Does Sleep Mediate the Association between School Pressure, Physical Activity, Screen Time, and Psychological Symptoms in Early Adolescents? A 12-Country Study

Round 1

Reviewer 1 Report

Title:      Does sleep mediate the association between school pressure, physical activity, screen time and psychological symptoms in early adolescents? A 12 country study.

Summary:  Using a large sample (>49000 children), the relationship between sleep, school pressure, screen time, physical activity and psychological symptoms was analyzed. Data support that higher physical activity is protective against psychological symptoms and increased screen time and school pressure was associated with higher rates of psychological concerns. Sleep (duration and presence of difficulties falling asleep) mediated the relationship between all variables with the exception of physical activity and psychological symptoms. This would support interventions targeting improving sleep and increasing physical activity as a means to address high rates of psychological concerns in this population.

Review

Introduction

Overall, well written and concise. No significant concerns.

Methods

Description of measure and how it was administered was clearly written and easy to follow.

Given that sleep quality is comprised of 7 variables per the author, is it accurate to state that sleep onset difficulties truly serve as a proxy for sleep quality? The wording of this should be tempered slightly given that you are looking at only one variable.

With my limited understanding of statistical analyses, the approach appears congruent with good practice; however, would support independent review of statistical analyses if deemed warranted by other reviewers and/or editors.

Results

Results were clearly presented.

Discussion

Appreciated how the impact of sleep variables were captured (2^nd paragraph in discussion, “per increasing hour of sleep on week- or weekend days, the score for psychological symptoms decreased…”). Made it clear and tangible.

It would be important to highlight that while results were statistically significant, the clinical significance remains to be seen. As an example, are there any studies suggesting that the difference of 4.56 minutes of sleep duration per day (under results section, 3.3.2) makes any functional difference during the day? Does the decrease of psychological symptoms of 0.13 and 0.04 on a 5 point scale (under discussion, paragraph 2) change the presentation of students during the day?

The authors propose an intervention focusing on the variables outside of sleep. Given the mediating effects of sleep, are there are any data to suggest that targeting sleep could improve the other variables? That is, what is the role of sleep hygiene in such an intervention?

Author Response

We wish to thank the editor and reviewer for their insightful comments and to give us the opportunity to revise our manuscript. We have addressed the reviewers’ comments chronologically in our response below, and have highlighted the changes in the manuscript. We hope to have sufficiently addressed the reviewer.

Introduction: Overall, well written and concise. No significant concerns.

Methods: Description of measure and how it was administered was clearly written and easy to follow.

1.      Given that sleep quality is comprised of 7 variables per the author, is it accurate to state that sleep onset difficulties truly serve as a proxy for sleep quality? The wording of this should be tempered slightly given that you are looking at only one variable.

We have added this concern as a limitation to this study.

Page 11, line 210 : ‘The HBSC study collects data on a variety of themes, each theme is therefore questioned as briefly as possible.’

Page 11 , line 214: ‘Sleep onset difficulties were used as a proxy for sleep quality, while sleep quality is defined by several components. Ideally, all seven components of sleep quality (sleep quality, sleep latency, sleep duration, habitual sleep efficiency, sleep disturbances, sleeping medication use, and daytime dysfunction) according to Buysse, Reynolds, Monk, Berman, & Kupfer (1989), should be measured.’

With my limited understanding of statistical analyses, the approach appears congruent with good practice; however, would support independent review of statistical analyses if deemed warranted by other reviewers and/or editors.

Results: Results were clearly presented.

Discussion: Appreciated how the impact of sleep variables were captured (2nd paragraph in discussion, “per increasing hour of sleep on week- or weekend days, the score for psychological symptoms decreased…”). Made it clear and tangible.

We wish to thank the reviewer for this positive feedback.

2.      It would be important to highlight that while results were statistically significant, the clinical significance remains to be seen. As an example, are there any studies suggesting that the difference of 4.56 minutes of sleep duration per day (under results section, 3.3.2) makes any functional difference during the day? Does the decrease of psychological symptoms of 0.13 and 0.04 on a 5 point scale (under discussion, paragraph 2) change the presentation of students during the day?

A difference of 4.5 minutes of sleep is indeed not considered as clinically significant. The clinical significance of the decrease of psychological symptoms of 0.13 and 0.04 points on a 5 point scale is difficult to estimate, as there are no norm values for this scale.

This has been added as a limitation to the study.

Page 11 ,  line 206: ‘Secondly, although all associations between the independent variables and sleep duration were found to be significant, the magnitude of the associations was rather small. It is possible that statistical significance was reached for such small associations because of the large sample size. In addition, although statistically significant, there is no clinical relevance of i.e. 4.56 minutes of extra sleep. Furthermore, there are no norm values for the scale of psychological symptoms, which makes the estimation of the clinical significance of the results (i.e. decrease of psychological symptoms of 0.13 and 0.04 on a 5 point scale) difficult. To avoid this, future research could use the Depression Anxiety Stress Scales (DASS), as this scale does provide cut-off scores.’

3.      The authors propose an intervention focusing on the variables outside of sleep. Given the mediating effects of sleep, are there are any data to suggest that targeting sleep could improve the other variables? That is, what is the role of sleep hygiene in such an intervention?

Focussing on sleep hygiene would certainly be valuable in an intervention to improve mental well-being in adolescents. Adequate sleep is associated with more physical activity and reduced stress in adolescents. We have added this to the discussion.

Page 11, line 188: “In addition, directly focusing on  sleep hygiene would be a valuable addition to mental health promotion programs  as well, as adequate sleep is in its turn also associated with higher levels of physical activity, which is associated with better mental health, and with reduced stress levels (43).”

Reviewer 2 Report

The study examines the role of sleep quality: duration and sleep onset in the association of school pressure, physical activity and screen-time with psychological symptoms in early adolescents. 

It is based on questionnaires of 49403 children in 12 countries participating in the WHO ‘Health Behaviour in School-aged Children’ 2013/2014 study. Data are analysed in Department of Public Health, Faculty of Medicine and Health Sciences, Ghent University, Belgium. 

The topic of sleep quality, life style and psychological symptoms in adolescents is of current and general interest.

Overall the quality of the study is satisfactory. The statistical data analysis is sound and the results are significant. The conclusion is of interest in the research field. Consequently it deserves publication in its present form. 

Author Response

We wish to thank the editor and reviewer for their insightful comments and to give us the opportunity to revise our manuscript. We have addressed the reviewers’ comments chronologically in our response below, and have highlighted the changes in the manuscript. We hope to have sufficiently addressed the reviewer.

Response to reviewer 2

The study examines the role of sleep quality: duration and sleep onset in the association of school pressure, physical activity and screen-time with psychological symptoms in early adolescents.

It is based on questionnaires of 49403 children in 12 countries participating in the WHO ‘Health Behaviour in School-aged Children’ 2013/2014 study. Data are analysed in Department of Public Health, Faculty of Medicine and Health Sciences, Ghent University, Belgium.

The topic of sleep quality, life style and psychological symptoms in adolescents is of current and general interest.

Overall the quality of the study is satisfactory. The statistical data analysis is sound and the results are significant. The conclusion is of interest in the research field. Consequently it deserves publication in its present form.

The authors strongly appreciate this positive feedback of the reviewer.

Reviewer 3 Report

The present study reported a multilevel analysis of data from 49,403 students from 12 countries to assess whether sleep duration or sleep onset latency (SOL; a marker of sleep quality) mediates the relationship between school pressures, physical activity, and screen time on the outcome variable of psychological symptoms.  With the exception of the association of physical activity and psychological symptoms via sleep duration, both sleep duration and sleep quality produced significant indirect pathways to the outcome.  The authors concluded that the three predictors are potentially modifiable direct factors in the treatment of psychological problems and, indirectly, via sleep.

The study’s epidemiological perspective, having a large sample size that was inclusive of adolescents from a variety of European countries, is a strength.

Make explicit in the introduction the relevance of this study to readers of this journal (relate to environmental research or public health).

Method

Why were these 12 countries selected of the 40, and did it make sense to combine them?  I assume the goal was to assess students in Europe.  Is there a reason for excluding some European countries?  Since country was one of the levels analyzed, provide the reader with statistical evidence that it made sense to combine them (that there weren’t significant country differences in the reported variables). 

Along that same line, country differences in climate and daylight hours might have contributed to sleep duration or onset (or physical activity or screen time, for that matter), strengthening the need to demonstrate why that level was dropped from the analysis.

I would like to see much more information about the measures.  Although I realize they are described elsewhere in the literature, readers need to know psychometric properties, whether they have been validated, and what exactly they measure.  “Psychological symptoms” seems to be a vague term. 

Justify the use of variables that are based on a single item.  The time of year and context (home? school?) in which data were collected also may be helpful information.

Results

Throughout the Results section, B and β seem to be used interchangeably (text and table), so I’m not clear on whether the authors are reporting standardized or unstandardized coefficients.  Please clarify.

I realize that journal preferences vary, but it does not seem helpful to duplicate in the text numbers that are already in tables. 

How many participants were excluded due to missing data?

Descriptive and correlational data are scattered throughout the results, but moving the information to a single table would be helpful.

Discussion

Line 126 indicates that sleep onset latency “is the key aspect of the sleep experience.”  I pulled the Cairns et al. 2014 study and did not find this claim in this source.  While it is true that sleep quality often shows the strongest associations with outcomes (there are many studies that can be cited to support this), other work has found that sleep duration may also have significant contributions. 

Line 135 claims that the present study indicates the pattern of findings confirmed “on a European level” the findings of an earlier study from Norway.  This seems like a strong statement to make based on the results of one study.

The authors acknowledge that the indirect effect coefficients for sleep duration were small, although significant.  They were very small, and the significance may be due to the tremendous power of a sample size that large.  Perhaps this could be addressed as a limitation.

Other limitations include the self-reported measures and single-item variables. 

There are a few minor spelling or grammatical errors that I assume will be caught during copy editing.

Author Response

Response to reviewer 3

We wish to thank the editor and reviewer for their insightful comments and to give us the opportunity to revise our manuscript. We have addressed the reviewers’ comments chronologically in our response below, and have highlighted the changes in the manuscript. We hope to have sufficiently addressed the reviewer.

The study’s epidemiological perspective, having a large sample size that was inclusive of adolescents from a variety of European countries, is a strength.

1.      Make explicit in the introduction the relevance of this study to readers of this journal (relate to environmental research or public health).

The following was added to the introduction to make the relevance of the study more clear to readers of this journal.

Page, line 6-7: “Given the high prevalence of mental health problems in European adolescents, the prevention of these problems is an important public health issue.”

2.      Why were these 12 countries selected of the 40, and did it make sense to combine them?  I assume the goal was to assess students in Europe.  Is there a reason for excluding some European countries? 

Within the HBSC study, some questions are mandatory (all participating countries must include them), others are optional. The variables that were used to calculate sleep duration used in this paper, were part of an optional package ‘psychological health’. We included all countries that had all data that were needed for the purpose of this paper. To clarify this, the following was added to the method:

Page 2, line 43: ‘Every participating country can choose to add an optional package of questions to this questionnaire.’

Page 3, line 57-58: ‘As sleep duration was part of an optional package within the HBSC-study, only countries that had questioned all above-mentioned variables were included in this study.’

3.      Since country was one of the levels analyzed, provide the reader with statistical evidence that it made sense to combine them (that there weren’t significant country differences in the reported variables).  

Only 2 levels could be added to the analyses (Multilevel mediation analyses conducted in R). Explained variance was assessed for all levels. It made sense to only include the level of classes (7% explained variance) in the analyses and not the level of country or school, as the explained variance for country and school were both smaller than 1% (0.8%).

We have added the percentages of explained variance to the methods in this paper.

Page 7, line 7: ‘Only the clustering of individuals within classes was taken into account, as it had the largest explained variance of the three levels (7% for class-level vs. 0.8% for school- and country-level).’

4.      Along that same line, country differences in climate and daylight hours might have contributed to sleep duration or onset (or physical activity or screen time, for that matter), strengthening the need to demonstrate why that level was dropped from the analysis.

We agree that climate and daylight hours might influence sleep duration and the other variables PA en ST. But as it was only possible to add two levels in the analyses, we decided to add the class level in the analyses based on the higher explained variance in the dependent variable psychological symptoms (see also answer on comment 3).

5.      I would like to see much more information about the measures.  Although I realize they are described elsewhere in the literature, readers need to know psychometric properties, whether they have been validated, and what exactly they measure.  “Psychological symptoms” seems to be a vague term.  

For the variable ‘psychological symptoms’ three psychological symptoms of the HBSC Symptom Check List (HBSC-SCL) were used. The HBSC-SCL consists of 4 questions on somatic and 4 questions on psychological symptom load.  As the HBSC study refers to the symptoms as ‘psychological symptoms’ we decided to use this term as well.

We have added information about validation of the used measures in Table 1.

6.      Justify the use of variables that are based on a single item. 

For this study we used the available data from the HBSC study. We have added the fact that some variables were based on single items as a limitation of the study.

Page 12, line 208: ‘Some variables (such as school pressure or sleep onset difficulties) were therefore constrainedly based on a single item.’

7.      The time of year and context (home? school?) in which data were collected also may be helpful information. 

As described in line 49, adolescents completed this questionnaire at school: ‘Adolescents complete the international standard questionnaire developed by the research network of HBSC in the classroom after instruction by a teacher or a trained interviewer’

Data collection happened between October 2013 and December 2014, different time of year for each included country. If necessary, an overview for each country could be added. We have added this information to the methods in this paper.

Page 3, line 49: ‘Data collection happened between October 2013 and December 2014, during a different time of year for each included country.’

8.      Throughout the Results section, B and β seem to be used interchangeably (text and table), so I’m not clear on whether the authors are reporting standardized or unstandardized coefficients.  Please clarify.

We used standardized coefficients (β), this has been clarified in the results section (page 8).

9.      I realize that journal preferences vary, but it does not seem helpful to duplicate in the text numbers that are already in tables.  

We prefer to keep the numbers in the text, as we think it would be difficult to follow the results section without mentioning the numbers.

10.   How many participants were excluded due to missing data?

We added the information to the methods of the paper.

Page 7, line 14: ‘9433 participants were excluded due to missing data.’

11.   Descriptive and correlational data are scattered throughout the results, but moving the information to a single table would be helpful. 

All information is summarized in Table 2.

Page 9, line 100: Table 2

12.   Line 126 indicates that sleep onset latency “is the key aspect of the sleep experience.”  I pulled the Cairns et al. 2014 study and did not find this claim in this source.  While it is true that sleep quality often shows the strongest associations with outcomes (there are many studies that can be cited to support this), other work has found that sleep duration may also have significant contributions.  

Cairns et al. was indeed not the right source. We apologize for this. We intended to refer to following research: Bei B, Wiley JF, Allen NB, Trinder J. A cognitive vulnerability model of sleep and mood in adolescents under naturalistically restricted and extended sleep opportunities. Sleep. 2015;38(3):453.

We agree that sleep duration is an important aspect of sleep perception as well. We have deleted the following on page 10, line 129: ‘, which could be due to the fact that sleep onset is considered to be the key aspect of sleep experience (33) ‘

Line 135 claims that the present study indicates the pattern of findings confirmed “on a European level” the findings of an earlier study from Norway.  This seems like a strong statement to make based on the results of one study.

We changed the claim to ‘broader European Level’.

Page 10, line 135: ‘In a previous Norwegian study, school stress was positively related to adolescents’ psychological symptoms as well (35), which is now confirmed on a broader European level.’

13.   The authors acknowledge that the indirect effect coefficients for sleep duration were small, although significant.  They were very small, and the significance may be due to the tremendous power of a sample size that large.  Perhaps this could be addressed as a limitation.

This was added as a limitation as suggested.

Page 12, line 207: ‘Secondly, although all associations between the independent variables and sleep duration were found to be significant, the magnitude of the associations was rather small. It is possible that statistical significance was reached for such small associations because of the large sample size.’

Other limitations include the self-reported measures and single-item variables. 

This was added as a limitation.

Page 12, Line 215: ‘Finally, this research was limited by the measurements used in the self-reported HBSC-study. The HBSC study collects data on a variety of themes, each theme is therefore questioned as briefly as possible. Some variables (such as school pressure or sleep onset difficulties) were therefore constrainedly based on a single item.’

There are a few minor spelling or grammatical errors that I assume will be caught during copy editing.

Reviewer 4 Report

Author Response

Response to reviewer 4

The research is well conducted for the publication of the journal. Writing is also fine.

We wish to thank the editor and reviewer for their insightful comments and to give us the opportunity to revise our manuscript. The authors strongly appreciate this positive feedback of the reviewer.

Round 2

Reviewer 3 Report

I have read the authors' comments in response to my questions, and my concerns have been sufficiently addressed.