STRENGTHS AND LIMITATIONS OF THIS STUDY
-
The main strength of this study is its application of component data analysis, a valuable method for understanding the intricacies of time constraints, to examine the relationship between 24-hour activity behaviour and emotional and behavioural problems of left-behind children (LBC).
-
This study’s participants displayed remarkable compliance with the accelerometer use protocol, instilling confidence in the accuracy and consistency of the study measurements.
-
Due to its cross-sectional nature, the study did not allow us to establish a cause-and-effect relationship between 24-hour activity behaviour and emotional and behavioural problems.
-
The study’s sample, comprising elementary and middle school LBCs from Ningbo, Zhejiang Province, China, introduces a potential limitation in terms of generalisability.
Introduction
As industrialisation progresses and economic disparities between urban and rural areas in developing countries expand, an increasing number of workers are opting to migrate from their registered residential places in pursuit of better-paying jobs. This rural-to-urban migration has, unfortunately, led to the separation of workers from their children, as destitute living conditions in urban areas make it impractical to bring the children to the cities.1 These children are usually called ‘‘left-behind children’ (LBC), defined as children under the age of 16 residing in their hometown and being nurtured by caregivers other than their parents for a duration exceeding 12 months.2 The global number of LBC is not precisely estimated but is believed to be in the hundreds of millions. In China, more than one-third of rural children are considered left behind.3 Additionally, estimates point to 27% in the Philippines, 36% in Ecuador and over 40% in rural South Africa.4 The disruption in parent–child attachment caused by long-term separation from parents, along with insufficient monitoring and education, creates an interwoven situation that makes LBC prone to emotional and behavioural problems.5 When contrasted with their non-left-behind counterparts, LBC are more likely to experience a profound sense of loneliness, heightened anxiety and depression, and a decreased display of prosocial behaviours.5–8 The detrimental effects of emotional and behavioural problems transcend the boundaries of childhood, continuing to impact the lives of LBC well into their adult years. This enduring influence manifests in various forms, including persistent mental health struggles, engagement in criminal behaviour and substance abuse.9 10 It is crucial to understand the protective factors associated with emotional and behavioural problems of LBC.
Among the myriad factors impacting the emotional and behavioural problems of LBC, physical activity (PA) has emerged as a focal point of scholarly investigation. Several studies have emphasised that actively participating in PA can contribute to positive outcomes such as improved social adaptation, heightened feelings of happiness, and a reduced likelihood of mood disorders.11 12 Moreover, there is a significant link between children’s sedentary behaviour (SED) and symptoms of depression and anxiety.13 14 Particularly, the time spent on screen-based sedentary activities is associated with unfavourable behavioural conduct poor self-esteem in children.15 16 Time spent sleeping is also important for children’s emotional regulation, internalising and externalising behavioural problems.17 18 Previous research has predominantly focused on exploring the independent associations between PA, SED, sleep, and emotional and behavioural problems. Nonetheless, these behaviours are inherently codependent. From a 24-hour behaviour perspective, each day encompasses a finite allocation of movement behaviours, encompassing SED, light PA (LPA), moderate-to-vigorous PA (MVPA) and sleep.19 Altering the duration of one behaviour mandates a compensatory adjustment in other behaviours. Hence, it is crucial only to examine the impact of a singular behaviour on emotional and behavioural problems but also to evaluate the comprehensive impact arising from the amalgamation of various behaviours.
Compositional data analysis, developed for exploring interdependent data confined within a finite whole, provides a robust framework for examining the potential health implications associated with redistributing time across different movement behaviours.20 To date, some studies have used this statistical approach to explore the relationship between 24-hour activity behaviour and parameter such as body mass index (BMI),21 cardiometabolic biomarkers,22 motor skills23 and mental health24 in children. However, to our knowledge, no study has applied compositional data analysis to evaluate the potential relationship between time spent in various movement behaviours and emotional and behavioural problems specifically in LBC. In order to implement effective targeted interventions, gaining a deeper understanding of the associations between movement behaviours and emotional and behavioural problems is imperative. Therefore, this study aimed to use compositional data analysis to investigate associations between reallocations of time among time-use behaviours (SED, LPA, MVPA and sleep) and the manifestation of emotional and behavioural problems in LBC. We hypothesised that substituting SED with LPA and MVPA led to decreased scores for internalisation and externalisation problems, along with increased scores for prosocial behaviour in LBC.
Methods
We conducted a cross-sectional study and followed the Strengthening the Reporting of Observational Studies in Epidemiology checklist for reporting observational studies (online supplemental table 1).25
Supplemental material
Participant and data collection
In this study, LBC refers to children under the age of 16 residing in their hometown, with one or both parents migrating to other cities for more than 6 months. The study draws from a convenience sample, specifically selecting one elementary and one middle school for LBC from Ningbo, Zhejiang Province, situated in southern China. According to the Statistical Bulletin of Education Development in Zhejiang Province in 2023, there are 4938 elementary and middle schools in the province.26 These schools were selected based on their involvement in providing intensive education for children left at home as their parents work outside the hometown. Through our assessment of the schools’ basic situation, we discovered that over half of the enrolled children attending these schools are classified as LBC. During the sampling process, two classes were randomly selected from each grade in elementary school (grades 4–6, with students aged 9–12 years) and middle school (grades 7–9, with students aged 13–16 years), taking into account the potential difficulties that children below the fourth grade might encounter in completing the questionnaire. The study encompassed a total of 472 students distributed among 12 classes, all of whom were invited to participate and provide informed assent. 458 students, with a response rate of 97.0%, obtained consent from their parents or guardians to join the study. Spanning from February to May 2023, a total of 16 weeks were dedicated to the execution of this study. Each grade in both elementary and middle schools was allocated a dedicated 2-week period for data collection. At the commencement of the first week, research assistants distributed accelerometers to the students, ensuring proper wear through detailed instructions. Subsequently, in the second week, these research assistants collected the accelerometers and administered questionnaires to the students, focusing on aspects of emotional and behavioural problems and sociodemographic information (eg, age, gender and BMI).
Measurements
Waist-worn GT3X+triaxial accelerometers (ActiGraph, Pensacola, Florida, USA) were used to objectively measure children’s 24-hour activity behaviours (ie, SED, LPA, MVPA and sleep). The comprehensive algorithm and the validity of the accelerometer device have been thoroughly detailed in previous literature.27 Children were instructed to wear the accelerometers on their right hip, fastened with an elastic belt, for a duration of eight consecutive days, including during nighttime sleep. Due to the non-waterproof nature of the device, it was emphasised that removal was necessary during water-based activities such as bathing and swimming. The data were initialised, downloaded and processed using the ActiLife V.6.13.4 software program. A sampling rate of 30 Hz and 15 s epochs was used. To be included in the analytical sample, children’s accelerometer data needed to show valid wear for a minimum of 10 hours per day, covering at least 3 weekdays and 1 weekend day. Any periods of 60 consecutive minutes with zero counts were treated as non-wear and omitted from the analysis.28 Accelerometry-derived activity intensity levels were determined using Zhu cut-points: sedentary time (0–99 counts/min), LPA (100–2799 counts/min), moderate PA (2800–3999 counts/min) and vigorous PA (VPA, ≥4000 counts/min).29 The estimation of sleep duration employed a polysomnography-validated accelerometer algorithm, which relied on the distribution of change in the z-angle (ie, corresponding to the axis positioned perpendicular to the skin surface).30
The Strengths and Difficulty Questionnaire (SDQ) was used to measure the children’s emotional and behavioural problems.31 The Chinese version of the SDQ has good validity and reliability and has been widely used in China.32 It consists of five subscales: emotional symptoms, conduct problems, hyperactivity inattention, peer problems and prosocial behaviour. Each item is scored on a 3-point Likert scale (0=‘not true’, 1=‘somewhat true’, 2=‘certainly true’). The SDQ is extensively used, and in community samples, it is recommended for the broad constructs of internalising problems (emotional symptoms and peer problems), externalising problems (conduct problems and hyperactivity inattention) and prosocial behaviour.33 Higher scores indicate greater severity of internalising and externalising problems while the opposite is true for prosocial behaviour. Internal consistency of the SDQ in our sample was Cronbach’s α=0.793, 0.806 and 0.737 for internalising problems, externalising problems and prosocial behaviour, respectively.
A self-assessment questionnaire on sociodemographic characteristics was conducted to collect basic information on the participants, such as age, gender, BMI, sibling, family economic status, parental education levels and left-behind type. Family economic status was evaluated by dividing the annual household income by the number of individuals in the household, leading to the classification of families into low-income, middle-income and high-income categories.34 The classification of left-behind types comprised ‘one-migrating-parent children’ and ‘two-migrating-parents children’. The former denoted children cared for by a single parent during the left-behind period while the latter indicated children under the care of grandparents or other relatives during the same period.
Statistical analysis
Statistical analyses were conducted by using the R Statistical Software system (V.4.2.2) with a pre-established alpha level of 0.05. Descriptive statistics (mean and percentiles) were calculated to describe the characteristics of the participants. Compositional descriptive statistics, including compositional geometric means and variation matrix, were calculated using the R packages compositions and robcompositions.35 Compositional variation matrices were used to communicate the variation of the calculated pairwise log ratios (eg, the variation of ln (SB/MVPA)). A variation coefficient closer to 0 would indicate that there is higher codependency between two behaviours.
The associations between the activity composition and emotional and behavioural problems were explored using multiple linear regression models. Movement behaviours were expressed as isometric log-ratio co-ordinates (using the default ilr () transformation) and used as explanatory variables in the linear models. Sociodemographic covariates (age, gender, BMI, sibling, family economic status, parental education levels and left-behind type) were also included as adjusted variables. The outcome variables were internalising problems, externalising problems and prosocial behaviour. A χ2 type II test of the multiple linear regression models was used to assess the significance of the activity composition.36 Because parameter estimates for ilrs are difficult to interpret in the context of units of change in the raw behaviours, compositional isotemporal substitution was implemented. This facilitated the estimation of expected changes in children’s emotional and behavioural problem scores through a 15 min reallocation across all potential combinations of sleep, SED, LPA and MVPA.37 The procedure was repeated for all significant substitution effects between sleep, SED, LPA and MVPA for increments of 15, 30, 45 and 60 min. Predictions were plotted for reallocations of −60 to 60 min to aid interpretation where appropriate.
Patient and public involvement
None.
Results
Out of the original 458 participants, 26 (5.7%) had insufficient accelerometer data (less than 4 days of valid data), and 8 participants (1.7%) were missing data related to emotional and behavioural problems. After excluding these participants, 424 individuals remained. Among them, 275 (64.9%) were categorised as LBC and were eligible for further analysis. The characteristics of the analysed sample are presented in table 1. The participants had an average age of 12.9 years (SD=2.0), with 44% being boys. A majority of the participants (63%) have siblings, and 48% were identified as children with two migrating parents. The geometric means for different activities were as follows: sleep (596 min/day), SED (592 min/day), LPA (206 min/day) and MVPA (46 min/day). Sleep and SED each account for 41% of the entire 24-hour day while LPA takes up 14%, and MVPA comprises 3%. The mean scores for internalising problems, externalising problems and prosocial behaviour were 6.6 (SD=2.8), 6.7 (SD=2.6) and 5.9 (SD=1.9), respectively (online supplemental table 2) conveys the matrix of compositional variation, with coefficients ranging from 0.303 to 1.064. The highest coefficient was the log ratio variance between LPA and MVPA (ln=1.064) and the smallest was the log ratio variance between SED and LPA (ln=0.303).
Supplemental material
Participant characteristics
In the multiple regression analysis (online supplemental table 3), after adjusting for age, gender, BMI, sibling, family economic status, parental education levels and left-behind type, the proportion of time spent in sleep and LPA were negatively associated with internalising (β sleep=−3.63, p<0.01; β LPA=−6.96, p<0.001) and externalising problems (β sleep=−3.04, p<0.05; β LPA=−3.49, p<0.001). Time allocated to SED relative to the other behaviours was positively associated with internalising (β SED=4.24, p<0.001) and externalising problems (β SED=4.37, p<0.001). Negative correlations were observed between time spent in SED and LPA relative to other behaviours and prosocial behaviour (β SED=−2.14, p<0.05; β LPA=−2.08, p<0.001). Conversely, time spent in MVPA relative to other behaviours exhibited a positive association with prosocial behaviour (β MVPA=3.05, p<0.001).
Table 2 displays the predicted difference in internalising and externalising problems and prosocial behaviour when 15 min was reallocated from the behaviours in the columns to the behaviours in the rows, keeping the remaining behaviours constant. Adjusting for covariates, a 15 min reallocation from SED to sleep was associated with LBC experiencing 0.53 units lower internalisation problems and 0.31 units lower externalisation problems. After reallocating 15 min from SED to LPA, there was a decrease of 0.46 units in internalisation problems and a decrease of 0.23 units in externalisation problems. Moreover, reallocating the same duration from LPA to MVPA was linked to a 0.40-unit decrease in internalisation problems. Finally, reallocating 15 min from sleep, SED or LPA to MVPA was associated with predicted increases in prosocial behaviour of 0.81 units, 0.75 units and 0.82 units, respectively. None of the other possible time reallocations between movement behaviours showed statistical significance.
Estimated difference (95% CI) in emotional and behaviour problems associated with the reallocation of 15 min
Figures 1–3 show predicted differences in emotional and behavioural problem outcomes when incremental durations of time (ranging from −60 to 60 min) were added or subtracted from the most influential activity behaviour in table 2 and redistributed from/to one other activity, keeping all remaining activities constant. Reallocations from SED to sleep and LPA were associated with lower internalising problems, whereas reallocations from sleep and LPA to SED were associated with higher internalising problems. Estimated differences in internalising problems showed a non-linear association with an increase in time reallocated between LPA and MVPA. The positive differences associated with reallocating away from MVPA on internalising problems were greater than the negative differences associated with reallocating to MVPA (figure 1). Reallocations from SED to sleep and LPA were associated with lower externalising problems, whereas reallocations from sleep and LPA to SED were associated with higher externalising problems (figure 2). Reallocations from sleep, SED and LPA to MVPA were associated with better prosocial behaviour, but the estimated differences were non-linear. The predicted differences in prosocial behaviour were greater when MVPA was replaced with either sleep, ST or LPA, rather than when the time was reallocated to MVPA (figure 3).
Change in the estimated difference in internalising problems during the 24 hours activity time reallocation. LPA, light physical activity; MVPA, moderate-to-vigorous physical activity; SED, sedentary behaviour.
Change in the estimated difference in externalising problems during the 24 hours activity time reallocation. LPA, light physical activity; SED, sedentary behaviour.
Change in the estimated difference in prosocial behaviour during the 24 hours activity time reallocation. LPA, light physical activity; MVPA, moderate-to-vigorous physical activity; SED, sedentary behaviour.
Discussion
Through the component data analysis, this study delved into the intricate relationship between the distribution of 24-hour activity behaviour and the emotional and behavioural problems faced by LBC in China. The research had two main goals: (1) To explore the connection between the 24-hour activity behaviours of LBC and their internalising and externalising problems, along with prosocial behaviours and (2) To assess how redistributing time among activities correlates with changes in internalising and externalising problems and prosocial behaviour. The component regression analysis revealed that an increased allocation of time to sleep and LPA correlated with diminished internalising and externalising problems. Conversely, a higher duration of SED was associated with heightened internalising and externalising problems. Additionally, the time proportion devoted to SED and LPA demonstrated a negative correlation with the prosocial behaviour of LBC while a greater proportion of MVPA significantly fostered the development of their prosocial behaviour.
The literature extensively documents the connection between insufficient sleep and externalising problems like attention deficit hyperactivity disorder in children.38 39 Additionally, meta-analyses covering a wide age range, from 7 to 79 years, lay the groundwork for suggesting that insufficient sleep could also be linked to internalising problems, such as negative mood or challenges in emotion regulation, among children.40 This highlights the essential role of sleep in promoting emotional and behavioural problems among children. The beneficial impact of PA on children’s emotional and behavioural problems can be understood through three dimensions: neurobiology, psychosociology and behavioural mechanisms.41 Chinese scholars have conducted research indicating that PA can effectively alleviate internalising and externalising problems in LBC.11 12 However, an interesting finding in this study is the negative correlation between LPA and the prosocial behaviour of LBC. Notably, the accelerometer-based data used to measure PA lacked the ability to differentiate the context, such as whether the activities were done individually or in a group setting. Engaging in PA with others may be a more accurate predictor of their impact on prosocial behaviour.42 Extended periods of SED have been linked to an increase in internalising and externalising problems, along with diminished prosocial behaviour in LBC. Results from a longitudinal survey in the UK align with these findings, indicating that children spending over 3 hours daily on TV or video games were more likely to report emotional and peer relationship difficulties 2 years later.43 A systematic review investigating the relationship between SED and emotional and behavioural problems in children and adolescents establishes a direct positive correlation between screen-based sedentary time and heightened hyperactivity/inattention and internalisation problems.16 The association between SED and elevated internalising and externalising problems, as well as compromised prosocial behaviour, in LBC may highlight the potential negative impacts of excessive screen time. To enhance our understanding of the correlation between PA, SED, and emotional and behavioural problems in LBC, future research might consider incorporating ecological momentary assessment or using device-based pattern recognition methods. These approaches could help identify the context of PA and specific types of SED, offering a more nuanced and accurate portrayal of the relationship between active behaviour and emotional and behavioural problems.44
Isotemporal substitution revealed that substituting an average of 15 min of sleep or LPA for an equal duration of SED was associated with a noteworthy decrease in both internalisation and externalisation problem scores among LBC. These findings emphasise the crucial role of sleep and LPA in the prevention, control and intervention of internalising and externalising problems in this group. The log-ratio variance matrix in this study reveals a significant likelihood of conversion between SED and LPA (ln SED/LPA=0.303). This implies that actions such as encouraging students to leave their seats between classes or engaging in LPA like standing in place stretching during school hours could facilitate a positive transition between these behaviours. Moreover, reducing sedentary activities such as studying and screen time after school and substituting them with LPA may further support this conversion.45 The substitution of MVPA with LPA is believed to correlate with decreased internalisation problems in LBC. Physiological mechanisms suggest that MVPA, as opposed to LPA, may enhance the activation of the ascending reticular activating system responsible for arousal and sustaining wakefulness. This heightened activation leads to cortical excitation to a certain extent, triggering the inhibitory mechanism of the ascending reticular activating system.46 Consequently, there is a reduction in afferent stimulation to the body, resulting in diminished relaxation responses and emotional symptoms. There was an association between replacing MVPA with sleep, SED and LPA and improved prosocial behaviour in LBC. In contrast to sleep, SED and LPA, MVPA demands higher energy consumption and elicits more pronounced physiological and hormonal changes in the human body. This heightened level of PA plays a crucial role in improving emotional and behavioural problems, including boosting self-esteem, self-worth and mitigating feelings of depression and anxiety.47 Additionally, children engaging in MVPA, including activities like sports and structured exercises, may experience peer involvement and support from significant others. These social aspects are positively associated with enhanced prosocial behaviours.48
According to the component analysis of 24-hour activity behaviour and emotional and behavioural problems in LBC, a notable finding emerges: the mutual substitution effect of MVPA and other behaviours displays an asymmetry. Specifically, when MVPA replaces sleep, SED or LPA, there is a slow decrease in internalising problems and a gradual increase in prosocial behaviours. Conversely, substituting in the opposite direction results in rapid changes in the opposite direction. The observation of this asymmetry in health outcomes when substituting MVPA for other active behaviours aligns with previous findings.49 50 This discrepancy can be attributed to the relatively modest proportion of MVPA within the overall activity component, representing 45.56 min or 3.16% of the total activity time. Notably, subtracting 15 min from MVPA equates to a substantial reduction of nearly one-third (32.9%) of MVPA time in this study. In contrast, a 15 min reduction in sleep, SED or LPA time corresponds to only 2.5%, 2.5% and 7.3% of their respective total activity durations. Consequently, a 15 min alteration significantly impacts MVPA while minimally affecting other activities, leading to an asymmetry in their isotemporal substitution effects. The finding also implies that reducing SED could be crucial for enhancing the MVPA time among LBC. Alternatively, while maintaining the current level of MVPA, allocating additional time to sleep and LPA can compensate for the insufficient total MVPA time. This approach seeks to support the healthy development of LBC’s emotional and behavioural problems in all dimensions.
One of the major strengths of this study is its utilisation of compositional data analysis methods, a valuable approach to account for the constrained nature of time. Another commendable aspect is the study’s robustness, as evidenced by the comprehensive analyses that effectively control for numerous potential confounding variables. Furthermore, it is noteworthy that participants in the study exhibited remarkable adherence to the accelerometer use protocol, surpassing 20 hours per day. This high level of compliance underscores the reliability of the data obtained, providing confidence in the accuracy and consistency of the study’s measurements. Despite its strengths, it is important to recognise certain limitations in this study. First, as the research hinges on cross-sectional data, caution is warranted in making causal inferences. Second, the study’s sample, comprising elementary and middle school LBC from Ningbo, Zhejiang Province, China, introduces a potential limitation in terms of generalisability. Finally, this study did not assess the context and specific type of SED engaged in by participants. Different forms of sedentary activities could have varied effects on the observed relationships,51 and the absence of this detailed context is an important limitation to keep in mind when interpreting the findings.
Conclusions
In this study, compositional data analysis was employed to investigate the relationship between activity composition and emotional and behavioural problems in LBC. The findings indicate that redistributing time from SED to sleep and LPA is linked to improved internalisation and externalisation problem scores. Additionally, the substitution of MVPA for any other behaviours is positively associated with enhanced prosocial behaviours. It is worth noting that the differences observed were non-linear, with reallocations of time away from MVPA showing stronger adverse associations with internalising problems and prosocial behaviours. The study results suggest that, in the context of ample sleep, replacing SED with an active recess or incorporating diverse forms of LPA, and alternatively, integrating MVPA into both school and extracurricular physical endeavours, may contribute to the improvement of emotional and behavioural issues in LBC. Future studies should longitudinally examine associations between activity behaviours and emotional and behavioural problems within a representative sample of the target population.
Data availability statement
Data are available on reasonable request.
Ethics statements
Patient consent for publication
Ethics approval
This study involves human participants and was approved by the Ethics Committee of Zhejiang Normal University (No: ZSRT2022048). Participants gave informed consent to participate in the study before taking part.
This post was originally published on https://bmjopen.bmj.com