Abstract

Background

Psychological distress (PD) is a major risk factor for mental health among middle-aged and older adults and affects their quality of life and well-being. This study aimed to examine the evolution of PD with age and the relative importance of its determinants, issues that have been insufficiently studied.

Methods

We used longitudinal data obtained from 17-wave social surveys conducted in Japan from 2005 to 2021, to track 34,128 individuals (16,555 men and 17,573 women) born between 1946 and 1955. We defined PD as a Kessler 6 score (range: 0鈥�24)鈥夆墺鈥�5 and estimated fixed-effects regression models to examine the evolution of its proportion with age. We also conducted a mediation analysis to examine the relative importance of specific mediators such as self-rated health (SRH), activities of daily living (ADL), and social participation, in the association between age and PD.

Results

Regression model results confirmed an increase in PD with age. Poor SRH, issues with ADL, and no social participation were key mediators of aging on PD, accounting for 34.2% (95% confidence interval [CI]: 21.0鈥�47.3%), 13.7% (95% CI: 8.2鈥�19.3%), and 10.5% (95% CI: 8.0鈥�13.0%), respectively; consequently increasing PD between 50 and 75 years.

Conclusion

The results suggest the need for policy support to encourage middle-aged and older adults to promote health and increase social participation in order to prevent depression while aging.

Background

Psychological distress (PD) is a major risk for mental health among middle-aged or older adults and affects their quality of life and well-being [1,2,3]. As people age, they face life events that affect their mental health such as coping with health problems, caregiving for their parents, and losing family members. Many people adequately adjust to these life events; however, some may experience feelings of anxiety, social isolation, or loneliness [4, 5]. When these feelings are serious and persistent, they can lead to PD and other mental illnesses [6, 7], although their impact may depend on their severity and the used definitions.

Several studies have examined determinants and correlates of PD in various contexts. Among others, self-rated health (SRH), which can be used as an indicator for general health status or as a mediator between medical diagnoses and anxiety/depression, has been closely linked to PD and other psychosocial measures [8,9,10]. Researchers have focused more specifically on activities of daily living (ADL) as an important indicator for physical disability; they reported that ADL problems increase the risk of psychological problems, including suicidal ideation [11, 12].

In addition to these health-related variables, socioeconomic factors such as income and employment status have been considered potential determinants of PD [13,14,15,16]. Conventional measures of poverty or economic hardship have also been reported to negatively affect mental health; thus, focus has been placed on the impact of retirement or the transition from work to retirement among middle-aged and older adults [17,18,19]. In addition to these socioeconomic factors, family relationships are closely related to PD. Studies have reported that widowhood and divorce tend to have depressive effects, especially among men [20,21,22]. In recent years, parental caregiving has also been stressed as a key and imminent risk factor for PD among family caregivers, despite formal long-term care services being available [23, 24]. Additionally, an increasing number of studies have stressed the importance of social participation in maintaining mental health later in life [25,26,27]. Participation in social activities is postulated to be crucial for successful aging, as social interactions increase older adults鈥� chances of obtaining social support [28].

Based on the findings of previous studies, this study attempted to provide new insights into PD among middle-aged and older adults in two ways. First, unlike many studies that have used cross-sectional or longitudinal data with a small number of waves, we examined the evolution of PD with age by tracking a cohort born between 1946 and 1955 in Japan over 17 waves since 2005. In this data setup, we also controlled for individual-level time-invariant attributes such as sex, birth year, educational attainment, and innate traits, thereby helping us precisely capture the response of PD to events in later life. It is reasonable to hypothesize that aging will be accompanied by poorer SRH, increased issues concerning ADL, higher risks of spousal loss, caregiving burdens, low income, job loss, and limited opportunities for social participation, all of which are expected to negatively impact PD. If this hypothesis is validated, then these factors can be interpreted as mediators of the adverse effects of aging on PD.

Second, we considered several factors (i.e., SRH, ADL problems, marital status, caregiving, low income, job status, and social participation) that were expected to affect PD and compared their relative importance in terms of their impact on PD evolution. Previous studies have focused on a single or limited number of variables in different contexts, leaving their relative importance unknown. Some studies [1, 29] have compared associations with PD across various factors, but their analyses were mainly based on cross-sectional data. We examined the age evolution of each factor and compared its impact on PD evolution across factors within the framework of mediation analysis using structural equation modeling (SEM) [30, 31].

Methods

Study samples

We used data obtained from a nationwide 17-wave panel survey, 鈥淭he Longitudinal Survey of Middle-Aged and Older Adults,鈥� conducted by the Japanese Ministry of Health, Labor, and Welfare (MHLW) each year from 2005 to 2021. Japan鈥檚 Statistics Law requires surveys to be reviewed from statistical, legal, ethical, and other perspectives. Survey data were obtained from the MHLW with official permission; therefore, ethical approval was not required.

The survey began with a cohort of people aged 50鈥�59 years (born between 1946 and 1955) in the first wave. Samples in the first wave were collected nationwide from individuals aged 50鈥�59 years in November 2005 using a two-stage random sampling procedure. First, 2,515 of the 5,280 districts, which were initially randomly selected from approximately 940,000 national census districts, were randomly selected. Second, depending on the population size of each district, 40,877 residents aged 50鈥�59 years as of October 30, 2005 were randomly selected. In total, 34,240 individuals responded to the survey (response rate: 83.8%). The second to 17th waves were conducted in early November of each year from 2006 to 2021 with no additional sampling. By the 17th wave, 18,999 individuals remained (average attrition rate in each wave: 3.6%). In the first five waves, the questionnaire was manually distributed to the participants鈥� homes, completed by the participants by early November, and manually collected thereafter. In the sixth wave and later, the questionnaire was individually mailed to the respondents, who were asked to mail it back within a week. We used unbalanced longitudinal data from 400,440 observations of 34,128 individuals in our study (16,555 men and 17,573 women).

Variables

Psychological distress

We used the Kessler 6 (K6) scores to measure PD [32, 33]. The reliability and validity of this tool have been demonstrated in Japanese samples [34]. First, we obtained the respondents鈥� assessments of PD using the six items of the K6 scale as follows: 鈥淒uring the past 30 days, about how often did you feel 1) nervous, 2) hopeless, 3) restless or fidgety, 4) so depressed that nothing could cheer you up, 5) that everything was an effort, and 6) worthless?鈥� These items are rated on a 5-point scale, from 0 (never) to 4 (all of the time). We then calculated the sum of the reported scores (range: 0鈥�24) and defined this as the K6 score. Higher K6 scores indicate higher levels of PD. The Cronbach鈥檚 alpha coefficient for this sample was 0.899. We defied PD by K6 scores鈥夆墺鈥�5, which indicated mood/anxiety disorder in a Japanese sample, as validated by previous studies [33, 35].

Potential mediators

We considered seven potential mediators of the effects of age on PD: (1) poor SRH, (2) any issues in ADL, (3) loss of spouse, (4) caregiving for family members, (5) low income, (6) no paid job, and (7) no social participation, all of which were expected to become more prevalent with age and enhance the probability of PD. Binary variables were constructed for each factor.

For SRH, the respondents were asked to rate their current health condition as follows: 1 (very good), 2 (good), 3 (somewhat good), 4 (somewhat poor), 5 (poor), and 6 (very poor). We constructed a binary variable for poor SRH by allocating 1 to those who chose 5 or 6, and 0 to others. We confirmed that the results remained largely unaffected with a wider definition of poor SRH, which included 4 in addition to 5 and 6. We also considered ADL problems based on the participants鈥� subjective assessments. We constructed a binary variable for ADL problems by allocating 1 to those who answered that they currently needed assistance in at least one of the 10 ADLs (walking, getting into and out of bed, getting into and out of a chair, putting on and taking off clothes, washing hands and face, eating, using the bathroom, taking a bath, going up and down stairs, or carrying out shopping). Regarding marital status, we constructed a binary variable for loss of spouse by allocating 1 to those who answered that they had no spouse. For the caregiving of family members, we constructed a binary variable by allocating 1 to those who answered that they were providing care to at least one family member. We constructed a binary variable for no paid jobs by allocating 1 to those who answered that they did not typically have any paid job. The definition of no paid job included retirement, unemployment, and housework. For low income, we first adjusted household spending, which was used as a proxy for household income, for household size by dividing it by the square root of the number of household members [36], and then evaluated the adjusted household spending at 2020 consumer prices. Finally, we constructed a binary variable for low income by allocating 1 to the lowest tertile of real household-adjusted household spending and 0 otherwise. The lowest tertile was 1.291听million JPY, equivalent to approximately 9,800 USD.

To measure social participation, respondents were required to indicate whether they participated in each of the following six types of social activities: (1) hobbies or entertainment, (2) sports or physical exercise, (3) community activities, (4) childcare support or educational or cultural activities, (5) support for the elderly, and (6) others (multiple answers permitted). We constructed a binary variable for social participation by allocating 1 to respondents who reported participating in at least one of the six types of social participation and 0 otherwise.

Analytic strategy

For the descriptive analysis, we examined how the proportion of PD evolved with age in three ways. First, we depicted its evolution using a pooled sample without adjustment. Second, we compared the evolution across different birth-year cohorts. Third, we compared these values across different survey years. The second and third analyses highlighted the cohort and period effects, respectively.

For regression analysis, we estimated two linear fixed-effects models, Models 1 and 2, to explain the probability of PD. The fixed-effect models can control for individual-level time-invariant attributes (such as sex, birth year, educational attainment, and innate traits), even if they are unobserved or unobservable [37, 38]. In this model setting for Models 1 and 2, all variables were mean-centered for each individual over the estimation period.

Model 1 used a set of binary variables for each age (reference age: 50 years) and survey year (reference wave: 1) to predict an individual鈥檚 PD.

$$\:PD = \alpha \: + \sum\limits_{a = 51}^{75} {{\beta _a}D{A_a}} + \sum\limits_{w = 2}^{17} {{\gamma _w}D{W_w} + e + \:\varepsilon \:}$$

Here, DA_a and DW_w indicate binary variables for age a and wave w, respectively; 伪 is an intercept, e represents individual-level fixed effects, and 蔚 is an error term. We did not assume any specific form of the age function to avoid arbitrariness in the age-PD relationship. We defined the sum of estimated coefficients on each age variable, that is, $\:\sum\:_{a=51}^{75}{\beta\:}_{a}$, as the total age effect over the ages of 50鈥�75 years. This means that $\:\sum\:_{a=51}^{75}{\beta\:}_{a}/25$ corresponds to the per-year average age effect over 25 years (from the age of 50 years).

To conduct a mediation analysis with SEM [30, 31], Model 2 consisted of (1) the main equation to explain PD by a set of binary variables for each age, potential mediator, and survey year and (2) seven auxiliary equations to explain the probabilities of each potential mediator by a set of binary variables for each age and survey year.

$$\:PD = {\alpha _0} + \sum\limits_{a = 51}^{75} {{\beta _{oa}}D{A_a}} + \sum\limits_{m = 1}^7 {{\delta _m}D{M_m}} + \sum\limits_{w = 2}^{17} {{\gamma _w}D{W_w} + {e_0} + {\varepsilon _0},}$$

$$\:D{M_1} = {\alpha _1} + \sum\limits_{a = 51}^{75} {{\beta _{1a}}D{A_a}} + \sum\limits_{w = 2}^{17} {{\gamma _{1w}}D{W_w} + {e_1} + {\varepsilon _1},}$$

$$\cdots$$

$$\:D{M_7} = {\alpha _7} + \sum\limits_{a = 51}^{75} {{\beta _{7a}}D{A_a}} + \sum\limits_{w = 2}^{17} {{\gamma _{7w}}D{W_w} + {e_7} + {\varepsilon _7},}$$

all of which are simultaneously estimated.

The total age effect on PD mediated by mediator m (m鈥�=鈥�1, 2, 鈥�, 7) over the age range of 50鈥�75 years is equal to

$$\:{\delta _m}\sum\limits_{a = 51}^{75} {{\beta _{ma}},} \:$$

and the total, total mediated, and unmediated effects of age on PD are given by

$$\:\sum\limits_{a = 51}^{75} {{\beta _{0a}}} + \sum\limits_{m = 1}^7 {\left( {{\delta _m}\sum\limits_{a = 51}^{75} {{\beta _{ma}}} } \right),} \sum\limits_{m = 1}^7 {\left( {{\delta _m}\sum\limits_{a = 51}^{75} {{\beta _{ma}}} } \right),} \:{\rm{and}}\sum\limits_{a = 51}^{75} {{\beta _{0a}},}$$

respectively; these become the per-year averages if divided by 25.

Based on the results obtained from this SEM estimation, we computed the proportion of the effect mediated by each mediator and all mediators to the total age effect, along with their 95% confidence intervals (CIs). We also computed the proportion of unmediated effects as residuals. Furthermore, we estimated Model 2 separately for men and women and compared the results. The Stata software package (Release 17) was used for all the statistical analyses.

We could not exclude the possibility that participants with serious PD would drop out of the survey, which could lead to biased estimations. Of the 34,128 individuals who entered the survey in the first wave, only 53.4%, that is, 18,208 individuals (8,380 men and 9,828 women), remained in the survey through the final (17th ) wave. However, it was technically difficult to fully control for attrition bias within the framework of the current statistical analysis. Hence, instead of directly addressing the attrition bias issue, we examined how the estimation results would have been affected if we had focused on individuals who remained until the final wave. This comparison is expected to help us speculate what potential attrition biases would look like.

Results

Figure听1 depicts how the proportion of individuals with PD evolved with age, using the pooled sample without any adjustment. After peaking in the mid-50s, the proportion diminished gradually until the mid-60s, followed by a modest rise. However, as depicted in Fig.听2, the different birth-year cohorts exhibited different age patterns. The curves were generally higher in younger cohorts, indicating higher PD levels at the same age. Figure听3 depicts that the evolution of PD also depended heavily on the survey year; the curves shifted rightward in an almost parallel manner as the survey year became more recent. The results presented in Figs.听1, 2 and 3 underscore the need to control for both cohort and period (survey year) effects to capture the evolution of PD with age.

Fig. 1

Evolution of psychological distress with age: pooled sample

Fig. 2

Evolution of psychological distress with age by birth year cohort

Fig. 3

Evolution of psychological distress with age by survey wave

Table听1; Fig.听4 compare the key estimation results between Model 1 and the main equation used to explain PD in Model 2. The results of Model 1 show that the proportion of PD almost consistently increased with age, with its lower end of 95% CI being above one over 52鈥�59 and 70鈥�75 years. In contrast, the Model 2 results revealed no association between age and PD. Meanwhile, Model 2 results indicated that all potential mediators, except for low income, were positively associated with PD. Specifically, poor SRH corresponded to a 15.7-percentage-point higher probability of developing PD. ADL problems and caregiving were closely associated with poor SRH in terms of the magnitude of their association with PD.

Table 1 Estimation results of fixed-effects regression models to explain probability of psychological distress^a

Fig. 4

Evolution of psychological distress with age: not adjusted (Model 1) versus adjusted (Model 2)

Note. Shadowed areas indicate 95% CIs

Figure听5 compares the age evolution across potential mediators based on the results of seven auxiliary equations (available upon request from the author) to explain the prevalence of each potential mediator. The probability of lack of social participation increased most remarkably with age, followed by no paid jobs, poor SRH, and ADL problems. The jumps in the probability of having no paid job at the ages of 60 and 65 years reflected mandatory retirement ages. Meanwhile, increases with age in the probability of having no spouse and having a low income were relatively limited. The probability of caregiving declined modestly with age from around the age of 60 years, after a gradual increase.

Fig. 5

Evolution of potential mediating factors with age

The evolution of the magnitude of the effect mediated by each mediator over age was determined by (1) the association of each mediator with PD, which is reported in the main equation of Model 2 (Table听2), and (2) the association of each mediator with each age in the auxiliary equations of Model 2. The estimated evolution is shown in Fig.听6. The mediating roles of poor SRH, issues affecting ADL, lack of social participation, and lack of paid jobs increased with age. Others, in which no spouse, low income, or caregiving were combined, exhibited limited mediating effects. The magnitude of the unmediated effect declined slightly between the ages of 60 and 65 years; however, this effect, which corresponds to the adjusted age effect reported in Table听1, was not associated with PD.

Table 2 Estimated proportions of age effect on psychological distress mediated by each potential mediator over ages 50鈥�75 years^a

Fig. 6

Contribution from each mediator to change in proportion of psychological distress since age 50

Table听2 reports the estimated proportions of the effect mediated by each mediator in the total age effect on PD over the ages of 50 to 75 years. Poor SRH was a key mediator which accounted for 34.2% (95% CI: 21.0鈥�47.3%) of the age effect on PD. Issues affecting ADL, no social participation, and lack of paid job consequently followed in terms of the magnitude of the mediating effect. Meanwhile, no spouse, caregiving, and low income had limited mediating effects. Specifically, it is noteworthy that caregiving had less mediating effect than lack of social activity and paid job, although the latter two had larger association with PD, as reported in Table听1. The total proportion of the effect of age on PD mediated by the seven mediators was 63.6% (95% CI: 47.4鈥�79.2%). The proportion of the age effect not mediated by any of the seven mediators was 36.4%, and its 95% CI included zero. This was consistent with the results of the adjusted age effect illustrated in the right panel of Fig.听4.

Table听2 compares the results of men and women. The total proportion of the age effect on PD mediated by the seven mediators was larger in women (73.4%; 95% CI: 51.8鈥�94.9%) than in men (44.7%; 95% CI: 25.2鈥�64.5%). Poor SRH, ADL problems, no paid jobs, and no social participation had more or similar mediating effects in women than in men.

Finally, Table听3 reports the estimation results obtained from individuals who remained in the survey until the final wave. These results were similar to those presented in Table听2. However, for men, the proportion of the age effect on PD mediated by all factors (36.5%) was lower than that for the entire sample reported in Table听2 (44.7%), while there was no substantial difference for women. These results imply that the proportion of the age effect on PD mediated by health and lifestyle factors may have been underestimated in the main analysis owing to attrition bias for men.

Table 3 Estimated proportions of age effect on psychological distress mediated by each potential mediator over ages 50鈥�75 years for respondents who participated in survey through the 17th wave^a

Discussion

We examined the evolution of PD with age using 17-wave longitudinal data of middle-aged and older adults in Japan. The key findings are summarized below along with their practical and policy implications.

First, our findings confirmed that PD increased with age, even after controlling for individual-level, time-invariant attributes and period (survey year) effects. An increase in PD with age suggests that mental health deterioration may be a key risk factor for the well-being of middle-aged and older adults, in line with the findings of previous studies [1,2,3]. We also observed substantial cohort and period effects, suggesting that caution should be exercised when interpreting the results obtained from cross-sectional or pooled datasets.

Second, the increase in PD was substantially attributable to health-related and socioeconomic/demographic factors, and age did not matter after controlling for these factors. The importance of these mediating factors implies that policy interventions to mitigate the impact of age on mediating factors can protect middle-aged or older adults from experiencing age-related mental health deterioration. It should also be noted that the proportion of the effect of age on PD mediated by these seven factors was much higher among women than among men. Notably, SRH, issues affecting ADL, and social participation had greater mediating effects in women. Although these differences between the sexes must be explored further, the results imply that women can absorb the negative impacts of aging more easily than men if they successfully manage age-related changes in health and social activities. In other words, men are more directly exposed to the negative effects of aging.

Third, unlike previous studies [2,3,4,5,6,7,8,9,10,11,12,13,14,15,16,17,18,19,20,21,22,23,24,25,26,27,28], this study revealed the relative importance of each factor linking age to PD. The key mediator was SRH followed by ADL, confirming the importance of general health and functional disability, both of which deteriorate consistently with age. SRH and ALD, if combined, account for nearly half of the impact of age on PD in those aged 50鈥�76 years. Another noteworthy finding was that social participation mediated a substantial proportion (10.5%) of the impact. The lack of contribution in social participation to the increase in PD was followed by poor SRH and ADL problems, as the prevalence of social participation declined more remarkably with age than with any other mediator. Local authorities should encourage residents to participate in community work and other social activities to enhance their psychological well-being. The results also demonstrated the importance of having a paid job, as retirement mediated the adverse impact of age on PD. Combined with the favorable impact of social participation, this observation underscores the importance of maintaining social relationships for mental health in later life. Caregiving did not have a substantial impact on the progression of PD, although its imminent shock to mental health was substantial.

This study has several limitations. First, we focused on within-individual variations within the framework of a fixed-effects analysis. We controlled for individual-level, time-invariant attributes, including unobservable attributes. However, we disregarded between-individual variations. Thus, we could not capture an overall picture of the association between PD and age or other variables.

Second, to simplify the analysis, we assumed that the association between each mediator and PD was constant over time. However, the impact on PD may vary over time, and people may gradually adapt to adverse life events after experiencing substantial shock at their onset. In this case, the estimated impact on PD may have been underestimated at the onset of the shock and overestimated over subsequent periods.

Third, we did not fully control for the attrition bias. We cannot exclude the possibility that the estimated pace of PD deterioration may have been underestimated, because participants who became physically and/or mentally unhealthier were likely to have left the survey by the last wave. In addition, comparing the results between Tables听2 and 3 implies that the proportion of the age impact on PD mediated by health and lifestyle factors may be underestimated in men.

Fourth, while we focused on seven factors as potential mediators linking age to PD, we cannot rule out the possibility that there might be other relevant mediators, such as relationships with family members other than spouses and health behaviors. However, the inclusion of these potential mediators further underscores the argument that age might not directly affect PD in the assessed population.

Despite these limitations, this study provides new insights into the evolution of PD with age and its determinants among middle-aged or older adults by controlling for individual-level, time-invariant attributes and period effects, as well as by comparing the relative importance of mediators linking age to PD.

Conclusions

This study confirmed an increase in PD with age and found that poor SRH, ADL problems, and lack of social participation were key mediators of aging with an increase in PD. These results suggest the need for policy support to encourage middle-aged and older adults to promote health and social participation to prevent depression while aging.