Left-behind Experience, Parent-child Communication and Psychological Resilience: a Structural Equation Modeling Analysis

Background: Rapid urbanization has caused left-behind children to become a social problem in China. Psychological resilience is an important mediating/intermediate variable. This study aims to examine the role of parental migration status and parent communication on the psychological and related behavioral status of left-behind children through psychological resilience. Methods: A cross-sectional survey was conducted in Anhui Province of China, and a questionnaire survey was conducted with 1992 teens using the Chinese version of the Parent-Adolescent Communication Scale, Connor-Davidson Resilience Scale, and Strength and Diculties Questionnaires. Results: Compared with the never left-behind group, left behind children had relatively lower PACS, CD-RISC and SDQ scores. Absence of parents is related with poorer psychological resilience, while good parent communication is related with better psychological resilience. Better psychological resilience is related with fewer psychological problems among different parental migration status. The currently left-behind status demonstrated a negative inuence on psychological resilience, while never left-behind status had a positive effect. Conclusions: Interventions are needed to enhance psychological resilience of left-behind children to prevent psychological and related behavioral problems.

how parental migration status and parent-child communication play a role in the mental health of leftbehind children through psychological resilience. This study stands out from prior studies by using structural equation modeling (SEM) in the analysis of three parental migration status models, thus enabling the comparison of associations between the different outcomes and the measurement of indirect effects.

Methods
Participants and procedure A cross-sectional survey was conducted in April 2018. The participants were recruited from Wuwei and Nanling of Anhui province, which are relatively underdeveloped counties of southeastern China and have large numbers of left-behind children. We randomly selected two towns from each county, and then randomly selected one primary school and one middle school from each town. In each school, students were recruited if they met the following criteria 1) enrolled in either grade 5 to 6 in primary school or Grade 7 to 8 in junior middle school, 2) were 11-17 years old, 3) had a local household registration in either the results of con rmatory factor analysis found an acceptable t of a model (χ 2 = 2597.81, df = 146, χ 2 /df = 17.79, RMSEA = 0.07, CFI = 0.88) [27][28].

Measurement of children's psychological resilience
The psychological resilience of children was measured using the Chinese version of the Connor-Davidson Resilience Scale (CD-RISC) [29]. The English version consists of 25 items which assess psychological well-being in ve dimensions: tenacity, tolerance of negative affect, positively acceptance of change, control and spiritual in uences. The Chinese version kept these 25 items, and revised items in three dimensions, namely tenacity, strength, and optimism. Each item is scored on a 5-point Likert scale from 1 (never) to 5 (often). Each dimension was measured by the summed score of its items as a sub-scale. A total score was calculated by summing the scores of three dimensions (thus range from 25 to 125). Higher scores indicate a better psychological resilience. By one study carried out in the south province of China, the Cronbach alpha of this Chinese version scale is 0.928, and test-retest reliability (r=0.812, P 0.001) and split-half reliability (r=0.890,P 0.001) were good [30].

Measurement of psychological and behavioral status of children
The psychological and behavioral status of students were evaluated by the Chinese version of Strength and Di culties Questionnaires (SDQ) [31][32]. The scale contains ve dimensions: emotional symptoms, conduct problems, hyperactivity, peer problem, and pro-social behavior. Each dimension has ve items, and each item is scored on a 3-point Likert scale from 0 (totally noncompliant) to 2 (fully compliant). The total di culties score is the sum of all dimension scores except the pro-social dimension, and a higher score re ects increased severity of emotional and behavioral problems. The reliability test from a study in the southeast province in China, showed that the Cronbach coe cient was 0.79, and the validity test showed the differences of the factors and total scores between abnormal sample and normal sample were signi cant (t=2.07-6.31, P<0.05) [32].

Covariates
Covariates in this study included gender, age, grade, sibling and self-reported family economic status relative to others in their community (much better off/better off, the same, poorer/much poorer). Most current studies divide the dissemination of their survey into two categories: left-behind children and never left-behind children. In order to have a more sensitive comparison of the status of parental migration, children in our study were divided into three groups: G1 (currently left-behind children); G2 (previously leftbehind children); G3 (never left-behind children). They were asked to answer two questions: "has your father (or mother) taken a job away from your hometown and been absent for over six months?" The options were "yes, currently absent," "yes, previously absent," and "no, never." If one or both parents were currently absent, the student was de ned as G1; if one or both parents were previously absent, the student was de ned as G2; and if neither parent was ever away, the student was G3.

Statistical analysis
Data were analyzed using SPSS 23.0(SPSS Inc., Chicago, IL, USA). First, sample characteristics were compared by chi-square test (for categorical variables) or analyses of variance (for continuous variables) among the three groups of children with different parental migration statuses. Second, structural equation models were used to evaluate the mediation hypothesis, and test that parental migration status and parent communication mediate psychological and related behavioral status through psychological resilience. Because these scores were not normally distributed, the differences in PACS and CD-RISC scores among the three parental migration status groups were compared by Kruskal-Wallis test. Third, we set up three mediated models divided by parental migration status (G1, G2 and G3) in order to explore the effect of parental migration status and parent-child communication on mental health through psychological resilience. These models were tested with the SPSS AMOS22.0 software using the Maximum Likelihood (ML) iteration procedure. The t indices for a good model included:1 ratios of χ 2 value to the degrees of freedom of between 2 and 5; 2) comparative t index (CFI) and Tucker-Lewis Index (TLI) >0.95; and 3) root mean square error of approximation (RMSEA) < 0.05. Data were examined for normal distribution and statistical tests were two-tailed.

Results
Of the 1992 participants, 1251 were in G1, 473 were in G2, and 268 were in G3. Demographic characteristics are shown in Table 1. More than half of the subjects were male. The ages of the three groups are similar. More than half of the subjects were in grades 7 or 8. More than 60% reported their income level as "the same" as their community. Over 65% had siblings. About 70% of subjects described their parent's education level as secondary school or below. (Table 1)  There were signi cant differences in the PACS, CD-RISC, and SDQ scores among the three groups of children with different parental migration status ( Table 2). The total PACS scores were both highest in G3 (mother 56.02 ± 8.91; father 57.27 ± 10.11) and lowest in G2 (mother 54.07 ± 9.21; father 54.85 ± 10.52). The total CD-RISC score was highest in G3 (85.45 ± 17.14), and lowest in G1 (81.26 ± 15.83). The total SDQ score was highest in G1 (12.74 ± 5.37), and lowest in G3 (11.13 ± 5.09). The SEM results showed that the t for each of the three models were acceptable. All of CFI and TLI > 0.93, and RMSEA < 0.074 (see Table 3). Figure 1-3 show the direct effects of parental migration status and parent-child communication on psychological resilience. In G1, currently left-behind status exhibited a negative effect (SSCs=-0.08, p < 0.001) on CD-RISC, and PACS scores (father SSCs = 0.21, p < 0.001; mother SSCs = 0.23, p < 0.001) were positively associated with CD-RISC scores. Currently left-behind status had no signi cant correlation to PACS scores (p > 0.05). In G2, PACS scores (father SSCs = 0.21, p < 0.001; mother SSCs = 0.23, p < 0.001) were positively associated with CD-RISC scores. Previously leftbehind status has a negative effect (SSCs=-0.06, p < 0.001) on PACS scores, and no signi cant correlation to CD-RISC scores (p > 0.05). In G3, never left-behind status plays a positive role (SSCs = 0.06, p < 0.01) on CD-RISC scores, and PACS scores (father SSCs = 0.20, p < 0.001; mother SSCs = 0.23, p < 0.001) were positively associated with CD-RISC. Never left-behind status had no signi cant correlation to PACS (p > 0.05). In all three models, the direct effect of CD-RISC on SDQ was signi cant (SSCs=-0.21, p < 0.001)  [22]. It is worth noting that children whose parents were previously absent demonstrated the lowest PACS scores, this is the rst study to report such a nding. It is possible that the experience of parental absence at a young age affects the formation of effective parent-child communication processes. This study also demonstrated that children with currently migrating parents have the lowest CD-RISC and the highest SDQ scores. This is consistent with other research ndings, namely that left-behind children were signi cantly more likely to display externalizing and internalizing problems [33][34].
The structural equation model shows how parental migration status and parent-child communication affect the children's psychological and related behavioral characteristics through psychological resilience. Our ndings suggest that psychological resilience is a mediating factor among parental migration status, parent communication and psychological problems. The absence of parents has a negative effect on psychological resilience, while good parent communication has a positive effect. Our data also suggests that better psychological resilience can reduce psychological problems among different parental migration statuses. It seems that resilience may play a key role in helping left-behind children to maintain psychological wellbeing. There are few studies which use structural equation models to analyze the mediating role of psychological resilience, and this is the rst study to report such a nding by comparison of the three parental migration status models. Ye et al. reported that resilience was found to be a protective factor for depressive symptoms and also mitigated the effects of peer victimization on depressive symptoms among rural-to-urban migrant children in China [35]. Therefore, resilience-based interventions might be useful to enhance the mental health of left-behind children, especially for currently left-behind children.
Our results showed that currently left-behind status had a negative in uence on psychological resilience, while never left-behind status had a positive effect on psychological resilience. These results suggest that separation from parents do play a role in children's mental health. This is echoed in other research on the subject: children who were separated from parents at a younger age had more symptoms of anxiety and depression [36][37]. While most studies in this area divided participants into left-behind children and never left-behind children, this study divided participants into currently left behind, previously left-behind and never left-behind groups. This allows for greater speci city in different statuses of parental migration.
This study demonstrated that parent-child communication has almost the same effect on psychological resilience under different parental migration status, and this conforms with previous research on this subject. We also found that communication with mothers was slightly more correlated with psychological resilience than was communication with fathers. Communication between parents and children is usually considered as an important factor, and similar ndings have also been reported in other studies. Van et al. (2015) reported that parent-child communication was a promising factor to focus on in interventions aimed at preventing mental illness, and Elgar et al. (2013) reported that parent-child communication during family dinners had 13-30% positive effect on mental health [38][39].
Contrary to prior studies, the current study shows that there is no consistent correlation between parental migration status and parent-child communication. We found only that being previously left-behind had a slightly negative correlation to parent-child communication. Most of prior studies indicated that the absence of parents does effect communication duration and frequency, and creating a lack of social support for their children and leading to psychological problems [1,17,40]. We speculate that children who experience being left behind in their early childhood have changes in their ability to communicate with their parents. And currently left-behind children are relatively older, and have a xed communication with their parents. As better communication conditions (such as phone calls, video chats, etc.) in China, communication between migrant parents and left-behind children is easier now relative to the past [23]. So lack of timely communication might be have less signi cant in uence as previously reported, and this topic warrants further research. However, this study did not measure the exact time and length of parental absence among children. Therefore, it is impossible to further judge the effect of separation length on parent-child communication. One study supports our ndings to some extent: Hedenbro and Rydelius (2019) indicates that early child-mother-father communication was related with children's social competence at the age of 15 [41].

Limitations
Firstly, this study did not measure the time and duration of parental absence, so we cannot compare the differences among children who were left behind at different stages of childhood. Secondly, the Parent-Adolescent Communication Scale measures subjective feelings, and children who are left behind for a long time will have a bias in this area. They may mistakenly report their present status of communication with their parents is normal and good, as it may be better than it was when their parent was absent, even if it does not live up to more common standards for parental communication among those whose parents have never migrated or migrated for shorter periods of time. Thirdly, our study considered only a limited range of potential determinants: we did not explore areas such as children's relationship with their caretakers and other related factors (i.e., family social capital, et al.)

Conclusion
This manuscript highlights that psychological resilience is the key mediating factor associated with parental migration status and parent-child communication. Better psychological resilience is related with fewer psychological problems among different parental migration status. To promote the health leftbehind children, interventions are needed to enhance psychological resilience, which may prevent psychological and related behavioral problems.