Introduction

With the acceleration of urbanization and economic development, a number of Chinese residents have chosen to migrate for work in order to seek better job opportunities. However, the long-lasting and highly peculiar system in China is a major obstacle to labor mobility, as it strictly restricts people’s entitlement to basic social benefits, which leaves them facing a number of challenges, including health care, education, instability of work, and inability to access services, outside of one’s birthplace; thus, people have to leave their children in their hometown or entrust them to other relatives and guardians (Wang & Pan, 2020). Under this context, the “left-behind children” (LBC) phenomenon has emerged in China. According to the latest definition of the State Council of China, left-behind children are rural children (rural household registrations) younger than 16 years of age who have two migrant parents or one migrant parent and another parent who has no guardianship ability (China, 2016). According to data, there were 6.436 million LBC in China at the end of 2020.

Due to the changes in the family upbringing structure, LBC show lower levels of academic performance, psychological well-being, character building, and other characteristics (Lin, 2003, Liu et al., 2018). (Lempinen et al., 2018) collected data in 1989, 1999, 2005, and 2014 from 3,749 participants and found that 20% of children reported often feeling lonely. At the same time, their physical growth, social interaction, and especially mental health are adversely affected (Zhang, 2014, Wright & Levitt, 2014). They face the common issue of parental absence caused by parents working outside the home for long periods, which breaks the normal family structure and affection. The characteristics of this parental absence often generate considerable negative effects on LBC emotional adaptations (Cheng et al, 2010, Wen & Lin, 2012), such as loneliness, depression, anxiety, and other unpleasant emotions or negative emotions (Cicchetti et al., 2010, Fan, 2011), among which loneliness appears most frequently (Antia et al., 2020).

Loneliness has been described as a perceived deficiency in one’s social relationships and the accompanying emotional distress, such as sadness, emptiness, or longing (Asher, 2003). Loneliness can be divided into emotional and social loneliness (Weiss, 1973, Cosan, 2014, Ditommaso, 1997). Emotional loneliness is marked by the absence of intimate emotional attachment with others; it leaves a feeling of emptiness, longing, and alertness. When children lose their parents as attachment figures, they are at risk of experiencing emotional loneliness (Ouellette, 2004). Loneliness not only affects children’ s cortisol level, sleep quality, and physical health but also often leads to depression, anxiety, and cognitive decline in children and reduces their level of self-esteem and subjective well-being (Liu et al., 2010, Maes et al., 2016), thus influencing their mental health. More severe cases may even lead to suicidal behavior (Simon & Walker, 2018, Mishra et al., 2018). An increasing amount of evidence has shown that LBC is closely related to loneliness, and the loneliness level of LBC is higher than that of non-LBC (NLBC) (Tang et al., 2019, Ren et al., 2017, Ji & Zhuo, 2017). This phenomenon has become a significant social problem and has received extensive attention from the government and media (Zhang, 2010).

Previous meta-analyses focusing on LBC in China suggested that parental migration had negative impacts on children’s mental health; however, there is still a lack of systematic research on the differences in loneliness between LBC and NLBC. Considering the harmful impact of loneliness on the development of LBC, it is crucial to conduct a systematic and comprehensive meta-analysis to synthesize and compare the loneliness status of LBC and NLBC and to gain a better understanding of the current status of LBC mental health. This study uses a meta-analysis to statistically combine the outcomes of several studies in order to explore the relationship between LBC and loneliness and provide a reference for promoting the mental health of LBC.

Materials and methods

Search strategy and study selection

The first author (XY) and the third author (LH) conducted an extensive automated search of electronic articles in the China National Knowledge Infrastructure, Wanfang, VIP Journals of Chinese Science, PubMed, and Web of Science (Science Citation Index, Social Sciences Citation Index, and Chinese Science Citation Database) to identify literature published from its establishment to September 30, 2022. The following search strategy and keywords were used: (“left-behind children” OR “left-behind students” OR “left-behind teenagers” OR “stay-at-home children”) AND (“loneliness”). All research was limited to English- and Chinese language articles.

The steps of literature screening were as follows:

  • (i) Delete duplicate studies; (ii) read titles and abstracts for preliminary literature screening; (iii) read full text for secondary screening; (iv) obtain the final set of included studies.

Eligibility criteria

Two researchers (the first two authors: XY and LXL) independently selected and reviewed the articles if they met the following criteria: (i) the research participants were LBC; (ii) the articles were empirical investigations of Chinese left-behind children’s loneliness; (iii) NLBC were set up as a control group; (iv) the articles provided sufficient statistic information for the calculation or estimation of effect sizes (e.g., as sample size, mean value, standard deviations, t-value, F-value, or p-value).

Studies were excluded if any of the following criteria were met: (i) case reports, reviews, lectures, editorials, conference articles or correspondence letters; (ii) duplicate studies, with only the most recent studies with the most complete dataset being ultimately selected; (iii) no control group, or the norm was used as the standard control; (iv) the original data were incomplete, or the supplementary data were not available.

Statistical analysis

We conducted a meta-analysis using Revman and Stata software (Revman 5.4 version). We used the odd ratio (OR) to assess the difference in loneliness between LBC and NLBC. Because of the different score units used by the papers, we calculated the standardized mean difference (SMD) to assess the effect size. The confidence interval was set at 95%. P < 0.05 was considered statistically significant. We also conducted a sub-group analysis based on different grades, guardians, and measurement tools The heterogeneity among the results was tested using the I2 statistic and was considered to be statistically significant at I2 greater than 50%. The fixed-effects model was used if there was no heterogeneity; otherwise, we used the random-effects model. A funnel plot was utilized to detect potential publication bias, If the scatter in the funnel plot is symmetrical, there is no publication bias. Finally, the forest map was drawn using GraphPad Prism 9.

Results

Data extraction

The following information was extracted from each eligible study: the first author, publication year, measurement tool, and total LBC and NLBC numbers. The data for each mean and standard deviation of the loneliness score of the LBC were also extracted according to education level, guardianship type, and measurement tool. A total of 300 potentially relevant articles were initially screened in the databases based on our literature searching strategy. 62 of which were subsequently removed due to duplication; 238 papers were selected for further screening. After screening based on titles, abstracts, and full texts, we included 23 studies in the final analysis (Fig. 1).

Fig. 1
figure 1

PRISMA diagram of included studies in the meta-analysis

Full size image

Characteristics of the studies and methodological quality

A total of 23 articles (Fan et al., 2016, 2017, Xie et al., 2015, Fan et al., 2014, Fan & Wu, 2020, Zhao, 2013, Zhao et al., 2013, Fan et al., 2009, Fan, 2011, Zhang, 2011a, b, Liu et al., 2008, Hou & Xu, 2008, Zhang et al., 2022, Yang et al., 2016, Zhang et al., 2014, Ling et al., 2012, Yue & Lu, 2015, Wang et al., Shen , 2011, Yue et al., 2014, Li et al., 2020, Xiong et al., 2019) were included in this study, all of which were cross-sectional studies, with a maximum sample size of 1,606 and a minimum sample size of 124. The age and grade ranges of the sample were 4–16 years old and grades 3–9, respectively, and a total of 6,678 LBC were included. All studies screened subjects and participants. The characteristics of the 23 studies included in this review are shown in Table 1.

Table 1 Characteristics of the studies and methodological quality
Full size table

Methodological quality assessment for all included studies was performed using the quantity table modified by (Downs & Black, 1998) and (Ferro & Speechley, 2009). The scale originally had 15 items, “Do the staff, locations, and facilities of the study patients represent the treatment that the majority of the patients receive?” was deleted from the revised Quality Assessment Scale by Ferro and Speehley because it is not suitable for assessing LBC with loneliness. The revised assessment scale has 14 items on four aspects of methodological quality: the (i) report (e.g., “Are the assumptions / objectives of the study clearly described?”); (ii) external validity (e.g.,“Are participants asked to participate in a study representing the entire population? How are they being recruited?”); (iii) internal validity (e.g.,“Are the primary outcome measures used valid and reliable?”); (iv) efficacy (“Does the study provide sample size or efficacy calculations to detect differences in probability values by chance < 0.05?” ).Each entry is scored as 0 (no/uncertain) or 1 (yes). The total score ranges from 0 to 14, with a higher score reflecting better literature quality. The 23 studies were scored, and the results showed that 4 studies had an average quality index, while the remaining 9 studies all scored 10 points and above, with a good quality index, indicating that the quality of the included literature was good.

Loneliness measurement tools

The scales used in the included analysis were all loneliness scales. The most common loneliness measurement scales are the Children’s Loneliness Scale (CLS) and the UCLA Loneliness Scale. The CLS is a child loneliness scale compiled in 1984 by Asher. The scale contains 24 items, 16 of which are used to measure children’s loneliness, with 10 positive scoring questions and 6 reverse scoring questions. Additionally, eight questions about children’s preferences were inserted to detect whether children lie when filling out the questionnaire. The items are scored on a 5-point scale, with 1 indicating “never” and 5 indicating “always”. The higher the final score, the stronger the loneliness.

The UCLA scale is a one-dimensional loneliness scale compiled by Russell et al. 1987 (Liu, 1999) to measure loneliness as the difference between the desired and real levels of social interaction. There are 20 items in the full scale, rated on a 5-point scales, among which 9 items are scored in reverse order. The higher the score, the higher the degree of loneliness. In most experimental studies in China, loneliness scale compiled by foreign scholars (such as the UCLA loneliness scale) have been used as the evaluation standard of experiments, but this scale is more suitable for adolescents rather than children compared with the CLS. Therefore, most related studies on loneliness among LBC at home and abroad use the CLS scale.

Effect size and homogeneity tests

The obtained mean effect size of the 23 studies included in this meta-analysis for exploring the effectiveness of levels of loneliness was SMD = 0.22, with a 95% confidence interval (CI) of 0.15–0.28. The results indicated that the loneliness level of LBC was higher than that of NLBC (Z = 6.41, P < 0.001). The heterogeneity among the 23 studies was relatively large (I2 = 65%, P < 0.01). According to the result of the random effects model, we found that the reasons for the heterogeneity may be complicated, and we believe that further research is indispensable (Fig. 2).

Fig. 2
figure 2

Forest map of loneliness analysis

Full size image

Subgroup Analysis

For further analysis and to identify sources of heterogeneity, we performed subgroup analysis in terms of the grades, guardianship types, and measurement tools of the LBC. The results showed high heterogeneity in both the primary school group (I2 = 63%, P < 0.001) and the junior high school group (I2 = 72%, P < 0.001). The total effect size between the two groups (I2 = 0%, P = 0.93) indicated that there was no significant difference in the effect size between the primary and middle school groups; that is, there was no significant difference in the degree of loneliness between the two groups.

In terms of the guardianship types, the SMD scores for other relative guardianship were higher than those for single-parent guardianship (SMD = 0.24, 95% CI = 0.17, 0.31), and the total effect size (I2 = 78.5%, P = 0.03) between the two groups indicated significant differences; that is, there were differences between the LBC of single-parent guardianship and other relative guardianship. According to the analysis results, LBC under single-parent guardianship were less lonely than those under the guardianship of other relatives. Moreover, the total effect size between the two groups of measurement tools (I2 = 0%, P = 0.62) indicated that there was no significant difference in the effect size between the CLS and UCLA, meaning there was no difference in the degree of loneliness measured by different measurement tools (Table 2).

Table 2 Results of subgroup analysis
Full size table

The results of the subgroup analysis showed that the degree of loneliness among LBC in both primary and middle school groups was high, which may mean that there were other regulatory variables in the assessment of loneliness between the two groups. Guardianship type subgroup analysis results showed that the loneliness of LBC under single-parent guardianship was lower than that of LBC under other relative guardianship. The study results are consistent with those of (Huang & Li, 2007) and others, who believed that parents play an irreplaceable role in childhood, even if only one parent is present, as they provide more dedicated care than other caregivers, offer more psychological comfort, and can better promote children’s mental health.

Publication bias

A funnel plot was applied to evaluate the publication bias of studies. Analysis of publication bias showed symmetric distribution (Fig. 3). Taken together, the calculated overall effect size seems to be rather robust and show a lack of publication bias.

Fig. 3
figure 3

Publication-bias funnel plot

Full size image

Discussion

The results of this meta-analysis suggest that the loneliness level of LBC is higher than that of NLBC (Z = 6.41, P < 0.001), and the detection rate of loneliness among LBC was 22%. The government has repeatedly stressed the need to establish and improve the care service system and rescue and protection mechanism for LBC to ensure their mental health, indicating that the mental health of LBC has been of highly concern to the country and society (China, 2016). Our study showed that LBC had worse loneliness issues than NLBC, the results are similar to previous studies. According to previous studies, the level of loneliness of LBC is affected by many factors, such as peer relationships, school atmosphere, frequency of contact with parents, and age at separation from parents. The improvement of peer relationships will reduce LBC’s level of loneliness; the higher the peer acceptance of children and the number of close friends, the lower the level of loneliness (Uruk & Demir, 2003). In addition, a good school atmosphere is negatively related to the loneliness of LBC, with the support of teachers and students being protective factors against the emotional behavior problems of LBC. In addition, (Zhang & Feng, 2011) showed that the better the teacher–student relationship, the lower the loneliness level of LBC. High-frequency, high-quality interaction between parent and child can improve adolescent emotional adaptation; otherwise, it is not conducive to adolescent mental and emotional health, which can lead to the emergence of loneliness (Larranaga et al., 2016). The younger the LBC are separated from their parents, the more loneliness symptoms appear (Zhang et al., 2018); likewise, the longer the separation from their parents, the more negative emotions appear (Lu et al., 2015). Loneliness can have a range of detrimental effects. Evidence suggests that a chronic and painful state of loneliness may contribute to a constellation of psychosocial problems, including depression, low self-esteem, conduct problems, and suicide attempts (Chai et al., 2019). These findings also explain why the loneliness of LBC must be given high attention.

Previous studies found that LBC with other relative guardianship reported higher levels of loneliness than those with single-parent guardianship (Yue & Lu, 2015), whereas other studies found no difference in loneliness between these two groups (Su et al., 2013). Our results indicate that LBC under other relative guardianship reported higher levels of loneliness than those under single-parent guardianship, similar to the results of (Wu et al., 2019). The following factors may explain this phenomenon. First, Since one or both parents are absent, the interaction between LBC and their parents is not conducive to the formation of a secure attachment relationship, which will negatively affect the physical and mental development of LBC (Al-Yagon et al., 2016). Based on the empirical analysis of logit model, (Liu et al., 2020) divided LBC into different types and found that children left behind by two parents were more lonely than those left behind by one parent. Second, for single LBC, the guardian not only assume multiple social roles, such as raising children, household care, labor production, and supporting the elderly, but also worries about the safety of the partner, resulting in mental pressure, which leads to a lack of patience for childrearing, with the parenting being simple, rough, easy to anger, and more often abusive (Ye & Wang, 2006). All of the above conditions may adversely affect the psychology of LBC.

Comparing the loneliness scores between primary and junior high school LBC, we could not find any significant differences. This means that loneliness does not decrease with grade or age. Considerable evidence has confirmed that the psychological problems of left-behind children are long-term. Because childhood and adolescence are critical periods for psychological and physical development, disrupted parental attachment during these critical periods may result in long-term psychopathological consequences that do not dissipate over time (Xu et al., 2018).

In addition, according to I2 of the subgroup analysis, we found that the heterogeneity was reduced only by the type of care, so we can speculate that different types of care may be the source of heterogeneity, which may be due to different guardians leading to different LBC loneliness scores, ultimately leading to a higher heterogeneous combined effect size. Additionally, according to the results of the funnel plot, the research results were relatively stable, and there was no obvious publication bias.

In summary, many LBC have a high levels of loneliness. The results suggest that social authorities and public health organizations should be aware that children left behind have a high level of loneliness. There is an urgent need for more care services to assess and manage these children. If possible, a psychological guidance plan should be developed according to the level of loneliness problems of each child.

Limitations

There are some limitations that need to be considered in the current meta-analysis. First, language bias should not be ignored. All included studies were Chinese studies because there were no English-language studies on this issue. Second, although publication bias was not detected in most of our results, it could still exist because only published articles were included. Third, there was high heterogeneity in this meta-analysis. In addition, the impact of gender and other unclear potential covariates on the loneliness of LBC could not be estimated because these data were unavailable in the eligible articles. Despite its limitations, this study has important implications for future research and practice to promote the health of LBC. This study can help bring more attention to LBC from society and academia, so as to formulate relevant policies to solve LBC’s mental health problems, reduce their loneliness and other issues, and promote their positive development.

Conclusion

In summary, the findings of this meta-analysis suggest that (i) LBC have a high level of loneliness, which requires the attention of relevant departments; (ii) there is no significant difference between the loneliness of primary and junior middle school LBC; (iii) other relative guardianship is more likely to cause loneliness in LBC than single-parent guardianship. Future research should develop individualized teaching and corresponding intervention measures for LBC of different ages and guardianship types; and (iv) there are still some limitations to this study; thus, further and more extensive research is needed in this field.