Retirement and mental health: does social participation mitigate the association? A fixed-effects longitudinal analysis

Population aging is occurring worldwide and an increasing number of older adults are engaging in work in late life. Japan is experiencing the fastest rate of population ageing in the world. In Japan, the labor force participation rate among people aged 65 years or older was 21.3% in 2014, which ranked 7th among all OECD countries [1], and the population of older workers increased from 4.8 million in 2004 to 6.8 million in 2014 [2]. To maintain both the quality of life of each older individual and the financial sustainability of healthcare and long-term care systems, preventing physical and mental impairments is the key goals of public health measures for the aged [3]. Specifically, the maintenance of mental health counts as it could be strongly affected by the stressful life events that are likely to happen in older ages. Among those life events, in this study, we focused on retirement, the major events that most people experience.

Transition from work to retirement is a major life event in most people’s lives. However, there is controversy about the impact of retirement on mental health [4]. While many studies done in European countries suggest that retirement is beneficial to mental health [5–8] or has no health impact [9–11], results from some studies in Asian populations indicate the opposite, i.e. the transition to retirement is detrimental to mental health [12, 13]. These geographic differences in the association between retirement and mental health may be partly due to the differences in employment conditions and how retirements are culturally perceived by people in each region.

There are several potential mechanisms through which retirement may have adverse impact on mental health. Above and beyond the immediate reduction in income, social role theory posits that retirement results in the loss of many non-financial benefits of work, such as opportunities for health-promoting social contacts and access to social support that can buffer the adverse effects of stress [14–16]. Activity theory also posits that the transition to retirement results in reduced wellbeing for individuals for whom work provided meaning in life [17].

Moreover, the impact of retirement on mental health may not be universal and could vary between individuals depending on his/her occupational class or social situation (e.g. the presence of social support in the home environment from a marital partner) [4, 6, 18, 19]. In addition, social participation after retirement may compensate for any adverse effect of retirement stemming from loss of sense of meaning in life, which has been suggested to be linked to mental health [4, 12, 13, 20]. However, the potential impact of other types of social participation such as hobby clubs remains unknown.

In the present study, we sought to evaluate [1] the association between changes in working status and changes in depressive symptoms in a cohort of community-dwelling Japanese older adults, [2] whether the association can be mediated by changes in social contacts and social support, [3] the interaction between changes in working status and occupational class, changes in marital status, and social participation after retirement.

When including interaction terms (Additional file 1 : Table S1), we did not find any statistically clear evidence on that the association between changes in working status and GDS scores were altered by occupational class (Model 3), though there were tendency that the relative increase in GDS score among men who were continuously retired were smaller among high occupational classes than lower occupational class (β = −0.08, 95% CI: -0.22–0.07) (Fig. 1). There was no significant interaction between changes in working status and occupational class among women. We observed no significant interactions between transition to retirement and changes in marital status among men and women (Model 4).

Analyses using the sub-dataset showed consistent results: significant increases in GDS score among men who transitioned to retirement (β = 0.40, 95% CI: 0.12–0.68) as well as men who were continuously retired (β = 0.29, 95% CI: 0.11–0.47) compared to men who kept working (Model 5 in Additional file 2: Table S2). Women who transitioned to retirement reported marginally significant increases in GDS scores (β = 0.32, 95% CI: -0.04–0.68, p = 0.09).

Although statistical evidence was weak but there was a trend among men and women that increase in depressive symptoms was less among those who participated in recreational social participation (β = −0.22, 95% CI: -0.72–0.29 for men, and β = −0.28, 95% CI: -0.92–0.37 for women) (Model 6 in Additional file 3: Table S3, Fig. 2 ). After additional adjustment for whether or not subjects have a sense of meaning in life (Model 7 in Additional file 3: Table S3), the magnitude of interaction effect between transition to retirement and recreational social participation among men considerably reduced (β = −0.05, 95% CI: -0.54–0.44), whereas it did not change among women (β = −0.27, 95% CI: -0.89–0.35). Notably, having a sense of meaning in life was strongly associated with increased depressive symptoms among men and women (β = −2.66, 95% CI: -3.00–-2.32 for men, and β = −1.84, 95% CI: -2.11-1.57 for women).

We summarize our key findings. First, the transition to retirement in both men and women and sustained retirement in men were associated with increased depressive symptoms among Japanese older adults. Second, the associations between retirement and increased depressive symptoms remained statistically significant after additionally adjusting for changes in social contacts and social support, suggesting that they did not mediate the impact of retirement on mental health at least among Japanese older adults. Third, there was a statistically non-significant trend that men from higher occupational class backgrounds appeared to be less influenced by sustained retirement, whereas we did not find such a trend among women. Fourth, we found no interaction between changes in working status and changes in marital status. Finally, there was a statistically non-significant trend that men and women who reported recreational social participation after retirement appeared to be less adversely impacted by retirement. Importantly, the magnitude of interaction effect between transition to retirement and post-retirement recreational social participation substantially reduced only among men after additionally adjusting for a sense of meaning in life. This, if causal, may suggest that having a sense of meaning in life through social participation may be crucial in mitigating the adverse impact of retirement on mental health – particularly for Japanese men.

In our data, nearly a quarter of male and female subjects indicated depressive symptoms at baseline defined by the GDS score of 5 or more. Another population-based study in Japan targeting people over the age of 64 reported that prevalence of those who scored more than 5 in GDS-15 was 23.8%, which is compatible to our data [26].

Our finding that retirement may increase depressive symptoms is inconsistent with many studies conducted in European countries [5–11], whereas a previous study in another Asian population supports our result [12, 13]. We speculate that, for Japanese workers (men especially), the loss of non-financial benefits of work outweighs the other benefits of retirement such as a relief from work-related stress overall. Interestingly, we found that changes in social contacts and social supports failed to explain the association between retirement and depressive symptoms. Although convincing evidence is lacking, the adverse effect of retirement may be attributable to a loss of other benefits of work including sense of meaning in life. Indeed, the literature suggests that many Japanese workers tend to find their purpose in life from work [27]. Retirement among Japanese workers may thus result in a loss of a sense of meaning in life.

Among Japanese workers who experienced the post-War period of rapid economic growth, employment was characterized by long hours of work as well as a shared sense of sacrifice with the goal of rebuilding the nation in the aftermath of the destruction wrought by the Second World War [28]. The participants in JAGES study -- all of whom were aged over 65 years at baseline in 2010 -- belong to this cohort. In return for their sacrifice of leisure and family life, Japanese workers of this generation were rewarded with a life-time employment guarantee. Commitment to the company thereby molded the identity of an entire generation of Japanese workers. It is in this context that we may interpret the findings of the present study. The transition to retirement for many Japanese workers entails the severing of links with work colleagues as well as loss of a sense of meaning in life. It is important to note at the same time, however, that the post-War lifetime employment guarantee has broken down in Japanese society in the aftermath of the 1991 economic bubble collapse. Our findings may therefore not apply to future generations of Japanese workers.

Our results also suggest that retirement may be more toxic among Japanese men from lower occupational class backgrounds compared to managerial/professional occupations. One explanation is that men from high SES backgrounds may have extra social roles besides work-related social roles, and therefore, may be less affected by a loss of work-related role due to retirement. Additionally, there may be some health benefits of retirement for the managerial/professional class. Indeed, mortality trends in Japan by occupational class indicate that suicide rates in these groups have risen sharply in the past two decades, possibly as a result of the pressures stemming from economic stagnation [29, 30]. It is possible that retirement can be somewhat beneficial for those workers as it relieves them of the work-related distress. The observed wide confidence intervals for the interaction terms between changes in working status and occupational class among women may be due to the small number of women who were from high-occupational background. Indeed, among female JAGES subjects in 2010, only 12.6% were categorized as high occupational class.

Our findings also indicate the trend that participation in recreational activity could mitigate the increase in depressive symptoms among Japanese retired men and women but perhaps through different mechanisms. Results from the model additionally adjusting for a sense of meaning in life suggests that recreational social participation may benefit retired Japanese men by compensating for a loss of a sense of meaning in life due to retirement. However, this was not the case for retired women. We speculate that recreational social participation for these retired women had differential meaning such as increased opportunities for informal socializing, rather than provision of a sense of meaning in life. Indeed, Takagi et al. also argued that there was a gender difference in mechanisms through which social participation enhances mental health [24]. We found no significant interaction between transition to retirement and social participation with roles including volunteerism, which was inconsistent with preceding findings [12, 13]. Participation in activities involving roles may not be completely voluntary and some participants might have been feeling a psychological distress due to “role strain”, namely a pressure to accomplish their roles [20, 31]. This may explain the non-significant interaction effect of social participation with roles and its wide variance.

Ours is, to our knowledge, the first study to evaluate the association between changes in working status and changes in depressive symptoms using data of Japanese older people. Strengths of our study include the large sample size of older adults with repeated measures of both outcomes and exposures, and the use of first-difference models to account for all time-invariant confounding factors. Nonetheless, several limitations should be noted. First, there is a possibility of reverse causation, such that workers are more likely to seek retirement when they feel depressed [32]. Although we partly addressed this issue by excluding subjects who showed depression at baseline from the analyses, future studies should specify reasons for retirement to permit causal inference. Second, unmeasured confounding factors may also exist. Even though we controlled for all observed/unobserved time-invariant confounding factors and several possible time-variant confounding factors, there may be other stressful life events or circumstances, which causes changes in working status and changes in depressive symptoms. Third, some of our measurements were admittedly crude. We used changes in received social supports from “friends” as a measurement of changes in work-related social relationships. However, it is obvious that social relationships people obtain from work are not necessarily reported as “friends”. Fourth, we evaluated social participation at the same time with our final outcome. Since we hypothesized that post-retirement social participation could mitigate the possibly adverse impact of retirement on mental health, it is ideal to measure social participation before measuring post-retirement mental health. Fifth, generalizability of our findings is limited. As the effect of retirement on mental health is presumably culture dependent, our findings may not be generalizable to other countries or future generations. Lastly, our single item measurement of “a sense of meaning in life” is not validated, and therefore, it may not successfully capture a subject’s sense of meaning in life.

Retirement may increase depressive symptoms among Japanese older people, possibly men with lower occupational class backgrounds in particular. An important public health implication of our study is that there is a possibility that encouraging recreational social participation may compensate for a loss of a sense of meaning in life due to retirement and mitigate the potentially adverse effects of retirement on mental health of Japanese older men.