Association between media use and poor sleep quality among senior high school students: a cross-sectional study

Background : Poor sleep quality (PSQ) is an increasing health problem among adolescents. Mobile phones and portable media devices have become a part of children’s lives and may affect their sleep duration and quality. This study aimed to explore the prevalence of PSQ and identify the association between media use and PSQ among adolescents studying in high school grades 10-12. Methods: This cross-sectional study was conducted in central Thailand. A multi-stage sampling technique was used to enroll 777 adolescents from eight schools from August to October 2016. The research instruments comprised factors of demographics and consumption behaviors and the Pittsburgh Sleep Quality Index questionnaire. Multivariable logistic regression was used to calculate adjusted odds ratios (OR adj) and 95% confidence intervals (CI). Results: Prevalence of PSQ was 56.24%. The study subjects were mostly 16-17 years old (67.82%) and female (70.39%). Multivariable logistic regression, after controlling for possible confounders, revealed an increased odds of PSQ among those who used a social media device (OR=1.34, 95%CI=0.97-1.87), and showed a higher proportion of social media use in the PSQ group. Conclusion: A surveillance system to detect media use and PSQ should be conducted accompanied by knowledge sharing on media use among parents, teachers and adolescents. To determine causal relationships, further longitudinal studies will be required to test the association between media users and PSQ. This study may also provide some implications for health promotion on sleep quality of senior high school students.


Introduction
Sleep is an essential part of life and plays important roles in physical and mental health. 1,2 Adolescents experience significant changes to the body and mind associated with sex hormones. 3 Insufficient sleep has been one of the most important public health problems among adolescents. Concerning the aspect of sleep, a few studies have found that poor sleep quality (PSQ) was associated with the amount of daytime sleep, exhaustion, weight gain, obesity, impaired memory and motor vehicle accidents. [4][5][6] PSQ is currently a widespread issue in most societies. The prevalence of PSQ among adolescents was reported to range from 32 to 62% 7-11 reflecting a wider range of PSQ prevalence. In Thailand, the prevalence of PSQ among adolescents was reportedly 32 to 48%. 7, 8 Insufficient sleep not only impacts at a personal level, but also can cause major impact on a larger scale through a high burden of non-communicable diseases, 12 many events such as motor vehicle crashes, 13 workplace accidents, increased mortality and reduced quality of life. 14 Media use such as watching TV and using electronic devices are activities that cause PSQ among children and adolescents. Especially among school age group, having a TV in the bedroom can disturb sleep resulting in decreased sleep duration and insufficient sleep. In addition, media use may increase the activity of physiological arousal, inadequate sleep hygiene practice and difficulty falling asleep. 11 Some studies have shown the association of media use related to PSQ. [15][16][17] Hence, the present study aimed to seek the prevalence of PSQ, justify the definition of PSQ by using the Pittburgh Sleep Qaulity Index (PSQI) for poor and good sleepers and determine its association with media use among senior high school students in Ratchaburi Province, Thailand. Provinces of Western Thailand which are composed of Kanchanaburi, Phetchaburi, Prachuap Khiri Khan, Tak and Ratchaburi, geographical region and academic area are similar so the authors selected Ratchaburi Province as the area of study because of its characteristics as a proxy of western provinces of Thailand. Ratchaburi is located on the bank of the Mae Klong river and one of the western provinces of Thailand with an area of about 5,196 square kilometer. It lies 80 km west of Bangkok, and borders Myanmar to the west with the Tenasserim Hills as a natural border containing a population of 871,714 and density of 170 km 2 in 2017. 18,19 Methods Study design and participants A cross-sectional study was carried out between August and October 2016 to explore PSQ and identify the association between media use and PSQ occurrence among senior high school students in Ratchaburi Province, Thailand.

Sample size and sampling technique
The sample size was calculated using a formula to estimate the population proportion with specified absolute precision 20 according to the following assumption: 32% of PSQ among adolescents (P), 7 with 95% confidence interval and 5% specified absolute precision (d). As a multistage sampling technique was employed to identify study subjects, a design effect of 2 was used. The calculated sample size totaled 709. Also, approximately 10% was added to adjust for nonresponses. Thus, the final sample size was at least 777.
A multi-stage stratified sampling technique was used to identify study subjects from senior high schools in Ratchaburi Province (design effect of 2 and added 10% for non-response). Schools were stratified by student numbers, namely, extra large (>2,500), large (1,500-2,499) and medium (500-1,499). We randomly selected at least one school from the list of three school categories: urban and rural public schools and private schools. The selection of schools was based on a list of REVISED Amendments from Version 1 In this final version of our manuscript, and after reading the comments and recommendations of the reviewer, we highlighted some words to add and revise, namely the 95%CI of PSQ, justify the definition of PSQ by using the Pittburgh Sleep Quality Index (PSQI) for poor and good sleepers, provinces of Thailand, design effect of 2 and added 10% for nonresponse, the data were verified, using the enter method, it depicted the difference of the media use in the PSQ group, as shown in Table 5.
Any further responses from the reviewers can be found at the end of the article schools obtained from the Provincial Education Office and willingness of school administrators to participate in the study. For each of the schools, the student sample size was calculated proportional to the size of the schools.

Ethical approval
The study was conducted in accordance with the ethical principles in the Declaration of Helsinki, and the protocol was reviewed and approved by the Human Research Ethical Review Committee of the Faculty of Public Health, Mahidol University (COA. No. MUPH 2016-097). The purpose of this study was explained to school principals and teachers of the target schools. Permission was obtained from these schools and students; written informed consent was obtained from the student's parents or legal guardians after informing them of the study details (the objectives of study, methods and protection of human rights). Parents or legal guardians were told that participating in the survey was voluntary and that the survey would remain anonymous. Confidentiality was maintained throughout the study using anonymous technique (schools and respondents were identified by code numbers to ensure confidentiality and the results were analyzed as a whole group).

Procedures
Study population was senior high school students grades 10-12 during the educational year 2016 in Ratchaburi province.

Inclusion criteria
a) Students who studied in grades 10-12.
b) Students who studied in high schools that were under the control of the secondary education service area office 8, Ratchaburi province c) Students who were willing to participate in the study and provided the written informed consent. d) Students who provided the written informed consent signed by their parents or legal guardians.
Exclusion criteria a) Students who were absent from school on a period of data collection.
b) Students who were chronically ill during the time of study.
Researchers contacted the educational administrators and the teachers for data collection. The paper-based questionnaire was provided for the participants to fill data at the free time from studying at their school. Researcher and research assistants explained the details of questionnaire and answered the questions from participants. This process was approximately 40 minutes. Information was collected using a self-administered, anonymous questionnaire comprising three parts, namely, demographics, consumption behaviors relating to sleep quality, sleep quality assessment and mediaused evaluation. A copy of the questionnaire can be found in the Extended data. 35 Sleep quality was evaluated using the Pittsburgh Sleep Quality Index (PSQI) translated to Thai with a cutoff point of scores > 5 was classified as poor sleepers and ≤5 was classified as good sleepers. [20][21][22][23] Reliability was tested revealing a Cronbach's alpha of 0.86.

Data analysis
The data were verified, encoded and processed for statistical analysis by using SPSS for Windows, Version 18). Categorical variables were given as frequency and percentage, crude odds ratio (OR c ), 95%CI of OR and p-value. Moreover, numerical variables were expressed as mean, median, minimum and maximum, standard deviation and quartile deviation. Univariate analysis was performed using univariable logistic regression to differentiate proportional exposures between poor and good sleepers for categorical variables. Adjusted odds ratio (OR adj ) and 95%CI of OR were calculated from multivariable logistic regression to examine associations between media use and PSQ occurrence, adjusted for potential confounders using the enter method. All statistics were performed using two-sided tests, and the criteria of p <0.05 was judged to be statistically significant.

PSQ and associated factors
The prevalence of PSQ was 56.24% (95%CI: 52.75-59.74). Using univariable logistic regression analysis, associated demographic factors of PSQ among adolescents included illness history during the last month, coffee and tea consumption, reading, annoyance, poor ventilation, stress, depression and sleep duration (p < 0.05), as shown in Table 2. In case of media use, we found an association between social media use and PSQ (OR = 1.53, 95%CI = 1.13-2.08), as shown in Table 3. Using multivariable logistic regression analysis, regarding association between social media use and PSQ among adolescents (adjusted for potential confounders), social media users were 1.34 times at risk compared with those of nonusers (OR = 1.34, 95%CI = 0.97-1.87) but without significance, as shown in Table 4. Comparing PSQ and good sleep quality (GSQ) groups, the most commonly activity before bedtime was social media (44.56%, 37.38%) and television watching (20.78%, 30.29%) respectively. Further, we found a higher proportion of social media use, it depicted the difference of the media use in the PSQ group, as shown in Table 5.

Discussion
Our findings demonstrated that PSQ prevalence rate was about 56% higher than related studies conducted in Thailand. 7, 8 Evidence from related studies on PSQ among college students showed PSQ prevalence was approximately from 32 to 62%. 7-11 The difference of PSQ occurrence might have stemmed from various factors, namely, environment, lifestyle, household characteristics, social media and activities, health behaviors etc. Univariable analysis showed that social media use played a critical role in the development of PSQ among adolescents (OR = 1.53, p = 0.006). However multivariable logistic regression analysis did not indicate significant differences (OR = 1.34, 95%CI = 0.97-1.87). Some studies indicated adolescents who used social media before bedtime had lower sleep efficiency. [15][16][17][24][25][26][27] Mobile phone use among young students for daily calling, using e-mail, text messaging and social network services were associated with short sleep duration, PSQ, excessive daytime sleepiness and presenting insomnia symptoms. 15,28,29 Higher frequency and volume of social media use had significantly greater odds of having sleep disturbance among young adults, 24,26 while one study showed a better sleep quality among users. 30 The present study showed the prevalence of social media use before bedtime in the PSQ group was approximately 44.56%. One half of social media users spent over 2 hours per day. The average time for social media use was 3.58 hours per day, and this might have affected sleep pattern. A related study showed users who spent 0.5 to 2 hours per day on social media were more likely to have poor sleep than those of spent less than 0.5 hours. 31 In addition, the meta-analysis studies reported social media users before bed were more likely to have insufficient sleep and tended to have PSQ. 27,32 Some related studies have suggested blue light emitted from smart phones might disturb sleep. 33,34 Therefore, monitoring social media use among adolescents, and cooperating with parents, caregivers, teachers and the adolescents themselves is recommended to reduce PSQ problems.

Study limitations
This study encountered a few limitations that need to be addressed. First, cross-sectional surveys reduced the ability of the study to make direct causal inferences. Second, these data apply only to those aged 14-19 years as the study subjects;  therefore, they could not represent all adolescents. Moreover, data collection might have excluded subjects absent from schools. Finally, all data were based using a self-report method subject to recall bias.

Conclusion
PSQ surveillance systems should be established along with knowledge sharing programs regarding associated factors of PSQ among adolescents with their parents and teachers. We recommend that the use of media and the presence of media equipment in bedroom should be limited. This may be beneficial to sleep quality.

Data availability
Underlying data OSF: Association between media use and poor sleep quality among senior high school students: a cross-sectional study. https://doi.org/10.17605/OSF.IO/KV2BJ. 35 This project includes the following underlying data.

-SAV Dataset
Extended data OSF: Association between media use and poor sleep quality among senior high school students: a cross-sectional study. https://doi.org/10.17605/OSF.IO/KV2BJ. 35 This project includes the following extended data. -

Sumattana Glangkarn
Faculty of Public Health, Mahasarakham University, Kantarawichai, Thailand I have read the version 2 that has been updated from the version 1 comments. It also reveals the limitations of the study due to it being a cross-sectional study and the subjects were not representative of all adolescents. The follow up study should be recommended for further research and expand the target population.

If applicable, is the statistical analysis and its interpretation appropriate? Yes
Are all the source data underlying the results available to ensure full reproducibility? Yes Are the conclusions drawn adequately supported by the results? Yes This manuscript addresses a very important issue for solving the problem of the adolescent group. However, some components should be considered and added up for fulfilling the content to implication or implementation. I have the following comments: The gap of knowledge for the association between media use and poor sleep quality (PSQ) among senior high school students should be identified clearly. For example, the magnitude of related factors for such an association, in previous studies, have not been justified based on some limitations (conditions). This study can add something more regarding detecting the precision or properly-related factors compared to previous studies 1.
Media use is a key factor -I would like to know how it is associated with the PSQ in this study, but the authors didn't mention the factor in the manuscript. Therefore, the media use should be identified clearly in the methods part.

2.
Potential confounders which are included in the multi-variable analysis step should also be identified in the methods section. According to such confounders, some studies ( For explicitly identifying the magnitude of estimation, the 95% CI of the PSQ prevalence should be added.

4.
In Table 5, I am confused regarding PSQ and good sleep quality (GSQ) for each type of media use. Given that they are opposites, a table comparing them may be difficult to interpret and lead to misunderstanding. Also, the total of each row doesn't add up to 100%, could you please explain how you want to want to present this table, and how you calculate the percentage of both (PSQ and GSQ) in each type of the media use? I suggest that the column of GSQ should be removed from the table, because the remainder of PSQ can be interpreted as GSQ. However, each row cannot be calculated total as 100%, so please describe more for this table.

5.
According to the results in Table 3, 4 and 5, data management should be mentioned and described in the part of data analysis, regarding how social media is considered and included into the final model.

6.
Issues with the statistical method issues should be mentioned, in particular the finding that is not statistically significant. For example, is the standard logistic regression appropriate for this data or not? As students from the same room may be correlated, while some students from the different room may be independent. Therefore, the statistical methods for correlating the outcome, like the generalised estimating equation (GEE) and logistic regression, should maybe be analysed based on identifying the cluster as with the room 7.
Finally, even if the data of this study was collected before the transmission of the COVID-19, the issue of media use can change, and affect the topic. I think the author should mention 8.
this point in the discussion or limitation part. quality (PSQ) among senior high school students should be identified clearly. For example, the magnitude of related factors for such an association, in previous studies, have not been justified based on some limitations (conditions). This study can add something more regarding detecting the precision or properly-related factors compared to previous studies calculated total as 100%, so please describe more for this table.
Response Table 5 showed the proportion of media use among the PSQ and the GSQ. We would like to seek the difference of the media use in both groups.