insecurity.

To our knowledge there are no existing studies that have examined the impact of childhood and young adulthood psychological distress on the association between work and common mental disorders in midlife. This paper uses data from the 1958 Birth Cohort20 to study the association between both self-reported work characteristics and work characteristics derived from a job exposure matrix and diagnoses of depressive episode and generalised anxiety disorder at the age of 45 years taking account of earlier psychological distress in childhood, in adolescence and in young adulthood. We hypothesise that psychological distress in childhood and early adulthood predicts adverse work characteristics (low decision latitude, high job demands, low work social support and high job insecurity) in mid-adulthood and that this association remains after adjustment for current adult psychiatric diagnoses. Secondly, we hypothesise that work characteristics and mental health will be associated in midlife and that these associations will not be fully explained by childhood and early adulthood psychological distress.

METHODS

Study population

The 1958 Birth Cohort commenced as a perinatal mortality survey that included 98% of all births in England, Scotland and Wales during a week in March 1958.21 The cohort members have been followed up at ages 7, 11, 16, 23, 33 and 42, with a biomedical follow-up at age 45. Data were obtained from parents and schools (teachers and doctors) on participants at ages 7, 11 and 16 years and through personal interviews at ages 23, 33 and 42 years. A total of 11 405 participants were interviewed at age 33 and 11 419 at age 42.20 After exclusions for death, emigration, permanent refusal, the armed forces and long-term non-contacts, 11 971 participants still in contact with the study at age 45 were invited to a nurse-led biomedical assessment including a computer assisted personal interview. The achieved sample was 9377 with a response rate of 72% of the contacted sample, representing 59% of the eligible sample. Ethical approval for the biomedical survey was given by the South East Multi-Centre Research Ethics Committee (MREC).

Assessment of work characteristics

Karasek’s job strain model has two orthogonal dimensions: decision latitude (comprised of decision authority and skill discretion) and job demands.2 Job strain is the combination of high demands and low decision latitude; the high job strain response (high demands and low decision latitude) was compared with low job strain (the other response categories combined). Work social support was added as a third dimension to the model.3 In this study at 45 years of age, decision latitude was measured by six items (three each on skill discretion and decision authority), job demands by four items, and work social support by three items in a self-completion questionnaire. These items were derived from the Whitehall II Study questionnaire22 based on Karasek’s Job Content Questionnaire.2 These items were selected for the current study because they correlated most highly with total subscale scores in the Whitehall II Study.22 There was good reliability for each of the subscales: Cronbach’s α was 0.79 for decision latitude, 0.66 for job demands and 0.81 for work social support. Scores on these work characteristic scales were divided into tertiles representing high, medium and low scores. Job insecurity was measured by a four-point scale: “How secure do you feel your present job is? Very secure, Secure, Not very secure, Very insecure”,6 which was dichotomised into secure versus insecure.

Job exposure matrix

A job exposure matrix was constructed using occupation classified by socioeconomic group at 42 years of age using the standard occupational classification (socioeconomic group, SEG) which takes occupational group, employment status and size of establishment to sort individuals with similar social and economic status into 17 groups.23 We assigned mean values for decision latitude, job demands and job insecurity to each individual based on their SEG occupational group mean. Scales were derived for job demands (mean 7.52, SE 0.28), decision latitude (mean 13.42, SE 1.48) and job insecurity (mean 2.15, SE 0.76). This method of assigning the mean is crude because it ignores the variance in scores within socioeconomic group.

Sociodemographic and health covariates

Social position in adulthood was based on housing tenure at age 45 years, a key measure of material circumstances classifying people according to whether they owned their housing or lived in public housing and other residential arrangements.24 At age 7 years housing tenure was measured with categories of owner/occupier, council rented, private renting/other. Marital status at 45 years was classified as married/remarried, single and separated/divorced/widowed. Educational attainment at 33 years was grouped into three hierarchical categories: no formal educational qualifications, ‘O’ levels (lower secondary education) and ‘A’ levels or higher (higher secondary education). Physical ill health was measured by a standard dichotomous question on long-standing illness at 42 years.

Adulthood psychological distress

Psychological distress at ages 23 and 33 years was measured using the Malaise Inventory,25 a reliable and valid measure of emotional disturbance and somatic symptoms in this study.26 27 The Malaise Inventory provides a total malaise scale score and a dichotomous measure of high versus low psychological distress; 7 and above indicates high levels of psychological distress.28

Childhood psychological distress

Psychological distress at ages 7 and 11 was measured using the teacher-rated Bristol Social Adjustment Guides29 comprising 146 items of behaviour, belonging to one of 12 separate syndromes. Reliable scales were derived for internalising at 7, internalising at 11, externalising at 7 and externalising at 11 in this sample.27 For each scale, a score in the top 13% defined a case of psychological distress, the lowest 50% were not cases and the remainder were borderline based on earlier studies.27 30 Internalising and externalising behaviours at age 16 were measured using the valid and reliable teacher version of the Rutter Scales.31 For each scale the top 13% were defined as psychologically distressed.27 30 A scale of the cumulative number of times psychological distress was reported for internalising problems at 7, 11 and 16 years was devised (ranging from 0 to 6).

Depressive episode and generalised anxiety disorder at age 45

Depressive episode and generalised anxiety disorder in the previous week were measured by the Revised Clinical Interview Schedule32 administered by a nurse using a computer assisted personal interview at age 45 years. Diagnoses were derived according to standard algorithms for ICD-10 diagnoses. ICD-10 diagnoses were grouped together to produce non-comorbid diagnoses for “depressive episode”, “generalised anxiety disorder” and a summary measure of “any diagnosis” that included “generalised anxiety disorder”, “depressive episode”, “any phobia” (excluding specific phobias) and “panic disorder” and including any comorbid disorders.

Statistical analysis

Initially, logistic regression analyses were conducted to examine the univariate associations between sociodemographic factors and work characteristics and depressive episode, anxiety disorder and any diagnosis, adjusting for sex. The longitudinal associations of childhood internalising and externalising disorders and adulthood malaise with work characteristics at age 45 were examined separately using multinomial logistic regression analyses adjusting for sex, marital status, housing tenure, long-standing illness, educational qualifications and any diagnosis at age 45; linear regression was used to examine these longitudinal associations for the work characteristics scales. For all analyses interactions between sex and each predictor variable were tested and results were reported separately by sex where significant. Longitudinal associations between childhood disorders and adulthood socioeconomic group at 42 years were examined using logistic regression, adjusting for sex. Cross-sectional associations between work characteristics and mental health at 45 were examined using logistic regression analyses, adjusting for sex, marital status, housing tenure, long-standing illness and educational qualifications. To examine these associations taking prior psychological distress into account, further adjustments were made for any internalising problems in childhood and malaise at 23; interactions between the work characteristics and prior psychological health measures were additionally tested. Linear regressions examined cross-sectional associations between the work characteristics scales and depressive episode, anxiety disorder and any diagnosis at 45 years.

Multiple imputation was used to address missing data in the analyses, using the ICE programme in STATA. All psychological health, sociodemographic and work variables reported in this paper were included in the imputation equations; employment status at 33 and father’s social class at 7 and own social class at 42 were also included as they were significantly associated with attrition.27 All living participants were included in the imputation, but analyses were conducted only for those who participated in the study at age 45 and were in paid employment (n = 8243). For the dependent variables, eight imputed values were used in the analyses of depressive episode at 45 years, 16 for generalised anxiety disorder at 45 and 31 for any diagnosis at 45. Missing data on the independent variables ranged from 0.2% to 13%, except for externalising at 16 (23%) and internalising at 16 (22%). Five imputation cycles were run and analyses indicated that the measures were stable across the imputations. In order to address attrition, inverse probability weights were then estimated from a logistic regression model predicting participation in the study at age 45. Sex and all of the independent variables used in the imputation equation, except those measured at 45, and all significant two-way interactions were used as predictors in this logistic regression. The weight was applied to all analyses in this paper.

RESULTS

There were 4281 men and 3962 women in the sample in paid employment at age 45. The prevalence of depressive episode, generalised anxiety disorder and any diagnosis by sociodemographic factors is reported in table 1. Living in rented accommodation and long-standing illness were associated with higher rates of depressive episode, generalised anxiety disorder and any diagnosis. Being single, separated, divorced or widowed was associated with higher rates of any diagnosis. Having no qualifications at age 33 was associated with higher rates of generalised anxiety disorder and any diagnosis.

In general the most adverse tertile of each work characteristic was associated with a higher prevalence of depressive episode, generalised anxiety disorder and any diagnosis (table 1). Few differences in prevalence of disorder were observed by socioeconomic group: “professional workers: employees” and “unskilled manual workers” had higher prevalence of any diagnosis.

Internalising behaviours were associated with an increased risk of reporting low decision latitude and low work social support in adulthood after full adjustment including depressive or anxiety disorder at age 45 (table 2). Conversely, internalising behaviours in childhood were associated with a decreased risk of reporting high levels of demands in adulthood. Externalising problems in childhood were not associated with adult work characteristics (results not reported).

Similar analyses examining the work characteristics scales for socioeconomic groups found that internalising problems in childhood were associated with low job demands, low decision latitude and higher job insecurity (table 3). Further analysis of the association between childhood internalising problems and socioeconomic group at 42 years confirmed the selection of individuals with childhood distress into lower status jobs (table 4). Individuals with distress were more likely to be “personal service workers”, “foremen and supervisors – manual”, “skilled manual workers”, “semi-skilled manual workers” and “unskilled manual workers”. Individuals with distress were less likely to be “employers or managers”, “professional workers” and “intermediate non-manual workers”.

Malaise at 23 was associated with an increased risk of reporting high demands, low decision latitude, high job strain, low support and high job insecurity at age 45 after full adjustment including depressive or anxiety disorder at age 45 (table 2). There was little difference in the magnitude of the effects of malaise at 23 and 33 years. In contrast, analyses of the associations between malaise at 23 and the work characteristics scales for socioeconomic groups found no significant associations; malaise at 33 was significantly associated with low latitude but not with low demands or high job insecurity.

The cross-sectional associations between work characteristics and depressive episode, generalised anxiety disorder and any diagnosis are reported in table 5. High levels of job demands were associated with 4.5 times the risk of depressive episode and high job strain and job insecurity were associated with nearly twice the risk of depressive episode. There was increased risk for generalised anxiety disorder for high levels of job demands (for males only), low levels of support, low decision latitude, high job strain and job insecurity. Low decision latitude was statistically significantly related to any diagnosis but was not related to depressive episode. In general, the associations of work characteristics with psychiatric outcomes did not differ by diagnosis of depressive episode or generalised anxiety disorder. In the job exposure matrix no significant cross-sectional associations were found between the mean scale scores for demands, decision latitude and job insecurity and diagnoses at 45 years.

Cross-sectional associations between work characteristics and any diagnosis at 45 were then adjusted for prior malaise at 23 and any internalising problems in childhood and sex (where not stratified by sex) (table 6). Adjustment of analyses of work characteristics and any diagnosis at 45 years for internalising problems resulted in minimal changes to the odds ratios. Further adjustment for malaise at age 23 reduced the size of the odds ratios by a relatively small amount; all the odds ratios remained significant (table 6). Interactions between work characteristics and both internalising problems and malaise at 23 years were not significant. The only statistically significant interaction was between sex and demands with larger associations in men than in women.

DISCUSSION

We found that mental ill health in childhood predicted adverse work characteristics in midlife for both the self-reported work characteristics and the job exposure matrix. Mental ill health in childhood was associated with lower status occupations. Adverse self-reported work characteristics were associated with increased risk of depressive episode and generalised anxiety disorder and any diagnosis in mid-adulthood. These associations were not found with the job exposure matrix. The associations between self-reported work characteristics and any diagnosis at 45 years remained significant after adjustment for internalising problems in childhood and malaise at 23 years although they were diminished slightly in magnitude.

Child and early adult psychological distress and midlife work characteristics

Internalising behaviours in childhood and psychological distress in early adulthood affect reported work characteristics in midlife as we hypothesised. The effect of distress measured at 23 or 33 years was similar suggesting that psychological distress earlier in the occupational career does not have a more powerful influence on work trajectories than later distress. It may be that both reporting distress and poor working conditions are subject to response bias in that negative affectivity33 is associated with a tendency to report negatively on health and working conditions across the life course. Psychological distress may tend towards negative reports of working conditions. However, an alternative explanation is possible because internalising behaviours in childhood also predict work characteristics on the job exposure matrix. Psychological distress may select people into less advantageous working conditions or mean that promotion into more favourable work is less likely. Stronger associations for childhood distress and the job exposure matrix compared to malaise at 23 years indicate that health selection is taking place during adolescence at the stage of higher education rather than young adulthood. These analyses suggest that childhood psychological distress is related to a lack of upward social mobility that might be linked to less advantaged working conditions in less advantaged occupations. Current working conditions could be a mediator by which early psychological distress perpetuates or evolves into later distress. The lack of interactions between psychological distress and work characteristics and midlife disorder does not support the idea that early psychological distress may be an indicator of personal vulnerability to the effects of work on mental health in midlife, but it is possible that if we used other work characteristics such as effort–reward imbalance, organisational justice or emotional demands at work we might have shown an interaction with early psychological distress as a vulnerability factor.

Two mechanisms have been put forward for the influence of mental health on work: perceptual mechanisms and the environmental change mechanism.19 The perceptual mechanism implies that prior psychological distress will negatively influence current perceptions of the working environment (gloomy perception mechanism). This explanation encompasses not only prior psychological distress but also dispositional traits such as neuroticism or negative affectivity. The environmental change mechanism suggests that prior mental illness may influence either selection into occupations or, within an occupation, may influence working relationships such that the person is no longer treated the same as those without psychological distress. It has been demonstrated that unhealthy people tended to report support, demands and control in more negative terms;34 these associations could relate to changes in perception although secular changes in the workplace might also differentially affect unhealthy people. No evidence was found in de Lange’s studies for the drift hypothesis whereby workers, due to illness, get increasingly unfavourable jobs. However, it is unlikely that previous research has studied a long enough period to identify drift effects.35 It seems possible in this study that psychological distress selects people into increasingly disadvantaged working conditions over the life course. Many adults with depressive and anxiety disorders have also experienced psychological distress in childhood when educational achievement may be compromised, leading to unemployment and limited occupational opportunities associated with unfavourable work characteristics.36 Also within jobs the experience of mental ill health may alter the perceptions and attitudes of co-workers and supervisors.37 Spells of sickness absence and reduced functioning at work might either elicit sympathy or irritation among co-workers with repercussions on reported social support. Supervisors may reduce the workload of psychologically distressed employees, give them less demanding and more trivial tasks, with less responsibility and less opportunities for control over work. This would be best followed up in qualitative studies.

Internalising behaviours and job demands

The finding of internalising behaviours in childhood predicting low job demands is interesting because it is opposite in direction to other associations. This is consistent with childhood internalising problems selecting people into less demanding jobs, particularly for males. Repeated mental ill health in childhood is associated with poor educational attainment and selection into jobs with little responsibility and low demands.38 Alternatively, young people with high levels of anxiety may choose to go into less demanding jobs. While high levels of job demands are a risk factor for mental ill health, so are low demands, which may be found in low status jobs with few expectations.39 It is surprising that we found no association between externalising problems and work characteristics. It may be that participants with more severe externalising problems have either dropped out of the cohort or may no longer be working at 45 years, thus limiting the distribution of externalising problems in the cohort and reducing the likelihood of finding associations with work characteristics.

Early adult distress, but not childhood mental ill health, was associated with current job insecurity. Job insecurity has been related to pessimism, financial difficulties and lack of work social support.6 It is hard to know whether low social support and financial difficulties explain job insecurity or whether they are all concomitants of a threat of job loss. Despite an association with neuroticism40 in a previous study, the current study’s findings suggest that job insecurity is more of a response to current stressors than an indicator of chronic anxiety.

Earlier psychological distress only partially explains adult associations of work and current mental health

In cross-sectional analyses at 45 years work characteristics are substantially associated with psychiatric diagnoses, showing equal magnitude of effects for depressive episode and generalised anxiety disorder. This confirms the first part of our second hypothesis. Naturally, causal associations cannot be inferred from these associations, but they are in keeping with longitudinal studies that have found that work characteristics predict later mental health.7^–15 It might be postulated, in keeping with criticisms of studies on work and coronary heart disease,41 that the associations of work characteristics and diagnoses at 45 years are artefactual, being the result of an association with earlier mental ill health influencing reporting of work characteristics in mid-adulthood and past mental ill health influencing current mental ill health. However, it is not the case that earlier psychological distress explained the mid-adulthood association of work characteristics and diagnoses at 45 years; the associations remained substantial in size after adjustment for psychological distress in earlier adulthood, confirming the second part of our second hypothesis. However, earlier psychological distress does contribute to the cross-sectional association of work characteristics with adulthood psychiatric diagnoses. Although prior psychological distress influences the association between perceptions of working conditions and mental health outcomes, as in other studies,17 the associations of working conditions and mental health are unlikely to be explained by response bias. On the other hand, there was no association between the work characteristics based on the job exposure matrix and depressive and anxiety disorders at midlife. This could be interpreted as saying this association was artefactual based on depressive and anxiety disorders making it more likely for people to report poor working conditions. However, it is probably not a fair test of this association because these contrasting findings reflect the lack of variance in these mean scale scores used in the job exposure matrix; the mean score for the individual’s socioeconomic group for the scale has been assigned, ignoring the variance of responses for the group.

Limitations

The strength of these findings compared with previous studies is that psychological distress was measured longitudinally on several occasions across the life course rather than contemporaneously using a neuroticism scale and the outcomes were interview-rated psychiatric diagnoses rather than self-report questionnaires. There are limitations to this study. Attrition of the population since birth, particularly between 16 and 23 years, has preferentially lost participants of less advantaged social position and some of those with childhood disorders from the cohort,42 although imputation may have dealt with some of this potential bias. In general, cohort attrition is likely to have weakened the associations between childhood psychological disorders and the midlife measures as those with more severe disorders will have left the cohort, thus the effects of childhood psychological disorders may be underestimated. We used an abbreviated version of the work characteristics questions; we might have seen stronger associations between work stressors and midlife disorders if we had been able to use a more extensive work stressor questionnaire. There were limited numbers of participants with depressive episode in mid-adulthood. All of these factors may have tended to weaken the strength of associations between work and psychological distress.

CONCLUSIONS

Overall, the findings suggest that childhood and early adulthood distress influences working conditions in mid-life through a process of health selection. The exact nature of the mechanism by which earlier psychological distress predicts work characteristics can be investigated in future studies with more detailed work environment measures. One possibility is that early academic problems, analogous to later work functioning, may have contributed to childhood internalising problems. Although the exact mechanisms for these effects still need to be established through further research, this study suggests that more attention should be paid to the consequences of childhood psychological distress for the individual’s occupational trajectory and subsequent working environment.