Study participants

This longitudinal study targeted students enrolled in grades 1 through 4 in Shanghai’s primary schools. Given the primary education system in Shanghai spans from grade 1 to 5 and the necessity for a year-long follow-up, the study was confined to students in grades 1 to 4. Shanghai had a total of 680 primary schools in 2021. For this study, five schools were contacted, and three schools agreed to participate. These three schools were located in two different administrative districts of Shanghai and were selected using convenience sampling based on their willingness to participate. These schools were contacted directly, and their participation was confirmed after discussions with school administrators.

The target sample size was based on the complexity of the cross-lagged panel model, requiring 10 to 20 participants per free parameter. With approximately 25 free parameters, the required sample size was estimated at 250–500 participants. To account for attrition, this was increased by 20%, resulting in a recommended sample size of 600 participants. Within each school, we initially targeted 4 (grades) × 3 (classes per grade) × 40 (students per class), for an estimated total of 1440 potential participants. After the follow-up and data cleaning processes, a final sample of 680 students was retained for analysis, which exceeds the minimum sample size required to achieve sufficient statistical power.

Ethical approval for this study was obtained from the Shanghai University of Sport Scientific Research Ethics Committee. The initial approval (approval number 102772021RT084) covered the baseline data collection, while an additional approval (approval number 102772022RT032) was granted to account for the follow-up component of the study. Consent was fully informed and obtained in written form from both the students and their guardians. Additionally, the principals of the participating schools provided written informed consent on behalf of their institutions. All procedures were conducted in accordance with ethical standards, ensuring the protection of all participants’ rights and privacy.

Measurements

The survey instrument comprised two sections, encompassing a total of 19 items. The initial section solicited demographic information, including school affiliation, grade, class, student ID, sex, date of birth, parental education levels, and socioeconomic status, amounting to 9 items. The subsequent section probed into the primary variables of interest: a) 24-hour movement behaviours, including MVPA (2 items), MSE (1 item), ST (2 items), and sleep duration (4 items); b) an assessment of LS (1 item).

The MVPA, MSE and ST items were adapted from the "Children and Adolescents Physical Fitness Index Survey Questionnaire (Student Version) [26]." The original questionnaire was developed in Chinese, and as the sample population in this study consisted of Chinese children, no translation was necessary. To reduce the burden on participants, particularly children, only 1 or 2 items from each construct were selected. These items were carefully chosen to reflect the core dimensions of physical activity and sedentary behaviour while ensuring clarity and manageability for younger respondents. The adapted items were validated using the same sample in this study, ensuring that the reliability (CR = 0.943) and structural validity (AVE = 0.615) remained satisfactory for our specific research context [26].

MVPA was assessed by asking, “In the past week, how many days from Monday to Friday did you engage in physical activity for at least 60 minutes?” and “In the past week, how many days from Saturday to Sunday did you engage in physical activity for at least 60 minutes?” PA was defined as “any physical activity that increases your breathing and heart rate,” aligning with the concept of MVPA. Participants responded with a range from 1 = 0 day to 6 = 5 days and from 1 = 0 day to 3 = 2 days, respectively.

MSE was surveyed through the question, "On how many days in the past week did you engage in MSEs such as push-ups, sit-ups, pull-ups, weightlifting, etc.?" Participants responded with a range from 1 = 0 day to 8 = 7 days.

ST was estimated based on "In the past week, how long per day from Monday to Friday did you engage in the following things in your free time? (1) Watching television; (2) Using cell phones, tablets, and other electronic mobile smart terminal products; (3) Using a laptop or desktop computer for Internet chatting, browsing pages, sending and receiving emails, doing homework, etc;" and "In the past week, how long per day from Saturday to Sunday did you engage in the following things in your free time? (1) Watching television; (2) Using cell phones, tablets, and other electronic mobile smart terminal products; (3) Using a laptop or desktop computer for Internet chatting, browsing pages, sending and receiving emails, doing homework, etc;" Participants responded with "1 = none, 2 = 0.5 hours, 3 = 1 hour, 4 = 2 hours, and 5 = more than 3 hours."

Sleep duration was calculated using the first four questions of the Pittsburgh Sleep Quality Index (PSQI), which assess usual bedtime, wake-up time, sleep latency, and sleep duration. Participants were asked to report their average bedtime and wake-up time over the past week, as well as the time it usually takes them to fall asleep. These responses were used to calculate their total sleep duration. The reliability and validity of the PSQI have been extensively validated in the Chinese population, demonstrating its suitability for this study [27].

LS was measured employing the Cantril Ladder, asks respondents to think of a ladder, with the most satisfied life for them being a 10 and the most unsatisfied life being a 0. The single-item life satisfaction scale has shown moderate to high reliability and validity across various samples and cultural contexts [28–30].

Parental education levels and socioeconomic status (SES) were included as covariates in the analysis, alongside sex and age, due to their potential influence on children’s life satisfaction and movement behaviours [31–35]. Parental education levels were assessed using a self-administered questionnaire, where participants reported the highest level of education attained by their father and mother, selecting from the following categories: “Below primary school level”, “Primary school”, “Junior high school”, “High school”, “Post-secondary school”, and “Bachelor level or higher”. Socioeconomic status was evaluated based on participants’ self-perceived economic condition. Respondents were asked to select the category that best described their family’s socioeconomic status from the following options: “Extremely wealthy”, “Relatively wealthy”, “Average”, “Not very well off”, and “Not at all well off”.

Data collection

To mitigate common method biases, procedural controls were implemented throughout the survey process. Participants were assured of the survey’s anonymity and confidentiality prior to distribution. During administration, students were encouraged to respond independently, with surveyors and teachers providing clarification on queries without influencing responses.

The survey was administered using the “https://www.wjx.cn/” platform. In the classroom, students were provided with a personal ID and the questionnaire link, which they then completed at home with parental assistance. The survey could be accessed and completed using any internet-enabled device, such as a computer, tablet, or smartphone. The initial survey (baseline) was conducted between 08/06/2021 and 14/06/2021, and the follow-up survey was administered from 15/06/2022 to 21/06/2022. For the follow-up, school teachers facilitated contact with students or their parents to ensure participation. Out of 879 collected questionnaires, 683 were deemed valid, yielding a validity rate of 77.70%.

Statistical analysis

Data entry and matching of two tests were performed using Excel 2016, followed by importing the data into SPSS 24.0 for descriptive analysis. Numerical variables were described using means ± standard deviation, while frequencies (percentages) were utilized for categorical variables. Missing data were handled by listwise deletion to ensure the integrity of the analyses. Subsequently, the Amos software was employed to model the cross-lagged relationships among MVPA, MSE, ST, sleep duration, and LS.

In the cross-lagged panel model (CLPM), each variable (MVPA, MSE, ST, sleep duration, and LS) was treated as both a predictor and an outcome at different time points. The autoregressive paths were specified to capture the stability of each variable across two time points (e.g., MVPA at Time 1 predicting MVPA at Time 2), while the cross-lagged paths were included to examine the reciprocal effects between different variables over time (e.g., MVPA at Time 1 predicting LS at Time 2, and vice versa).

In addition to constructing models for the overall sample, models for samples of different sex and grade levels were considered, with covariates such as sex, age, parental education, and socioeconomic status included to control for their potential confounding effects. During model construction, the Bollen-Stine bootstrap method was applied to address biases in multivariate normality, enhancing the validity of the model assessment, particularly in smaller subsamples [36]. Additionally, 5,000 resamples were generated via bootstrap methods, enabling precise determination of 95% confidence intervals for indirect effects [37, 38]. The structural equation model’s integrity is further reinforced by established metrics including the model goodness-of-fit index (GFI), adjusted goodness-of-fit index (AGFI), comparative fit index (CFI), and root-mean-square error of approximation (RMSEA), comprising a robust validation framework. The structural equation model’s fit was assessed using GFI and AGFI (≥ 0.90 for good fit), CFI (≥ 0.95 for excellent, ≥ 0.90 for acceptable), and RMSEA (≤ 0.06 for excellent, ≤ 0.08 for acceptable). Statistical significance was set as p < 0.05 (two sided).

In our analysis, the structure of the model, with its complex relationships and multiple paths, led to fit indices such as Chi-square (χ²), TLI, and SRMR indicating a perfect fit due to minimal degrees of freedom. Given the complexity and specific characteristics of our data, these indices are less informative in this context. Therefore, we focused on alternative fit indices (GFI, AGFI, CFI, RMSEA), which are more appropriate for assessing model performance under these conditions and demonstrate a satisfactory fit to the data.

MVPA

The cross-lagged analysis showed no significant predictive relationship between MVPA and LS in the overall sample, suggesting that other moderating factors might influence the relationship. This finding is consistent with Bae et al. (2017) who did not find a significant relationship between vigorous PA and LS [39]. This finding could be due to a variety of factors, including individual differences in interests, physical abilities, and personal preferences for types of activities. According to Headey et al. (2014), children’s happiness and LS are more closely linked to familial and social relationships rather than specific exercise types [40]. While numerous studies indicated that PA generally correlates positively with LS [11, 13], they often examine this relationship in isolation, neglecting the potential impact of ST and sleep duration on LS. The beneficial influence of MVPA on LS could be counterbalanced by the detrimental effects of excessive ST and inadequate sleep duration. This may also be attributed to differences in measurement methods. For instance, although both Alsubaie (2021) and this study utilized self-reported questionnaires to measure PA and LS, Alsubaie categorized responses to PA into three groups (physically inactive, moderately active, and highly active) and bifurcated LS responses into two categories ("satisfied" and "unsatisfied") [11]. In contrast, this study employed an eight-level self-reported scale (PA: 0–7 days) and an eleven-level scale (LS: 0–10 points). Therefore, variations in the design and granularity of the scales likely contributed to differing findings regarding the relationship between PA and LS.

When looking at sex- and grade- split, the relationship between MVPA and LS becomes more distinct. For females, MVPA positively influenced LS, whereas the influence was not significant in males. Similarly, in the study of Wold et al. (2013) [41], females benefited more than males in soccer. This could be attributed to the social and emotional benefits of PA, such as improved self-esteem, body image, perceived health and social interaction, which are particularly significant for females [41–43]. Furthermore, males and females may engage in physical activities with different motivations and levels of engagement, where males might prioritize competition and physical prowess, and females might seek social interaction and emotional well-being [44]. In older children (Level 2), a similar positive trend was observed. However, this positive association was not as pronounced in younger children. Older children might derive more satisfaction from MVPA due to increased autonomy, the development of specific skills, and the social aspects of participating in sports and physical activities [45, 46].

MSE

The cross-lagged model indicated that MSE did not significantly predict LS, suggesting a limited direct impact of these activities on children’s life satisfaction. Currently, less evidence can be used to prove the relationship between MSE and LS in children. The lack of a significant predictive relationship between MSE and LS among children can be attributed to several factors. Firstly, children’s perception of LS is influenced more by social and environmental factors than by specific types of PA. Furthermore, the enjoyment factor in physical activities is a significant determinant of LS among children. Activities that are perceived as fun and engaging are more likely to contribute to their overall happiness, as opposed to structured MSE, which might not always be enjoyable for them [47]. Lastly, according to the recommendations of the 24-hour Movement Guidelines, children and adolescents aged 5–17 should engage in MSE for no less than three days per week [16]. However, in the follow-up tests, the average MSE frequency for all children was 2.46 days. This may have resulted in an insufficient positive impact of MSE, thereby leading to a non-significant path coefficient between MSE and LS.

When dissected by sex, the relationship between MSE and LS showed notable differences. Specifically, for females, MSE was negatively associated with LS. This could be due to several factors. Firstly, social expectations and stereotypes around physical activities for females might influence their enjoyment and engagement in MSE [48]. Secondly, the competitive nature of these exercises could impose stress and pressure, particularly on females, affecting their overall LS [49]. This suggests that the way MSE is structured and perceived by females could be crucial in determining its impact on their well-being. The study also revealed that the impact of MSE on LS varied significantly between younger (Level 1) and older (Level 2) children. In younger children, MSE positively correlated with LS. This age group might perceive MSE more as play and an opportunity for physical exploration, which can be enjoyable and contribute to a sense of achievement [50]. In contrast, for older children, the relationship between MSE and LS was negative. As children grow, MSE might become more structured, competitive, and performance-focused, potentially leading to increased pressure and reduced intrinsic enjoyment, thereby negatively impacting LS. Further study should clarify the great difference in the relationship between MSE and LS.

ST

The cross-lagged analysis found no significant relationship between ST and LS across sexes and grade levels, indicating that screen time may not critically affect life satisfaction. The impact of ST on LS likely varies widely based on the type of screen activity (educational, entertainment, social interaction), the context (alone, with family or friends), and individual differences in preferences and tendencies towards screen use [51–53]. Furthermore, ST only exerts a negative impact on mental health when exceeding 120 minutes [53]. However, in the baseline assessment of this study, the average ST among students was 90.97 minutes, significantly below this threshold, potentially resulting in no substantial effect on LS.

Sleep duration

The cross-lagged model identified a positive correlation between sleep duration and LS, consistent across all sex and grade subgroups, highlighting sleep’s importance for well-being. This is consistent with previous research, where longer sleep duration is associated with higher LS [54, 55]. This relationship can be attributed to several key factors. Adequate sleep is essential for maintaining good physical health in children, including growth, immune function, and metabolic regulation. The secretion of growth hormones during sleep, particularly in children, contributes to better physical health, leading to increased activity levels and an overall improved quality of life [56, 57]. Furthermore, sleep’s critical contribution to cognitive processes, such as memory consolidation, learning, and attention, leads to improved concentration, problem-solving abilities, and academic performance in children [23]. These cognitive benefits foster a sense of achievement and self-esteem, crucial for LS. Additionally, adequate sleep plays a significant role in mood and emotional regulation, with well-rested children experiencing more stable emotions [21], thus contributing to overall LS. The study’s findings highlight the universal importance of adequate sleep duration, as adequate sleep influences not only emotional regulation and cognitive performance but also social interactions and behaviours. Children with sufficient sleep are better equipped for social engagement, reducing conflicts and enhancing social competence, which are essential for LS [57, 58]. Moreover, chronic sleep deprivation can lead to long-term health issues, including obesity, diabetes, and cardiovascular problems [59], underscoring the need for establishing good sleep habits in childhood to promote healthier lifestyles and reduce the risk of chronic diseases, thereby contributing to higher LS in the long term.

Study strengths and limitations

The application of cross-lagged panel analysis allowed us to explore the reciprocal and temporal relationships between movement behaviours and LS more comprehensively, offering a nuanced understanding of how these factors interact over time in different subgroups of children. However, the study faces limitations, including its convenience sampling from three Shanghai primary schools, limiting broader applicability. The reliance on self-reported data for PA, ST, and sleep duration could introduce response biases, and the single-item measure for LS, despite its reliability and validity, may not fully capture the multidimensional nature of this construct. Furthermore, the limited number of time points reduces the ability to establish causality, and there is a potential influence of unmeasured variables such as cultural factors, mental health status, and environmental influence.

A key limitation of cross-lagged analysis is its reliance on correlational data, which cannot establish definitive causality and may be influenced by unmeasured confounders. The method assumes linear and stable relationships over time, which may not capture the complexity of real-life dynamics. Additionally, the chosen time intervals may not align perfectly with the actual changes, potentially biasing the estimates. To address these limitations, we selected appropriate time intervals and controlled for key covariates, but future studies could use longer follow-up periods, more measurement points, and broader covariates to better capture these dynamics.