Life Satisfaction and Depression in the Oldest Old: A Longitudinal Study

Abstract

This study aimed to analyze the effect of individual differences and family variables on life satisfaction and depression in the oldest old compared with the young-old. A total of 1,799 cases from an 8-year period of the Korean Welfare Panel Study (2006–2013) were analyzed. A key finding was that life satisfaction significantly increased with time for the two groups of older adults while depression decreased. Moreover, family relationship satisfaction significantly affected both life satisfaction and depression in both groups. However, its impact was stronger for the oldest old. Finally, individual difference variables, that is, objective life conditions, such as gender, education, and religion, did not have a significant impact on life satisfaction or depression in the oldest old. The results suggest that the oldest old not only face death but also experience continuous growth from a gerotranscendence perspective.

Keywords

oldest old life satisfaction depressive symptoms gerotranscendence longitudinalstudy

Research on older adults has predominantly assumed this age-group to be homogeneous. Population aging, however, may suggest the necessity of further segmenting the group. Specifically, older adults in their 60s, 70s, 80s, and 90s may not be only biologically heterogeneous but also psychosocially different (Gonyea, 2010). This might explain the inconsistent findings of studies of psychosocial development of older adults (Berg, Hoffman, Hassing, McClearn, & Johansson, 2009; Chou & Chi, 1999; Heo, 2017; Kaup et al., 2016; Lee, 2014; McCamish-Svensson, Samuelsson, Hagberg, Svensson, & Dehlin, 1999; Menec, 2003; Park, Son, & Bae, 2009; Sung, 2013). If there are psychosocially heterogeneous or other types of subgroups within the older adult group, the social policies for older adults may need to be thoroughly reviewed (Chen & Jordan, 2018). Nevertheless, the concept of segmenting the older adult group in gerontology research is still in its infancy. Compared with other developing or developed countries, this issue might be more serious in South Korea because of its fast-growing older adult population.

In South Korea, the average life expectancy was 61.9 years for men and 70.4 for women as of 1980, which increased to 79.7 and 85.7 for men and women, respectively, by 2017; this is a respective increase of 17.8 and 15.3 years over the course of a mere 37 years (Statistics Korea, 2018a). It is predicted that population aging will accelerate further. The aging rate of South Korean society is one of the highest in the world. By 2026, South Korea is expected to become a super-aged society, with more than 20% of the population aged older than 65 years; in 2015, the oldest old (individuals aged 80 years or older) accounted for 2.6% of the total population, which is expected to increase to 14% by 2050 (Statistics Korea, 2018b).

This study focused on the oldest old within the older adult group. We particularly investigated psychosocial aspects, such as life satisfaction and depression, in relation to family factors. Thus, the purpose of the study was to analyze the effect of family factors on life satisfaction and depression in the oldest old compared with the young-old. We used longitudinal panel data over an 8-year period (2006–2013) and defined the oldest old as those aged 80 years or older, which is consistent with the definition used in previous social sciences and medical studies of the oldest old (Berg et al., 2009; Kassem et al., 2018; Oates, Berlowitz, Glickman, Silliman, & Borzecki, 2007). However, given the increasing number of older adults in the population, researchers may need to redefine the oldest old as those aged 85 years or older, a change already adopted by the U.S. Census Bureau.

Psychosocial Development Among Older Adults

In relation to the psychosocial aspects of older adults, Erikson’s (1959) argument is worth noting. He defined the course of human psychosocial development as a process taking place throughout one’s lifetime. He especially argued that after age 60, psychosocial development is completed through ego-integration and the acceptance of one’s own life. If positive ego-integration is not achieved, one may finish life in despair. This perspective has been further expanded in the concept of “gerotranscendence” as an additional life stage (Tornstam, 2011, p. 166). According to this theory, the perspective of seeing the world through materialism and rationalism transforms into a more cosmic and transcendent perspective in the oldest old stage, which may increase life satisfaction (Tornstam, 2011).

Understanding the processes underlying human psychosocial development, however, differs to some extent between the West and the East. Confucius discussed the developmental process of humans by age groups in “The Analects of Confucius,” where age 60 is “Isun,” meaning that one can understand whatever one hears by oneself, and age 70 is “Jongsim,” meaning that one does not break the law even if one acts up to one’s resolution as a saint (Choi, 2008, p. 16). Regarding human aging, the East and the West show similar perspectives: Humans grow continuously throughout the course of their lives and understand the importance of the subjective experience of life acceptance as they get older. Previous studies of the oldest old also included subjective aspects, such as retention of life purpose, sense of identity, daily hassles, life purpose, emotional support, social network quality, and self-rated overall health, as key variables that affect well-being (Berg et al., 2009; Browne-Yung, Walker, & Luszcz, 2017; Jeon & Dunkle, 2009; Krause, 2004).

In contrast to the West, variables related to life satisfaction and depression in old age that have received great attention in the East are family factors, including family structure and family relationships (Chai & Jun, 2017). This perspective may be associated with Confucianism, which emphasized the filial duty to parents. A study of the oldest old in China showed that the level of subjective well-being was higher among those living with their spouse or children than among those living alone (Chen & Short, 2008). Conversely, Western studies reported that social support from close friends, rather than family support, and other related variables were significant (Browne-Yung et al., 2017; Cho, Martin, & Poon, 2015; Jeon & Dunkle, 2009; Krause, 2004; McCamish-Svensson et al., 1999). These studies indicate that life satisfaction and depression can vary within the oldest old age-group depending on social customs, traditions, and cultural background.

In fact, across the West and the East, there have been diverse academic attempts to investigate the oldest old age-group. Longevity has been their main interest. With respect to the oldest old group, most of the studies have focused on understanding ways to increase life expectancy. Consequently, the psychosocial aspects of the oldest old group could not receive reasonable attention. This might explain the difficulty in finding studies that compared psychosocial characteristics between older adult groups by subdividing them. Furthermore, a few longitudinal studies of the oldest old have been conducted.

Life Satisfaction and Depression

With older adults being studied as a homogeneous rather than heterogeneous group, the results of the longitudinal studies of life satisfaction and depressive symptoms of older adults have shown mixed results. Some studies have shown that as the overall life satisfaction of older adult increases, the growth rate gradually changes to a parabolic form (e.g., McCamish-Svensson et al., 1999; Park et al., 2009). In contrast, there are studies that reported decreased life satisfaction as people aged (e.g., Berg et al., 2009; Chou & Chi, 1999; Heo, 2017; Menec, 2003). Regarding the depression levels of older adults, most of the previous longitudinal studies revealed that depressive symptoms gradually increased as people aged. Kaup et al. (2016) and Lee (2014) reported that the level of depressive symptoms of older adults tended to increase over time. However, a recent study of the oldest old (aged 80 years and older) found that depressive symptoms showed improvement at yearly interviews between 2006 and 2013 (Sung, 2013). In sum, these studies rarely examined older adults as a heterogeneous group and, thus, are limited despite using a longitudinal design. Furthermore, they did not consider a theoretical perspective that could explain human being as a pursuit of continuous development and growth. This might explain the inconsistent findings reported in these studies.

When reviewing the literature on the factors affecting life satisfaction in the oldest old, it is apparent that previous studies emphasized not only objective dimensions (such as sociodemographic characteristics, health, and financial status) but also subjective dimensions. In particular, the research findings in the West have indicated that life-related attitudes (such as a sense of identity, life purpose, and resilience) affect mental health (Browne-Yung et al., 2017; Jeon & Dunkle, 2009). According to Krause (2004), the emotional support from close friends can mitigate the negative influence of traumatic events on the overall life satisfaction of the oldest old. Cho et al. (2015) also reported that a higher frequency of conversation and interaction with nonfamily members improved the life satisfaction of the oldest old.

Therefore, the research on the life satisfaction of the oldest old in the West highlights the strong influence of social relationships, such as individuals’ attitude toward life and friends. On the other hand, research on the oldest old in Asian countries such as Korea and China, which have a Confucian culture, highlights family relations, such as those with spouses and children, as the main contributing factors of well-being. As discussed earlier, this is because the significance of the family remains strong until late old age. For example, Yeung and Fung (2007) examined the contribution of family members and friends on life satisfaction of Hong Kong Chinese older adults. They reported that family support, compared with friend support, more likely affected the life satisfaction of older adults. They also revealed that emotional support from family members appeared to be more helpful to improve life satisfaction. Other studies also showed that close family relations were significant predictors of depressive symptoms in the family-oriented culture of Asia (Chou & Chi, 2003; Lee, Ruan, & Lai, 2005). Older adults who received social support from family members showed fewer depressive symptoms over time.

In addition to the depressive symptoms of the oldest old, there is also a body of research examining the effects of objective dimensions (i.e., sociodemographic, physical, and financial characteristics) by dividing older adults into age groups. Ried and Planas (2002) analyzed older adults’ level of depression according to their gender and age and found that the oldest old women were more depressed than the oldest old men, and the oldest old were more depressed than the young-old. Women are more likely to feel lonely than men because they spend more time alone (Brittain et al., 2017). Regarding health, the higher the age and the lower the subjective level of health, the higher the level of depression. However, factors such as financial changes, changes in family relationships, and other psychosocial factors related to late old age were not considered. Smith, Borchelt, Maier, and Jopp (2002) analyzed the relationship between health status and well-being by classifying older adults into young-old and oldest old. The results showed that the level of well-being of the young-old group was significantly higher than that of the oldest old group. They found that chronic diseases and functional impairments in particular (e.g., vision, hearing, and mobility impairments) significantly limited the well-being of the oldest old group, which suggests that among objective conditions, physical health needs to be considered as a factor influencing the well-being of the oldest old.

Considering the limitations of previous studies, studies investigating life satisfaction and depressive symptoms of the Asian oldest old should take objective dimensions (i.e., sociodemographic, physical, and financial characteristics) as well as family factors into consideration. Furthermore, adopting a longitudinal perspective will permit exploring potential cause-and-effect associations.

Methods

Sample and Procedure

Longitudinal data from the Korean Welfare Panel Study (KWPS) from Wave 1 (2006) to Wave 8 (2013) were used to examine the pattern of life satisfaction and depression and their causal factors in the oldest old. The study was conducted by a joint team from the National Institute of Health and Social Welfare (Korea Institute for Health and Social Affairs) and Seoul National University. KWSP is secondary data collected through a nationwide random-stratified sampling method, resulting in a representative sample of Koreans (Nam, Moon, & Lee, 2012). The survey was completed by face-to-face interviews between investigators and respondents from panel households. The survey questionnaires included items on sociodemographic characteristics, economic activity, social welfare service uses, personal history, family relationships, mental health, and so forth (Korea Institute for Health and Social Affairs & Seoul National University, 2013).

There were 667 persons aged 80 years or older in 2006. Of the 667 cases, 89 cases were excluded from the analysis due to missing dependent variables (i.e., life satisfaction and depression). Another 248 cases were also not included in the analysis because they did not meet the sample selection criteria, which required participating in the survey more than four times during the 8 years (Wave 1 to Wave 8). However, even though the final sample comprised 330 cases, the analysis of cases may be different due to different response pattern in each year.

We performed a chi-square analysis to test homogeneity between the included and the excluded participants from Wave 1 to Wave 8. As shown in Table 1, there was no significant difference between the two groups. In fact, individual growth models require a minimum of three repeated observations to analyze personal change over time to make sure of a curve. However, we applied stricter criteria to select cases with a minimum of four repeated observations. In addition, for comparison purposes, we also analyzed 1,469 young-old persons (aged 65–69 years) using the same selection procedure.

Table 1.

Characteristics of the Oldest Old (80+).

	Year	2006	2007	2008	2009	2010	2011	2012	2013	χ²
Variables	n	320	311	322	317	274	236	193	159
Gender (%)	Male	33.4	33.8	34.2	33.4	35.0	33.5	34.2	34.0	0.258
Gender (%)	Female	66.6	66.2	65.8	66.6	65.0	66.5	65.8	66.0	0.258
Education (%)	No	63.8	63.7	63.4	65.0	63.9	63.6	63.7	64.2	3.44
	Middle	29.4	28.9	30.4	28.7	29.6	30.1	31.1	30.8
	High	4.0	4.2	3.1	3.2	2.9	3.0	2.6	2.5
	College	2.8	3.2	3.1	3.2	3.7	3.4	2.6	2.5
Spouses (%)	Yes	34.4	34.1	33.5	30.0	28.8	27.2	25.4	26.4	16.41
Spouses (%)	None	65.6	65.9	66.5	70.0	71.2	72.8	74.6	73.6	16.41
Religion (%)	Yes	53.1	57.2	54.7	49.5	56.2	53.4	57.0	53.5	5.27
Religion (%)	None	46.9	42.8	45.3	50.5	43.8	46.6	43.0	46.5	5.27
Age	M (SD)	83.4 (3.11)	84.3 (3.06)	85.3 (3.08)	86.4 (3.11)	87.1 (2.88)	87.9 (2.79)	88.9 (2.85)	89.7 (2.65)	–
Life satisfaction	M (SD)	2.75 (0.62)	2.99 (0.59)	2.98 (0.56)	2.94 (0.59)	3.07 (0.56)	3.14 (0.54)	3.08 (0.51)	3.20 (0.54)	–
Depression	M (SD)	9.03 (6.31)	8.06 (5.80)	7.59 (5.58)	7.42 (5.80)	8.01 (6.58)	7.5 (6.09)	6.51 (5.45)	6.85 (5.38)	–

Note. Life satisfaction = 5-point scale ranging from 1 (very dissatisfaction) to 5 (very satisfaction), Depression = 4-point scale ranging from 0 (once a week) to 3 (6 days or more). Korean Center for Epidemiological Studies Depression scale was used by sum of all items.

Measures

Life satisfaction

Life satisfaction was measured with seven items (i.e., health, economic condition, residential environment, job, social relationships, leisure life, and overall satisfaction) rated on a 5-point Likert scale ranging from 1 (very dissatisfied) to 5 (very satisfied). The mean value of the seven item scores was used as the life satisfaction score, with higher scores indicating higher life satisfaction. Cronbach’s alpha for this scale was .75 or higher over the 8 years. The value of average variance extracted was .540, which met the minimum standard value (.5; Chin, 1998). The composite reliability value was also satisfied for convergent validity at .852. Discriminant validity was assured by factor analysis (Choi, 2009).

Depression

For depressive symptoms, the study used the Korean version of the Center for Epidemiological Studies Depression (CES-D) scale with 11 items (poor appetite, as good as other people, depressed, felt difficulty, insomnia, lonely, live without complaint, people were unfriendly, sad, people disliked me, fearful). The Korean CES-D was translated from English into Korean and tested by Nam and Lee (Cho & Kim, 1993). The CES-D scale has shown high internal consistency and validity and has been used in several studies (Andreescu, Chang, Mulsant, & Ganguli, 2008; Huang et al., 2011). The Korean CES-D scale has appropriate test–retest reliability (.68 over several weeks), internal consistency (.89 to .93), and concurrent validity (Cho & Kim, 1993; Moon et al., 2017). Items are rated on a 4-point scale: 0 (very few/once a week), 1 (sometimes/2 to 3 days a week), 2 (often/4 to 5 days a week), and 3 (always/6 days or more). The depression score is calculated by summing all item scores, and higher scores indicate higher depression levels. Scores range from 0 (lowest) to 60 (highest). Cronbach’s alpha in this study was .83 or higher over the 8 years.

Independent variables

This study used time-variant variables (Level 1), where time was 2006 (0) to 2013 (7). Physical health change was assessed by measuring chronic disease and perceived health. Chronic disease was classified as no chronic disease (0), changed chronic disease (1), continuous chronic disease (2). Perceived health was measured using a self-report assessment on a 5-point scale: very unhealthy (1), unhealthy (2), normal (3), healthy (4), very healthy (5). Economic change consisted of public pension—none (0), changed (1), continued (2)—and poverty household. Poverty household was measured by income of less than 60% of the national median income with a range from 0 to 8 years, and higher scores meant poorer economic status. Family relationship change consisted of marital status and family relationship satisfaction. Marital status was measured by spouse loss (0), continued partnerless (1), continued partner (2). Family relationships were measured using a 7-point Likert scale—very dissatisfied (1), dissatisfied (2), slightly dissatisfied (3), neutral (4), slightly satisfied (5), satisfied (6), very satisfied (7). Higher scores indicated greater satisfaction with family relationships. A change in receiving social service was measured using living assistance support service and caregiving service use. Living assistance support service included old age pension, medical expense support, free meals, material support, and home-delivered meal service—none (0), changed (1), continued use (2). Caregiving service use included home service and visiting nursing—none (0), yes (1).

There were also time-invariant variables (Level 2), that is, independent variables indicating individual differences that were found to not change over time, including sociodemographic characteristics and personal psychological resources. Sociodemographic characteristics included gender—male (0), female (1); participation in religious activity—none (0), yes (1); education—none (0), middle school (1), high school (2), college (3); and adverse childhood events such as the loss of a parent early in life—none (0), yes (1) and school interruption as a result of poverty—none (0), yes (1). Self-esteem was considered a psychological resource. According to previous research, self-esteem may change or remain stable over time (Pullmann, Allik, & Realo, 2009). The present study analyzed the level of self-esteem for each age-group and found no change over 8 years. Self-esteem, therefore, was taken as a time-invariant variable and coded as the value at Wave 1 (2006). For example, the mean self-esteem score of the oldest old group at each time point (Wave 1 to Wave 8) was 2.72, 2.72, 2.73, 2.74, 2.73, 2.70, 2.68, 2.69, and 2.72, respectively. Self-esteem was assessed using the self-esteem scale developed by Rosenberg (1979), which contains 10 items rated on a 4-point Likert type scale: 1 (disagree), 2 (neutral), 3 (somewhat agree), 4 (strongly agree). High scores indicated higher self-esteem. The items include items such as “I feel that I’m a person of worth,” “I feel that I have a number of good qualities,” and “At times I think I am no good at all” and so on. Cronbach’s alpha at Wave 1 was .79.

Analytic Plan

This study used an individual growth model to analyze how specific variables changed over time for each group, using STATA version 13.0 statistical software. The individual growth model was found to be suitable to analyze the patterns of specific variables of individuals over time. The data set for this study was longitudinal, and variables were repeatedly measured at the individual level at various time points (Wave 1 to Wave 8). It had a nested multilevel data structure in which repeated measurements for each individual were recorded. Therefore, the Level 1 model was a within-subjects model based on repeated measures for each individual. The Level 2 model was a between-subjects model that functionalized the relationship between the initial values estimated from the Level 1 model, individual characteristics, and background variables (Kashy, Donnellan, Burt, & McGue, 2008). This methodology can overcome the limitations of the conventional regression analysis method in that it can resolve autocorrelation issues that can occur when using repeated measurements on individuals (Chen & Cohen, 2006). In addition, because the constant term and the slope parameters vary according to the Level 2 variables, which imply individual differences, this methodology functions as a random coefficient model between the multilevel models.

Moreover, because this study used longitudinal data from 2006 to 2013, the attrition of participants occurred over time. As presented in Table 1, the number of cases for the final analysis decreased from the first year to the last year. The sample consisted of 320 participants in Wave 1. In Wave 8, however, only 159 older adults remained. Case attrition could have resulted from the respondent’s death, health problems, behavioral problems, moving, and so on. Similarly, Jeon and Dunkle (2009) found that their sample of older adults also decreased from Wave 1 (193 cases) to Wave 4 (155 cases). They indicated that causes of the attrition included participant’s health problems, death, behavioral problems, and so on. Case attrition is known as a general limitation associated with longitudinal studies of older adults.

In the case of the oldest old group, only 136 (41.2%) of the 330 individuals had participated in all surveys from the first to the eighth year. The remaining 194 subjects (58.5%) participated at least four times. The individual growth model used in this study is robust against differences in the time of measurements or measurement values between individuals, unlike the latent variable growth model or structural equation model (Kashy et al., 2008). However, it was necessary to closely investigate the effect of attrition over time on the model and assess whether the analysis results indicated a change.

Therefore, this study evaluated whether the effects of the factors influencing the subjective well-being of older adults were significant after attrition cases had been considered. Referring to previous studies (Jeon & Dunkle, 2009; Park et al., 2009) that quantified the attrition factor in growth model analysis, participants were divided into two groups: those who completed all 8 yearly surveys (0) and those who did not complete all surveys (1). The estimated coefficients after controlling for attrition were found to be somewhat lower than the estimated coefficients without controlling for attrition. This finding suggests that the measurement values could have been biased by the responses of participants completing the survey all 8 years; therefore, they should be adjusted.

Results

Characteristics of the Oldest Old Group and the Young-Old Group

Sample characteristics of the oldest old are shown in Table 1. The respondents were predominantly women (more than 65% over the 8 years). Most respondents had a low education level. Less than 10% had education beyond high school or a college degree. Marital status changed over time; 34.4% lived with a spouse at Wave 1, but the percentage decreased over time, with only 26.4% of the oldest old having spouses at Wave 8. However, the percentage of persons indicating religious activity was consistent over the 8 years. About 54% of the oldest old on average indicated religious activity. The average age of the oldest old was 83.4 at Wave 1 and 89.7 at Wave 8. The mean life satisfaction increased from 2.75 at Wave 1 to 3.20 at Wave 8, and depression decreased from 9.03 to 6.85. Chi-square tests showed no significant differences in characteristics from Wave 1 to Wave 8, suggesting that this group was homogeneous over the 8-year study period.

Characteristics of the young-old sample are presented in Table 2. About 60% of the respondents were female, and more than half had a middle school education level, indicating that the education level of the young-old group was higher than that of the oldest old group. About two thirds of the young-old had spouses unlike the oldest old, but the number of young-old with spouses also decreased over time. The proportion of young-old with religious activity was also consistent at about 60% on average. Based on the result of the chi-square test of characteristics, there was a significant difference in gender, education, marital status, religion, life satisfaction, and depression between the oldest old and the young-old groups (p < .001). The result indicates that the two groups had heterogeneous characteristics. Furthermore, the average age of the young-old was 66.46 at Wave 1 and 73.44 at Wave 8. The mean life satisfaction had slightly increased over time from 2.87 to 3.22 for Wave 1 and Wave 8, respectively, and depressive symptoms decreased over time from 7.28 to 5.28 for Wave 1 and Wave 8, respectively. The overall depression level of the young-old was lower than that of the oldest old.

Table 2.

Characteristics of the Young-Old (65–69 Years Old).

	Year	2006	2007	2008	2009	2010	2011	2012	2013	χ²
Variables	n	1,420	1,397	1,433	1,430	1,408	1,344	1,322	1,279	χ²
Gender (%)	Male	40.14	40.30	39.85	39.93	39.13	38.47	38.06	37.69	4.06
Gender (%)	Female	59.86	59.70	60.15	60.07	60.87	61.53	61.94	62.31	4.06
Education (%)	No	21.69	21.62	21.56	21.75	21.88	22.32	22.60	22.20	2.05
	Middle	60.85	60.99	61.41	61.19	61.29	61.16	60.96	61.22
	High	11.97	11.88	11.58	11.40	11.36	11.16	11.34	11.49
	College	5.49	5.51	5.44	5.66	5.47	5.36	5.11	5.08
Spouses (%)	Yes	73.59	72.30	70.41	69.37	68.11	66.15	64.64	63.41	57.49***
Spouses (%)	None	26.41	27.70	29.59	30.63	31.89	33.85	35.36	36.59	57.49***
Religion (%)	Yes	59.48	63.77	60.71	58.46	61.72	61.31	63.06	62.47	13.23
Religion (%)	None	40.52	36.23	39.29	41.54	38.28	38.69	36.94	37.53	13.23
Age	M (SD)	66.46 (1.74)	67.47 (1.73)	68.46 (1.73)	69.47 (1.73)	70.45 (1.73)	71.45 (1.72)	72.45 (1.72)	73.44 (1.72)	–
Life satisfaction	M (SD)	2.87 (0.62)	3.02 (0.60)	3.04 (0.56)	3.05 (0.58)	3.16 (0.51)	3.19 (0.52)	3.18 (0.50)	3.22 (0.52)	–
Depression	M (SD)	7.28 (6.43)	7.07 (6.10)	6.39 (5.48)	6.16 (5.65)	6.18 (5.74)	5.72 (5.38)	5.04 (4.92)	5.28 (5.50)	–

***p < .001.

Trajectories of Life Satisfaction

To examine the pattern of life satisfaction over eight time points and individual differences, this study used an unconditional linear growth model. Table 3 shows individuals’ change patterns of life satisfaction. The intercept of life satisfaction at Wave 1 was 2.847 and the linear slope .046 (p < .001). The average slope of life satisfaction increased by .046 per year. The variance of the intercept and slope was found to be statistically significant, implying that individual differences varied significantly between the respondents. The correlation of the intercept and slope had a negative effect (b = –.0112), which indicated that the linear slope of the oldest old persons with higher life satisfaction at Wave 1 was more likely to be slower than that of the oldest old persons with lower life satisfaction at Wave 1. The results of the intraclass correlation coefficient (ICC) analysis showed that individual differences at Level 2 explained 46.6% (ICC = .166/(.166+.190) = .466) of the total variance in life satisfaction. The results imply that life satisfaction in the oldest old group is likely to increase with time, and their trajectories of life satisfaction can be explained by individual differences such as gender and education.

Table 3.

Results of Unconditional Model Analysis (Life Satisfaction; N = 1,779).

	Young-old (65–69)		Oldest old (80+)
	Coefficient	SE	Coefficient	SE
Fixed effects
Intercept	2.935***	0.014	2.847***	0.027
Linear slope	.044***	.002	.046***	.005
Random effects
Variance of intercept	0.207***		0.166***
Variance of slope	.002***		.002***
Variance of residual	0.157		0.190
Covariance	–0.013		–0.0112
ICC	0.569		0.466

Note. ICC = intraclass correlation coefficient.

***p < .001.

Trajectories of Depression

This study also analyzed the pattern of depression as another dependent variable (Table 4). For the oldest old group, the intercept of depression at Wave 1 was 8.45 and the linear slope –.207 (p < .001). This means that depression was likely to decrease over time, and the average slope of depression decreased .207 per year. The variance of the intercept and slope was statistically significant and so were individual differences. The correlation of the intercept and the slope had a negative effect (b = –.733). The reduction rate of depression of the oldest old with higher depression at Wave 1 was more likely to be slower than that of the oldest old persons with lower depression at Wave 1. This means that the depressed symptoms could have lasted for a long time among the oldest old who had higher depression at Wave 1. The results of the ICC analysis showed that individual differences at Level 2 explained 38.1% (ICC = 13.73/(13.73 + 22.26) = .381) of the total variance in depression. This indicates that depression in the oldest old group is likely to decrease with time, and their trajectories of depression can be explained by individual differences, such as gender and education.

Table 4.

Results of Unconditional Model Analysis (Depression; N = 1,779).

	Young-old (65–69)		Oldest old (80+)
	Coefficient	SE	Coefficient	SE
Fixed effects
Intercept	7.217***	0.135	8.45***	0.271
Linear slope	–.295***	.022	.207***	.590
Random effects
Variance of intercept	18.62***		13.73***
Variance of slope	.254***		.282***
Variance of residual	18.03		22.26
Covariance	–1.259		–0.733
ICC	0.508		0.381

Note. ICC = intraclass correlation coefficient.

***p < .001.

The Longitudinal Relationships Among the Changes in Life Satisfaction and Independent Variables

Once the unconditional linear growth model was tested, a conditional growth model was developed to consider additional predictors such as fixed variables (e.g., gender) and time-varying variables (e.g., family relationship). This study also examined the quadratic change effect by growth curve modeling. As a result, the variable of the time square had no significant effect statistically in both the oldest old and the young-old group (p < .360, p < .196). In the case of the oldest old group, the values of Akaike information criterion (AIC; 1921.16) and Bayesian information criterion (BIC; 2055.82) in the linear growth model were smaller than those of the growth curve model (AIC: 1922.32; BIC: 2062.37). The smallest value of AIC or BIC can be interpreted to have better model fit (Burnham & Anderson, 2004); that is why the linear growth model is selected for the final model of the study. Table 5 shows the first model of the slope of life satisfaction. The slope of life satisfaction of the oldest old group was significantly associated with time, self-reported health, poverty level, family relationship satisfaction, and self-esteem. In terms of Level 1 variables, the estimated coefficient for time was .031 (p < .001), and the slope of life satisfaction increased significantly when the oldest old persons were satisfied with family relationships (b = .118, p < .001) and perceived themselves as healthy (b = .178, p < .001). The results revealed that trajectories of life satisfaction increased with time, good family relationships, and higher perceived health. On the other hand, the findings indicated that chronic disease was not statistically associated with the slope of life satisfaction. For economic change of the oldest old group, the level of life satisfaction decreased by .032 when the period of poverty increased by 1 year (p < .001). Economic condition was still a meaningful factor for the oldest old.

Table 5.

Results of Conditional Growth Model for the Slope of Life Satisfaction (N = 1,732).

	Young-old (65–69)		Oldest old (80+)
	Coefficient	SE	Coefficient	SE
Fixed effects
Intercept	1.828 ***	0.092	1.988 ***	0.196
Attrition (0 = no, 1 = yes)	–0.144 ***	0.019	–0.122 **	0.036
Level 1 (within individual)
Family RS	0.103 ***	0.004	0.118 ***	0.010
Time	0.03 ***	0.002	0.031 ***	0.007
Chronic disease (reference: 0 = no)
Change (1)	–0.094	0.052	–0.147	0.102
Continued (2)	–0.141 **	0.053	–0.186	0.104
Self-RH	0.144 ***	0.006	0.178 ***	0.013
Public pension (reference: 0 = no)
Change (1)	–0.005	0.02	–0.010	0.066
Continued (2)	0.073 ***	0.019	–0.018	0.142
Poverty duration	–0.028 ***	0.003	–0.032 ***	0.007
Marital status (reference: 0 = loss)
Continued partnerless (1)	–0.011	0.026	0.046	0.054
Continued partner (2)	–0.019	0.024	–0.089	0.061
Living SS (reference: 0 = no)
Change (1)	–0.024	0.036	–0.011	0.084
Continued (2)	–0.063	0.045	–0.041	0.093
Care GS (0 = no, 1 = yes)	–0.029	0.018	–0.058	0.035
Level 2 (individual difference)
Gender (0 = male, 1 = female)	0.066 ***	0.019	–0.025	0.047
Religion (0 = no, 1 = yes)	0.037 **	0.011	0.024	0.025
Education	0.029 **	0.011	0.028	0.025
Lost PC (0 = no, 1 = yes)	–0.018	0.016	–0.030	0.036
School IP (0 = no, 1 = yes)	–0.001	0.014	–0.038	0.037
Self-esteem	0.161 ***	0.018	0.112 **	0.039
χ²	2,829.47***		610.51***
Random effects	Est	SE	Est	SE
Intercept^a	0.091***	0.008	0.101***	0.023
Slope^b	.002***	.0003	.003***	.001
χ²	930.49***		110.93***

Note. Family RS = family relationship satisfaction; Self-RH = self-reported health; Living SS = living support service use; Care GS = caregiving service use; Lost PC = loss of a parent in childhood; School IP = school interruption from poverty.

^aThe variance of intercept in Level 2 (individual difference).

^bThe variance of slope in Level 2 (individual difference).

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

Further, in Level 2, self-esteem only had a significant effect on the slope of life satisfaction (b = .112, p < .001). This means that older adults with higher self-esteem were more likely to have increasingly higher life satisfaction when compared with those with lower self-esteem. The sociodemographic characteristics like gender, religious activity, and education level had significant effects on the slope of life satisfaction. Females had a higher slope of life satisfaction than males (b = .066, p < .001), and those who had religious activity also had a higher slope than those who did not (b = .037, p < .01). Meanwhile, the rate of life satisfaction appeared to be positively associated with higher levels of education (b = .029, p < .01).

In sum, in comparison with the oldest old, for the young-old, objective life conditions still had a significant impact on the slope of life satisfaction as subjective factors did. For example, unlike the oldest old, chronic disease was significantly associated with the slope of life satisfaction in the young-old group. The young-old adults who were consistently eligible for a public pension had a higher slope of life satisfaction than those who were not. These results indicate that objective life conditions are important factors for the young-old group.

The Longitudinal Relationships Among the Changes in Depression and Independent Variables

Table 6 presents the second conditional growth model for the slope of depression. Like the conditional growth model for life satisfaction, the quadratic change effect for changes in depression was tested by growth curve modeling. The variable of “time square” appeared not to be significantly associated with the slope of depression in both old groups (p < .709, p < .311). In the case of the oldest old group, the values of AIC (9767.53) and BIC (9902.19) in the linear growth model were also smaller than those of the growth curve model (AIC: 9769.39; BIC: 9909.44). The linear growth model was taken as the final model of the study. First, family relationship satisfaction (b = –1.144, p < .001), time (b = –.153, p < .05), self-reported health (b = –1.364, p < .001), and caregiving service use (b = .874, p < .05) were associated with the slope of depression for the oldest old group. The slope of depression decreased significantly when the oldest old persons were satisfied with family relationships and perceived themselves as healthy.

Table 6.

Results of Conditional Growth Model for the Slope of Depression (N = 1,732).

	Young-old (65–69)		Oldest old (80+)
	Coefficient	SE	Coefficient	SE
Fixed effects
Intercept	17.57***	0.869	20.81***	2.137
Attrition (0 = no, 1 = yes)	1.157***	0.2	0.812**	0.394
Level 1 (within individual)
Family RS	–0.876***	0.042	–1.144***	0.109
Time	–0.240***	0.026	–0.153***	0.07
Chronic disease (reference: 0 = no)
Change (1)	0.942	0.548	–0.174	1.114
Continued (2)	1.208 *	0.553	0.314	1.127
Self-RH	–1.071***	0.06	–1.364***	0.147
Public pension (reference: 0 = no)
Change (1)	–0.096	0.211	–0.203	0.739
Continued (2)	–0.233	0.206	2.057	1.553
Poverty duration	0.112***	0.029	0.122	0.076
Marital status (reference: 0 = loss)
Continued partnerless (1)	–0.508	0.270	–0.163	0.592
Continued partner (2)	–1.554***	0.250	–0.806	0.673
Living SS (reference: 0 = no)
Change (1)	0.103	0.373	–0.275	0.904
Continued (2)	0.658	0.467	0.112	1.010
Care GS (0 = no, 1 = yes)	0.239	0.191	0.874 *	0.391
Level 2 (individual difference)
Gender (0 = male, 1 = female)	0.957***	0.195	0.238	0.514
Religion (0 = no, 1 = yes)	–0.318**	0.117	–0.212	0.284
Education	0.028	0.117	–0.147	0.276
Lost PC (0 = no, 1 = yes)	0.753***	0.166	–0.262	0.400
School IP (0 = no, 1 = yes)	0.262	0.165	0.415	0.409
Self-esteem	–1.577***	0.186	–1.583**	0.433
χ²	1,850.63***		358.38***
Random effects	Est	SE	Est	SE
Intercept^a	10.08***	0.922	6.451***	2.387
Slope^b	.286***	.038	.123***	.113
χ²	806.48***		72.27***

^aThe variance of intercept in Level 2 (individual difference).

^bThe variance of slope in Level 2 (individual difference).

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

However, the oldest old who had received caregiving services, such as home service and visiting nursing, showed a higher level of depression than those who had not. Usually, those with difficulty in performing activities of daily living or instrumental activities of daily living due to physical constraints need to receive caregiving services. Physical constraints may negatively affect depression among the oldest old (Jeon & Dunkle, 2009). However, this was not derived from an analysis of causal relationships between the related variables; so further research is required. In terms of the individual differences model (Level 2), only self-esteem was associated with the slope of depression, as in the first conditional growth model. This shows that subjective cognitions such as self-reported health, family relationship satisfaction, and self-esteem are significant factors affecting depression for the oldest old.

Discussion

Previous literature on older adults has generally assumed this age-group to be homogeneous, and life satisfaction and depression levels of the older adults have been studied with cross-sectional data rather than longitudinal methods. However, the present study empirically analyzed the pattern of life satisfaction and depression of the oldest old (80 years old or older) and its causal factors compared with the young-old (65 to 69 years old) over time using longitudinal data from the KWPS from Wave 1 (2006) to Wave 8 (2013).

This study yielded several major findings. First, there was no difference in the trajectory of life satisfaction improvement and depression reduction between the two groups. However, the young-old appears to have significantly higher life satisfaction and lower depression compared those in the oldest old. For example, the oldest old group’s mean depression score over 8 years was 7.73 (SD = 5.96) but that of the young-old group was 6.16 (SD = 5.72). This finding is consistent with the results of previous studies, which showed that the oldest old group tends to have a higher level of depression than does the young-old group (Blazer, Burchett, Service, & George, 1991; Stallones, Marx, & Garrity, 1990). Although the finding of the present study was obtained by considering old adults as a heterogeneous group, it is consistent with the results of previous studies, which considered old adults as a homogeneous group (e.g., McCamish-Svensson et al., 1999; Park et al., 2009). In fact, previous studies about the changes in life satisfaction and depressive symptoms among older adults over time have produced mixed results. Some studies found that the level of life satisfaction of older adults decreases (e.g., Chou & Chi, 1999; Menec, 2003) and the level of depression increases (Henderson et al., 1997; Lee, 2014; Sung, 2013) over time. However, apart from our study, there have been no comparative studies that considered older adults as a heterogeneous group.

Second, this study shows that family relationship satisfaction significantly affects the life satisfaction and depression of both groups based on the conditional growth model. However, an interesting point is that the effect size of the family factor worked differently for the two groups. In addition, we further analyzed the data to determine the effect size using Cohen’s index.

Cohen’s index is used to compare effect sizes of influencing factors, similar to using p values and their significance. Usually, a value from .02 to .15 indicates a small effect, from .15 to .35 a medium effect, and from .35 or higher a large effect (Cohen, 1992). For the oldest old group in this study, the effect sizes of family relationship satisfaction on life satisfaction and depression models were .250 and .124, respectively, which were the highest among the independent variables. For the young-old group, the effect sizes of the variable of time were the highest in the life satisfaction and depression models (.252 and .244, respectively). Self-esteem and family relationships satisfaction had the second- and the third-highest effect sizes in the life satisfaction and depression models, respectively. This finding can be understood in relation to the socioemotional selectivity theory, in which Carstensen, Fung, and Charles (2003) argued that older adults tend to move toward emotionally meaningful goals than do younger people, and they are likely to focus on the present than the unknown future. In addition, the oldest old is likely to reduce social network as they age; therefore, their networks are composed of emotionally close family relationships including spouses, parents, siblings, and children. Particularly, their relationship with children could strongly affect well-being among the oldest old.

Third, life satisfaction and depression in the oldest old group might be explained by factors related to the subjective perception of life (i.e., family relationship satisfaction or self-esteem) rather than objective conditions of life (i.e., gender, marital status, education, chronic disease). This finding implies that the rate of change in well-being increases when the oldest old perceive themselves as healthy and have high self-esteem. These results are supported by the gerotranscendence perspective that developmental adaptation process continues with age (Gondo, Arai, & Hirose, 2014; Tornstam, 2011). However, for the young-old, the objective conditions of life had an effect on their life satisfaction and depression. That is, marital status, which is dependent upon having a spouse, was found to be a significant factor. In addition, variables, such as persistent chronic diseases, receiving a public pension, receiving a livelihood support service, gender, religion, and level of education, also appeared to be significant. In sum, the oldest old may more likely be affected by the subjective cognitive domain rather compared with the young-old.

Finally, based on the gerotranscendence theory, life satisfaction and depression appeared to continuously improve in both groups. However, for the oldest old, subjective domain variables, such as family relationship satisfaction, self-reported health, and self-esteem, significantly affected life satisfaction and depression, while individual difference variables, that is, objective life conditions, such as gender, education, and religion, did not have a significant impact. The results suggest that the oldest old not only face death but also experience continuous growth from a gerotranscendence perspective. The continuous growth of the oldest old appeared to be affected by subjective factors but not by objective life conditions, which were significant for the young-old.

This study has some implications. First, older adults are not a homogeneous group. At the least, the oldest old and the young-old should be differently perceived with respect to gerontology policies and programs. In addition to the current policies, separate programs for the oldest old should be developed. For example, based on the gerotranscendence theory, well-dying programs should be actively expanded (Wang, Lin, & Hsieh, 2011). Further, the gerotranscendence theory, as well as the socioemotional selectivity theory, should be more actively accepted and applied in counseling services. On the other hand, financial interventions, such as job creation, should be more focused for the young-old group.

In addition, the results of the individual growth model and effect size analysis showed that the level of family relationship satisfaction significantly influenced the changes in life satisfaction and depression of the oldest old group over time. These findings offer a persuasive foundation that interventions related to family factors and the development of gerontological services and social policies for the oldest old are needed, especially in Asian countries with a still strong heritage of Confucian ideology.

Despite the study findings, this study has some limitations. First, variables such as social support and social participation activity types and frequencies, which could affect life satisfaction and depression in older adults, were not included in the research model. This is a limitation of using a secondary data set. Second, in analyzing the 8-year longitudinal data of the KWPS, attrition of analyzed cases occurred over time. Accordingly, despite finding no significant difference in the characteristics from Wave 1 to Wave 8 based on the homogeneity test, by including a control variable indicating whether participants dropped out of the survey, there is still a bias of dropout.

Footnotes

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 received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Minhong Lee

Author Biographies

Sang-Woo Lee, PhD, is a senior researcher of the Division for Social Policy Research at Korea National Council on Social Welfare, Seoul, Korea. His research interests are caregiving for older adults, mental health for older adults, and social innovation in social service settings.

Jae-Sung Choi, PhD, is a professor at the Department of Social Welfare at Yonsei University at Seoul, Korea. His research interests include social service planning & evaluation, non-profit management, older adults issues, and long-term care.

Minhong Lee, PhD, is a professor at the Department of Social Welfare at Dong-Eui University in South Korea and currently serve as the editor-in-chief of the Journal of Korean Gerontological Society. His research interests are caregiving issues for older adults with disabilities, elder abuse, and integrated program development and evaluation for the disabled older adults.

References

Andreescu

Chang

C. H.

Mulsant

B. H.

Ganguli

(2008). Twelve-year depressive symptom trajectories and their predictors in a community sample of older adults. International Psychogeriatrics, 20(2), 221–236. doi:10.1017/S1041610207006667

Berg

A. I.

Hoffman

Hassing

L. B.

McClearn

G. E.

Johansson

(2009). What matters, and what matters most, for change in life satisfaction in the oldest-old? A study over 6 years among individuals 80+. Aging and Mental Health, 13(2), 191–201. doi:10.1080/13607860802342227

Blazer

Burchett

Service

George

L. K.

(1991). The association of age and depression among the elderly: An epidemiologic exploration. Journal of Gerontology, 46(6), M210–M215.

Brittain

Kingston

Davies

Collerton

Robinson

L. A.

Kirkwood

T. B.

Jagger

(2017). An investigation into the patterns of loneliness and loss in the oldest old – Newcastle 85+ study. Ageing & Society, 37(1), 39–62. doi:10.1017/S0144686X15001142

Browne-Yung

Walker

R. B.

Luszcz

M. A.

(2017). An examination of resilience and coping in the oldest old using life narrative method. The Gerontologist, 57(2), 282–291. doi:10.1093/geront/gnv137

Burnham

K. P.

Anderson

D. R.

(2004). Multimodel inference: Understanding AIC and BIC in model selection. Sociological Methods & Research, 33(2), 261–304.

Carstensen

L. L.

Fung

H. H.

Charles

S. T.

(2003). Socioemotional selectivity theory and the regulation of emotion in the second half of life. Motivation and Emotion, 27(2), 103–123. doi:10.1023/A:1024569803230

Chai

H. W.

Jun

H. J.

(2017). Relationship between ties with adult children and life satisfaction among the middle-aged, the young-old, and the oldest-old Korean adults. The International Journal of Aging and Human Development, 85(4), 354–376. doi:10.1177/0091415016685834

Chen

Short

S. E.

(2008). Household context and subjective well-being among the oldest old in China. Journal of Family Issues, 29(10), 1379–1403. doi:10.1177/0192513X07313602

10.

Chen

Cohen

(2006). Using individual growth model to analyze the change in quality of life from adolescence to adulthood. Health and Quality of Life Outcomes, 4(1), 410. doi:10.1186/1477-7525-4-10

11.

Chen

Jordan

L. P.

(2018). Intergenerational support and life satisfaction of young-, old- and oldest-old adults in China. Aging & Mental Health, 22(3), 412–420. doi:10.1080/13607863.2016.1261798

12.

Chin

W. W.

(1998). Issues and opinion on structural equation modeling. MIS Quarterly: Management Information Systems, 22(1), 7–16.

13.

Cho

Martin

Poon

L. W.

(2015). Successful aging and subjective well-being among oldest-old adults. The Gerontologist, 55(1), 132–143. doi:10.1093/geront/gnu074

14.

Cho

M. J.

Kim

K. H.

(1993). Diagnostic validity of the CED-D (Korean version) in the assessment of DSM-III-R major depression. Journal of Korean Neuropsychiatric Association, 32(3), 381–399.

15.

Choi

B. M.

(2008). A comparison of human development stages between Confucius and Maslow. Journal of Far East Social Welfare, 4, 7–28 (in Korean).

16.

Choi

J. S.

(2009). The effects of job security of workers with severe disability on life satisfaction: Mediating effect of job satisfaction and moderating effect of discrimination in the workplace. Social Welfare Policy, 36(4), 343–367 (in Korean).

17.

Chou

K. L.

Chi

(1999). Determinants of life satisfaction in Hong Kong Chinese elderly: A longitudinal study. Aging & Mental Health, 3(4), 328–335. doi:10.1080/1360786995610

18.

Chou

K. L.

Chi

(2003). Reciprocal relationship between social support and depressive symptoms among Chinese elderly. Aging & Mental Health, 7(3), 224–231. doi:10.1080/136031000101210

19.

Cohen

(1992). A power primer. Psychological Bulletin, 11(1), 155–159.

20.

Erikson

E. H.

(1959). Identity and the life cycle. New York, NY: Norton.

21.

Gondo

Arai

Hirose

(2014). Wellbeing in the oldest old and centenarians in Japan. In B. L. Thomas, Kirkwood & C. L. Cooper (Eds.), Wellbeing in Later Life: A complete reference guide (vol 4, pp. 1–12). John Wiley & Sons, Inc. doi: 10.1002/9781118539415.wbwell113

22.

Gonyea

J. G.

(2010). The policy challenges of a larger and more diverse oldest old population. In Hudson

R. B.

(Ed.), The new politics of old age policy (3rd ed., pp. 155–182). Baltimore, MD: Johns Hopkins University Press.

23.

Henderson

A. S.

Korten

A. E.

Jacomb

P. A.

Mackinnon

A. J.

Jorm

A. F.

Christensen

Rodgers

(1997). The course of depression in the elderly: A longitudinal community-based study in Australia. Psychological Medicine, 27(01), 119–129.

24.

Heo

W. G.

(2017). Life satisfaction development trajectory and health characteristics factor for the elderly. The Korean Journal of Social Welfare Policy, 44(2), 297–318 (in Korean).

25.

Huang

J. F.

Wong

R. H.

Chen

C. C.

Mao

I. F.

Huang

C. C.

Chang

W. H.

Wang

(2011). Trajectory of depression symptoms and related factors in later life: A population based study. Journal of Affective Disorders, 133, 499–508. doi:10.1016/j.jad.2011.04.048

26.

Jeon

H. S.

Dunkle

R. E.

(2009). Stress and depression among the oldest-old: A longitudinal analysis. Research on Aging, 31(6), 661–687. doi:10.1177/0164027509343541

27.

Kashy

D. A.

Donnellan

M. B.

Burt

S. A.

McGue

(2008). Growth curve models for indistinguishable dyads using multilevel modeling and structural equation modeling: The case of adolescent twins’ conflict with their mothers. Developmental Psychology, 44(2), 316. doi:10.1037/0012-1649.44.2.316

28.

Kassem

A. M.

Ganguli

Yaffe

Hanlon

J. T.

Lopez

O. L.

Wilson

J. W.

, Study of Osteoporotic Fractures (SOF) Research Group. (2018). Anxiety symptoms and risk of dementia and mild cognitive impairment in the oldest old women. Aging & Mental Health, 22(4), 474–482. doi:10.1080/13607863.2016.1274370

29.

Kaup

A. R.

Byers

A. L.

Falvey

Simonsick

E. M.

Satterfield

Ayonayon

H. N.

Yaffe

(2016). Trajectories of depressive symptoms in older adults and risk of dementia. JAMA Psychiatry, 73(5), 525–531. doi:10.1001/jamapsychiatry.2016.0004

30.

Korea Institute for Health and Social Affairs & Seoul National University. (2013). Korean Welfare Panel Study: User’s guide (in Korean). Seoul, South Korea: Author.

31.

Krause

(2004). Lifetime trauma, emotional support, and life satisfaction among older adults. The Gerontologist, 44(5), 615–623. doi:10.1093/geront/44.5.615

32.

Lee

(2014). A longitudinal data analysis on depressive symptoms of Korean older adults. The Journal of the Population Association of Korea, 37(1), 10–130 (in Korean).

33.

Lee

R. P.

Ruan

Lai

(2005). Social structure and support networks in Beijing and Hong Kong. Social Networks, 27(3), 249–274. doi:10.1016/j.socnet.2005.04.001

34.

McCamish-Svensson

Samuelsson

Hagberg

Svensson

Dehlin

(1999). Social relationships and health as predictors of life satisfaction in advanced old age: Results from a Swedish longitudinal study. The International Journal of Aging and Human Development, 48(4), 301–324. doi:10.2190/GX0K-565H-08FB-XF5G

35.

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), S74–S82. doi:10.1093/geronb/58.2.S74

36.

Moon

J. R.

Huh

Song

Kang

I.-S.

Park

S. W.

Chang

S.-A.

Jun

T.-G.

(2017). The center for epidemiologic studies depression scale is an adequate screening instrument for depression and anxiety disorder in adults with congenital heart disease. Health and Quality of Life Outcomes, 15, 176–182. doi:10.1186/s12955-017-0747-0

37.

Nam

S. H.

Moon

S. W.

Lee

K. J.

(2012). The 2012 Korea Welfare Panel Survey: In-depth report (in Korean). Seoul, South Korea: Korea Institute for Health and Social Affairs.

38.

Oates

D. J.

Berlowitz

D. R.

Glickman

M. E.

Silliman

R. A.

Borzecki

A. M.

(2007). Blood pressure and survival in the oldest old. Journal of the American Geriatrics Society, 55(3), 383–388. doi:10.1111/j.1532-5415.2007.01069.x

39.

Park

Son

Bae

(2009). A longitudinal analysis on the prospective change of life satisfaction among elderly. Social Science Research Review, 25(3), 1–24 (in Korean).

40.

Pullmann

Allik

Realo

(2009). Global self-esteem across the life span: A cross-sectional comparison between representative and self-selected Internet samples. Experimental Aging Research, 35(1), 20–44. doi:10.1080/03610730802544708

41.

Ried

L. D.

Planas

L. G.

(2002). Aging, health, and depressive symptoms: Are women and men different? Journal of Women's Health, 11(9), 813–824. doi:10.1089/15409990260430963

42.

Rosenberg

(1979). Conceiving the self. New York, NY: Basic Books.

43.

Smith

Borchelt

Maier

Jopp

(2002). Health and well-being in the young old and oldest old. Journal of Social Issues, 58(4), 715–732. doi:10.1111/1540-4560.00286

44.

Stallones

Marx

M. B.

Garrity

T. F.

(1990). Prevalence and correlates of depressive symptoms among older US adults. American Journal of Preventive Medicine, 6(5), 295–303.

45.

Statistics Korea. (2018a). Life table. Retrieved from http://kostat.go.kr/portal/korea/kor_nw/1/2/7/index.board?bmode=read&aSeq=371860

46.

Statistics Korea. (2018b). Population projections. Retrieved from http://kosis.kr/statHtml/statHtml.do?orgId=101&tblId=DT_1BPA002&conn_path=I

47.

Sung

(2013). Effects of family and household’s economic factors on depression: Panel regression analysis using fixed-effect model. Social Science Research, 24(3), 207–228 (in Korean).

48.

Tornstam

(2011). Maturing into gerotranscendence. Journal of Transpersonal Psychology, 43(2), 166–180.

49.

Wang

J. J.

Lin

Y. H.

Hsieh

L. Y.

(2011). Effects of gerotranscendence support group on gerotranscendence perspective, depression, and life satisfaction of institutionalized elders. Aging & Mental Health, 15(5), 580–586. doi:10.1080/13607863.2010.543663

50.

Yeung

G. T.

Fung

H. H.

(2007). Social support and life satisfaction among Hong Kong Chinese older adults: Family first? European Journal of Ageing, 4(4), 219–227.