Education in A New Place and Social Support Are Needed to Improve The Health of North Korean Defectors: A Cross-Sectional Study

Background: The number of North Korean defectors (NKDs) escaping to South Korea has increased. The health status of NKDs is an essential factor for a successful settlement into South Korean society. However, no studies have been conducted on the health status of NKDs in terms of education and social support. The aim of this study was to determine the effects of education and social support on the self-rated health status among NKDs. Methods: This study utilized data gained from face-to-face interviews with 126 NKDs. A multivariable logistic regression and path analysis were performed to assess the effects of education in South Korea and social support with their self-rated health status and to explore the complex relationships between direct and indirect effects of the variables. Results: NKDs who did not experience regular education in South Korea responded that they were in poor health compared to their counterpart (OR=5.78). Although a direct association between education in South Korea and self-rated health was not shown, there was an indirect path from education in South Korea to self-rated health through social support. Conclusions: Participation in regular education is important for the health status of NKDs. Social policies and NKD assistance programs should consider and reect the combination of education and social support interventions relevant to the health status of NKDs.

morbidity and mortality are well established [5][6][7]. High education is shown to protect against a decline in self-rated health (SRH) [8,9], chronic disease [10], and mortality [11]. As another essential component of human life, social support has been known to be associated with health and well-being. Social isolation entails a mortality risk comparable in magnitude to that of smoking, a sedentary lifestyle, obesity, and alcohol abuse [12].
Previous evidence suggests that education and social relationships interact with each other to affect health. Some research in this area has found that the protective effects of social support on health are stronger among individuals with high education levels [13,14]. These studies have demonstrated that those with high levels of education bene t most from the protective qualities of social support, and the social support-health connection is almost nonexistent among those with lower levels of education.
Conversely, one study found that the protective qualities of social support are stronger among individuals who are less educated [15]. Yet, additional empirical evidence has demonstrated that social support is predictive of worse, rather than better health outcomes among individuals with a low level of education [13,14].
Education is not only a determinant of health but also an opportunity to form social relationships.
Especially for NKDs, being a part of a social community in South Korean society, such as a regular education in South Korea, provides them with good opportunities to receive social support and may have an important health-promoting impact. Previous studies [16,17] showed that facilitating the visit of educational institutions for refugees is an essential factor for successful social inclusion.
However, literature is scarce on the exact nature of the links among education in South Korea, social support, and self-rated health status among NKDs. We hypothesized that education in South Korea, social support, and self-rated health status would be directly associated with each other. We also hypothesized that education in South Korea would be directly and indirectly associated with self-rated health status through social support. Therefore, the aims of this study was to identify the individual effects of education and social support on the self-rated health status, respectively, and then to further assess the path from education in South Korea to self-rated health status through social support among NKDs.

Methods
This was a cross-sectional study to assess the association of education and social support with selfrated health. It was approved by the Institutional Review Board of the Seoul National University School of Dentistry (No.S-D20170041).

Participants
The study participants were recruited with cooperation from the Hana Center, alternative schools for NKDs and the Catholic Bishop's Conference of Korea National Reconciliation Committee in Seoul and Gyeonggi Province. 60 participants were required for each group with alpha 0.05, power 0.8, and difference of prevalence rate 25%. In consideration of the drop-out rate, we included 69 participants in each group. A total of 137 participants were recruited from November 2017 to February 2018, and the purpose and method of this study were explained to them. All the participants voluntarily agreed to join this study and gave their written consent. The criteria for selection and exclusion were as follows.
-Selection criteria · those who completed education at the Hana center · those who can read and answer the questionnaire by themselves or who can listen to the explanation of the questionnaire and answers · those willing to participate in this study -Exclusion Criteria · those with a severe mental illness condition · those who want to quit Variables The dependent variable was self-rated health status. It was measured with a 5-point scale (very good, good, moderate, poor, and very poor) and reclassi ed into two groups (good/moderate and poor).
Independent variables were education in South Korea and social support. Education in South Korea was reclassi ed into 1) college or more, 2) middle/high school, and 3) no education experience in South Korea. Social support score was measured by the Korean version of the Duke-UNC social support questionnaire which was developed by Suh et al. [18] The tool consists of 13 questions with a 5-point scale. After summing up, the score ranges from 13 to 65 points, and a higher score indicates more social support. The reliability of the Korean version of the Duke-UNC social support questionnaire was a Cronbach's =0.89 at the time of its development. Social support was categorized as a lower/higher group based on the median score (34 point).
Confounding variables were age (continuous), gender (male/female), living with family (yes/no), residence period in South Korea (continuous), smoking (yes/no), and drinking (yes/no). Residence period in South Korea was calculated by months from entrance date into South Korea to October 2017 when the data collection was completed.

Statistical analysis
After excluding missing values, 126 participants were included in the nal sample for analysis. The characteristics of the participants were analyzed using frequency, percent, means, and standard deviations. Chi-square tests for categorical variables and independent t-tests for continuous variables were used to assess the associations of education in South Korea and confounders with subjective poor health ( Table 1). The association between social support and self-rated health status (Supplementary Table 1 and Fig. 1) and the association between social support and education in South Korea (Supplementary Table 2 and Fig. 2) were evaluated by analysis of variance (ANOVA) and analysis of covariance (ANCOVA) controlling for confounders, respectively. The association between education in South Korea and subjective poor health was assessed using logistic regression models adjusting for the above mentioned confounders (Model 1 in Table 2). Social support was regarded as a mediator, and logistic regression analysis was done controlling for confounders in model 1 and social support (Model 2 in Table 2). The role of the social support was evaluated by the percent (%) excess odd explained, which was calculated as [(OR adj confounders -OR adj confounders + mediator ) / (OR adj confounders -1) X 100 (%)] in this study [19]. The % excess odd explained represents the degree to which a mediator explains the relationship between education in South Korea and subjective poor health. Finally, to assess the direct and indirect effects of the education of NKDs in South Korea on their health, we set up a model that hypothesizes that education in South Korea affects their SRH directly and indirectly by their perceived social support. A path analysis was conducted using AMOS 23.0 [20]. Path coe cients were estimated by maximum likelihood estimation. Maximum Likelihood Bootstrapping was conducted to test the signi cance of the indirect effect with bootstrap samples of 2,000 and 95% bias-corrected con dence intervals. Model t was evaluated with proper indices.  Additionally, 78.6% of the participants completed a regular educational curriculum in South Korea. The average score of social support was 33.9 with a minimum score of 13 and a maximum score of 51 points. Subjective health conditions showed that "good" and "moderate" were 38.9%, respectively, which surpassed "poor" (22.2%).
ANOVA and ANCOVA results are presented in Tables 2 and 3. The association between social support and self-rated health status is shown in Table 2. Social support scores increased as self-rated health status deteriorated. Individuals with at least a college degree reported a lower social support score (that is, a high social support status) than those who had less than a college degree (Table 3). In an adjusted logistic regression analysis controlling for age, gender, living with family, residence period in South Korea, and smoking and drinking, those who did not experience a regular educational curriculum in South Korea showed 5.78 odds of self-reported poor health (Model 1 in Table 4, p-value = 0.024). This association was attenuated to 3.95 in model 2 (p-value = 0.096). The % excess odd explained for no educational experience in South Korea was 38.23% when social support was included in model 2. We evaluated the path model of direct and indirect effects between the three variables of interest, namely education in South Korea, social support, and self-rated health status along with a control variable age.
As shown in Fig. 1 and Table 5, the path from Education in South Korea to SRH was not signi cant, in other words, the direct effect of education was not supported (p = 0.064). Paths from education to social support (β = 0.30, p = 0.001) and from social support to health (β = 0.30, p = 0.001) were all statistically signi cant, and the strength was big as the effect of age on health (β =-0.31, p = 0.001). Thus, education in South Korea did not directly affect SRH but did indirectly affect it through social Support. The t of the path model was assessed using the indices: χ 2 /df = 2.64; Normed Fit Index (NFI) = 0.945; Comparative Fit Index (CFI) = 0.961; and Standardized Root Mean Square Residual (SRMR) = 0.0107. All the indices were above the recommended criteria for good or t [21].

Discussion
This study analyzed the association of receiving education in South Korea and social support with selfreported health status among NKDs and the link among those three variables. This is the rst study that observed the independent main effect of receiving education in South Korea on health among NKDs. Our ndings indicate that receiving education in South Korea is positively associated with self-rated health status, even when potential confounding factors are taken into consideration. Like other populations, educational attainment is a signi cant predictor of morbidity, mortality, and other indices of physical health among NKDs [22][23][24]. Receiving education in South Korea among NKDs is also signi cantly associated with social support. Those with college education in South Korea, relative to those without, had a higher level of social support. It is evident that education in South Korea is associated with psychological aspects of health that may be considered as an important part of social relationships.
Our ndings also indicate that social support is positively associated with better health. The buffering hypothesis suggests that social support affects health by attenuating the physiological effects of psychosocial stress [25]. Social support has been shown to have both buffering effects and direct bene cial effects on general morbidity [26,27]. Evidence suggests that socioeconomic status (SES) and social relationships can interact with each other to affect health. Some studies have found that the protective qualities of social support are stronger among individuals with a low SES [15,28]. Social support is particularly important for individuals with a lower SES. However, some research reports that the protective effects of social support on health are stronger among individuals with a high SES [13,14,29]. This study's nding further supports that not only receiving education in South Korea is associated with the health status of NKDs-independence of socio-demographic factors and health behaviors-but that also social support has an important role in the association between education in South Korea and selfrated health among NKDs. However, we are not clear on the exact pathway from education to health which remains to be elucidated.
The hypothesis that is the most interesting to us is the potential effects of receiving education in South Korea on self-rated health either through social support or not. Our nding showed that education in South Korea did not directly affect their health, but education leads to social support, which affects their health. In other words, education affects health indirectly through social support. This result is not consistent with other studies [5][6][7][8][9] that argued a direct education effect on health. However, the signi cant path from education to social support and the one from social support to health were in agreement with previous studies. Previous evidence has supported that greater social support is associated with more education [30,31]. Formal education may promote interpersonal skills and friendships and provide more resources (e.g., income and leisure time) to develop and maintain relationships [32]. Social support may also promote healthier lifestyles by enhancing individuals' motivation, normative environment, and access to information [33][34][35]. The signi cant indirect effect of education on health through social support is in the same line with studies showing that education and social relationships interact with each other to affect health [13,14]. It is noteworthy that the strength of this indirect effect of education (0.25) and direct effect of social support (0.30) on health is comparable to the strength of the age effect (-0.31) on health. There could be many other mediating factors between education and health. In this study, we proposed social support as one of the factors. Further study is recommended to discover other factors that link education and health. Therefore, based on our ndings, there is a need for a policy enforcing education and social integration activities in the NKD group for their health.
This study has limitations. First, the selection of study participants was restricted. There are 33247 NKDs throughout South Korea. However, the participants in this study were selected by a snowball sampling method mainly from the residents in the Seoul and Gyeonggi metropolitan areas. Therefore, the study participants may not be considered as a representative sample of the NKD population. This led to a wide range for the con dence interval in the results of the logistic regression analysis in this study. Therefore, the interpretation of the study results should be performed carefully. However, considering the limited opportunities to contact NKDs and restrictions on information collection from NKDs, these limitations are inevitable. Second, because this study was a cross-sectional survey research, it is di cult to infer causality among the variables. A well-designed longitudinal follow-up study will be needed to nd underlying factors affecting the health status of NKDs on which proper health promotion and intervention strategies and programs can be developed. Third, although the questionnaires used in this study had proven validity and reliability, it was still challenging for some NKDs with different social and cultural backgrounds to fully understand and answer the questionnaires. It is necessary to develop tools to better measure the health status of this vulnerable group.

Conclusions
Despite the aforementioned limitations, this study has some vital ndings. The NKDs encounter challenges in adapting to a new political and social system in South Korea. To assist NKDs to overcome these challenges and improve their health condition, social integration policies and interventions focusing on enrollment in regular education curriculums/programs and participating in social interaction activities are essential. Further research should develop related intervention programs for NKDs and assess their outcomes.

Consent for publication
Informed consent was obtained from all participants.

Availability of data and materials
The datasets used and/or analysed during the current study are available from the corresponding author on reasonable request.