1. Introduction
Healthy behaviors are learned in adolescence [
1]. Healthy eating, physical activity (PA), and positive self-perceptions are important for quality of life and physical health [
2]. Screen-based behaviors (e.g., television viewing and using social networking sites) are key leisure activities among adolescents [
3], and the cumulative time spent engaged in these behaviors is often thought of as a youth risk behavior [
4,
5,
6]. Population estimates show that between 40% and 80% of young people exceed screen time recommendations [
7,
8].
Since the establishment of recommendations for pediatric screen times [
3], an increase in the number and volume of sedentary activities negatively impacting positive behaviors [
9] and overall wellbeing [
10] has been observed. These effects include less time for PA and sleep [
9], poorer academic performance [
11], and higher risk of obesity and becoming overweight [
12]. Screen-based behaviors may have negative psychological effects [
3], and may further impact health through behavioral mediators [
3] such as unhealthy eating and PA habits [
13,
14].
Chinese youth experience many cultural influences that may promote screen-based behaviors more than for their western peers. For example, many adolescents in China and other eastern Asian countries attend “cram schools” (i.e., test preparation centers) or employ private teachers to enhance their academic achievement [
15]. The additional time spent in class, tutoring, or studying may limit the available time for PA or sleep and promote screen-based behaviors such as information seeking on the internet, which may negatively affect academic performance [
15]. A study conducted in Beijing, China, showed that using computers, watching TV and playing e-games for more than two hours per day influenced middle school students to become overweight [
16]. Other studies indicated that, for a number of reasons, Chinese teenagers did not spend enough time on PA [
17,
18,
19], which may have led to obesity or certain mental diseases [
18,
20]. For Chinese adolescents aged from seven to 18, the prevalence of obesity increased from 1.63% in 1991 to 5.99% in 2011 (2.36% to 7.27% for boys and 1.40% to 4.64% for girls) [
21].
Unfortunately, very few studies examining the impacts of screen-based behaviors have been conducted in Asian adolescents. Therefore, the aim of this study was to determine the amount of time adolescents in Wuhan, China, spent on screen-based behaviors, and the associations of this with adiposity, unhealthy eating behaviors, sleep, PA, academic performance, anxiety, self-esteem, and life satisfaction.
2. Materials and Methods
2.1. Study Population
A cross-sectional survey was conducted at two schools, a middle school and a high school, in Wuhan, Hubei, China, during the late spring/early summer of 2016. The study was conducted in accordance with the Declaration of Helsinki, and the protocol was approved by the Ethics Committee of Wuhan University (Project identification code 2016031269). All adolescents (n = 3059) enrolled in grades 7–12 were invited to participate in the study via a recruitment letter and consent form sent home to parents; written informed consent was obtained from a parent or guardian. For those providing consent, a survey was sent home to be completed and returned to school. Of those who consented, 149 respondents did not return the questionnaire, and 282 respondents did not provide essential information (i.e., grade, sex, eating behaviors, sleeping time and screen time variables). Two respondents older than 19 years of age and a respondent younger than 12 were excluded; this resulted in a final sample size of 2625 youths aged 13–18 years (86% of those contacted). All procedures were approved by the Wuhan University Ethics Board and the medical school district administrators.
2.2. Measures
The school and grade of each student was recorded by study staff; date of birth, sex, height and weight were self-reported. Age, body mass index (BMI: weight in kg/height in m
2), and sex and age standardized BMI (BMI z-score) were calculated [
22].
Students were asked how many hours a day they usually spent watching television, playing e-games, receiving news or study materials from electronic devices, using social media sites or apps, and watching videos both on school days and on non-school days. Total hours per week of overall screen-based behaviors was calculated per type of screen-based behaviors. Response options referred to daily use (≤1, 2–3, 3–4, and >4 h/day). We collapsed “use, but not daily” and “do not use” [
2] into four categories: “never”, “not every day”, “<1 h, 2–4 h”, and “>4 h daily”.
Academic performance was approximated by using an informal ranking scale based on students’ self-reports of scores on the last cumulative examination in their grade; options were top 20%, 20–40%, 40–60%, 60–80%, and lowest 20%. Answer selections were collapsed into two categories split at the median for analysis; the top 40% were coded as 1 and the last 60% were coded as 0.
Regarding unhealthy eating behaviors, students were asked how many times per week they skipped breakfast or had a late-night snack (after dinner). Answer options were none, 1, 2–3, 4–5, or all 7 days [
23]. A dichotomous variable was created for skipping breakfast, to reflect consumption on all 7 days (coded as 1), as opposed to all other options (coded as 0) [
24]. For late-night snacking, we created a dichotomous variable; participants who never snacked at night were coded as 1, and participants who reported snacking at night were coded as 0.
Regarding sleep duration, students were asked how many hours they typically slept at night on school days. The response options were <4, 5, 6, 7, 8, 9, or >10 h. A dichotomous variable was created with respondents who slept <8 h per night on school days being coded as 0, and those sleeping ≥8 h per night being coded as 1 [
25].
Since strenuous PA is strongly [
26] and independently associated with markers of cardio metabolic health [
27], and can be more reliability assessed than light or moderate PA [
28], we assessed only strenuous PA [
28]. Students were asked whether they engaged in strenuous activity, equal to or more than three days a week (Yes = 1, No = 0). Strenuous activity was defined as sports, games, or dance that made them breathe hard, made their legs feel tired, or made them sweat [
10].
The Middle School Student Mental Health Scale developed by Wang [
29] was used to assess adolescents’ levels of anxiety. Participants were asked to quantify their anxiety during the previous seven days. The scale is comprised of six items scored on a five-point Likert-type scale. Responses were summed to derive an anxiety score ranging from 6 to 30 (with higher scores indicating higher levels of anxiety) [
29].
The Satisfaction With Life Scale [
30] was used to measure life satisfaction of participants. Students completed a seven-point Likert-type response scale in response to five items; answers were summed to create a life satisfaction score for each student. The possible range of scores was 5–35, with 20 representing the mid-point (5–19 indicating dissatisfied; 21–35 indicating satisfied) [
31].
The Rosenberg Self-Esteem Scale [
32] was used to assess participants’ self-esteem. Students answered 10 questions using a four-point Likert-type scale. Five items were positively worded and five were negatively worded. Items were summed with the negatively worded items reverse-coded to produce a self-esteem score from 0 to 30 (with higher scores indicating higher self-esteem) [
33].
2.3. Statistical Analysis
All analyses were conducted using Stata 14 (StataCorp LLC, College Station, TX, USA). The distributions of each of the continuous outcome variables were assessed for normality. Because BMI was positively skewed, age and sex adjusted BMI z-scores [
22] were calculated and dichotomized to overweight/obese (=0) or normal/underweight (=1). Because anxiety was positively skewed, a log transformation was applied.
Pearson’s chi-squared test (χ2) was conducted to examine differences in dependent variables by gender. Forced entry logistic regression analyses were used to examine the association between screen time (hours/week) and BMI z-score, unhealthy eating behaviors, sleep, PA, and academic performance-dependent variables. Ordinal least squares linear regression analyses were used to examine the association between screen time and psychological states the dependent variable. Models were adjusted for age and sex. Distributions and frequencies for each category of variables were examined, and unstandardized regression coefficients (b), standard error (SE), odds ratios (OR), and p-values were calculated, where appropriate, to determine the relationships between screen time and the dependent variables.
4. Discussion
In this study, we found that for Chinese adolescents, more time spent watching television, on social networking sites, and videos may be negatively associated with academic performance; however, the association between academic performance and receiving news and study materials from electronic devices was not statistically significant. One possible explanation is that Chinese students may use electronic devices in “cram schools”, which may neutralize the negative effect of screen-based behaviors. Additionally, our results suggest no negative relationships between screen-based behaviors and BMI z-score. Viewing only social networking sites was significantly associated with BMI z-score, but only on non-school days. More time on such sites was associated with being underweight or normal weight. Although studies using both self-reported and objective measures of height and weight indicate that sedentary behaviors are associated with excess weight in children and adolescents [
34,
35], we did not observe an association between screen-based behaviors and weight status after adjustment for covariates (sex and grade).
A possible explanation for these findings is related to the associations between social networking sites and body image concerns among adolescents [
36,
37]. For example, Tiggemann and Slater [
37] indicated that adolescent female Facebook users report more appearance concerns and dieting behavior than non-users, and that this link intensified with the amount of time spent on Facebook. Appearance comparison may explain the association between use of social network sites and lower BMI z-score. Social networking sites allow users to post photos and compare their appearance with others, placing users at a higher risk of body dissatisfaction [
2]. Moreover, using SNS is not always a sedentary behavior. Students could also spend time on SNS when they are walking, for example; this may be another reason why we did not observe an association between screen-based behaviors and weight status.
Our results suggest that the amount of time reported playing electronic games and watching television is positively associated with unhealthy eating habits, although this relationship was not linear. These findings corroborate previous research indicating that sedentary behavior, particularly screen time, predicts unhealthy eating behavior [
38,
39]. Previous studies have suggested that the type rather than volume of sedentary behavior may be more important in explaining unhealthy eating behaviors [
40,
41,
42]. For example, Borghese et al. [
40] suggested that time spent watching TV is more strongly associated with unhealthy eating choices than total sedentary time is. Our results confirm this by showing that television viewing on school days and playing electronic games on non-school days were both related to snacking at night. Our data also suggest a link between electronic games and skipping breakfast among youth, both on school days and non-school days.
Our results suggest that higher reported time spent watching TV may be associated with greater sleep duration on non-school days. Few studies have previously shown that screen time affects sleep duration, suggesting a potentially novel finding in Chinese youth. However, playing electronic games may lead to less sleep. We found that watching television on school days and getting electronic news or study materials on non-school days were negatively associated with PA, perhaps because of displacement of other activities. However, using social networking sites was positively associated with PA. Although this is contrary to previously reported findings in western youth [
9], it is consistent with the relationship we found between the use of social networking sites and BMI z-score. Our results suggest that using social networking sites may be a type of sedentary behavior not associated with low PA and a higher BMI among adolescents. Future studies are needed to better elucidate the associations between BMI, PA, and the use of social networking sites.
In this survey, anxiety levels, life satisfaction, and self-esteem levels were independently related to screen-based behaviors, especially on school days. Use of social networking sites and behaviors related to news or study materials from electronic devices were significantly associated with higher levels of anxiety in a whole week. Our results contribute to a relatively small body of literature that suggests relationships between screen-based behaviors and markers of poor mental health in adolescents. If our results were observed over a longer period of time, they could suggest greater anxiety, lower life satisfaction, and lower self-esteem in youth who spend excessive amounts of time engaged in screen-based behaviors [
43], which may negatively affect their academic performance.
Childhood overweight, obesity, and insufficient PA are among the most pressing public health concerns today [
44,
45]. Several factors contribute to the imbalance between energy intake and energy expenditure that influence weight gain [
44]. PA and screen-based behavior habits begin to form in youth [
45]. Therefore, encouraging and building healthy habits regarding PA and screen-based behaviors is important to both the current and future health of children and adolescents [
46,
47,
48].
There are some limitations to the present study. The sample is limited to one place and may not be representative of other groups of adolescents. All data were self-reported; therefore, reporting bias could have influenced the results. Assessment of PA was limited to a single item to reduce participant burden. Unfortunately, this approach allowed only for the assessment of vigorous (strenuous) PA at the exclusion of light or moderate activities, which is a limitation of the present work. That the weight status was dichotomized may have caused potential problems given that being underweight was a risk factor for certain diseases. Moreover, the cross-sectional design limits the ability to determine causal relationships. Thus, future longitudinal studies are needed. Finally, non-validated measures in our questionnaire aiming to measure the outcome variables (i.e., breakfast skipping, snacking at night, sleep duration and academic performance) may raise potential issues related to reliability. More refined and precise measures of unhealthy eating behavior would be desirable in future studies.
The strengths of this study include the large sample of Chinese adolescents, the robust data quality assurance procedures, and controlling for variables with known relationships to the dependent variables (i.e., age and gender). Unique to this analysis are scales developed for Chinese youth, previously piloted in another sample.