1. The Need for Research

In Korean society in the past, grandparents participated in the process of socialization of their grandchildren in the role of grandchildren's emotional stability, protection and discipline [1]. However, as the number of working women increases and women become more active in society, the role of caring for young children is expanding not only to the mother but also to the grandparents and the father. In particular, child care is often left to the child's grandmother or grandmother-in-law rather than to the child care facilities or others because of the problems of economic burden and trust in caregivers [2]. It is reported that grandmother is providing quality service at a much lower cost than any proxy service, and plays an important role in women's employment [3]. According to the 2014 survey of Korean older persons, 21.8% of elderly people aged 65 and over living together with their married children said that the reason of living together with their married children is to provide their children with housekeeping/grandchild raising help, and there were some differences by regions (8.7~49.0%) [4]. The elderly people aged 65 and over who have been directly brought up their grandchildren under 10 years old for the past year were 6.4%, which is significantly higher than 5.3% in 2008 [5].

As the function of socialization in the home weakens, grandparents have played important roles as tutor for the discipline of grandchildren, emotional supporter supporting the emotional security of grandchildren, family integrator maintaining and connecting the intergenerational relationship, and a guardian of their grandchildren [6]. Thus, improving the quality of life (QoL) of grandparents is a very important task not only for personal reasons but it affects the QoL of the family members that the grandparents belong to, the economic activity of the country, and the fertility rate as well.

In the meantime, the elderly are undergoing changes such as weakening of strength, loss of function, and occurrence of chronic diseases in physical changes [7], while grandchildren's care work for grandparents in their old ages is further deteriorating their physical health, and can cause negative consequences such as stress, depression, and physical burden [8]. On the other hand, there are some reports that raising grandchildren is an experience that gives fun and pleasure to the plain life of old age, and experiences that make them feel important in the family and society [9]. In this way, grandparents' care should be grasped and approached in a multifaceted way because it can have a positive and negative impact on the QoL of grandparents. Regarding the QoL in old age, physical activity positively affects the QoL regardless of age, activity level, and health [10]. In addition, parks are associated with increasing physical activity by providing a safe environment for elderly people who like to walk a short distance and allowing them to spend more time in the park [11], and parks are a crucial condition for the interpersonal leisure satisfaction of elderly people and improving their QoL [12]. To improve the QoL for the elderly, it is necessary to consider the environment that increases physical activity and leisure activities, and in particular, in case of grandparents caring for grandchildren, the community environment related to caring should also be considered. Furthermore, it has been reported that formal social support, such as services provided by institutions and local communities as community characteristics, can reduce depression in grandparents [13]. In Korea, some local governments have supported grandparents who raise their grandchildren by providing policy grants since 2011 [14].

An ecological model is an approach in which human behavior related to health is influenced by the environment and focuses on the interaction of its influences [15]. McLeroy and colleagues presented factors affecting human behaviors as five multi-layered levels including intrapersonal factors, interpersonal factors, organizational factors, community factors, and public policy factors [15]. Previous studies based on ecological models have applied ecological models to health promotion behaviors and leisure activities of elderly people using elderly general welfare centers or elderly people living in free elderly facilities [12, 16], but studies on the QoL of grandparents who raise their grandchildren based on ecological models are scarce. Thus, based on an ecological model, this study attempts to identify the specific predictors of the QoL of grandparents who raise their grandchildren by examining intrapersonal, interpersonal, and community factors.

2. Purpose of Research

This study attempts to investigate the factors affecting the QoL of grandparents caring for their grandchildren based on an ecological model, and following are the specific objectives.

First, this study aims to explore grandparents' intrapersonal factors (demographic characteristics, health related characteristics), interpersonal factors (family characteristics, social support system), community factors and QoL. Second, this study aims to investigate the difference of grandparents' QoL according to intrapersonal, interpersonal, and community factors and their relevance. Third, this study aims to identify intrapersonal, interpersonal, and community factors affecting the QoL of grandparents.

1. Research Design

This study is a descriptive correlation study using the raw data of the Korean Longitudinal Study of Aging (KLoSA) and secondary data from the National Statistical Office to investigate the factors affecting the QoL of grandparents caring for their grandchildren.

2. Research Participants and Data Collection Methods

The Korean Longitudinal Study of Aging (KLoSA), which was used in this study, was the data collected for the elderly people aged 45 years and over nationally excluding Jeju Island, for the purpose of investigating the actual conditions and behavior patterns of elderly people [17]. The contents of the survey were about the population, family, health status, employment, income and consumption, assets, subjective expectation and QoL, and when the panel subject died, the other family members were surveyed.

KLoSA's survey method was the Computer Aided Personal Interview (CAPI), interviewers with the notebooks visited the sample households and conducted computer interviews. The sampling frame is the enumeration districts of the Population and Housing Census in 2005, and 261,237 general area and apartment enumeration districts excluding the island area and the facility unit enumeration districts are determined as the sampling unit enumeration districts. Before sampling the enumeration district, the whole areas were stratified by 15 metropolitan cities, and then stratified by local government administrative districts called "Dong", "Eup", or "myeon", and finally stratified into the general house and the apartment enumeration districts within each administrative district. After that, the systematic sampling method was applied to extract the number of assigned samples. When the sampling area changes as the panel survey continued, the preliminary sample enumeration districts were utilized to keep 1000 sample enumeration districts. The 5th KLoSA sample in 2014 was 9,431 people, and 7,467 people were surveyed (response rate, 79.2%), including 438 deaths and 7,029 survivors. Among 7,029 living participants, 262 participants were answered "Yes" to the question "Have you helped raise your grandchildren under the age of 10?", and 38 people were excluded because of suspected dementia, the final number of participants included in this study was 224. As results of Mini-Mental State Examination (MMSE) consisted 19 items, if the cognitive function score is 24 points or more, it can be regarded as normal; the score of 18 points or more and 23 or less indicates a decrease in cognitive function; if the score belows 17 points, it can be suspected as dementia [18].

In addition, this study used local newspapers in the area and the National Statistical Office (NSO) data where the participants reside, which have the potential to affect the QoL of grandparents. The data on the grandchild caretaker support project was provided by the Gwangju Daily Newspaper [14], and the nursery supply rate by region was provided by the research report of the Korea Institute of Child Care and Education [19], and the number of libraries, the number of children's parks, the number of public sports facilities, the number of public heath institutions, and the number of the elderly leisure welfare facilities were provided by the data of the NSO [20, 21, 22, 23, 24].

The KLoSA data are the data downloadable from the employment survey site with personal information deleted, and the data of the NSO are publicly available. Therefore, this study was conducted after the approval of the study ethics committee (IRB No. 2016-0026), ensuring anonymity and confidentiality without harm to the participants.

3. Selection and Definition of Research Variables

4. Data Analysis Methods

In this study, the relationship between the QoL of grandparents and the ecological factors was analyzed using the PASW SPSS/WIN 23.0 program, and multilevel analysis was performed using STATA 13.1.

First, grandparents' QoL and intrapersonal factors (sociodemographic characteristics, health related characteristics), interpersonal factors (family characteristics, social support system), and community factors were analyzed using descriptive statistics such as percentages and averages.

Second, the difference and relationship among grandparents' intrapersonal factors, interpersonal factors, and community factors and QoL were analyzed using independent t-test, ANOVA, Pearson's correlation coefficient, and posttest of ANOVA was conducted using Scheffé method.

Third, multi-level regression analysis was conducted to identify intrapersonal factors, interpersonal factors, and community factors that affect the QoL of grandparents. This study conducted a two-level regression analysis. First, a null model, which includes only constants without independent variables, was established, and then the individual level model, which includes individual level predictor variables including intrapersonal and interpersonal factors (Model 1), and individual-community level model, which includes predictive variables at the individual level and the community level (Model 2), were established to analyze the fitness of the models and the fixed and random effects at the individual and community level.

1. Characteristics of the Participants

The participants of this study were 224 grandparents who had taken care of their grandchildren. The average age was 67.86±7.06, and 86 participants (38.4%) were below 64 years old and 138 participants (61.6%) were over 65 years old. There were more women than men (84.4%), 53.6 % resided in metropolitan cities and 46.4% residing in small and medium cities and rural areas.

In the health characteristics, 51 participants (22.8%) were satisfied with the subjective health status, while 111 participants (49.6%) perceived to be normal 62 participants (27.7%) perceived to be worse, and 107 participants (47.8%) had regular exercise. Also, the grip strength index was 21.74±6.34kg and the depression score was 3.08±2.64 points. The cognitive function score was 25.27±3.33, which was normal, but 70 participants (31.3%) of the participants had cognitive function score of 23 or less, demonstrating the cognitive function was decreased (Table 1).

Table 1 Characteristics of Intrapersonal Factors and Quality of Life according to Intrapersonal Factors (N=224) Click for larger image

2. Characteristics of Interpersonal Variables

In terms of the characteristics related to the family, the gross income of households was 2,750.12±2,251.08 (10,000 won) last year, and it was confirmed that the income gap per household was large, and 99 participants (44.2%) were living with their children. 98.2% of the children had occupations, and the number of children in care was 1.74±0.99. The mean duration of care for grandchildren was 3.41±4.25 years, with an average time of 59.39±59.13 hours per week.

In relation to the social support system, 141 participants (62.9%) frequently met close friends, but 83 participants (37.1%) rarely met close friends. It was confirmed that 47 participants (21%) had a religious meeting, 129 participants (57.6%) had a social gathering, and 13 participants (5.8%) had a cultural center or sports meeting that they could spend leisure time together. In the last year, 107 participants (47.8%) had travel, tourism, and outing experience, and 21 participants (9.4%) had participated in movies, performances and concerts (Table 2).

Table 2 Characteristics of Interpersonal Factors and Quality of Life according to Interpersonal Factors (N=224) Click for larger image

3. Characteristics of Community Variables

Among the characteristics of the 15 cities and provinces, there was one place (6.7%) in which there was a grandchild caretaker support project, and the average childcare center supply rate by region was 55.13±5.29%. The number of public libraries and small libraries compared to the population was 100.95±77.36, and the average number of children's parks per 100,000 population was 21.43±6.21. In addition, the number of public sports facilities per one million population was 364.97±168.18, and the number of public health centers, substation health centers, and health clinics combined per 100,000 population was 5.60±8.74. It was confirmed that the number of elderly welfare centers, senior citizens centers and silver colleges for the leisure of elderly people are 843.15±554.61 per 100,000 elderly people (Table 3).

Table 3 Characteristics of Community Factors and Quality of Life according to Community Factors (A=15, N=224) Click for larger image

4. Relationship between Major Variables and QoL

The score of average QoL of grandparents was 62.63±15.53 points (Table 1), and according to the residence area of the grandparents, the score was 62.00±17.47 points in large cities, 63.29±12.47 points in medium-sized and small-sized cities, and 63.64±14.98 points in eupmyeon districts. The relationship between individual characteristics, community characteristics, and QoL is shown in Tables 1, 2, and 3. Age was not statistically significant, and the QoL of men was higher than that of women (t=2.25, p=.026). With regard to the health related characteristics, the higher the perceived health status, the higher the QoL (F=29.81, p<.001), and the QoL was high when exercising regularly (t=3.75, p<.001). In addition, the grip strength index (r=.30, p<.001), the depression score (r=-.39, p<.001), and the cognitive function (r=.306, p=.001) were shown to be statistically significant. In terms of family-related characteristics, the higher the total gross income of households, the higher the QoL (r=.16, p=.020), whether to live with the children, whether or not for the child to have a job, and the number of grandchildren they cared for, caring period and weekly caring time did not show statistically significant difference. In addition, the more frequently they met close friends (t=2.87, p=.005) and the higher they participated in religious meetings (t=3.06, p=.003), their QoL was found to be high, and when they went on a trip or an outing last year (t=4.91, p<.001) and when watching movies and performances (t=2.46, p=.015), their QoL was found to be significantly high.

In terms of community characteristics, the QoL was significantly high when there was a grandchild caretaker support project (t=5.63, p<.001), and the supply rate of childcare centers was higher (r=.24, p<.001), the QoL was significantly high (r=.24, p<.001). In addition, the QoL was high when there were a large number of children's parks which children can use with their grandparents (r=.21, p=.001), and the more elderly welfare facilities available to the elderly over 65 years, the QoL was significantly high (r=.13, p=.047).

5. Effects of Individual and Community Factors on the QoL of Grandparents Caring for Grandchildren

The null model, the individual level model (Model 1), and the individual-community level model (Model 2) were sequentially tested to confirm the factors affecting the QoL of grandparents caring for their grandchildren. In the univariate analysis using significance level .05, variables showing significant differences in grandparents' QoL were included as the independent variables to be included in the multi-level analysis. Age was not significant in univariate analysis but was included in the final model as demographic characteristics. The Variance Inflation Factor (VIF) value for the correlation between independent variables before the multilevel analysis was 1.10~6.64, which was less than 10, confirming that there was no problem in multi-collinearity.

The null model is a model without explanatory variables and verifies whether there is regional difference in the QoL of grandparents. To examine the random effects of the null model, the two-level variance (τ=12.34) indicating the difference in QoL between regions was statistically significant (χ²=3.92, p=.024). Intra-class Correlation Coefficient (ICC) was 0.051 {12.34/227.72+12.34}, and 5.1% of the total variance was caused by regional differences. In other words, the QoL of grandparents caring for grandchildren can be influenced not only by individual factors but also by regional differences.

Model 1 is a model in which only independent variables at the individual level are input. We identified a regression model with explanatory variables such as age, gender, subjective health status, presence of regular exercise, grip strength index, depression score, cognitive function score, household gross income of last year, meeting with friendly people, religious meetings, last year's travel or sightseeing, experience of outing, participation in movies, performances, and concerts. As a result of analyzing the fixed effect on the QoL in model 1, the QoL was lower as the subjective health condition was worse (β=-9.82, p<.001), and the QoL was higher as household income increased (β=0.01, p<.001), there was a religious meeting (β =4.47, p=.043), and they went on a trip or outing last year (β=4.07, p=.035). Other variables including age, gender, regular exercise status, grip strength index, depression and cognitive function scores, the number of times they meet close friends, and participation in movies and performances last year were not found to be statistically significant.

In the random effect of Model 1, the p-value was .055 as a result of the likelihood ratio test of comparing between the model with random effect and the model without random effect, and there was no statistically significant difference at the significance level of .05. However, the ICC was still 6.34% even with individual level variables, it indicated the need for Model 2 to account for differences between regions. Therefore, the results of model 2 are obtained by additionally inputting the regional level variable as an independent variable. In the random effects of Model 2, the variance of the region was reduced to nearly zero, and the value of reduced variance was not statistically significant (p>.999). In other words, it can be said that the input of the community variables explained the difference between the regions well.

As for the fixed effects of Model 2, cases such as the subjective health status, last year's household gross income, religious gatherings, and travel or outings last year that had shown statistical significance when individual level variables were applied, were still statistically significant. At the community level, the higher the number of children's parks per 100,000 population (β=0.47, p=.035), the higher the QoL, but the grandchild caretaker support project, the nursery supply rate by region and the number of elderly leisure welfare facilities were not statistically significant (Table 4).

Table 4 Factors related to the Quality of Life among Grandparents Caring for Grandchildren Using Multilevel Analysis (N=224) Click for larger image