Insights into adolescent sleep and mental health in rural area of Northwestern China

Abstract

Adolescents in affluent nations have experienced a decline in sleep duration, associated with adverse outcomes such as behavioral issues and health concerns. However, the connection between sleep and mental well-being during adolescence, particularly in developing regions like rural China, remains underexplored. A cross-sectional study of 18,516 adolescents in 124 junior high schools in Ningxia, China, utilized the strengths and difficulties questionnaire to assess mental health. The findings highlighted a complex, nonlinear link between sleep duration and mental health, with a U-shaped trend for overall difficulties and an inverse U-shape for prosocial behavior. The study also explored potential mechanisms behind these relationships, suggesting that time allocation to activities such as screen time and outdoor activities could mediate the effects of sleep duration on mental health. Longer sleep durations could lead to less screen time and more outdoor engagement, both of which positively affect mental well-being. Balanced sleep duration is crucial for adolescent mental health. The study calls for interventions to improve sleep hygiene and mental health services in rural areas, emphasizing the need for policy support to address sleep deprivation and its impact on mental well-being.

Introduction

An increasing body of literature underscores the pivotal role of sleep in adolescent health and development^1,2. Over the past century, adolescents in affluent nations have witnessed a significant decline in sleep duration³. This trend has been linked to various adverse outcomes, including excessive daytime sleepiness⁴, reduced daytime alertness⁵, diminished academic performance⁶, increased risk of myopia⁷, behavioral issues such as aggression, theft, smoking, and alcohol consumption⁸, as well as health concerns like obesity⁹ and injuries in school and sports settings¹⁰. Though awareness of the relationship between sleep and mental health is growing, particularly in affluent nations, rural areas have often been overlooked in research. Our study addresses this gap by focusing on rural adolescents, an understudied demographic, to reveal the impact of sleep on adolescent development in these regions.

Mental health problems represent a pressing global public health challenge during adolescence¹¹. Notably, the onset of mental illness in adolescence can have long-term effects, leading to adverse outcomes like substance addiction¹²and reduced work productivity¹³. Consequently, investigating the link between sleep and mental health during adolescence is imperative for mitigating the risk of severe mental illnesses later in life.

The limited existing research on the relationship between sleep and mental health among adolescents exhibits several key limitations. Firstly, the majority of studies conducted in developed countries, such as the United States¹⁴, Japan¹⁵, and Sweden¹⁶, have indicated a correlation between sleep duration and mental health issues. However, there is a notable dearth of research exploring this relationship among adolescents in developing environments, particularly in China and rural areas. Moreover, previous investigations in less developed nations often lacked comprehensive data on rural regions¹⁷, with small sample sizes and a failure to consider the impact of sleep duration on various aspects of mental health. Additionally, research from both developed and less developed regions has tended to focus predominantly on specific mental health conditions, such as depressive symptoms or anxiety, potentially overlooking other prevalent issues like conduct problems or prosocial difficulties^18,19.

Rural China presents an ideal context to address these aforementioned limitations. Globally, 10–20% of children and adolescents suffer from mental health issues¹¹, with rates in China ranging from 10 to 57% among school-aged children²⁰. Given that approximately 65% of China’s school population resides in rural areas^21,22, understanding child mental health in these regions is of utmost importance. However, research on sleep duration among adolescents, particularly regarding its association with mental health, remains scarce in rural China²³. In China’s vast northwest, students’ long commutes due to the region’s expansive geography often reduce their sleep time²⁴. Moreover, electronic device use, especially late-night screen time, affects rural students’ sleep^25,26, potentially delaying sleep onset and affecting quality, despite lower device prevalence than in urban areas²⁷. These factors, combined with the general scarcity of research, highlights a pressing need for large-scale empirical studies to analyze the relationship between sleep duration and the multifaceted dimensions of mental health in rural Chinese communities.

This study sets out to fulfill three key goals: to delineate the patterns of sleep duration and mental health in rural Chinese adolescents using the Strengths and Difficulties Questionnaires; to examine how sleep duration correlates with mental health outcomes; and to uncover the underlying mechanisms by which sleep can affect the mental well-being of adolescents.

The remainder of this paper is structured as follows: We begin by presenting the data and describing the variables. Subsequently, we introduce the econometric models utilized and present the estimation results and accompanying discussion. Finally, we synthesize our findings and propose relevant policy implications.

Methods

Ethical approval and consent to participate

Research ethical approval for this study was granted by the Institutional Review Boards of Stanford University in Palo Alto, USA (Protocol 52514). In addition, each sample school’s local education bureaus and principals granted permission to conduct the in-school survey. Written informed consent was provided from at least one parent for each child participant. Student participants gave oral consent during the survey and could withdraw from the study anytime. The Declaration of Helsinki was closely adhered to throughout the study.

Compliance with guidelines and regulations

In conducting this study, we ensured that all methods were performed in accordance with the relevant guidelines and regulations. The study design, data collection, and analysis were carried out following the ethical principles outlined in the Declaration of Helsinki for research with human participants. Personal data is anonymized to protect the privacy and confidentiality of the collected data. Any potential conflicts of interest were disclosed and managed in accordance with the guidelines provided by the Institutional Review Boards of Stanford University.

Setting

The data used in this study was collected during a survey among rural students in Ningxia province in the fall of 2019. Ningxia is a comparatively poor area located in northwest China with a population of 6.95 million and an area of 66 thousand square kilometers. Ningxia’s gross domestic product (GDP) per capita was lower than the national average (of USD 12,458), at about USD 9623 in 2019. Ningxia Province was 28th of 31 provincial administrative regions in China, according to the China Statistical Yearbook 2019.

Sampling

The study employed a two-stage cluster sampling method at the school level. First, from an initial pool of 273 secondary schools in Ningxia province, we narrowed down the sample to 124 rural schools by excluding urban schools and those with fewer than 40 grade 7 students. Additionally, four schools declined to participate. In the second stage, for efficiency, we excluded classes with more than 70 students. In schools with multiple grade 7 or grade 8 classes, one class per grade was randomly selected. All students within the chosen classes were included in the study, without further sampling at the individual level. This approach resulted in a sample of 20,375 students from 474 classes across 124 schools. After excluding cases with missing data, we analyzed a final sample of 18,516 students.

Data collection

Questionnaires

The survey conducted in October 2019 asked students about demographic details such as age, gender, and parental age. It also inquired about their daily routines and free time. To assess mental health and well- being, participants completed the Strengths and Difficulties Questionnaire (SDQ).

Socio-demographic characteristics

Demographic data in our study encompassed characteristics of both students and their parents, with the majority of variables recorded on a binary scale. Student age was self-reported in years. The following binary codes applied to the student demographic variables: Grade 8 students were coded as 1 (Grade = 1), males as 1 (Gender = 1), only child in the family as 1 (Only child = 1), Han ethnicity as 1 (Ethnicity = 1), boarding students as 1 (Boarding student = 1), and students with divorced parents as 1 (Parents divorced = 1). Parental demographic information included their age in years, educational attainment with a senior high school education or higher coded as 1 (Senior high school education or above = 1), and parental migration status with 1 indicating at least one parent who had migrated for work (Parent migrated = 1). We assessed family assets using a binary scale based on 12 items from the National Household Income and Expenditure Survey. Students indicated whether they possessed each asset, with 1 representing ownership and 0 indicating they did not. The assets included cars, motorcycles or electric vehicles, refrigerators, washing machines, heating systems, range hoods, flush toilets, tractors, large agricultural machinery, computers, internet access, and ownership of property in county-level cities or above. We employed Principal Component Analysis to derive a family asset index, which we then categorized into three terciles: bottom, middle, and top.

Mathematics assessment

A standardized mathematics assessment, lasting 35 min and designed by local education experts for students in the seventh and eighth grades, was also administered to measure academic performance. The scores obtained from this assessment were subsequently normalized for analysis.

Sleep duration

Two self-reported items were listed in the questionnaire to collect data on sleep duration on weekdays: (1) When do you usually get up on weekdays? (2) When do you usually go to bed on weekdays? The response is provided in a 24-h time frame, with the precise minute. We subsequently determine the sleep duration by computing the time difference between their get-up time and bedtime.

Mental health outcomes

The assessment of mental health outcomes was conducted by utilizing the self-reported version of the Strengths and Difficulties Questionnaire (SDQ). The SDQ is believed to be a widely used instrument for identifying mental and behavioral health problems for children and adolescents aged 4–16 years²⁴. The Chinese version of the Strengths and Difficulties Questionnaire (SDQ) has demonstrated exceptional internal consistency reliability, with a Cronbach’s alpha coefficient of 0.927, and high test–retest reliability, indicated by a Pearson’s correlation coefficient of 0.719²⁵. This instrument has been widely utilized in studies assessing the mental health of students^28–30.

The Strengths and Difficulties Questionnaire (SDQ) has 25 items spread across five subscales: emotional symptoms, conduct problems, hyperactivity, peer problems, and prosocial behavior. Prosocial behavior is considered a Strength, while the other four contribute to the Total Difficulties score. Each item has three response options (0 = Not True, 1 = Somewhat True, 2 = Certainly True), resulting in a Strength score from 0 to 10 and a Total Difficulties score from 0 to 40. Based on Chinese norms, a Total Difficulties score above 19 is “abnormal.” The “abnormal” range for the Prosocial Score is below 4. In this study, both Total Difficulties and the Prosocial Score will be used as dependent variables.

Statistical methods

The literature first provides descriptive statistics on student demographics, family backgrounds, student-reported sleep duration, and mathematics scores. We then applied a quadratic regression model to examine the link between sleep duration and mental health (measured by SDQ scores), controlling for various student and family factors: age, grade, gender, only child status, boarding, parental marriage, migration, napping habits, household wealth, parental age, education level, and standard mathematics scores. The regression’s inflection point (− ${\beta }_1{/2\beta }_2$) is within the sleep duration range (S). A positive coefficient for the squared sleep term (${\beta }_2$) indicates a U-shaped curve, while a negative coefficient suggests an inverted U-shape.

$${\text{y}}_i=\alpha +{\beta }_1{\text{S}}_i+{\beta }_2{\text{S}}_{\text{i}}^2++{\epsilon }_i$$ 1

$${\text{y}}_i=\alpha +{\beta }_1{\text{S}}_i+{\beta }_2{\text{S}}_{\text{i}}^2+{\beta }_3{\mathit{CONTROL}}_i+{\epsilon }_i$$ 2

$${\text{y}}_i=\alpha +{\beta }_1{\text{S}}_i+{\beta }_2{\text{S}}_{\text{i}}^2+{\beta }_3{\mathit{CONTROL}}_i+{\beta }_4{\mathit{COUNTY}}_i+{\epsilon }_i$$ 3

where y is a standard continuous indicator for SDQ score of respondent i. S_i measures weekday sleep duration, calculated by get-up time and bedtime. ${\text{S}}_{\text{i}}^2$ is the squared sleep duration term. CONTROL_i includes the control variables listed in the regression model, COUNTY represents the fixed county effect. ${\epsilon }_i$ is a random error term. Here, i represents each of the observations. We adjust standard errors for clustering at the school level in all regression models except the basic one. Heterogeneity and mechanism analyses are used to to validate the sleep-duration-SDQ relationship. Analysis is performed using Stata18 (StataCorp, Texas, USA).

Results

Background characteristics

Table 1 contains a summary of the descriptive statistics for all samples. The mean SDQ Total Difficulties score was 12.52 (SD = 4.89), collected by summing emotional symptoms (mean = 3.35; SD = 2.23), conduct problems (mean = 2.53; SD = 1.64), hyperactivity (mean = 3.58; SD = 1.95), and peer problems (mean = 3.07; SD = 1.52). The mean Prosocial score of the total students was 7.11 (SD = 2.20). The average sleep duration on the weekdays was 8.25 hours (SD = 1.06).

Table 1.

Summary statistics of student individual and family characteristics.

Characteristics	Mean	Std. Dev	Min	Max	N
Dependent and independent variables
SDQ total difficulties score	12.52	4.89	0.0	33.0	18,516
SDQ emotional symptom score	3.35	2.23	0.0	10.0	18,516
SDQ conduct problems score	2.53	− 1.64	0.0	10.0	18,516
SDQ hyperactivity score	3.58	− 1.95	0.0	10.0	18,516
SDQ peer problems score	3.07	− 1.52	0.0	10.0	18,516
SDQ prosocial score	7.11	− 2.20	0.0	10.0	18,516
Weekdays sleep duration, hours	8.25	− 1.06	4.0	12.0	18,516
Individual and family characteristics
Age, years	13.42	− 1.00	9.8	16.0	18,516
Grade (grade 8 = 1)	0.50	− 0.50	0.0	1.0	18,516
Ethnicity (han = 1)	0.60	− 0.49	0.0	1.0	18,516
Gender (male = 1)	0.51	− 0.50	0.0	1.0	18,516
Only child (yes = 1)	0.07	− 0.25	0.0	1.0	18,516
Boarding student (yes = 1)	0.35	− 0.48	0.0	1.0	18,516
Parents divorced (yes = 1)	0.09	− 0.29	0.0	1.0	18,516
Parent migrated (yes = 1)	0.31	− 0.46	0.0	1.0	18,516
Nap at noon, hours	0.29	− 0.38	0.0	2.0	18,516
Household wealth is in the bottom thirda	0.33	− 0.47	0.0	1.0	18,516
Household wealth is in the middle thirda	0.33	− 0.47	0.0	1.0	18,516
Household wealth is in the top thirda	0.34	− 0.47	0.0	1.0	18,516
Father’s age, years	41.14	− 4.81	30.0	55.0	18,516
Mother’s age, years	38.65	− 4.66	30.0	55.0	18,516
Father has senior high school education or above (yes = 1)	0.09	− 0.28	0.0	1.0	18,516
Mother has senior high school education or above (yes = 1)	0.06	− 0.24	0.0	1.0	18,516
Standardized math score	0.03	− 0.99	− 2.8	2.5	18,516
Time allocation
Exercise during lunch, hours	0.16	− 0.28	0.0	2.0	18,516
Study after school, hours	1.01	− 0.70	0.0	2.0	18,516
Use screen devices after school, hoursb	0.74	− 0.91	0.0	6.0	18,516

^aWe used the possession of certain rural durable assets as a proxy of household wealth. First, we asked about the household’s ownership status of 12 assets, including cars, refrigerators, washing machines, computers and vice versa. If a household owned a specific asset, it was recorded as 1; otherwise, it was recorded as 0. Second, by using the principal components analysis, we calculated the scoring factors for 12 assets. Finally, there are three types of household wealth.

^b“Use screen devices after school” is calculated by summing up the time spend on cell phone, television and computer.

At the individual and family level, the characteristics of students and their family status were described as follows.The 18,516 students had an average age of 13, with males making up 51%. About 7% were only children, 35% were boarders, 9% had divorced parents, and 31% had at least one parent who migrated for work. Fathers averaged 41.14 years old, and mothers 38.65 years old. However, the proportion of parents with a senior high school education or above (around 12 years for school) was relatively low, at 9% for fathers and 7% for mothers.

With regard to time allocation, it can be observed that the average outdoor time during lunchtime was 0.16 h, the average study time after-school was 1.01 h, and the average electronic screen time, including mobile phones, television, and computers, was 0.74 h.

The association between sleep duration and mental health

Firstly, we preliminarily explore the correlation between sleep duration and mental health in terms of the Total Difficulties score and Prosocial score. Nonlinear relationships for Total Difficulties score and Prosocial scores are demonstrated in Fig. 1. The relationship between sleep duration and Total Difficulties score exhibits a U-shaped curve. While an inverse U-shaped curve indicates a divergent correlation between sleep duration and Prosocial score. Table 2 presents the regression results for sleep duration’s impact on mental health. The coefficient for sleep duration on the Total Difficulties score was significantly negative, and the squared term was significantly positive (columns 1–3). This suggests a U-shaped relationship, with a turning point at about 8.8 h. Below this threshold, longer sleep duration is associated with lower Total Difficulties scores, indicating better mental health. The effect reverses above the threshold. The regression coefficient for sleep duration on the Prosocial score was positive, while the squared term was negative (columns 4–6). This shows an inverse U-shaped relationship, peaking at about 6.6 h.

Fig. 1.

Fitting curve of relationship between sleep duration and SDQ.

Table 2.

Quadratic fitting regression model of sleep duration effects on SDQ.

	Total difficulties			Prosocial score
	1	2	3	4	5	6
Sleep duration	− 3.860***	− 3.259***	− 3.323***	0.471***	0.255*	0.246*
	− 0.328	− 0.331	− 0.325	− 0.145	− 0.139	− 0.142
Sleep duration square	0.225***	0.185***	0.190***	− 0.036***	− 0.019**	− 0.019**
	− 0.02	− 0.02	− 0.019	− 0.009	− 0.008	− 0.009
Constant	28.815***	25.753***	24.828***	5.680***	6.246***	6.658***
	− 1.347	− 1.577	− 1.507	− 0.593	− 0.662	− 0.698
Control variables	No	Yes	Yes	No	Yes	Yes
County fixed effects	No	No	Yes	No	No	Yes
R-squared	0.009	0.062	0.071	0.004	0.048	0.053
N	18,516	18,516	18,516	18,516	18,516	18,516
Inflection point (h)	8.6	8.8	8.8	6.6	6.7	6.6

Standard errors in parentheses, clustered at the school level.

Control variables were excluded in columns 1 and 4. County fixed effects were controlled in columns 3 and 6.

*p < 0.10, **p < 0.05, ***p < 0.01. P < 0.1 was considered statistically significant in this study.

Up to this point, increasing sleep duration improves Prosocial scores, signifying better behavior. However, very long sleep durations do not further enhance Prosocial conditions.

Figure 2’s marginal analysis mirrors the regression findings. The marginal utility of sleep duration less than 8.8 h appears to be negative for the Total Difficulties score, and with the sleep duration increasing, the negative effect approaches zero, and the positive effect occurs after passing the turning point. For Prosocial scores, the positive marginal utility of sleep is insignificant before 6.6 h, becoming significantly negative beyond this point.

Fig. 2.

Average marginal effects of sleep duration on SDQ.

Robustness check

By far, the regression results have shown that sleep duration has a U-shaped impact on the Total Difficulties score and an inverted U-shaped impact on the Prosocial score. To ensure the robustness of the model findings, we now consider several alternative subgroups to identify whether the results from our basic model are robust. Table 3 reports the justification of the robustness that was obtained from the estimation for alternative subgroups.

Table 3.

Robustness checks the impact of sleep duration on SDQ.

	Four subscales of total difficulties				keep samples with the SDQ score of 10–90%
	1	2	3	4	5	6
	Peer problems	Hyperactivity	Conduct problems	Emotional symptoms	Total difficulties	Prosocial score
Sleep duration	− 0.570***	− 0.680***	− 0.889***	− 1.185***	− 1.528***	0.159
	− 0.103	− 0.125	− 0.102	− 0.158	− 0.245	− 0.118
Sleep duration square	0.035***	0.037***	0.052***	0.065***	0.087***	− 0.012
	− 0.006	− 0.007	− 0.006	− 0.009	− 0.015	− 0.007
Constant	4.891***	6.012***	6.165***	7.769***	17.805***	5.891***
	− 0.519	− 0.552	− 0.491	− 0.722	− 1.14	− 0.558
Control variables	Yes	Yes	Yes	Yes	Yes	Yes
County fixed effects	Yes	Yes	Yes	Yes	Yes	Yes
R-squared	0.039	0.037	0.051	0.053	0.038	0.023
N	18,516	18,516	18,516	18,516	15,167	14,326

The control variable were included in all regression.

Standard errors in parentheses, clustered at the school level, fixed at county level.

*p < 0.10, **p < 0.05, ***p < 0.01. P < 0.1 was considered statistically significant in this study.

We first estimate robustness by using the four subscales of the Total Difficulties, including peer problems, hyperactivity, conduct problems, and emotional symptoms. The results in columns 1 to 4 suggest that the coefficient of each subscale-dependent variable is consistent with the basic estimation, which shows that sleep duration does have a curvilinear association with all four subscales.

Additionally, we enhance the assessment of resilience by selecting student participants whose SDQ scores were within the 10th and 90th percentiles in columns 5 and 6, respectively, for Total Difficulties score and Prosocial score. The outcomes were consistent with the full sample, showing a significant impact of sleep duration on Total Difficulties score at the 1% level. The Prosocial score’s correlation with sleep duration was not statistically significant. Robustness checks confirm that the findings hold for other subgroups.

Mechanism analysis

The findings above imply that sleep duration affects students’ Total Difficulties score and Prosocial score. In this section, our objective is to deliberate on possible mechanisms that may account for these effects and provide some illustrative evidence. Time allocation could be the potential explanation for the sleep duration effect. It has been observed that, in school-aged students, time allocation decisions among diverse activities have been found to have distinct effects both on physical and mental health^31–33.

Sufficient or longer sleep duration can limit time for sedentary behaviors, like excessive screen time, which has been linked to physical and psychological health issues, including obesity and depressive symptoms. Conversely, longer sleep allows for more outdoor activities, which can improve psychological well-being and happiness³⁴, and are more beneficial for mental health than nighttime activities, reducing the risk of depression symptoms³⁵.

We examine this potential mechanism using ordinary linear regression analysis. We construct three continuous variables (HT_i) the duration of students’ study after school, their time spent on screen devices, and their time spent on outdoor activities at noon. Independent variables and other control variables are identical to those specified in Eq. 3. The further analysis is as follows:

$${\text{HT}}_i=\alpha +{\beta }_1{\text{S}}_i+{\beta }_3{\mathit{CONTROL}}_i+{\beta }_4{\mathit{COUNTY}}_i+{\epsilon }_i$$ 4

As illustrated in Table 4, the results obtained through the utilization of ordinary linear regression are more direct. Columns 1 and 2 display the outcomes of the time spent on study after school and the time spent on screen devices. Column 3 provides the results of the time spent on outdoor activities.

Table 4.

Mechanism analyses of the influence of sleep duration on SDQ.

Dependent Variables	1	2	3
	Time studying after school	Time spending on screen devicesa	Time spending on outdoor activities at lunch time
Sleep duration	− 0.046***	− 0.056***	0.001
	− 0.006	− 0.01	− 0.003
Control variables	Yes	Yes	Yes
County fixed effects	Yes	Yes	Yes
Constant	1.630***	0.508***	0.317***
	− 0.1	− 0.152	− 0.047
R-square	0.206	0.127	0.124
N	18,516	18,516	18,516

The control variable were included in all regression.

Standard errors in parentheses, clustered at the school level, fixed at county level.

^a“Time spending on screen” is calculated by summing up the time spend on cell phone, television and computer. Questionnaire asked respondents to report how much time they actually spent on each screen devices.

*p < 0.10, **p < 0.05, ***p < 0.01. P < 0.1 was considered statistically significant in this study.

As presented in Table 4 , the ordinary linear regression analysis yields clear results. Columns 1 and 2 reveal significant negative coefficients for the time spent studying after school and using screen devices (P < 0.01), indicating that students who sleep longer tend to reduce their after-school study time and screen exposure. While column 3 indicates a positive, though not statistically significant, correlation between longer sleep duration and time spent on outdoor activities, it does not reach the threshold for statistical significance.

The results imply that extended sleep durations might correlate with decreased study and screen time among students, potentially leading to enhanced mental well-being. This suggests that strategic time management, particularly prioritizing sleep, could be a valuable strategy for boosting mental health.

Discussions and conclusion

This study examined the sleep duration and mental health of 18,516 lower secondary school students in rural Ningxia Province, revealing an average weekday sleep of 8.25 h. This exceeds the averages reported in the United States³⁶ and Hong Kong³⁷ but still falls short of the recommended 9 h for this age group^38,39.The longer sleep duration may be linked to the less developed economic status of the students’ rural families⁴⁰, yet various factors such as heavy homework and late dinners contribute to this sleep deficit. Parental influences, including emotional support and screen time behaviors, also play a role²³. The findings highlight global concerns about adolescent sleep and its impact on mental health⁴¹.

Furthermore, concerning students’ mental health, the mean score for SDQ Total Difficulties (mean = 12.52) was notably higher compared to similarly aged cohorts both within and outside of China. For instance, it surpassed the mean scores reported in the Netherlands (mean = 9)⁴² and Guangdong (mean = 11.90)⁴³, although it aligned closely with a rural sample from the northwestern province of Gansu (mean = 12.38)³⁰. This higher score may be due to the lower household incomes typical of our students, as seen in studies of south China’s migrant and rural students⁴³. Additionally, the over one-third left-behind status of our students could be linked to worse mental health, as previous research suggests⁴⁴.

In terms of the average Prosocial score (mean = 7.11) observed in our study, it was notably lower compared to findings from other research conducted abroad. This observation also aligned with another Chinese study conducted among lower secondary school students. This disparity may be partially attributed to the geographical and cultural diversity across different countries and regions, suggesting varying contexts that could influence prosocial behaviors, as exhibited in a research on global adolescent mental health measurement and the comparison among distinct cultures⁴⁵.

Our study uncovered a statistically significant nonlinear relationship between sleep duration and mental health, mirroring findings from previous research on school-aged students⁴⁶. Specifically, a U-shaped curve delineated the relationship between sleep duration and Total Difficulties score. We found that students in our sample experienced an average deficit in sleep, which correlated with a higher prevalence of emotional and behavioral issues⁴⁷, consistent with prior research indicating that shorter sleep duration compromises daytime functioning to some extent^4,48. Additionally, our observations align with earlier studies suggesting that adolescents residing in less developed areas are more vulnerable to emotional and behavioral disturbances⁴⁹

Our study reveals an inverted U-shaped link between sleep duration and Prosocial scores, indicating that moderate sleep is key for optimal prosocial behavior. Contrary to the belief that more sleep always improves prosocial actions, excessive sleep may not have the same benefits⁵⁰. Previous studies have shown that overuse of social media reducing sleep negatively affects prosocial behavior⁵¹, and morning-oriented individuals tend to be more prosocial⁵². However, some research contradicts these findings, suggesting no link between sleep duration and prosocial behavior in early childhood⁵³.

In summary, our results indicates that balanced sleep durations are crucial for better mental health, especially in rural China and underdeveloped areas. It highlights the need for sleep duration policies, aligning with the Chinese Ministry of Education’s efforts to combat homework-induced sleep deprivation³⁹. Our findings support the creation of sleep guidelines for junior high students.

We acknowledge several limitations inherent in our study. Firstly, the cross-sectional data limits causal conclusions about the link between sleep duration and mental health. Although we hint at a causal effect of sleep on mental health, the reverse could also be true, with conditions like depression potentially leading to more sleep⁵⁴. Longitudinal data are needed for a clearer causal picture. Additionally, due to constraints in tracking large cohorts of children’s sleep, this study relied on self-reported sleep duration data. Caution is warranted when interpreting these results, as the reliance on self-reported recall data, common in sleep and health studies^55,56, can introduce biases such as recall bias and social desirability effects, which are particularly pronounced in younger children^57,58. For more reliable sleep research, future studies should consider using technology for precise sleep tracking. Finally, to deepen our understanding, future research should consider diverse factors like gender, grade, and parental migration status.

This study, with a large sample, explored the link between sleep duration and mental health among rural Chinese students. It found higher rates of emotional and behavioral issues in these students compared to those in developed regions, both domestically and internationally, as assessed by the Strengths and Difficulties Questionnaire. The main finding was a nonlinear relationship: a U-shaped curve for sleep duration and Total Difficulties score, and an inverse U- shaped curve for Prosocial score. This study contributes to the expanding body of literature examining the relationship between sleep duration and multidimensional aspects of mental health and highlights the need for longitudinal research to determine causality and alternative sleep assessment methods.

The study findings underscore the urgent need for policies and interventions targeting sleep issues and mental health among adolescents in rural areas. To address these pressing concerns, strategies should simultaneously prioritize improving sleep hygiene and mental well-being. This may involve integrating sleep education into school curricula, providing resources and support for teachers, raising awareness among parents, and enhancing access to mental health services. By focusing on both sleep and mental health, we can better support the overall well-being of rural adolescents and mitigate the risk of long-term health consequences.

Author contributions

WL contributed to the data curation, writing and original draft preparation. HG contributed to the conception of this study, methodology, writing and editing, substantively revising the study, and funding for the research. XC contributed to the interpretation of the data. LZ contributed to the methodology and interpretation of the data. All authors read and approved the final manuscript.

Funding

The program was funded by the Central University Basic Research Expenses Special Funds Project (23ZYYB002), the Higher Education Discipline Innovation Project, Grant Number B16031. The study’s funder had no role in study design, data collection, data analysis, data interpretation, or report writing.

Data availability

The datasets used and/or analyzed during the current study are available from the corresponding author on reasonable request.

Declarations

Competing interests

The authors have no competing interests to declare that are relevant to the content of this article, including no financial competing interests such as funding, employment, or personal financial interests that could be perceived as influencing the work presented.