Introduction

The populations of Western countries are ageing. In Europe, the number of people of working age (15–64 years) for each person aged ≥65 years is projected to decrease from over three in 2016 to under two by 2080 (an increase in the old-age dependency ratio from 29.3% to 52.3%) [1]. In a bid to mitigate the economic challenges posed by an ageing population and the projected insufficiency of resources for pensions, governments are introducing policies and legislation to encourage people to remain in paid work to older ages.

Macro level measures to extend working life may however be limited in their efficacy and run the risk of widening social inequalities and the disability employment gap because they do not recognise the complex individual level barriers and facilitators that accumulate throughout the lifecourse to affect a person’s ability to extend their working life [2–4]. In particular, the capability of women to work to older ages is an especially pertinent issue in the light of the recent harmonisation of the pension age for men and women in many European countries. Government policies do not reflect that work is different for men and women across the lifecourse and that, in general, women’s work is more likely to be insecure, part-time, poorly remunerated, and more likely to be dovetailed with family and caring responsibilities [3, 5]. Moreover, although many people thrive on work and will continue to reap its physical, mental, social and financial benefits at older ages, it is not universally feasible to maintain all types of occupations well into the seventh or eighth decades of life and this capability varies not only individually but also regionally and between communities [6]. Different occupations are associated with variable levels of physical and/or mental strain on employees and what is acceptable without deleterious consequence to health or wellbeing will change throughout the working lifecourse. Low-paid older workers who are employed in manual occupations, or who experience disability or chronic morbidity, may face the double jeopardy of needing to earn for longer but being unable to remain in their established job [6].

It has been suggested that strategies to extend working life should be individually-focussed, recognising the heterogeneity of older workers, and enabling them to continue to work in a way that matches their capabilities and needs, whilst complementing their wider social context [7–9]. Health is a major reason for early exit from employment [10]; a focus on identification of the wider drivers of health-related job loss (HRJL), irrespective of specific illnesses or diagnoses, therefore offers the potential to inform the development and implementation of workplace strategies to encourage and enable work to older ages.

We used OVID to search Medline and Embase for papers describing predictors of HRJL in general population samples of older workers. A combination of free-text terms were used to identify HRJL because disability retirement is not a formal exit mechanism from employment in all countries (the UK amongst them [11]) and relevant articles might otherwise have been missed. Risk factors identified included: poor socioeconomic circumstances [2, 12–16], low educational attainment [2, 12, 14–16], and difficult financial circumstances [14, 17]; lifestyle risk factors and poor health behaviours (smoking [18–25], high alcohol intake [19, 20], low physical activity [22, 24, 26–28] and obesity [16, 18, 20, 27, 29, 30]); physical demands of work and ability to cope with them [13, 18, 24, 31–34]; mental job demands [32]; and psychosocial aspects of work including effort-reward imbalance, job demand and control [16, 24, 28, 33–38], and job satisfaction [32–34].

Only one study of 14,708 Dutch employees followed-up between 1999–2008 considered a wide panel of personal and employment related risk factors [24]; educational inequalities in health-related exit from work were shown to be partly mediated by health, lifestyle and work characteristics. Most of the studies cited above were from Scandinavia or other areas of mainland Europe, only two papers included data from the UK [2, 19]. Carr et al showed that workers with low socioeconomic position were at increased risk of HRJL at older ages in seven cohort studies (of which four were British); however, the only participant characteristics considered were age, sex, education, occupational grade, and self-rated health [2]. Hagger-Johnson et al showed that smoking, heavy drinking or a poor diet in midlife were risk factors for HRJL by early old age in 7,704 men and women in the Whitehall II study of civil servants; they called for future research into the mechanisms underlying their findings, and for consideration of gender differences [19].

To address this gap in knowledge we examined a broad panel of personal and employment-related risk factors for health-related job loss over two years of follow-up among participants in the Health and Employment After Fifty (HEAF) Study, UK.

Methods

Results

Characteristics of HEAF participants

Table 1 describes the participants by sex and employment status between baseline and 2-year follow-up; 75% of men and 66% of women were employed at some point. Table 1 shows that people who were employed at some point were on average more likely to be: younger; to have financial dependents; to mortgage rather than own their home outright; and to be doing some weekly physical activity. Men who worked were also more likely to be never smokers (although the proportion of current smokers was similar), and were more likely to be in the overweight BMI category (although the proportion who were obese was similar). Employed men were more likely to be married, and employed women were less likely to be married but more likely to be better educated, than their non-working counterparts.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics according to sex and employment status between HEAF baseline and 2-year follow-up.

https://doi.org/10.1371/journal.pone.0239383.t001

Table 1 also reveals sex differences in the characteristics of employed HEAF participants. With respect to socio-demographic and lifestyle characteristics, men were, on average, more likely than women to: be married; have qualifications higher than school level; earn ≥50% of household income; have financial dependents; be financially comfortable; have access to a private pension; have no weekly contact with friends/family outside the household; be in the overweight BMI category; be heavy drinkers or ex smokers. In terms of their work, men were more likely than women to: be self-employed; work nights; have a physically-demanding job; have a very short or very long entitlement to sick pay; and be eligible for an ill-health retirement pension. In contrast, women were more likely than men to report difficulties coping with the mental demands of their job.

Occupations by work pattern between baseline and 2-year follow-up

Fig 1 shows the percentage distribution of prevailing occupation (coded to 1-digit level of SOC2010) according to sex and work pattern over the 2-year period of follow-up. Sex differences are apparent; on the whole, men were more likely than women to be employed in skilled trades, in process, plant or machine operative roles, or in elementary occupations. In contrast, women were more likely than men to be employed in administrative and secretarial occupations, caring, leisure and other service roles, or in sales and customer service.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. SOC2010 1-digit prevailing* job code by work pattern between HEAF baseline and 2-year follow-up.

*Job codes are those prevailing at the time of the first job exit of the type indicated. Job code for first reported job between baseline and 2-year follow-up is graphed for people in work with no exits. HRJL: health-related job loss.

https://doi.org/10.1371/journal.pone.0239383.g001

Fig 1 also shows that men reporting HRJL were less likely to be employed as managers, directors and senior officials, in professional occupations, or in associate professional and technical occupations, than those who stopped working not for health reasons.

S3A and S3B Appendices expand on Fig 1 and show the distribution of prevailing occupational codes at the 3-digit level of SOC2010 by work pattern. The most frequently occurring individual occupations among men who left work for health-related reasons were ‘Road transport drivers’ and ‘Vehicle Trades’; for women these were ‘Teaching and Educational professionals’, ‘Nursing and Midwifery professionals’ and ‘Caring Personal Services’.

Longitudinal analysis of risk factors for health-related job loss

The 2-year survival analysis file included 2418 men and 2614 women of whom 108 and 176 experienced a HRJL respectively. Rates of HRJL per 1,000 person-years employed were 25.0 (95%CI 20.7, 30.2) for men and 38.3 (95%CI 33.0, 44.4) for women. Table 2 shows hazard ratios for HRJL for one risk factor at a time, adjusted for age. Characteristics associated with increased risk of HRJL among men and women were: close proximity to expected retirement; difficulty managing financially; no weekly physical activity; job dissatisfaction; job insecurity; and difficulty coping with the physical, and mental, demands of the job. In addition, owning one’s home outright, self-employment and lower physical work score were associated with reduced likelihood of HRJL among men. Women who were highly educated, were obese/severely obese, or were current smokers, were at increased risk of HRJL. No other job characteristics were associated with HRJL.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Risk factors for health-related job loss: one risk factor at a time, adjusted for age.

https://doi.org/10.1371/journal.pone.0239383.t002

In final mutually-adjusted models (Table 3), the important socio-demographic risk factors for HRJL were: close proximity to retirement (both sexes), and high educational level (women only). Important lifestyle risk factors were: no weekly physical activity among men, and current smoking and being in the overweight BMI category among women. Job dissatisfaction and difficulty coping with the physical demands of the job were employment related risk factors among men and women (Table 3 and Fig 2). However, difficulty coping with the mental demands of the job was only a risk factor among women. Occurrence of HRJL varied markedly according to number of employment related risk factors: for example, HRJL was only experienced by 2.0% of men and 3.4% of women who reported being satisfied with their job and coping with its physical and mental demands. In contrast, HRJL was experienced by 21.9% of men and 24.5% of women who were dissatisfied with their job, and not coping with its physical or mental demands.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Employment related risk factors for health-related job loss by sex.

HRJL: health-related job loss. Each plot was derived from the final mutually adjusted models for HRJL as presented in Table 3, with the exception of ‘coping with the mental demands of work’ among men; this plot was derived for illustrative purposes only from a model which included this work characteristic in addition to all of the variables included in the final model for men.

https://doi.org/10.1371/journal.pone.0239383.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Mutually adjusted hazard ratios for health-related job loss by sex.

https://doi.org/10.1371/journal.pone.0239383.t003

Discussion

In this contemporary prospective cohort study, we explored the sectors in which older workers in England are employed, described the characteristics of their employment and exits from it, and evaluated a wide range of potential risk factors for self-reported health-related job loss. A quarter of workers exited paid work during the two years of follow-up, with a quarter of them doing so for health reasons; women reported HRJL more frequently than men. As expected, the predominant occupations of male and female older workers differed, and the occupations exited relatively more frequently for health than non-health reasons also differed (these were: road transport drivers and vehicle trades occupations (men), and educational, health and social care occupations (women)). Job dissatisfaction and difficulty coping with the physical demands of work were the dominant risk factors for HRJL, independent of socio-demographic and lifestyle factors. Difficulty coping with the mental demands of work was an additional risk factor for women. Overall, HRJL was reported by 2.0% of men and 3.4% of women who were satisfied with their job and coping with its physical and mental demands, but by 21.9% of men and 24.5% of women who were dissatisfied and not coping with these demands.

Our work considers self-reported exit from employment mainly or partly for health reasons and we suggest that it is particularly valuable in doing so. In some countries, people who retire early on health grounds are eligible for a disability pension and are classified as having a “disability retirement”; indeed this type of exit from employment constituted the outcome variable in many of the studies that we identified in our literature review of risk factors for health-related job loss. However, “disability retirement” may only be the tip of the iceberg of health-related early exits from work, and it is not a formal mechanism of exit from employment in all countries (the UK amongst them [11]); disability pension provision in the UK is voluntary (financed privately by the individual or employer). This limits the generalisability of findings between different countries and it is also possible that different characteristics may be implicated as risk factors for a job loss that a person, rather than a set of state defined criteria, attributes to health reasons. People who would qualify for formal disability retirement or benefit can reasonably be expected to comprise a subset of those who self-report exit from work for health reasons. We feel that our consideration of a broad sample of people in whom health is implicated in their reason for stopping work, provides important insights for the development of interventional strategies to extend working life [2, 13].

It is no surprise that work differs in men and women [40] and yet macro level legislation such as changing the age of eligibility for a state pension takes no account of gender differences, or indeed any other heterogeneity in older workers. Our data pertain to a contemporary cohort of older workers in England who are progressing through their retirement transition at a time when recent legislation to raise and harmonise state pension age is taking effect: UK women born in the 1950s (HEAF participants were born between 1949 and 1963) have seen their age of eligibility for state pension rise from 60 to 66 years over one decade and further increases in state pension age are scheduled. HEAF women who reported HRJL during follow-up were particularly employed in educational, health and social care occupations. Employers can play a fundamental role in encouraging fuller working lives by enabling older workers to match their work with their life [42]. Employer led initiatives to retain older workers include [42]: encouraging and enabling flexible working which might include part-time or remote working, or variable start and finish times, which complement the caring responsibilities, physical capabilities and long-term health conditions that comprise the life context of each individual employee [43, 44]; providing opportunities for reskilling and redeployment to less physically demanding roles if desired by the employee; and listening to, engaging with, and responding to the needs of older workers so they feel a valued part of the workforce with opportunities and benefits equal to those of their younger colleagues. The relative importance of these employer based strategies for retaining older workers is likely to differ for employers with a mixed, predominantly male, or predominantly female workforce. Given that older working women are more likely to also have caring responsibilities than their male counterparts [3, 5], our study suggests that flexible working policies might be of particular importance for the prevention of HRJL in sectors with predominantly female workforces such as education, health and social care.

Relationships between socio-economic disadvantage and increased risk of health-related exit from work have been reported previously [2, 13]. The implication is a risk of widening social inequality as a consequence of universal policies and legislation to lengthen working life [2] if those from deprived backgrounds are unlikely to keep working until state pension age. HEAF men who rented rather than mortgaged or owned their home, or who were finding it difficult to manage financially, were at increased risk of HRJL in univariate analyses. However, no socio-economic marker remained important in our fully adjusted model for men, suggesting that proximity to retirement, physical activity, job satisfaction and difficulty coping with the physical demands of work may mediate the impact of socio-economic position on HRJL among HEAF men. Our results suggest a more nuanced relationship between socio-economic position and HRJL among women. In univariate analyses, HEAF women who reported struggling financially were at increased risk of HRJL. However, so were highly educated women, and education remained important in the final mutually adjusted model. Further investigation revealed that incidence of HRJL was particularly high (16%) among women who were highly educated but struggling financially; incidence was 10% among women with low education who were struggling financially. Incidence of HRJL among women who were financially comfortable varied little by education (6% and 5% among women with high or low education respectively). The group of highly educated women who were struggling financially were also more likely than other women to: be divorced/single; earn ≥50% of their household income; have financial dependents; not own their home; work for a large employer, for a short amount of time, on a temporary contract; report job insecurity, dissatisfaction and difficulty coping with work's physical and mental demands. No such effects were evident among men. Interestingly, another study that considered a wide panel of personal and employment-related characteristics of 14,708 Dutch employees [24], found that the effect of education as a predictor of exit from paid employment through disability benefits was partly mediated by health, lifestyle and work characteristics. Our findings appear to support this. The patterns observed in our study are consistent with the suggestion that the consequences of divorce for men are transient, but for women become chronic, resulting in long term loss of income and increased risk of poverty [45]. Our results also hint that these consequences may be particularly acute among women for whom lifetime expectation of socio-economic status (as reflected by education) is not matched by reality in later life. The interplay between educational level, financial circumstances, and marital status as predictors of HRJL merits further investigation.

Relationships between socio-economic position and HRJL are likely to be at least partly attributable to differences in lifestyle factors and health behaviours. Previous studies have shown that smoking [18–25], alcohol use [19, 20], physical inactivity [22, 24, 26–28], poor diet [19] and obesity [16, 18, 20, 27, 29, 30] increase the risk of exit from work through disability retirement or receipt of disability benefits. In our study, physical inactivity among men, and current smoking and overweight among women, were factors associated with HRJL in final models mutually adjusted for proximity to retirement, socio-economic position, job satisfaction and ability to cope with the demands of work. Diet was not assessed in the early phases of HEAF although a short food-frequency questionnaire has been included in more recent follow-ups. We asked about lifestyle risk factors using standard questionnaire tools but it is possible that these were not sensitive to identify a wider set of associations between lifestyle factors and HRJL. A healthy participant effect, common to most cohort studies [46], may have also influenced our results if it diminished the range of lifestyle exposures arising among the study participants and limited our ability to discriminate them as risk factors for HRJL; however, many lifestyle risk factors were broadly similar among those who did any work between baseline and follow-up and those who did none, lessening this concern.

Another way in which social factors might affect risk of HRJL is through the nature and demands of work. Men who were self-employed or whose work did not have heavy physical demands were less likely to report HRJL. Not coping with the physical demands of one’s work was an important risk factor for HRJL in our final models for both men and women, robust to adjustment for all other factors and attenuating the association with difficulty managing financially. Other researchers have reported similar results in studies of exit from work through disability retirement, receipt of disability benefits, or voluntary early retirement [13, 18, 24, 31–34]. It seems that physically-demanding work (more common among people from more deprived backgrounds) becomes more difficult at older ages and that individuals who perceive a mismatch between what their work requires of them and their physical capacity are more likely to experience HRJL. The perceived mismatch may be explained by a growing burden of age-related comorbidities, for example osteoarthritis or COPD, that impact on functional capability, or it may be the perception of their capability relative to perceived demands that changes over time; this requires exploration in future research.

Job dissatisfaction was a dominant risk factor for HRJL among both men and women independent of socio-demographic and lifestyle factors, findings consistent with those from European studies which have shown similar associations between job dissatisfaction [32–34] and other poor psychosocial work characteristics (effort-reward imbalance and low job control) [16, 24, 28, 33–38] and exit from work through disability retirement, receipt of disability benefits, or voluntary early retirement. Job satisfaction is a complex multi-faceted phenomenon incorporating perceived aspects of the work and its rewards as compared with its disadvantages. Implicitly, an individual rating their job satisfaction will also be incorporating, at least to some extent, their personal assessment of their health in relation to their job. We have previously shown in HEAF [47] that job dissatisfaction was more likely in younger male workers and that the main perceptions of work that affected job dissatisfaction were lack of appreciation and/or a feeling of achievement, and difficulty with colleagues at work and/or feeling unfairly criticised. Job insecurity and dissatisfaction with pay were more likely to cause dissatisfaction in the self-employed. Importantly however, our current findings suggest that if job satisfaction could be increased amongst older workers, this might enable longer working lives.

Our study had some limitations. First, full- or part-time working status was only ascertained at baseline so could not be incorporated in the longitudinal analysis file; this question has been reintroduced in recent follow-ups. Second, the information about HEAF participants is self-reported; face to face measures of physical capability and direct measurement of physical activity by wearable accelerometers would be valuable and a pilot study to assess the feasibility of this is underway. Third, we did not have a sufficient number of HRJL events to disaggregate analyses by permanent or temporary exits from employment, and moreover we could not be certain that exits that were not followed by a return to employment within only two years of follow-up would necessarily remain permanent. The ongoing annual longitudinal follow-ups of the HEAF cohort will generate valuable data on these older workers as they move through the retirement transition and we will be able to identify those who leave work permanently as opposed to those in whom there is a trajectory of work exits and re-entries. Fourth, we observed a bias towards healthier participants of higher socio-economic position in the sample of workers versus non-workers during the 2-year follow-up; a bias arising in most cohort studies [46]. Although the estimated incidence of HRJL might be lowered by this bias, the principal relationships that we have explored between participant characteristics and HRJL were internal to the sample. Our results should therefore only be biased if the relationships observed among responders were systematically different than among non-responders which seems unlikely. Fifth, our analyses were based on the sample of participants with complete information on the outcome and all explanatory variables used in each model (i.e. complete case analysis approach). We acknowledge that such an approach results in inefficient estimates and can lead to biased point estimates if the data are e.g. not missing at random. Yet, our analysis identified some strong associations between HRJL and many of the risk factors explored and the risk estimates presented in this study under complete case analysis models are plausible and in keeping with the results of previous studies, suggesting that if there is any effect of bias due to missingness, it is likely to be small. Future work will investigate the usefulness of multiple imputation techniques [48] in the HEAF study. Finally, we have not considered the interplay between specific morbidities and the characteristics identified as risk factors for HRJL in this analysis; however, this is an area for future work in HEAF given its linkage with the Clinical Practice Research Datalink at baseline.

Our study also has many strengths. First, HEAF is a contemporary, general population cohort of older working-age adults in England, widely geographically distributed, which strengthens the topicality and generalisability of our findings to the wider population. Second, HEAF is a large cohort study which has allowed us to explore sex differences in the employment characteristics of older workers. Third, the rich characterisation of HEAF cohort members enabled simultaneous investigation of a wide range of socio-demographic, lifestyle and employment characteristics as risk factors for HRJL; which few studies in the literature have achieved.

Whilst our findings begin to hint at the complexity of enabling work to older ages, there are some key messages for employers and policy-makers. It seems that employers would be well-advised to take a nuanced approach to retaining older workers, considering the needs of male and female workers separately and evaluating their perception of their physical and mental capability in relation to their assessment of the job demands. Attempts to measure, and improve, rates of job satisfaction could also lead to increased retention of older workers. Achieving this might, for example, involve consultation with older workers’ representatives and implementing suitable and inclusive policies around flexible working to enable caring responsibilities [43, 44]. For policy-makers, the message is clear that blanket policies of raising the age of eligibility for state pension and prohibiting age discrimination will not be universally effective at extending working lives: it is generally easier to remain in sedentary jobs than it is to stay working in those which are physically very demanding. Moreover, our results hint that the burden of HRJL among older workers will not be equal, and that those who are socioeconomically disadvantaged and struggling financially (for example as a consequence of the accumulated effects of lower pay, not owning their home, and a lack of private pension arrangements) will be more likely to need other governmental financial support through welfare benefits before they are old enough to claim their state pension. Going further however, what may be needed by policy-makers is a cultural shift from regarding capability to remain in work to older ages as a challenge for occupational health departments and employers to fix, to a recognition that a long working life is something that needs to be addressed and planned for by individuals, employers and public health policies throughout a person’s life [49, 50]. In order to work later in life, mental and physical capacity needs to be preserved and these are both influenced by factors that operate throughout the lifecourse [51]. Our results that risk of HRJL is related to physical activity among men, and smoking and BMI among women, show that lifecourse health behaviours influence work ability beyond the age of 50 years. Employers and policy-makers could usefully consider ways in which to promote healthy behaviours throughout the lifecourse both at work and outside of the workplace, thereby benefitting all.

In summary, our results emphasise that the employment contexts and characteristics of contemporary older male and female workers are different and that characteristics such as job satisfaction and perceived ability to cope with the physical and mental demands of work are key determinants of HRJL. The implications are twofold; first for future research, and second for workplace interventions and policies aimed at extending working life. We recommend that future studies of older workers collect data longitudinally with time-varying covariates and consider sex differences, and investigate the predictive, confounding, or mediating role of a wide panel of potential risk factors for HRJL, with psychosocial work characteristics key amongst these. A next step will be to explore the interplay between the important factors identified in this study with specific morbidities also known to have a marked impact on premature exit from the workforce such as musculoskeletal disorders and mental health conditions. Our results also support calls for the development of policies and workplace interventions which: take a flexible, person-centred approach to extended working life [8, 52], recognise the importance of psychosocial work characteristics, and acknowledge heterogeneity between employees, occupations, and employment settings.

Supporting information

Acknowledgments

We wish to thank the Clinical Practice Research Datalink and the 24 general practices that supported data collection; also, the staff of the MRC Lifecourse Epidemiology Unit who provided data entry and computing support (notably Vanessa Cox). Finally, we thank the HEAF participants for giving their time so generously to participate in the study.