Participants
The data were obtained from China’s National Adolescent Health Surveillance from 2014 to 2015. This is an annual school-based surveillance system involving adolescents and young adults from the same junior and senior high schools located in Xinxiang (central), Yangjiang (south), Chongqing (west) and Shenyang (north) areas. These areas are broadly representative of the average population within China in terms of economic development and demographic composition. Eight schools (2 junior and 2 senior high schools from urban and rural, respectively) in each geographic region were randomly selected.
A total sample of 15 713 students from grade 7–12 were invited to participate in the study. Of these students, 1 037 (6.6%) were excluded from the study because of (1) absence from school on the day of the survey or unwillingness to respond to the questionnaire, (2) missing data through fictitious or inconsistent responses. Finally, we received 14 676 (93.4%) effective questionnaires, including 7 017 males (47.8%) and 7 377 junior school students (50.3%). The students aged from 10 to 24 years (mean 15.20, SD 1.75). In addition, the participants of four regions were 3 968 (Xinxiang), 3 539 (Yangjiang), 4 007 (Chongqing), and 3 162 (Shenyang), respectively.
Procedure
The study ensured all the participants and their guardians are aware of the purpose and content of this investigation. If students agree to participate in the survey, they stayed in the classroom, and, if not, they were allowed to leave. Each center used an anonymous questionnaire. Completion of the self-reported questionnaire took approximately 25 minutes. A teacher was sitting in a place of the classroom but was unable to observe student responses. The investigators checked the accuracy and completeness of returned questionnaires in time to take out the unqualified questionnaire. Informed consent was sought from parents/ guardians of each student prior to completion. Approval for the design and data collection procedures was obtained by the Ethics Committee of Anhui Medical University. The detailed survey information can be found in our previous article [38].
Measures
Socio-demographic Information
A self-report questionnaire was used to collect socio-demographic information, including gender (male or female), grade (junior or senior middle school), urban/rural residency, household structure (only-child or more than one child), parents’ education level (less than junior middle school, junior middle school, senior middle school, college or more) and self-perceived family socioeconomic status (poor, general or good). The additional file provides a questionnaire (Additional file 1).
Psychological Symptoms and Screen Time
Psychological symptoms, including emotional symptoms (including depression and anxiety symptoms, e.g., ‘Do you always feel distressed?’), behavioural symptoms (including paranoid and hostile behaviors, e.g., ‘Do you always have the impulse to damage something?’) and social adaptation symptoms (including bad relationships with family and friends etc., e.g., ‘Could you always not be suited for school life?’), were evaluated by the psychological domain of the Multidimensional Sub-health Questionnaire of Adolescents (MSQA) [31]. The MSQA has been widely used in mainland China and reported by various groups to be a valid and reliable method to explore the current state of adolescents’ psychological health status [35, 36]. Cronbach’s alpha for the MSQA was 0.951 in the present study.
Screen time was measured by video time and video game time. The subjects reported video time and video game time using the following questions: “On an average school day, how many hours do you watch video (such as watching TV, mobile phone, MP4, DVD/VCD)? ” and “On an average school day, how many hours do you play video games (such as game consoles, computer games, mobile games)? ” A similar question was used to define the usual weekend video time and weekend video game time[37]. All of these questions have seven answer categories: ‘0 h’, ‘≤0.5 h’, ‘0.5–1 h’, ‘1–2 h’, ‘2–4 h’, ‘4–6 h’, and ‘>6 h’.
Sleep Variables
The questionnaire contained 2 questions concerning sleep duration[38]. The first question represents usual weekday sleep duration: “In the last month, how many hours of actual sleep do you usually get at night on weekdays? ” A similar question was used to define usual weekend sleep duration. Usual daily sleep duration was calculated as a weighted average of weekday and weekend sleep durations, using the formula: ([{usual weekday sleep duration}×5]+[{usual weekend sleep duration}×2])/7. Weekend catch-up sleep using the formula: ([usual weekend sleep duration]-[usual weekday sleep duration]). For statistical analyses purposes in the present study, sleep duration was collapsed into 6 categories: ‘<6 h’, ‘6–7 h’, ‘7–8 h’, ‘8–9 h’, ‘9–10 h’, and ‘≥10 h’. Weekend catch-up sleep was collapsed into 5 categories: ‘<0 h’, ‘0–1 h’, ‘1–2 h’, ‘2–3 h’, and ‘≥3 h’. Sleep quality were obtained from responses on a 4-point Likert scale to the item “In the past month, what do you think of your sleep quality? ” Response options were very good, good, poor, very poor[38].
NSSI
NSSI was assessed using the following question: ‘In the past 12 months, have you ever harmed yourself in a way that was deliberate, but not intended to take your life?’ Eight NSSI behaviors were presented, and the details of the questions were as follows: (1) Have you ever hit yourself?; (2) Have you ever pulled your hair yourself?; (3) Have you ever banged your head or fist against something?; (4) Have you ever pinched or scratched yourself?; (5) Have you ever bitten yourself?; (6) Have you ever cut or pierced yourself?; (7) Have you ever deliberately taken an overdose (e.g. of drugs, alcohol or smoking)?; and (8) Have you ever ingested a toxic substance or object? For those who confirmed that they had engaged in NSSI, the frequency of NSSI was investigated [31]. In the present study, the Cronbach’s alpha coefficient for the NSSI was 0.798.
Statistical Analysis
The statistical analyses were carried out by SPSS version 24.0 and R software version 3.6.1. Means and standard deviations of sleep duration were calculated separately for each gender group as well as for the grade group of participants. For comparisons between gender and grade groups, independent t-tests were conducted for sleep duration. Frequencies and percentages of NSSI in different groups were calculated. Pearson’s chi-squared test was used to examine differences in gender, grade, sleep duration and other variables between the adolescents reporting NSSI versus no NSSI. The univariate logistic regression analysis was used to examine the associations between covariates (e.g., gender, sleep duration and so on) with NSSI. Binomial logistic regression models were used to examine the associations of NSSI with daily sleep duration, weekday sleep duration, weekend sleep duration and weekend catch-up sleep, by adjustment for sociodemographic variables and so on. Gender and grade differences in this associations were examined. The locally estimated scatter plot smoothing (LOESS) method was used to help explore the associations of total NSSI number with daily sleep duration, weekday sleep duration, weekend sleep duration and weekend catch-up sleep. This method was especially useful to represent the nonlinear relation between variables, e.g., the U-shaped relation. GetData Graph Digitizer version 2.26 was used to find the nadir point. Binomial regression analysis was used to help test the relationship between total NSSI number and sleep duration by adjustment for sociodemographic variables, sleep quality, psychological symptoms, screen time. The level of significance was set at P<0.05.