Barthorpe, A. E., Winstone, L., Mars, B., & Moran, P. A. (2020).

Is
social media screen time really associated with poor adolescent
mental health? A Time Use Diary Study. Journal of Affective
Disorders, 274, 864-870. https://doi.org/10.1016/j.jad.2020.05.106

Peer reviewed version

Link to publication record on the Bristol Research Portal

This is the author accepted manuscript (AAM). The final published version (version of record) is available online
via Elsevier at https://doi.org/10.1016/j.jad.2020.05.106 . Please refer to any applicable terms of use of the
publisher

University of Bristol – Bristol Research Portal

This document is made available in accordance with publisher policies. Please cite only the
published version using the reference above. Full terms of use are available:
http://www.bristol.ac.uk/red/research-policy/pure/user-guides/brp-terms/
 Is social media screen time really associated with poor adolescent mental health? A
Time Use Diary Study.

Amber Barthorpe12, Lizzy Winstone134, Becky Mars4, Paul Moran4

Background
There is increasing concern regarding the potential impact of social media use on the mental health of
young people. Previous research has relied heavily on retrospective accounts of social media screen-
time. Yet recent evidence suggests that such self-report measures are unreliable, correlating poorly
with more objective measures of social media use. In principle, time use diaries provide a less biased
measure of social media use.

Methods
We analysed cross-sectional data from the Millennium Cohort Study to explore associations between
social media screen-time as recorded in time use diaries (TUD) and key mental health outcomes –
self-harm in the past year, depressive symptoms (Short Mood and Feelings Questionnaire), self-
esteem (shortened Rosenberg scale) – in adolescence. Social media TUD data were available for
4,032 participants (25.4% aged 13; 73.5% aged 14; 1.1% aged 15).

Results
Following adjustment for confounders, a greater amount of time spent on social media was associated
with an increased risk of self-harm (adjusted OR per 30-minute increase in weekday use: 1.13, 95%
CI 1.06 to 1.21) and depression (adjusted OR=1.12, 95%CI 1.07 to 1.17) and lower levels of self-
esteem (adjusted B = -0.12, 95%CI -0.20 to -0.04) in females. Findings were similar for weekday and
weekend use.

Limitations
The cross-sectional nature of the data limits inference in relation to the causal direction of these
associations.

Conclusions
Future research should examine the direction of the associations with self-harm and other mental
health outcomes and explore how adolescents engage with social media as well as how much time
they spend online.

Keywords: Social media; Adolescence; Self-harm; Depression; Self-esteem; Millennium Cohort

1
Joint first author
2
South London & Maudsley NHS Foundation Trust, UK
3
correspondence to Lizzy Winstone, Population Health Sciences, Bristol Medical School, University of Bristol,
Bristol BS8 2BN, UK; lizzy.winstone@bristol.ac.uk
4
Population Health Sciences, Bristol Medical School, University of Bristol, Bristol BS8 2BN, UK

1
Introduction

Social media is a relatively new phenomenon, though today’s adolescents have grown up in a world

in which it has always been present. Social media use amongst adolescents has risen rapidly over the

last 16 years since the launch of Facebook; according to Ofcom (2019) just under 70% of 12-15 year-

olds in the UK have a social media profile, with over a third agreeing that they found it difficult to

control their screen-time (Ofcom, 2019). Over the same period of time, rates of suicide, non-suicidal

self-harm and common mental disorder among young people in the UK have increased, particularly

among females (McManus et al, 2019; Morgan et al, 2017). Given the ramifications of poor mental

health in adolescence, in terms of academic attainment, employment and mental health in adulthood

(Lopez-Lopez et al, 2020), it is vital that we enhance understanding about the potential relationship

between social media use and adolescent mental health.

Findings from several studies suggest that social media use may be associated with poorer

mental health in adolescents. With poor mental health becoming increasingly prevalent among female

adolescents in particular, the question has been raised as to whether social media – used more

intensively by females – may be an important contributor to this gender difference (Gunnell et al,

2018). Using cross-sectional questionnaire data derived from the Millennium Cohort Study, Kelly et

al (2018) found that time spent on social media at age 14 was associated with depressive symptoms,

with a stronger association in females than males. Using Specification Curve Analysis of cross-

sectional data from the Millennium Cohort Study and Monitoring The Future, Orben and Przybylski

(2019a) also found a negative association between social media use and adolescent well-being,

although the effect size was small. Prior studies of the impact of social media on mental health have

largely focused on depression, anxiety and cognitive or affective wellbeing. To date, little research

exists examining an association between social media screen-time and risk of self-harm (George,

2019). Self-harm not only has immediate physical consequences but is associated with increased risk

of suicide (Muehlenkamp & Guttierrez, 2007). Levels of self-harm have risen substantially over

recent years, particularly among 13-16 year-old girls (Morgan et al, 2017), making the lack of

evidence about associations with social media use a key knowledge gap.

2
 Previous research has been limited by retrospective self-report estimates of time spent using

social media on a ‘typical’ week-day, with evidence indicating only moderate correlations with

objective measures such as apps or log files (Andrews at al, 2015) or ‘aggregated in-situ measures

such as experience sampling’ (Naab et al 2019). In the absence of accessibility of data from social

media companies themselves, usage data from apps likely represents the most reliable measure of

social media screen-time, but installation of apps is difficult to implement in large representative

samples and log files are difficult to amalgamate for all devices (Scharkow 2016). Time use diaries

(TUD), using methodology akin to experience sampling, may be a good compromise, whereby

respondents are asked to record their social media use (amongst other activities) in small units of time

over a specific day (Orben and Przybylski 2019b). TUD are subject to less recall bias, are more

specific rather than considering a ‘typical’ day, and likely introduce less response bias than

questionnaire measures which impose more limiting response categories on the respondent (Naab et al

2019).

The current study extends previous research by using TUD to assess social media use in a

large-scale representative sample of UK adolescents. We investigated associations between time spent

on social media and self-harm, as well as depressive symptoms and self-esteem, taking account of

prior mental health problems. We hypothesise that social media screen-time reported in TUD will be

negatively associated with mental health, based on previous research demonstrating this with

retrospective self-report measures of screen-time. Consistent with previous findings, we also expect

these associations to be more compelling in females than males.

3
Method

We utilised cross-sectional data from the sixth sweep (2015) of the nationally representative birth

cohort Millennium Cohort Study (MCS), when respondents were aged 13-15 years. At this sweep

interview data are available for 11,884 individual cohort members, with measures including mental

health and well-being and demographic information. The sample comprised 2,864 13-year-olds, 8,860

14-year-olds and 160 15-year-olds (50% female). In this sweep, 10,337 cohort members were also

invited to complete time use diaries (TUD) recording activities from a pre-defined list, performed

during two 24-hour periods; a randomly selected weekday and a weekend day. Diaries were recorded

on paper, online or via an app. The smallest unit of activity time depended on the modality of

collection; 1-minute units via the app and 10-minute units online or on paper. Data were then recoded

into 30-minute units for analysis and treated as a continuous variable. 4,642 participants completed

and returned time use diaries (TUD) (26% aged 13, 73% aged 14, 1% aged 15; 55% female). Social

media data was available for 4,032 participants (3,800 on a weekday and 3,814 on a weekend day).

Further information on the MCS is available from the Centre for Longitudinal Studies

(http://www.cls.ioe.ac.uk/mcs). Cohort members who completed the time use diary data and had data

on social media use were significantly younger and more likely to be female, White and from a higher

income household (see table 1).

Measures

Social media use Time use diaries (TUD) were used to record time spent on social media. Social

media use was assessed via the activity ‘browsing and updating on social networking sites, e.g.

Twitter, Facebook, BBM, Snapchat’. Associations were explored separately for weekday and

weekend use.

Outcome measures

1) Self-harm: Participants responded yes or no to the question ‘In the past year have you hurt yourself

on purpose in any way?’.

4
2) Depressive symptoms: Assessed using the Short Mood and Feelings questionnaire (SMFQ, Angold

et al, 1995). The SMFQ includes 13 negatively worded questions - such as ‘I cried a lot’ and ‘I felt

lonely’ - about the past two weeks with responses given on a three-point scale from 0 (not true); 1

(somewhat true); 2 (true). Summed scores ranged from 0 to 26, with higher scores indicating

increased depressive symptoms. For sensitivity analysis, a binary variable was also created with a cut-

off point of 11 or more suggested by previous research as indicative of high risk of depression

(McKenzie et al 2011).

3) Self-esteem: Assessed using the 5-item shortened Rosenberg scale measuring self-satisfaction

(Rosenberg, 1989). Respondents were asked to indicate agreement or disagreement with five

positively worded statements (for example ‘on the whole, I feel satisfied with myself’ and ‘I feel good

about myself’). Response options were Strongly agree, Agree, Disagree and Strongly disagree. To

facilitate interpretation the variables were recoded so that higher scores indicate greater self-esteem,

with summed scores ranging from 0 to 15.

Confounders

We adjusted for several potential confounding variables based on a-priori grounds and previously

published literature on self-harm, including age, ethnicity, socioeconomic status (equivalised income

quintiles) and parent-reported internalising symptoms at age 11. Internalising symptoms at age 11

were measured using the summed ‘peer relationship problems’ and ‘emotional symptoms’ subscales

of the Strengths and Difficulties Questionnaire (SDQ, Goodman, 2001).

Statistical analysis

Logistic and linear regression models were used to explore associations between time spent on social

media and mental health outcomes. In line with previous research indicating the association between

social media use and mental health may be stronger in females than males (e.g. Kelly et al 2018)

findings are presented separately by gender. Data were weighted according to MCS guidance. All

analyses were completed using STATA software version 15.

5
Results

Self-harm

Self-harm in the past year was reported by 14.7% of the sample (20.2% of females and 6.5% of males,

see table 2). Following adjustment for confounders, a greater amount of time spent on social media on

a weekday was associated with an increased risk of self-harm for females (adjusted OR per 30-minute

increase in social media use = 1.13, 95%CI 1.06 to 1.21), see table 3. For weekend use, effect

estimates were consistent with an increased risk of self-harm for both males and females, however

there was little statistical evidence for an association (the confidence intervals include the null)

(adjusted OR for females =1.04, 95%CI 0.98 to 1.10; adjusted OR for males = 1.06, 95%CI 0.97 to

1.16)

Depression

Females reported higher levels of depressive symptoms (Mean=6.78, SD =6.33) than males (Mean=

3.99, SD= 4.38, see table 2). . For females, increasing time spent on social media was associated with

a greater number of depressive symptoms (see table 4). Findings remained after adjusting for

confounders and were consistent for weekday and weekend use (adjusted β per 30-minute increase in

social media use: weekday use = 0.36, 95%CI 0.22 to 0.50; weekend use = 0.19, 95%CI 0.06 to 0.32).

There was little evidence for an association between time spent on social media and depression in

males. Results were consistent in sensitivity analysis with depression included as a binary outcome

(indicating high risk of depression) (supplementary table 1).

Self-esteem

Males reported higher levels of self-esteem (Mean=11.49, SD =2.54) than females (Mean= 9.92, SD=

2.88, see table 2). For females, increasing time spent on social media was associated with increased

6
risk of poor self-esteem (see table 5). Findings remained after adjusting for confounders and were

consistent for weekday and weekend use (adjusted β per 30-minute increase in social media use:

weekday use = -0.12, 95%CI -0.20 to -0.04; weekend use = -0.12, 95%CI -0.18 to -0.07). There was

little evidence for an association between time spent on social media and self-esteem in males.

7
Discussion

In this large population-based study of adolescents, using novel time use diary data, we found that a

greater amount of time spent on social media was associated with increased risk of self-harm and

depression and lower levels of self-esteem in females. Findings persisted after adjustment for

covariates, including prior mental health problems. This study provides important new evidence in

relation to the knowledge gap around potential ‘toxic’ effects of screen-time, as identified by the UK

Royal College of Paediatrics and Child Health (https://www.rcpch.ac.uk/resources/health-impacts-

screen-time-guide-clinicians-parents). The findings here show that 30-minute increments in exposure

were independently associated with a clear increase in depressive symptoms and risk of self-harm and

poor self-esteem.

Our findings are consistent with previous research demonstrating negative associations between social

media screen-time reported retrospectively and adolescent mental health and extends previous

findings on depression (Boers et al 2019), anxiety (Woods & Scott 2016), self-esteem (Boers et al

2019; Woods & Scott, 2016) and well-being (Booker et al 2018; Orben et al 2019) to demonstrate

associations between social media screen-time and increased risk of self-harm in adolescence. Gender

differences demonstrated here are also consistent across the literature, with evidence of negative

associations between social media screen-time and mental health generally more compelling in

females than males (Booker et al 2018; Kelly et al 2018). It should be noted however, that there may

have been less power to detect an association in males; females were over-represented in the TUD

sample, including when survey weights were applied. Results here, as elsewhere, indicate that female

adolescents report spending more time using social media and report poorer mental health than their

male peers. The reasons underlying the difference in screen-time are unclear and are likely to be

multifaceted. The increased risk of poor mental health in female adolescents compared to males is

already well established, and a variety of contributing factors have been identified, including both

biological (e.g. hormones) and environmental (e.g. sexual harassment, pressure to conform to societal

expectations of body image (Gunnell et al 2018; Kelly et al 2018; Kuehner 2017). While there is some

8
evidence that cyberbullying, inadequate sleep and lower levels of physical activity may mediate the

relationship between social media use and mental health in female adolescents, further research is

needed to enhance understanding of these gender differences (Viner et al 2019).

Whilst the direction of association cannot be determined in a cross-sectional study, the results

presented here are consistent with findings from previous research using longitudinal data from the

panel survey Understanding Society (Booker et al 2018), whereby hours spent ‘chatting or interacting

with friends through a social website’ on a school day at age 10 was longitudinally associated with

lower happiness and poorer mental well-being in females. Further analysis of the same data indicates

that the negative relationship between social media and life satisfaction is reciprocal (Orben et al

2019). However, an ecological momentary assessment study of US adolescents found little evidence

of a longitudinal association between social media use and depression or worry (Jensen et al 2019).

To our knowledge, there have been no prospective studies exploring the relationship between social

media use and self-harm. It is possible that young people in distress or self-harming may be more

likely to turn to social media for sources of comfort and support or for help-seeking information. It is

also possible that viewing self-harm or suicide related imagery posted by others on social media may

normalise such behaviour and thus increase the risk of self-harm (George 2019). Longitudinal studies

are required to better understand this relationship. It should also be noted that while this paper focuses

on direct effects, considering self-esteem as an outcome in its own right, complex associations

between self-esteem, depression and self-harm are likely to exist. Future longitudinal research should

consider whether the association between social media use and depression are mediated or moderated

by self-esteem, as well as investigating other potential pathways.

Previous research using this cohort has demonstrated negative associations between retrospectively

self-reported social media screen-time (assessed using a single-item measure) and well-being (Kelly

et al 2018; Orben and Przybylski 2019a). Findings have been inconsistent however, and another study

using TUD data from the same cohort found no meaningful association between wellbeing and a more

broadly defined measure of ‘digital engagement’ (including social media use but also general internet

9
use, watching TV or playing electronic games) (Orben and Przybylski 2019b). In addition to

including self-harm as a previously under-studied outcome, this study complements prior work by

using rich time use diary data, which is considered to be a more reliable measure than previously used

self-report survey items. We also focus specifically on the use of social media, distinct from other

forms of screen-time. The TUD data also allowed separation into weekend and weekday social media

use though few differences in mental health outcome associations were observed between the two

days.

The MCS is a methodologically rigorous population-based cohort study. Use of TUD to collect

information on social media screen-time is rare in such a large nationally representative sample. It has

previously been suggested that the reliability of retrospective self-report in measuring screen-time

may be compromised as a result of recall bias, perception of the middle response category as the

‘average’ use, or the cognitive burden associated with summing different online activities, across

different technologies (smartphones, tablets, laptops and PCs) and deliberating what to include in the

calculation (ambiguous activities could include fleeting checks of a notification, responding to a

WhatsApp message, watching a video on YouTube). These calculations, completed under time

pressure, may be further complicated when considering the use of social media simultaneously on

multiple devices (such as responding to online messages on a smartphone while watching a video on

YouTube) (Ellis, 2019). Time use diaries are an in-situ mode of data collection, whereby respondents’

recall is facilitated by the specificity of the day in question, as well as the format of selecting a pre-

defined activity to account for each small unit of time throughout the day. The absence of broad time-

based response categories in TUD means there is no social desirability associated with a perceived

‘normal’ or average amount of social media screen-time.

Despite the advantages noted above, the TUD mode implemented in the MCS is not without

limitations. Although respondents were encouraged to respond ‘in-situ’ throughout the day, it was

possible to do so retrospectively, which may increase recall bias. Most importantly, the TUD did not

allow for the reporting of multi-tasking. A scenario in which someone is watching television and

10
using social media on a smartphone simultaneously would require the participant to report only one of

these activities in a given time period. It is feasible that the use of social media on smartphones could

make it more likely to be perceived as a secondary activity if done while eating, travelling, at school

or watching a film – and thus under-recorded in a TUD. Where an activity appeared in the list of

options might also influence its likelihood of being selected (e.g. Krosnick et al 1987). In the list of

activities offered to participants, sleep, personal care, school, homework, housework, looking after

people or animals, eating or drinking, exercise, travelling, social and family time were all positioned

above digital media, including social media, and therefore may be more likely to be prioritised when

deciding between simultaneous activities (Ipsos MORI technical report, 2016). This may go some

way to explaining the lower levels of screen-time reported in TUD compared to retrospective self-

report in the MCS (see Orben and Przybylski, 2020 for a comparison).

The TUD category most exclusively related to social media use was ‘browsing and updating on social

networking sites, e.g. Twitter, Facebook, BBM, Snapchat’. However, an additional category included

‘email, instant messaging and texting’. We excluded this category from our TUD analysis due to the

conflation with emailing and other general mobile telephone use, but this may lead to an

underestimation of social media screen-time if messaging sites such as WhatsApp were excluded.

There may be some further overlap in definitions and interpretations of other activity categories

offered in the TUD. Social media apps are often used to play games, listen to music and watch videos.

Such complexities are likely to present challenges to respondents deciding how best to categorize

these activities, potentially compromising the accuracy of any self-report measure of social media

screen-time.

Without an objective measure of social media use, it is impossible to accurately assess the reliability

of TUD. Further development and evaluation of ways in which to directly capture this activity is

essential in improving knowledge of the relationship between screen-time and adolescent mental

health. The research field would further benefit from more longitudinal studies to examine the

direction of associations between screen-time and mental health, as well as examining mechanisms

11
underpinning the gender differences in social media screen-time, and in its relationship with mental

health.

Beyond distinguishing social media use from other forms of digital engagement, there is growing

consensus that future research must evolve beyond one-dimensional measures of social media screen-

time (whether self-reported retrospectively or in-situ) and adopt a more nuanced examination of

different social media apps and activities (Orben et al 2019; Scott & Woods 2019). Social media use

is not a homogenous activity, and a simplistic estimate of ‘social media’ screen-time is unlikely to be

the most informative variable in helping us to understand the relationship between mental health and

social media use. Social media use encompasses a very wide range of activity, including passive

lurking on strangers’ Facebook profiles, posting highly edited selfies on Instagram, using WhatsApp

to arrange a get together with friends, watching any number of videos on YouTube, reading the news

on Twitter and so much more. Different activities or apps are likely to have different relationships

with mental health (Verduyn et al 2017) and as social media and our use of it further evolves in

complexity, so too will the impacts on mental health.

Funding: LW is funded by National Institute of Health Research School for Public Health Research.

BM and PM are funded by National Institute of Health Research Biomedical Research Centre,

University Hospitals Bristol NHS Foundation Trust, Bristol, UK. Acknowledgements to Aase

Villadsen for advice regarding the Millennium Cohort Study data.

12
References

1. Andrews S, Ellis DA, Shaw H, Piwek L (2015) Beyond Self-Report: Tools to

Compare Estimated and Real-World Smartphone Use. PLoS ONE 10(10): e0139004.

doi:10.1371/journal. pone.0139004

2. Angold A. Costello. EJ, Messer SC, Pickles. A., Winder F., Silva D.,(1995) The

development of a short questionnaire for use in epidemiological studies of depression

in children and adolescents. International Journal of Methods in Psychiatric

Research.;5:237-49.

3. Boers E, Afzali MH, Newton N, Conrod P. Association of screen time and depression

in adolescence. JAMA pediatrics. 2019 Sep 1;173(9):853-9.

4. Booker CL, Kelly YJ, Sacker A. Gender differences in the associations between age

trends of social media interaction and well-being among 10-15 year olds in the UK.

BMC public health. 2018 Dec;18(1):321.

5. Ellis DA. Are smartphones really that bad? Improving the psychological measurement

of technology-related behaviors. Computers in Human Behavior. 2019 Aug 1;97:60-

6.

6. George M. The Importance of Social Media Content for Teens' Risks for Self-harm.

Journal of Adolescent Health. 2019 Jul 1;65(1):9-10.

13
7. Goodman R. Psychometric properties of the strengths and difficulties questionnaire.

Journal of the American Academy of Child & Adolescent Psychiatry. 2001 Nov

1;40(11):1337-45.

8. Gunnell, D, Kidger, J, Elvidge, H. Adolescent mental health in crisis. BMJ 2018

June;361:k2608.

9. Jensen M, George MJ, Russell MR, Odgers CL. Young adolescents’ digital

technology use and mental health symptoms: Little evidence of longitudinal or daily

linkages. Clinical Psychological Science. 2019 Nov;7(6):1416-33.

10. Kelly Y, Zilanawala A, Booker C, Sacker A. Social media use and adolescent mental

health: Findings from the UK Millennium Cohort Study. EClinicalMedicine. 2018

Dec 1;6:59-68.

11. Krosnick JA, Alwin DF. An evaluation of a cognitive theory of response-order effects

in survey measurement. Public Opinion Quarterly. 1987 Jan 1;51(2):201-19.

12. Kuehner C. Why is depression more common among women than among men?. The

Lancet Psychiatry. 2017 Feb 1;4(2):146-58.

13. Ipsos MORI. Millennium Cohort Study Sixth Sweep (MCS6): Technical Report.

2016. Prepared for the Centre for Longitudinal Studies, UCL Institute of Education.

14
14. López-López JA, Kwong AS, Washbrook E, Pearson RM, Tilling K, Fazel MS,

Kidger J, Hammerton G. Trajectories of depressive symptoms and adult educational

and employment outcomes. BJPsych Open. 2020 Jan;6(1).

15. McKenzie DP, Toumbourou JW, Forbes AB, Mackinnon AJ, McMorris BJ, Catalano

RF, Patton GC. Predicting future depression in adolescents using the Short Mood and

Feelings Questionnaire: a two-nation study. Journal of affective disorders. 2011 Nov

1;134(1-3):151-9.

16. McManus S, Gunnell D, Cooper C, et al. Prevalence of non-suicidal self-harm and

service contact in England, 2000-14: repeated cross-sectional surveys of the general

population. Lancet Psychiatry 2019; 6(7): 573-81.

17. Morgan C, Webb RT, Carr MJ, Kontopantelis E, Green J, Chew-Graham CA, Kapur

N, Ashcroft DM. Incidence, clinical management, and mortality risk following self

harm among children and adolescents: cohort study in primary care. bmj. 2017 Oct

18;359:j4351.

18. Muehlenkamp JJ, Guttierrez PM: Risk for suicide attempts among adolescents who

engage in non-suicidal self-injury. Arch Suicide Res. 2007, 11 (1): 69-82.

10.1080/13811110600992902.

15
19. Naab TK, Karnowski V, Schlütz D. Reporting mobile social media use: how survey

and experience sampling measures differ. Communication Methods and Measures.

2019 Apr 3;13(2):126-47.

20. Ofcom, 2019. Children and parents: Media use and attitudes report 2018.

https://www.ofcom.org.uk/__data/assets/pdf_file/0024/134907/Children-and-Parents-Media-

Use-and-Attitudes-2018.pdf (Accessed 16th August 2019)

21. Orben A, Dienlin T, Przybylski AK. Social media’s enduring effect on adolescent life

satisfaction. Proceedings of the National Academy of Sciences. 2019 May

21;116(21):10226-8.

22. Orben A, Przybylski AK. (a) The association between adolescent well-being and

digital technology use. Nature Human Behaviour. 2019 Feb;3(2):173.

23. Orben A, Przybylski AK. (b) Screens, teens, and psychological well-being: evidence

from three time-use-diary studies. Psychological science. 2019 May;30(5):682-96.

24. Orben A, Przybylski AK. Teenage sleep and technology engagement across the

week. PeerJ 2020 Jan ;8:e8427 https://doi.org/10.7717/peerj.8427

25. Rosenberg M. Society and the adolescent self-image. Rev. ed.Middletown, Conn:

Wesleyan University Press; 1989.

26. Scharkow M. The accuracy of self-reported internet use—A validation study using

client log data. Communication Methods and Measures. 2016 Jan 2;10(1):13-27.

16
27. Scott H, Woods HC. Understanding Links Between Social Media Use, Sleep and

Mental Health: Recent Progress and Current Challenges. Current Sleep Medicine

Reports. 2019 Jun 29:1-9.

28. University of London, Institute of Education, Centre for Longitudinal Studies.

Fitzsimons E. Millennium Cohort Study: sixth survey, 2015. UK Data Service; 2017

SN: 8156.

29. Verduyn P, Ybarra O, Résibois M, Jonides J, Kross E. Do social network sites

enhance or undermine subjective well‐being? A critical review. Social Issues and

Policy Review. 2017 Jan;11(1):274-302.

30. Woods HC, Scott H. # Sleepyteens: Social media use in adolescence is associated

with poor sleep quality, anxiety, depression and low self-esteem. Journal of

adolescence. 2016 Aug 1;51:41-9.

17
Table 1: description of analytic sample, demographics

Social media time use Social media time use χ2

diary data available diary data not available

N 4,032 7,852

Proportion aged:

13 years 25.4% 23.4% χ2=8.86, p=0.012

14 years 73.5% 75.1%

15 years 1.1% 1.5%

Proportion female 55.2% 47.2% χ2=67.95, p=0.000

Proportion non-White 15.4% 23.0% χ2=92.53, p=0.000

Proportion lowest 10.1% 20.8% χ2=214.06, p=0.000

income quintile

Proportion highest 29.9% 18.9% χ2=185.57, p=0.000

income quintile

Unweighted estimates

18
Table 2: description of analytic sample, social media screen-time and mental health outcomes by

gender

Females Males Overall

%/mean (95% CI) %/mean (95% CI) %/mean (95% CI)

N 2,225 1,807 4,032

Social media screen-time, weekday

<30 minutes 65.1% (62.9% to 81.9% (79.9% to 72.7% (71.2% to

67.2%) 83.7%) 74.2%)

30-60 minutes 13.6% (12.2% to 7.4% (6.2% to 8.8%) 10.8% (9.8% to

15.3%) 11.9%)

60-90 minutes 9.9% (8.7% to 11.3%) 5.2% (4.2% to 6.5%) 7.8% (7.0% to 8.7%)

90-120 minutes 4.6% (3.7% to 5.6%) 2.1% (1.5% to 3.0%) 3.5% (2.9% to 4.1%)

120 minutes or more 6.8% (5.7% to 8.0%) 3.3% (2.5% to 4.3%) 5.2% (4.5% to 6.0%)

Social media screen-time, weekend day

<30 minutes 59.3% (57.0% to 81.3% (79.3% to 69.1% (67.7% to

61.5%) 83.2%) 70.8%)

30-60 minutes 14.4% (12.9% to 7.7% (6.5% to 9.1%) 11.3% (10.3% to

16.1%) 12.5%)

60-90 minutes 11.9% (10.5% to 6.2% (5.1% to 7.5%) 9.3% (8.4% to 10.3%)

13.4%)

90-120 minutes 4.9% (4.0% to 6.0%) 2.0% (1.4% to 2.8%) 3.6% (3.0% to 4.3%)

120 minutes or more 9.5% (8.3% to 10.9%) 2.8% (2.1% to 3.7%) 6.4% (5.7% to 7.3%)

Proportion self-harmed 20.2% (18.6% to 6.5% (9.2% to 9.5%) 14.1% (13.0% to

in past year 21.9%) 15.2%)

Mean depression score 6.78; SD=6.33 (CI: 3.99; SD=4.38 (CI: 5.54; SD=5.71 (CI:

6.52 to 7.05) 3.79 to 4.19) 5.36 to 5.71)

19
 Mean self-esteem score 9.92; SD=2.88 (CI: 11.49; SD=2.54 10.62; SD=2.84

9.80 to 10.04) (11.37 to 11.61) (10.53 to 10.71)

Unweighted estimates

20
21
Table 3: Multivariable regressions, odds ratios for self-harm by social media use weekday and weekend day
Females Males

Unadjusted Adjusted for age, Additionally Unadjusted Adjusted for age, Additionally
income and adjusted for income and adjusted for
ethnicity previous ethnicity previous
internalising internalising
symptoms symptoms
OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Weekday social 1.14 (1.07 to 1.22) 1.14 (1.07 to 1.22) 1.13 (1.06 to 1.21) 0.88 (0.76 to 1.03) 0.91 (0.79 to 1.05) 0.92 (0.79 to 1.06)
media use
Males (n=1461);
Females (n=1812)
Weekend social 1.06 (1.00 to 1.12) 1.05 (0.99 to 1.12) 1.04 (0.98 to 1.10) 1.04 (0.95 to 1.15) 1.05 (0.96 to 1.16) 1.06 (0.97 to 1.16)
media use
Males (n=1453);
Females (n=1814)
Results are weighted estimates

22
Table 4 Multivariable regressions, coefficients for depression symptoms by social media use weekday and weekend day

Females Males

Unadjusted Adjusted for age, Additionally Unadjusted Adjusted for age, Additionally
income and adjusted for income and adjusted for
ethnicity previous ethnicity previous
internalising internalising
symptoms symptoms
β (95% CI) β (95% CI) β (95% CI) β (95% CI) β (95% CI) β (95% CI)
Weekday social media 0.43 (0.28 to 0.57) 0.39 (0.24 to 0.54) 0.36 (0.22 to 0.50) 0.10 (-0.11 to 0.31) 0.08 (-0.12 to 0.28) 0.10 (-0.10 to
use 0.30)
Males (n=1458);
Females (n=1810)
Weekend social media 0.25 (0.13 to 0.37) 0.22 (0.09 to 0.35) 0.19 (0.06 to 0.32) -0.01 (-0.22 to -0.02 (-0.24 to -0.01 (-0.23 to
use 0.21) 0.20) 0.20)
Males (n=1450);
Females (n=1814)
Results are weighted estimates.

23
Table 5 Multivariable regressions, coefficients for self-esteem scores by social media use weekday and weekend day

Females Males

Unadjusted Adjusted for age, Additionally Unadjusted Adjusted for age, Additionally
income and adjusted for income and adjusted for
ethnicity previous ethnicity previous
internalising internalising
symptoms symptoms
β (95% CI) β (95% CI) β (95% CI) β (95% CI) β (95% CI) β (95% CI)
Weekday social media -0.14 (-0.21 to - -0.13 (-0.20 to - -0.12 (-0.20 to - -0.00 (-0.16 to 0.01 (-0.14 to 0.16) 0.00 (-0.15 to 0.15)
use 0.06) 0.06) 0.04) 0.15)
Males (n=1447);
Females (n=1797)
Weekend social media -0.14 (-0.20 to - -0.13 (-0.19 to - -0.12 (-0.18 to - 0.05 (-0.05 to 0.15) 0.05 (-0.05 to 0.16) 0.05 (-0.05 to 0.15)
use 0.09) 0.08) 0.07)
Males (n=1443);
Females (n=1802)
Results are weighted estimates

24
Supplementary table 1: Multivariable regressions, odds ratios for depression by social media use weekday and weekend day

Females Males

Unadjusted Adjusted for age, Additionally Unadjusted Adjusted for age, Additionally
income and adjusted for income and adjusted for
ethnicity previous ethnicity previous
internalising internalising
symptoms symptoms
OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI) OR (95% CI)
Weekday social media 1.13 (1.08 to 1.19) 1.12 (1.07 to 1.18) 1.12 (1.07 to 1.17) 1.04 (0.91 to 1.19) 1.04 (0.92 to 1.19) 1.05 (0.92 to 1.20)
use
Males (n=1458);
Females (n=1810)
Weekend social media 1.09 (1.04 to 1.14) 1.08 (1.03 to 1.13) 1.08 (1.03 to 1.13) 1.05 (0.93 to 1.18) 1.05 (0.93 to 1.18) 1.05 (0.93 to 1.19)
use
Males (n=1450);
Females (n=1814)
Results are weighted estimates; Depression defined as SMFQ score of 11 or more.

25