The Role of Leisure Activities in the Relationship Between Marital Transition in Later Midlife and Psychological Well-Being Trajectories

Abstract

This study examined the levels and rates of changes in psychological well-being for middle-aged adults of different statuses or marital transitions. The moderating effects of different leisure activities were also tested. Longitudinal data on 1,270 persons aged 50 to 65 years at baseline from the Taiwan longitudinal study on aging were analyzed. Adults who were stably unmarried or unpartnered reported worse mental health at baseline, but their psychological well-being improved over time. The trajectory of depressive symptoms fluctuated markedly in adults who became widowed during our observation period. Engagement in physical, cognitive, or social activities was significantly associated with participants’ psychological well-being. Participation in religious activities was significantly associated with life satisfaction and decreased depressive symptoms for those undergoing bereavement. Findings from this study suggest that social and physical activities, among the four selected leisure activities, have the greatest association between decreasing depressive symptoms and increasing life satisfaction, respectively. Religious activities, in particular, may improve psychological well-being in bereaved middle-aged and older adults.

Keywords

leisure activity marital transitions psychological well-being longitudinal study multilevel modeling

Marital status has been strongly correlated with various measures of psychological well-being (Mastekaasa, 1992). Mastekaasa (1992) indicated that the relationship between marital status and psychological well-being could be the joint result of the social selection and causation processes. Intuitively, compared with a single individual, married people have various readily available types of support, which helps them go through stressful situations. Social selection posits that individuals may select or be selected into marital roles based on their health (e.g., happy, cheerful people are more attractive than depressed people; Goldman, 2001; Mastekaasa, 1992). Numerous studies have demonstrated that married or cohabitating individuals report higher levels of psychological well-being than those who are unmarried or single (i.e., not cohabiting or partnered; Horwitz, White, & Howell-White, 1996; Marks, 1996; Reneflot & Mamelund, 2011; Waite, 1995). In addition, several studies have found that bereavement has a substantial, negative effect on psychological well-being, including increased depressive symptoms (Sasson & Umberson, 2014), worsening life satisfaction (Chipperfield & Havens, 2001), and poorer mental health (Wade & Pevalin, 2004).

Despite extensive research documenting the importance of marital status on psychological well-being, less attention has been directed toward the examination of long-term observation of individuals who have experienced different marital transitions. For example, most research has only focused on two time points—preloss and postloss—when assessing the effect of widowhood on well-being (Barrett, 2000; Chipperfield & Havens, 2001; LaPierre, 2009; Winter, Lawton, Casten, & Sando, 2000). According to Yan, Huang, Huang, Wu, and Qin’s (2011) meta-analysis, the risk of depression in adults who had become widowed or had remained unmarried was 1.49 times and 1.32 times higher, respectively, than those who were stably married. However, it is unclear whether the impact of bereavement or singlehood on psychological well-being is stable or diminishes over time. Thus, there is a need to assess the long-term trajectory of psychological well-being in adults who have experienced various transitions in marriage and investigate the mechanisms that may buffer against any negative impacts of such marital transitions.

The activity theory of aging, developed by the gerontologist Robert J. Havighurst in 1963, indicates that successful aging occurs when older adults stay active and maintain social interactions. Subsequent studies also demonstrated the importance of participation in various types of activities on the well-being of middle-aged and older adults (Adams, Leibbrandt, & Moon, 2011; Lemon, Bengtson, & Peterson, 1972). In general, leisure activity has been categorized into three to four categories in geriatric studies (Crowe, Andel, Pedersen, Johansson, & Gatz, 2003; Menec, 2003; Verghese et al., 2003), including productive, cognitive (intellectual), physical, and social activity, although activities often fit into more than one of the categories, and the types of leisure activity may vary from study to study. The positive effect of engaging in social, productive activities on physical function, life satisfaction, happiness, and mortality has been proven in previous empirical studies (Glass, de Leon, Marottoli, & Berkman, 1999; Menec, 2003); similarly, participation in physical activities is associated with fewer depressive symptoms and higher life satisfaction (Kim, Choe, & Chae, 2009; Lee & Park, 2008; Menec, 2003; Zimmer & Lin, 1996). Over and above physical activities, Koenig et al. (1999) demonstrated in a longitudinal study that attending religious activities appears to have an advantage in terms of prolonging life expectancy. Neill and Khan (1999), moreover, also revealed the beneficial impact of religious social activities on the life satisfaction of older widowed women. In the data set used in the current study, about 90% of the participants had participated in one type of religious activity at least once in the previous month. Given the high prevalence of religious participation, in the present study, religious activity is therefore included in our examination of how leisure activities mediate between marital transition and psychological well-being.

The Current Study

While the existing research on this topic has been valuable in highlighting the correlates of psychological well-being, no research to our knowledge has examined the role of leisure activities in the interplay between marital transition and psychological well-being. Furthermore, guided by the life course perspective, we have realized how early events, such as marriage and divorce, determine an individual’s future health trajectory (Elder Jr., Johnson, & Crosnoe, 2003; Williams & Umberson, 2004). Thus, there is an indispensable need to explore these relationships within a longitudinal context. In this study, we took advantage of longitudinal data and a multilevel modeling (MLM) approach, which takes into account multiple observations over time and controls for time-varying covariates at the same time, to (a) investigate the effects of marital transitions in later midlife on older adults’ depressive symptoms and life satisfaction trajectories during older adulthood and (b) explore whether participation in various leisure activities differentially moderate the link between marital transitions and psychological well-being. We hypothesized that marital transitions in later midlife is associated with depressive symptoms and life satisfaction trajectories during older adulthood; additionally, leisure activites moderate the associations differentially for those with different marital transition status.

Method

Participants

Data were drawn from the Taiwan longitudinal study on aging (TLSA, 1989–2007), which is a panel survey first conducted in 1989 (participants aged 60 + years), with follow-up interviews conducted in 1993, 1996, 1999, 2003, and 2007. Full details on the recruitment procedures and characteristics of TLSA participants have been described in a previous study (Chiu, Wray, & Ofstedal, 2011). To examine the long-term impact of midlife transitions in marital status among middle-aged and older adults, in the present study, the participants were restricted to being less than 65 years at baseline so it would be possible to determine how marital transitions in later midlife affect psychological well-being trajectories in older age. Data on a total of 2,220 participants aged 50 to 65 years in 1996 were included. Of the 2,220 participants, 712 were excluded based on following criteria: (a) died or withdrew at follow-ups in 2003 and 2007 (n = 540), (b) lived in a long-term care facility during the observation period (n = 3), and (c) a proxy completed one or more interviews for the primary respondent during the observation period (n = 169). (d) In addition, to ensure a group that is stably married or partnered, participants who had been asked in 1996 asked if their marital status had changed in the past 10 years and who had remarried or repartnered or had become widowed between 1986 and 1996 (n = 228) were excluded. We also excluded adults who remarried or repartnered during follow-ups (n = 5) or became separated during follow-ups (n = 5) because these groups included only a very few cases. A detailed description of our sample is illustrated in Appendix A.

Measures

Marital transitions were classified according to participants’ marital or cohabiting partnership statuses, transitions, and durations. In this study, marital transitions were categorized as: “stably married/partnered” (n = 1,034), defined as those who reported being married or partnered before 1996 and throughout the entire observation period; “stably unmarried or widowed” (n = 76), defined as those who reported being unmarried or widowed before 1996 and throughout the entire observation period, and those who “became widowed during follow-ups” (n = 160), defined as those who reported married or partnered before 1996 and were going through bereavement during the observation period.

Psychological well-being was measured with two outcomes: depressive symptoms and life satisfaction. For both major outcomes, we used the responses of each participant in 1996, 1999, 2003, and 2007 to construct the trajectory of psychological well-being for each person. The depressive symptoms score is assessed by the commonly used 10-item Center for Epidemiologic Studies Depression Scale (Kohout, Berkman, Evans, & Cornoni-Huntley, 1993), with each symptom ranging from 0 to 3 points. Total scores of Center for Epidemiologic Studies Depression Scale ranged from 0 to 30, and Cronbach’s α were between 0.82 and 0.85. Life satisfaction was assessed using a 10-item scale, derived from the Life Satisfaction Index–Version A Scale (Liang, 1984). Each item was rated as yes or no (scored 1 or 0), with a total score ranging from 0 to 19. Kuder-Richardson reliability for the life satisfaction scale ranged from 0.71 to 0.74 over the four waves.

Leisure activities, as a time-varying variable, was defined according to previous research (Cheung, Ting, Chan, Ho, & Chan, 2009; Paillard-Borg, Wang, Winblad, & Fratiglioni, 2009; Zimmer & Lin, 1996). The TLSA assessed respondents’ frequency of participation in 10 to 12 leisure activities depending on the year of investigation. The frequency of participation was measured by the following question:
During the past 30 days, how many times did you participate in the following leisure activities, including gardening, walking, jogging, playing ball and mountain-climbing, reading novels, newspapers or books, playing chess or mahjong, going to church, reading the Bible, listening to or watching religious programs or worshipping, playing an instrument, painting, collecting postage stamps, going to concerts or performances, socializing, or joining a club?
and were coded with a 0 to 4 scale as: 0 = never, 1 = once a month or less, 2 = 2 or 3 times a month, 3 = 1 or 2 times a week, and 4 = almost every day. Participants were asked to describe their frequency of participation in each activity at each wave. To consistently define the participation in leisure activity, we used 10 leisure activities, which were assessed in all surveys, as the condition of leisure activity engagement in the present study. In addition, we grouped these activities into four types: physical (i.e., gardening, walking, jogging, playing ball, and mountain-climbing), cognitive (i.e., reading novels, newspapers or books, playing chess, or mahjong), religious (i.e., going to church, reading the Bible, listening to or watching religious programs, or worshipping), and social activities (i.e., socializing or joining a club), and recoded each type as a binary variable (i.e., never engaging in the activity was recoded as nonparticipation). For example, a respondent who had participated in one of the physical activities at least once a month would be classified as participating in a physical activity.

Covariates included “demographic factors” such as gender, age, years of education, job (yes or no), economic satisfaction (5-point Likert scale, extremely satisfied – not at all satisfied), and self-assessed health status (5-point Likert scale, very good – poor). Economic satisfaction was assessed by asking the participants “In general, are you satisfied with your current economic status?” The participants were also asked to estimate their general health status between five alternatives: very good, good, fair, not good, or poor. For both economic satisfaction and health status, a higher score indicated a more positive condition. Gender and years of education were measured at baseline; other variables were introduced as time-varying covariates.

Statistical Analysis

We used MLM to depict the rate of change in depressive symptoms and life satisfaction across different marital statuses over time, where time = 0 refers to the baseline (i.e., 1996) for the “stably married/partnered” and “stably unmarried/widowed”group; however, for those who “became widowed during follow-ups,” time = 0 refers to the time at which the death of the individual’s spouse or partner was reported. That is, the beginning observed wave of widowed respondents varies depends on the timing of the spouse’s death. For example, if the spouse died in 1997, the survey data in 1996 (Wave 1) were used as the first observation of the widowed respondent, and the time would be coded as “–2”; likewise, the time = 2 (observation in 1999, Wave 2), time = 6 (observation in 2003, Wave 3), and time = 10 (observation in 2007, Wave 4) were used to express the length of postwidowhood, respectively. To simplify the research design, the beginning wave of widowed respondents was determined by the nearest survey point prior to the death of spouse; namely, if the spouse died in 2000 (1 year after Wave 2), the observation in 1999 would be used as the first observation of the widowed respondent (time = −1).

We also employed a hierarchical design to explore whether participation in leisure activities moderated the link between marital status and mental health. Hence, the following models were conducted: In Model 0, we merely evaluated the time effect on depressive symptoms and life satisfaction. To visualize the variations in psychological well-being over time, we also depicted the trajectories of depressive symptoms and life satisfaction across 10 years for three marital transition groups based on the original score of each outcome. In Model 1, we added sociodemographic and self-rated health covariates to examine how these factors affected the depressive symptoms and life satisfaction trajectories. In Model 2, we assessed the impact of marital transition on depressive symptoms and life satisfaction trajectories while accounting for the aforementioned confounds. In Model 3, we tested the moderating effects of four categories of leisure activities by including the interaction term marital transition X leisure activity to test if the impact of each leisure activity would be similar across adults with different marital statuses. We used SAS 9.3 to perform all analyses, with a significance level set at α = 0.05.

Results

Participant Characteristics

Table 1 presents the demographic characteristics of our participants. In this study, 44.2% of those who were “stably married/partnered,” (hereinafter referred to as “stably married”) were female, while more than half of “Stably unmarried or widowed” (hereinafter referred to as “stably unmarried”) and “became widowed during follow-ups” (hereinafter referred to as “widowed”) were female, 61.84% and 77.5%, respectively. Widowed adults and those who were stably unmarried were disproportionately those who had received no formal education (43.75% and 31.58%, respectively). Adults who were “stably married” (M_age = 57.26) tended to be younger than those in other marital transition categories (mean age: 59.33 for unmarried; 62.59 for widowed, p < .001). Furthermore, the age of those who became widowed during our observation period and the age of their partners at the time of death were between age 62.59 and 66.59 years (not shown in table). In comparison with the working status across the three types of marital transition, widowed individuals had the lowest frequency of working for pay (29.11%). Among the three study groups, they have a similar distribution of health status and economic satisfaction with respect to the median.
Table 1.
Demographic and Leisure Activity Characteristics of Middle-Aged and Older Taiwanese Adults at Baseline, by Marital Transition Category.^a,b

Entire sample Stably married or partnered Stably unmarried or widowed Became widowed during follow-ups p ^c

(N = 1,270) (n = 1,034) (n = 76) (n = 160)

Gender, n (%) <.001

Male 642(50.6) 577(55.8) 29(38.2) 36(22.5)

Female 628(49.4) 457(44.2) 47(61.8) 124(77.5)

Mean age (years) 58.06(4.5) 57.26(4.20) 59.33(4.12) 62.59(4.47) <.001

Education, n (%) <.001

No formal education 362(28.5) 268(25.9) 24(31.6) 70(43.8)

Grade 1–6 613(48.3) 500(48.4) 37(48.7) 76(47.5)

Grade 7+ 295(23.2) 266(25.7) 15(19.7) 14(8.7)

Job, n (%) 685(54.02) 600(58.03) 39(51.32) 46(29.11) <.001

Economic satisfaction 3(0.88) 3(0.86) 3(0.93) 3(0.95) <.001**

Self-assessed health status 3(1.07) 3(1.08) 3(1.00) 3(1.01) .02*

Leisure activities, n (%)

Physical activities 962(75.8) 799(77.3) 64(84.2) 99(61.9) .8664

Cognitive activities 518(40.8) 446(43.1) 25(32.9) 47(29.4) <.001

Social activities 948(74.7) 786(76.0) 56(73.7) 106(66.3) .8824

Religious activities 1,104(86.9) 895(86.6) 64(84.2) 145(90.6) <.001

Depressive symptoms 4.53(5.12) 4.27(5.01) 6.55(6.25) 5.21(5.00) <.001

Life satisfaction 6.73(2.36) 6.87(2.31) 5.71(2.52) 6.30(2.45) <.001

a
Data are expressed as means (SD), median (SD), or n (%). The continuous variables (age, depressive symptom, and life satisfaction) are expressed as means (SD). The ordinal categorical variables (economic satisfaction and health status) are expressed as median (SD) on a 5-level scale (higher score indicates better health or greater satisfaction). The remaining variables are expressed as n (%).

b
For “stably married/partnered” and “stably unmarried/widowed,” the baseline information represents the data collected at Wave 1 (1996); for “became widowed during follow-ups,” we used the interview wave before the partner died as their baseline information. (e.g., if the partner was alive between 1996 and 1999 and died between 1999 and 2003, then we used the data in 1999 as their baseline information).

c
Based on ANOVA tests for continuous variables and a Wilcoxon-Mann Whitney test for ordinal categorical variables, and chi-square tests for all other nominal categorical variables among stably married or partnered, stably unmarried/widowed, and became widowed groups at baseline.

*
p < .05. p < .01.

In addition, in our sample of participants, over 50% had engaged in more than three types of activities. There was no difference found in terms of participation in physical activities or social activities across the three groups. At least 60% of all adults engaged in social and physical activities. Approximately 90% of the widowed adults, however, engaged in at least one type of religious activity in the past month at baseline, compared with other groups. Less than half of participants engaged in cognitive activities in the present study, especially widowed older adults (30%). In terms of psychological well-being at baseline, stably married participants had the lowest depression score (M_depression = 4.27) and highest life satisfaction (M_{Life Satisfaction} = 6.87), followed by widowed people (M_depression = 5.21; M_{Life Satisfaction} = 6.30), and stably unmarried participants had the lowest score in psychological well-being (M_depression = 6.55; M_{Life Satisfaction} = 5.71).

Effect of Marital Transitions on Psychological Well-Being

MLM analyses tested the predictors of depressive symptoms (Table 2) and life satisfaction (Table 3). We first examined the overall trajectory of depressive symptoms and life satisfaction on polynomial (i.e., linear, quadratic, and cubic) functions of time. We considered linear and quadratic models for the depressive symptoms and a linear slope for the life satisfaction model. As presented in Models 1 and 2 (both in Tables 2 and 3), the marital transition category was significantly associated with the model’s goodness of fit, Δχ²(6) = 19.1 (p < .01) and Δχ²(6) = 20.8 (p < .01), respectively, thereby suggesting that marital transitions had a significant effect on psychological well-being.
Table 2.
The Effect of Marital Transitions on Trajectories of Depressive Symptoms (N = 1,270).

Model 0^a Model 1^b Model 2^c Model 3^d

Fixed effect

Intercept 4.50* 9.62* 9.53* 10.68*

Time −0.12 −0.24* −0.25* −0.22*

Time 2 0.01* 0.02* 0.02* 0.02***

Age 0.05* 0.04 0.04

Male −0.91* −0.87* −0.74*

Education: No formal education 1.09* 1.09* 0.74

Education: Grades 1–6 0.09 0.09 −0.06

Work −0.31 −0.33* −0.41

Satisfied with financial situation −2.99* −2.98* −2.88*

Better self-assessment health −3.11* −3.10* −3.04*

Stably unmarried 1.63* 2.04*

Widowed 0.03 1.66

Stably unmarried × time −0.23* −0.21

Widowed × time 0.08 0.07

Physical activities −0.48**

Cognitive activities −0.39*

Social activities −0.75*

Religious activities −0.19

Stably unmarried × physical activities −0.45

Widowed × physical activities 0.63

Stably unmarried × cognitive activities 0.96

Widowed × cognitive activities −0.69

Stably unmarried × social activities −0.36

Widowed × social activities 0.44

Stably unmarried × religious activities −0.40

Widowed × religious activities −2.29

Random effect

Variation

Intercept (11) 10.91* 5.80* 5.61* 5.13*

Slope (22) 0.21 0.14 0.11 0.06

Curvature (33) 0.001 0.001 0.001 0.0004

Covariance

Intercept, slope (12) −0.22 −0.24 −0.19 −0.08

Intercept, curvature (13) 0.01 0.01 0.01 0.001

Slope, curvature (23) −0.01 −0.01 −0.01 0.003

Residual 14.37* 13.87* 13.93* 13.94*

Model fits

AIC 28648.6 27759.0 27747.9 27703.0

BIC 28700.0 27846.5 27856.0 27872.8

−2LL 28628.6 27725.0 27705.9 27637.0

Parameters 3 15 21 53

LR test

$Δ x^{2}$ (degree of freedom) 903.4(12)* 19.1(6) 68.9(32)***

Note. Stably unmarried = stably unmarried or widowed; Widowed = became widowed during follow-ups; − 2LL = −2 log likelihood; LR test = likelihood ratio test; AIC = Akaike information criterion; BIC = Bayesian information criterion.

a
Model 0 tested time effects only.

b
Model 1 tested the effects of time and the control variables.

c
Model 2 tested the effects of time, marital status, and the control variables.

d
Model 3 tested the effects of time, marital status, control variables, and leisure activity participation. We also tested the moderating effects of leisure activity participation on the link between marital status and our outcome variables.

*
p < .05. p < .01. p < .001.

Table 3.
The Effect of Marital Transitions on Trajectories of Life Satisfaction (N = 1,270).

Model 0^a Model 1^b Model 2^c Model 3^d

Fixed effect

Intercept 6.74 4.75* 4.87* 3.97*

Time 0.003 0.04* 0.02 0.01

Age −0.01 −0.005 −0.004

Male −0.06 −0.11 −0.22

Education: No formal education −0.65* −0.66* −0.27*

Education: Grades 1–6 −0.26 −0.26 −0.06

Work −0.02 −0.02 0.07

Satisfied with financial situation 1.96* 1.94* 1.87*

Better self-assessment health 1.00* 1.01* 0.97*

Stably unmarried −0.89* −1.04

Widowed −0.34* −1.37***

Stably unmarried × time 0.07* 0.06*

Widowed × time 0.005 −0.005

Physical activities 0.54*

Cognitive activities 0.38*

Social activities 0.35***

Religious activities 0.02

Stably unmarried × physical activities −0.06

Widowed × physical activities −0.004

Stably unmarried × cognitive activities −0.31

Widowed × cognitive activities 0.30

Stably unmarried × social activities 0.24

Widowed × social activities 0.14

Stably unmarried × religious activities 0.22

Widowed × religious activities 0.91*

Random effect

Intercept (11) 1.91* 0.76* 0.72* 0.63*

Slope (22) 0.01 0.002 0.002 0.001

Intercept, slope (12) −0.01* −0.01 −0.01 −0.003

Residual 3.75* 3.65* 3.64* 3.60*

Model fits

AIC 21459.3 20552.8 20540.0 20413.3

BIC 21490.2 20619.7 20627.5 20562.6

−2LL 21447.3 20526.8 20506.0 20355.3

Parameters 2 14 20 52

LR test

$Δ x^{2}$ (degree of freedom) – 920.5(12)* 20.8(6) 150.7(32)***

Note. Stably unmarried = stably unmarried or widowed; Widowed = became widowed during follow-ups;−2LL = −2 log likelihood; LR test = likelihood ratio test; AIC = Akaike information criterion; BIC = Bayesian information criterion.

a
Model 0 tested time effects only.

b
Model 1 tested the effects of time and the control variables.

c
Model 2 tested the effects of time, marital status, and the control variables.

d
Model 3 tested the effects of time, marital status, control variables, and leisure activity participation. We also tested the moderating effects of leisure activity participation on the link between marital status and our outcome variables.

*
p < .05. p < .01. p < .001.

As shown in Table 2, the depressive symptoms score for adults who became widowed* was higher than for those who were stably married. However, the difference was not significant (β_Widowed = 1.66, p = .059). Moreover, the linear trend of depressive symptoms for widowed adults was not clear, which suggested that their depressive symptoms might have varied during the study period. To visualize these findings, we plotted the average depressive symptoms in terms of the original score for depressive symptoms for each category of marital transitions. As presented in Figure 1, correspondingly, the trajectories depicting the change in the raw scores show that the widowed individuals had remarkably fluctuating scores throughout the study period. Regarding life satisfaction, the results in Table 3 indicate that individuals who became widowed had the lowest life satisfaction as compared with those in the other groups (β_Widowed = −1.37, p < .001; see Table 3). In the case of this group, it can be observed that their trajectory of life satisfaction appeared to decline before their partners passed away, but gradually increased after the death of their spouse (see Figure 2).
Figure 1.
Depressive symptom trajectories by marital transition type.

Figure 2.
Life satisfaction trajectories by marital transition type.

For those who were “stably unmarried,” the depressive symptoms score was significantly higher than it was in the case of those who were “stably married” (β_{Stably unmarried} = 1.63, p < .001), but the score significantly decreased over time (β_{Stably unmarried×time} =−0.23, p < .001). Even after controlling for leisure activities, these results remained unchanged (see Model 3, Table 2). Likewise, the trajectory of raw depressive symptom scores gradually decreased during the observation period and became lower than those who were “stably married” after 11 years of follow-up (see Figure 1). In addition, Model 2 in Table 3 indicates that life satisfaction in “stably unmarried” adults was lower than in those who were “stably married” at baseline. However, the differences in life satisfaction between these two groups narrowed over the observation period, β_{Stably unmarried} = −0.89 (p < .001) and β_{Stably unmarried − time} =0.07 (p < .05), respectively, consistent with the picture presented in Figure 2.

Effect of Participation in Leisure Activities on Psychological Well-Being

According to Models 2 and 3 (see Tables 2 and 3), the models’ goodness of fit both significantly increased, Δχ² (32) = 68.9, p < .001; Δχ² (32) = 150.7, p < .001, with the inclusion of leisure activities, thereby indicating that these activities can predict depressive symptoms and life satisfaction.

As presented in Table 2, except for religious activities, all leisure activities—and particularly social activities— significantly decreased depressive symptoms. The depressive symptoms score of those who participated in social activities was lower by 0.75 points as compared with those who did not participate in such activities. Compared with adults who did not engage in sports activities or cognitive activities, adults who engaged in one or both of these activities reported lower depressive symptoms, β_{Physical activities} = −0.48 (p < .01) and β_{Cognitive activities} = −0.39 (p < .05), respectively. This study also found that depressive symptoms in adults who became widowed decreased if they participated in religious activities, β_{Widowed − religious activities} = −2.29, p < .01; however, that effect was not observed in other marital transition categories. For example, although the depressive symptoms score of those who lost spouses within our observation period was higher than the “stably married” adults, β_Widowed = 1.66, p = .059, the average score of those who became widowed decreased by 0.63 points if they participated in religious activities.

Table 3 shows that participation in cognitive, social, and particularly, physical leisure activities, increased life satisfaction. In addition, the significant interaction between marital status and leisure activities indicated that the life satisfaction of those who lost spouses within our observation period was increased if they participated in religious activities, β_{Widowed × religious activities} = 0.91, p < .05. In other words, if adults who became widowed engaged in religious activities, the difference in life satisfaction between widowed and stably married narrowed considerably (to only 0.46 points).

Discussion

Few past studies have explored the long-term effects of marital transitions on psychological well-being. The handful of studies that have done so mainly focused on the mental health of those who became widowed at only one or two time points after their partners had died (Yan et al., 2011). Our study explored the trajectory of changes in depressive symptoms and life satisfaction over four-time points—a span of 11 years—across different marital transitions in middle-aged and older Taiwanese adults. In addition, during follow-ups, we examined whether engagement in leisure activities promoted the psychological well-being of those who became widowed, as well as of those who were stably unmarried or widowed before baseline.

Our results indicated that depressive symptoms trajectories were significantly different between those who were stably married or partnered and those who were stably unmarried or unpartnered. Despite the disadvantage of being single or unmarried (Horwitz et al., 1996; Marks, 1996; Reneflot & Mamelund, 2011; Waite, 1995), the trajectory and estimated time effect in the present study showed that the depressive symptoms for stably unmarried adults decreased over time. Adults who were unmarried or unpartnered reported higher depressive symptoms at baseline, but it is possible that they become used to the single life and are content with being single in later life. In terms of the widowed adults and depressive symptoms, although the difference in depressive symptoms between those who lost spouses and those who were stably married or partnered was not significant, it was marginally significant. These findings were partially consistent with previous research suggesting that those who were widowed experienced higher depressive symptoms and, in turn, higher mortality risk (Manzoli, Villari, Pirone, & Boccia, 2007). Interestingly, our results revealed that the trajectory of depressive symptoms in those who became widowed fluctuated markedly from 1 year before to 3 years after their spouses died. These fluctuations may be the result of heightened worry about their partners’ physical health before their spouses passed away that dissipated over time following their spouses’ deaths. Past studies also have indicated that depression is higher within 6 months after the death of partners and decreases over time (Winter et al., 2000).

In terms of marital transitions and life satisfaction, this study found that the life satisfaction of those who lost spouses within our observation period and those who were stably unmarried or unpartnered was lower than was the case for stably married or partnered adults. This finding is consistent with the work of Chipperfield and Havens (2001) who found that the life satisfaction of recently widowed individuals decreased significantly, regardless of gender. However, the impact of widowhood on men’s mental health was stronger than it was for women in their study. In addition, acquisition of a new partner improved life satisfaction in men but not in women. In contrast, our study did not find gender differences in the effects of marital transitions on life satisfaction. One explanation for the relatively attenuating life satisfaction in men is that men typically experience widowhood at a later age than women, thereby they have relatively less time to adapt to their loss (Barer, 1994; Chipperfield & Havens, 2001). Because the present study followed widowed participants’ life satisfaction before and after they had lost their spouses, our study accounted for the time during which they became widowed (the effect of recency of loss). This may be the possible reason for the lack of gender differences in the effects of widowhood in the present study. Although the life satisfaction of adults who lost spouses did not significantly increase after their partners died, we did observe that the difference in life satisfaction between those who lost partners and adults in other marital groups was markedly decreased. This finding is consistent with past studies showing no differences in life satisfaction between married and widowed adults while controlling for prior levels of life satisfaction (Bennett, 2005). Interestingly, our findings also showed that life satisfaction in stably unmarried or unpartnered adults gradually increased over time, implying that they adapted positively to single life. The potential explanations for the gradual improvement in psychological well-being in stable unmarried or unpartnered adults are worth examining in future research. Also, whether or not the different types of singlehood, such as divorced, widowed, or never-married, further shape the impact should be considered in future studies.

Engagement in various forms of leisure activities is a very important factor in successful aging (Rowe & Kahn, 1997). Many studies have confirmed that participation in leisure activities significantly benefits physical and psychological health (Glass et al., 1999; Menec, 2003). However, the relationship between the frequency of participating in activities and psychological well-being has not been consistent (Mulsant, Ganguli, & Seaberg, 1997). In general, frequent participation in leisure activities often results in better physical health, greater subjective happiness, and lower likelihood of depressive symptoms (Everard, Lach, Fisher, & Baum, 2000; Lu, 2011; Mendes de Leon, Glass, & Berkman, 2003; Menec, 2003; Tessier et al., 2007; Zimmer & Lin, 1996). Our results showed that engagement in social activities was most strongly linked to lower depressive symptoms. This might be because social activity provides both increased tangible resources and intangible resources (e.g., a sense of belonging and purpose; Carstensen, 1991). In contrast, life satisfaction was the highest in the adults who were involved in physical activities. One plausible explanation for this finding could be that physical activities provide a challenge and sense of accomplishment. Finally, although participation in religious activities did not significantly affect psychological well-being in stably married or partnered or stably unmarried adults, such participation did lessen the impact of the death of a spouse, reversing their increased depressive symptoms and improving their life satisfaction. Past studies also indicated that religious activities enabled widows to grow spiritually, understand the meaning of life, and develop new interpretations of death (Neill & Kahn, 1999).

The strengths of our study include long-term follow-ups, which allowed for the observations of the psychological well-being of the participants both before and after a marital transition. This makes it possible to ascertain both the lowest psychological status where interventions for susceptible groups can be targeted and the recovery timing in adults who have lost spouses. In addition, testing for moderating effects of four types of leisure activities on the relationship between marital transition and psychological well-being at the same time may provide intervention planners with useful information about what types of activities might be the most beneficial to older adults who are experiencing a marital transition. However, there were several limitations in this study. First, although leisure activities were classified according to previous studies, whether the same types of activities would have the same meaning across cultures is unknown. Second, we combined those who reported being unmarried from 1986 to 2007 (N = 27) or widowed before 1986 and throughout the entire observation period (N = 49) into the “stably unmarried or widowed,” due to the small sample size. After examining the heterogeneity between these two groups at baseline, we found those who had lost a spouse had lower rates of educational attainment (over seventh grade), a lower participation rate in cognitive activity, a higher participation rate in religious activity, and higher life satisfaction as compared with those who were signal, suggesting that our result estimating the effect of being stably unmarried on life satisfaction might have been underestimated. Third, because we tracked changes in marital status, we required our study participants to have completed all four surveys. Thus, the health status of our study sample is likely to be biased in favor of healthier respondents, compared with the full sample. However, our sensitivity analysis (data available upon request) comparing participants who were not survived till the last follow-up with those who were followed to the last observation showed no significant differences in the models, suggesting that this may not be an issue.

In the light of these strengths and limitations, it was found that those who remained stably unmarried or unpartnered improved their psychological well-being over time, both decreasing depressive symptoms and increasing life satisfaction, as compared with adults who remained married or partnered. Findings from this study showed that religious participation lessens the negative impact of widowhood on psychological well-being, suggesting the importance of support systems in helping widowed adults. In particular, because the trajectory of depressive symptoms in those who lost their partners fluctuated markedly for at least 3 years after their partners deaths, policy and practice planning should take into account the long-term effects of widowhood.

Appendix A. Illustration of Sample Selection

	Entire sample	Stably married or partnered	Stably unmarried or widowed	Became widowed during follow-ups	p ^c
Gender, n (%)					<.001**
Male	642(50.6)	577(55.8)	29(38.2)	36(22.5)
Female	628(49.4)	457(44.2)	47(61.8)	124(77.5)
Mean age (years)	58.06(4.5)	57.26(4.20)	59.33(4.12)	62.59(4.47)	<.001**
Education, n (%)					<.001**
No formal education	362(28.5)	268(25.9)	24(31.6)	70(43.8)
Grade 1–6	613(48.3)	500(48.4)	37(48.7)	76(47.5)
Grade 7+	295(23.2)	266(25.7)	15(19.7)	14(8.7)
Job, n (%)	685(54.02)	600(58.03)	39(51.32)	46(29.11)	<.001**
Economic satisfaction	3(0.88)	3(0.86)	3(0.93)	3(0.95)	<.001**
Self-assessed health status	3(1.07)	3(1.08)	3(1.00)	3(1.01)	.02*
Leisure activities, n (%)
Physical activities	962(75.8)	799(77.3)	64(84.2)	99(61.9)	.8664
Cognitive activities	518(40.8)	446(43.1)	25(32.9)	47(29.4)	<.001**
Social activities	948(74.7)	786(76.0)	56(73.7)	106(66.3)	.8824
Religious activities	1,104(86.9)	895(86.6)	64(84.2)	145(90.6)	<.001**
Depressive symptoms	4.53(5.12)	4.27(5.01)	6.55(6.25)	5.21(5.00)	<.001**
Life satisfaction	6.73(2.36)	6.87(2.31)	5.71(2.52)	6.30(2.45)	<.001**

	Model 0^a	Model 1^b	Model 2^c	Model 3^d
Fixed effect
Intercept	4.50***	9.62***	9.53***	10.68***
Time	−0.12**	−0.24***	−0.25***	−0.22***
Time 2	0.01***	0.02***	0.02***	0.02***
Age		0.05*	0.04	0.04
Male		−0.91***	−0.87***	−0.74***
Education: No formal education		1.09***	1.09***	0.74**
Education: Grades 1–6		0.09	0.09	−0.06
Work		−0.31	−0.33*	−0.41**
Satisfied with financial situation		−2.99***	−2.98***	−2.88***
Better self-assessment health		−3.11***	−3.10***	−3.04***
Stably unmarried			1.63***	2.04*
Widowed			0.03	1.66
Stably unmarried × time			−0.23***	−0.21**
Widowed × time			0.08	0.07
Physical activities				−0.48**
Cognitive activities				−0.39*
Social activities				−0.75***
Religious activities				−0.19
Stably unmarried × physical activities	−0.45
Widowed × physical activities	0.63
Stably unmarried × cognitive activities	0.96
Widowed × cognitive activities				−0.69
Stably unmarried × social activities				−0.36
Widowed × social activities				0.44
Stably unmarried × religious activities	−0.40
Widowed × religious activities				−2.29**
Random effect
Variation
Intercept (11)	10.91***	5.80***	5.61***	5.13***
Slope (22)	0.21	0.14	0.11	0.06
Curvature (33)	0.001	0.001	0.001	0.0004
Covariance
Intercept, slope (12)	−0.22	−0.24	−0.19	−0.08
Intercept, curvature (13)	0.01	0.01	0.01	0.001
Slope, curvature (23)	−0.01	−0.01	−0.01	0.003
Residual	14.37***	13.87***	13.93***	13.94***
Model fits
AIC	28648.6	27759.0	27747.9	27703.0
BIC	28700.0	27846.5	27856.0	27872.8
−2LL	28628.6	27725.0	27705.9	27637.0
Parameters	3	15	21	53
LR test
$Δ x^{2}$ (degree of freedom)		903.4(12)***	19.1(6)**	68.9(32)***

	Model 0^a	Model 1^b	Model 2^c	Model 3^d
Fixed effect
Intercept	6.74***	4.75***	4.87***	3.97***
Time	0.003	0.04***	0.02	0.01
Age		−0.01	−0.005	−0.004
Male		−0.06	−0.11	−0.22**
Education: No formal education		−0.65***	−0.66***	−0.27*
Education: Grades 1–6		−0.26**	−0.26**	−0.06
Work		−0.02	−0.02	0.07
Satisfied with financial situation		1.96***	1.94***	1.87***
Better self-assessment health		1.00***	1.01***	0.97***
Stably unmarried			−0.89***	−1.04**
Widowed			−0.34*	−1.37***
Stably unmarried × time			0.07*	0.06*
Widowed × time			0.005	−0.005
Physical activities				0.54***
Cognitive activities				0.38***
Social activities				0.35***
Religious activities				0.02
Stably unmarried × physical activities	−0.06
Widowed × physical activities				−0.004
Stably unmarried × cognitive activities	−0.31
Widowed × cognitive activities				0.30
Stably unmarried × social activities				0.24
Widowed × social activities				0.14
Stably unmarried × religious activities	0.22
Widowed × religious activities				0.91*
Random effect
Intercept (11)	1.91***	0.76***	0.72***	0.63***
Slope (22)	0.01	0.002	0.002	0.001
Intercept, slope (12)	−0.01*	−0.01	−0.01	−0.003
Residual	3.75***	3.65***	3.64***	3.60***
Model fits
AIC	21459.3	20552.8	20540.0	20413.3
BIC	21490.2	20619.7	20627.5	20562.6
−2LL	21447.3	20526.8	20506.0	20355.3
Parameters	2	14	20	52
LR test
$Δ x^{2}$ (degree of freedom)	–	920.5(12)***	20.8(6)**	150.7(32)***

Footnotes

Acknowledgments

The authors thank the Department of Health Promotion at the Ministry of Health and Welfare in Taiwan for permission to analyze the data from the Taiwan Longitudinal Study on Aging (TLSA).

Author Note

An earlier version of this paper was presented at the Annual Conference of Health and Society Association of Taiwan and Symposium of “Aging, Health and Society” in 2013.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by Research Center for Humanities and Social Sciences at National Cheng Kung University (C.-J. C., grant FD101037).

Author Biographies

Yi-Han Hu is a doctoral student in School of Public Health at Indiana University Bloomington, IN.

Ching-Ju Chiu is an associate professor of Gerontology at the National Cheng Kung University in Tainan, Taiwan.

Jen D. Wong is an assistant professor in the Department of Human Sciences at The Ohio State University.

Dai-Chan Lin is an assistant researcher of Gerontology at the National Cheng Kung University in Tainan, Taiwan.

Linda A. Wray is an associate professor of Biobehavioral Health at the Pennsylvania State University.

References

Adams

K. B.

Leibbrandt

Moon

(2011) A critical review of the literature on social and leisure activity and wellbeing in later life. Ageing and Society 31(04): 683–712.

Barer

B. M.

(1994) Men and women aging differently. The International Journal of Aging and Human Development 38(1): 29–40.

Barrett

A. E.

(2000) Marital trajectories and mental health. Journal of Health and Social Behavior 41(4): 451–464.

Bennett

K. M.

(2005) Psychological wellbeing in later life: The longitudinal effects of marriage, widowhood and marital status change. International Journal of Geriatric Psychiatry 20(3): 280–284.

Carstensen

L. L.

(1991) Selectivity theory: Social activity in life-span context. Annual Review of Gerontology and Geriatrics 11: 195–217.

Cheung

Ting

Chan

(2009) Leisure participation and health related quality of life of community dwelling elders in Hong Kong. Asian Journal of Gerontology and Geriatrics 4: 15–23.

Chipperfield

J. G.

Havens

(2001) Gender differences in the relationship between marital status transitions and life satisfaction in later life. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 56(3): 176–186.

Chiu

C.-J.

Wray

L. A.

Ofstedal

M. B.

(2011) Diabetes-related change in physical disability from midlife to older adulthood: Evidence from 1996–2003 survey of health and living status of the elderly in Taiwan. Diabetes Research and Clinical Practice 91(3): 413–423.

Crowe

Andel

Pedersen

N. L.

Johansson

Gatz

(2003) Does participation in leisure activities lead to reduced risk of Alzheimer's disease? A prospective study of Swedish twins. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 58(5): P249–P255.

10.

Elder Jr., G. H., Johnson, M. K., & Crosnoe, R. (2003). The emergence and development of life course theory. Handbook of the life course (pp. 3–19). New York, NY: Springer.

11.

Everard

K. M.

Lach

H. W.

Fisher

E. B.

Baum

M. C.

(2000) Relationship of activity and social support to the functional health of older adults. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 55(4): 208–212.

12.

Glass

T. A.

de Leon

C. M.

Marottoli

R. A.

Berkman

L. F.

(1999) Population based study of social and productive activities as predictors of survival among elderly Americans. British Medical Journal 319(7208): 478–483.

13.

Goldman

(2001) Social inequalities in health. Annals of the New York Academy of Sciences 954(1): 118–139.

14.

Havighurst

R. J.

(1963) Successful aging. Processes of Aging: Social and Psychological Perspectives 1: 299–320.

15.

Horwitz

A. V.

White

H. R.

Howell-White

(1996) Becoming married and mental health: A longitudinal study of a cohort of young adults. Journal of Marriage and Family 58: 895–907.

16.

Kim

J. I.

Choe

M. A.

Chae

Y. R.

(2009) Prevalence and predictors of geriatric depression in community-dwelling elderly. Asian Nursing Research 3(3): 121–129.

17.

Koenig

H. G.

Hays

J. C.

Larson

D. B.

George

L. K.

Cohen

H. J.

McCullough

M. E.

Blazer

D. G.

(1999) Does religious attendance prolong survival? A six-year follow-up study of 3,968 older adults. The Journals of Gerontology Series A: Biological Sciences and Medical Sciences 54(7): M370–M376.

18.

Kohout

F. J.

Berkman

L. F.

Evans

D. A.

Cornoni-Huntley

(1993) Two shorter forms of the CES-D depression symptoms index. Journal of Aging and Health 5(2): 179–193.

19.

LaPierre

T. A.

(2009) Marital status and depressive symptoms over time: Age and gender variations. Family Relations 58(4): 404–416.

20.

Lee

Park

(2008) Does physical activity moderate the association between depressive symptoms and disability in older adults? International Journal of Geriatric Psychiatry 23(3): 249–256.

21.

Lemon

B. W.

Bengtson

V. L.

Peterson

J. A.

(1972) An exploration of the activity theory of aging: Activity types and life satisfaction among in-movers to a retirement community. Journal of Gerontology 27(4): 511–523.

22.

Liang

(1984) Dimensions of the life satisfaction index A: A structural formulation. Journal of Gerontology 39(5): 613–622.

23.

(2011) Leisure experiences and depressive symptoms among Chinese older people: A national survey in Taiwan. Educational Gerontology 37(9): 753–771.

24.

Manzoli

Villari

Pirone

G. M.

Boccia

(2007) Marital status and mortality in the elderly: A systematic review and meta-analysis. Social Science & Medicine 64(1): 77–94.

25.

Marks

N. F.

(1996) Flying solo at midlife: Gender, marital status, and psychological well-being. Journal of Marriage and Family 58(4): 917–932.

26.

Mastekaasa

(1992) Marriage and psychological well-being: Some evidence on selection into marriage. Journal of Marriage and Family 54: 901–911.

27.

Mendes de Leon

C. F.

Glass

T. A.

Berkman

L. F.

(2003) Social engagement and disability in a community population of older adults: The New Haven EPESE. American Journal of Epidemiology 157(7): 633–642.

28.

Menec

V. H.

(2003) The relation between everyday activities and successful aging: A 6-year longitudinal study. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 58(2): 74–82.

29.

Mulsant

B. H.

Ganguli

Seaberg

E. C.

(1997) The relationship between self-rated health and depressive symptoms in an epidemiological sample of community-dwelling older adults. Journal of the American Geriatrics Society 45(8): 954–958.

30.

Neill

C. M.

Kahn

A. S.

(1999) The role of personal spirituality and religious social activity on the life satisfaction of older widowed women. Sex Roles 40(3–4): 319–329.

31.

Paillard-Borg

Wang

H.-X.

Winblad

Fratiglioni

(2009) Pattern of participation in leisure activities among older people in relation to their health conditions and contextual factors: A survey in a Swedish urban area. Ageing and Society 29(05): 803–821.

32.

Reneflot

Mamelund

S.-E.

(2011) The association between marital status and psychological well-being in Norway. European Sociological Review 28: 355–365.

33.

Rowe

J. W.

Kahn

R. L.

(1997) Successful aging. The Gerontologist 37(4): 433–440.

34.

Sasson

Umberson

D. J.

(2014) Widowhood and depression: New light on gender differences, selection, and psychological adjustment. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences 69(1): 135–145.

35.

Tessier

Vuillemin

Bertrais

Boini

Le Bihan

Oppert

J. M.

Briançon

(2007) Association between leisure-time physical activity and health-related quality of life changes over time. Preventive Medicine 44(3): 202–208.

36.

Verghese

Lipton

R. B.

Katz

M. J.

Hall

C. B.

Derby

C. A.

Kuslansky

Buschke

(2003) Leisure activities and the risk of dementia in the elderly. The New England Journal of Medicine 2003(348): 2508–2516.

37.

Wade

T. J.

Pevalin

D. J.

(2004) Marital transitions and mental health. Journal of Health and Social Behavior 45(2): 155–170.

38.

Waite

L. J.

(1995) Does marriage matter? Demography 32(4): 483–507.

39.

Williams, K., & Umberson, D. (2004). Marital status, marital transitions, and health: A gendered life course perspective. Journal of Health and Social Behavior, 45(1), 81–98.

40.

Winter

Lawton

M. P.

Casten

R. J.

Sando

R. L.

(2000) The relationship between external events and affect states in older people. International Journal of Aging and Human Development 50(2): 85–96.

41.

Yan

X. Y.

Huang

C.-Q.

W.-H.

Qin

(2011) Marital status and risk for late life depression: A meta-analysis of the published literature. Journal of International Medical Research 39(4): 1142–1154.

42.

Zimmer

Lin

H. S.

(1996) Leisure activity and well-being among the elderly in Taiwan: Testing hypotheses in an Asian setting. Journal of Cross-Cultural Gerontology 11(2): 167–186.

	Entire sample	Stably married or partnered	Stably unmarried or widowed	Became widowed during follow-ups	p ^c
	(N = 1,270)	(n = 1,034)	(n = 76)	(n = 160)	p ^c
Gender, n (%)					<.001**
Male	642(50.6)	577(55.8)	29(38.2)	36(22.5)
Female	628(49.4)	457(44.2)	47(61.8)	124(77.5)
Mean age (years)	58.06(4.5)	57.26(4.20)	59.33(4.12)	62.59(4.47)	<.001**
Education, n (%)					<.001**
No formal education	362(28.5)	268(25.9)	24(31.6)	70(43.8)
Grade 1–6	613(48.3)	500(48.4)	37(48.7)	76(47.5)
Grade 7+	295(23.2)	266(25.7)	15(19.7)	14(8.7)
Job, n (%)	685(54.02)	600(58.03)	39(51.32)	46(29.11)	<.001**
Economic satisfaction	3(0.88)	3(0.86)	3(0.93)	3(0.95)	<.001**
Self-assessed health status	3(1.07)	3(1.08)	3(1.00)	3(1.01)	.02*
Leisure activities, n (%)
Physical activities	962(75.8)	799(77.3)	64(84.2)	99(61.9)	.8664
Cognitive activities	518(40.8)	446(43.1)	25(32.9)	47(29.4)	<.001**
Social activities	948(74.7)	786(76.0)	56(73.7)	106(66.3)	.8824
Religious activities	1,104(86.9)	895(86.6)	64(84.2)	145(90.6)	<.001**
Depressive symptoms	4.53(5.12)	4.27(5.01)	6.55(6.25)	5.21(5.00)	<.001**
Life satisfaction	6.73(2.36)	6.87(2.31)	5.71(2.52)	6.30(2.45)	<.001**