We studied trajectories in functioning among Finnish women from midlife to retirement age and beyond. Our initial analyses focused on social class trajectories in physical functioning, as the participants become older. Our extended analyses examined whether employment status differentiation, i.e. continuous paid employment, transition to mandatory retirement or disability retirement would modify the class trajectories in functioning.
The main results can be summarised as follows. Firstly, among all women, physical functioning was better in the higher class than in the lower class and deteriorated faster in the lower class, leading to widening class trajectories by age 60 and plateauing after that. However, these overall trajectories obscured the divergent trajectories observed for the three employment statuses. Secondly, for continuously employed women, poorer functioning in the lower class deteriorated faster towards older age, leading to constant widening of the class trajectories. Thirdly, after transition to mandatory retirement at age 55 or older, functioning was also poorer in the lower class but deteriorated equally in both classes. Fourthly, after transition to disability retirement, the trajectories contrasted with the previous ones as functioning improved from initially poor level equally in both classes. Fifthly, simultaneous consideration of all three employment statuses showed convergent trajectories, with the disability-retired catching up the level of functioning among the employed and the mandatorily retired. Nevertheless, for each employment status group, the higher-class advantage in functioning either remained or strengthened.
The overall pattern of these results suggest the cumulative disadvantage model, with accumulating work exposures potentially contributing to ill health [21]. This finding has novel implications for understanding whether work is favourable or unfavourable towards to the end of work career [35]. In our study, exit from paid employment modified the class trajectories in functioning, allowing those who retire due to disability to catch up with those who follow the more favourable trajectories of higher-class employees remaining in work.
Interpretation
Health and functioning typically deteriorate over adult life [20, 36, 37]. In our study, that was the case for the continuously employed and the mandatorily retired women among whom functioning deteriorated clinically significantly in both classes. However, after transition to disability retirement the initially poor functioning improved and reached the level of the continuously employed and the mandatorily retired by age 65, a common mandatory retirement age. The improvement was substantial and highly clinically significant.
Our follow up was 12 years, with ages ranging from 40 at Phase 1 to 72 years at Phase 3. Some prior studies suggest improvement in health after transition to retirement in general or to mandatory retirement [21, 22, 38, 39], whereas some others suggest deterioration towards retirement and beyond [13, 16, 20, 22, 37, 38]. This mixed picture calls for further analyses to add our understanding of the health trajectories and their potential explanations. Considering the bearing of social class brings the two theoretical models into the analysis, i.e. age-as-leveller suggesting narrowing or stable and cumulative disadvantage suggesting widening class trajectories in health [9,10,11].
Following, firstly, the age-as-leveller model, relative equalisation of disadvantage, such as work and non-work exposures, as well as advancing biological frailty in older ages would lead to narrowing class trajectories in functioning [11]. Our cohort can be assumed to be relatively advantaged, as all participants were employed at the beginning of the follow-up and they shared benefits like occupational health care and high employment security, which might speak for narrower class inequalities. Nevertheless, the class trajectories in functioning widened for the employed and remained after transition to both types of retirement, with some narrowing after disability retirement. A number of prior studies have equally reported widening or stable class trajectories in health in late working age and early old age [13, 15, 16, 39]. The evidence for the age-as-leveller effect is limited and might be seen only towards later old age [14, 17], not covered by our study.
Following, secondly, the cumulative disadvantage model, the class inequalities in the trajectories in functioning would widen, as disadvantage accumulates in an unequal way [9, 10]. In our study, widening materialised clearly for women remaining employed. This may be related, in particular, to accumulating work exposures [21]. There were minor signs for widening also after mandatorily retirement, but no signs after disability retirement. Mandatory retirement follows a “normal” route to exit paid employment, with no necessary disadvantage, and this may lie behind the almost equal deterioration of functioning in both classes. The route to retirement due to work disability, in turn, is a lengthy process starting years before statutory retirement age. Work disability results from mismatch between individual characteristics, i.e. loss of health and functioning, in particular, as well as accumulating work exposures [30, 40]. This double nature of work (dis)ability may help understand the poor initial functioning among the disability-retired, its subsequent improvement as well as the stability of the class inequalities in the trajectories in our study. The improvement in functioning found after transition to disability retirement may be an indication of the benefits of the prior rehabilitation and treatment efforts as well as other benefits of being eligible for disability retirement equally in the lower and the higher class, as suggested also by a Swedish study [4].
Studies show some evidence for the cumulative disadvantage model but less for the age-as-leveller model [13, 14, 16, 19]. A reason for the unclear status of the theoretical models vis-à-vis evidence may relate to their general nature. The models predict overall socioeconomic inequalities in health trajectories [9,10,11], whereas subgroups behind and divergent routes to exit paid employment are not captured. Our analyses, based on three-way interaction between age, social class and employment status, suggest divergent trajectories for the subgroups. Transition to retirement, in particular, and social class, to a lesser extent, modified the trajectories in functioning among originally employed women from midlife to post-retirement age. Yet, the explanatory power of the two theoretical models may be better visible in later old age.
Examining the joint importance of age, social class and employment status for functioning, a major novel finding of our study was that retirement type, i.e. transition to either mandatory or disability retirement, modified the class trajectories in functioning during ageing. The most striking modifying effect was the contrast between the deterioration of functioning after mandatory retirement but improvement after disability retirement. There is some prior evidence from Britain on retirement types among non-manual classes that is in accordance with our findings [20]. Finally, considering simultaneously all trajectories, functioning converged between the continuously employed, the mandatorily retired and the disability-retired women, and reached an equal level of functioning by age 65. The prior evidence is also in accordance with the convergent trajectories that we found [20]. Throughout, the class inequalities in functioning among the studied women remained for the three employment statuses towards retirement age, suggesting that the inequalities in functioning are deep-rooted, like health inequalities in general [41].
Methodological considerations
Our data came from a relatively large prospective cohort, collected at Phases 1–3 among originally employed women. The study design was suitable for our purposes to examine class trajectories in functioning among employees facing transition to retirement. We measured SF-36 physical functioning, which is a reliable and validated instrument [31].
There were limitations as well. Firstly, non-response and attrition are inherent problems in follow-up surveys [42]. Response to our surveys was acceptable or good. Our non-response analyses have shown that the data are largely representative of the target population, with lower class responders somewhat underrepresented at baseline [25]. Further analyses including Phases 1–3 showed that the class difference in responding was minimal for ages below 60 and somewhat larger for ages above 60. Class mobility over Phases 1–3 was minor and more often upwards than downwards; overall, 90% remained in the same class. While lower class responders had poorer functioning than their higher-class counterparts, this difference was similar between the responders and the non-responders. Employment status distributions over age were largely similar in both classes, but after age 55, the proportion of disability-retired was higher in the lower class whereas the proportion of mandatorily retired was higher in the higher class. The effects of non-response and attrition are likely modest and conservative [42]. Nevertheless, we acknowledge these are potential sources of bias. Secondly, 80% of our respondents were women, which corresponds to the Finnish municipal sector. The female majority in our data allowed us to focus on women only, whereas many prior studies have been male-dominated or included exclusively men. Sensitivity analyses among men suggested somewhat less steep deterioration and narrower class inequalities in functioning compared to women. However, the number of men was too small for reliable analyses of the modifying effect of employment status. Thirdly, we adjusted for working hours as well as period and cohort in sensitivity analyses, but these had negligible effects on the trajectories. Fourthly, further sensitivity analyses used four social classes instead of two, but the results were practically identical. Only for employed women, four classes showed slightly stronger widening of the class trajectories in functioning. Fifthly, selective mortality may constrain the widening of the class trajectories, as those in lower classes have poorer health and higher mortality [16]. Maximal age among our participants was 72 years and this limits the number of deaths. Sixthly, our retirement variable derived from survey questionnaire, which lacked details on part-time retirement and diagnoses for disability retirement. Finally, our results cannot be directly generalised to other countries, the Finnish population or labour force at large, but they represent better public sector employees and, in particular, municipal employees.