Adolescents' screen time, sleep and mental health: literature review

Methods

Search strategy

We anticipated a large body of literature relating to RQ 1-4 (quantitative data), so we applied a sequential series of literature search strategies. While we expected some overlap in results, having different search strategies ensured we would not miss studies relevant to each question.

We used the following electronic bibliographic databases: CINAHL (EBSCO), ERIC (EBSCO), EMBASE (OVID), MEDLINE (OVID), PsycINFO (EBSCO), IBSS (Proquest), ASSIA (Proquest), Social Science Citation Index (Web of Science), and Emerging Sources Citation Index (Web of Science).

The key terms for the search strategies related to (i) the population (young people), (ii) mobile devices and related software (e.g. smartphone, social media, mobile apps, etc.), (iii) sleep outcomes (e.g. sleep duration, sleeplessness, night awakening etc.), and (iv) mental health. The Medline search string is included in Appendices A & B.

This broad electronic search strategy facilitated the identification of evidence from experimental and observational studies as well as evidence from literature reviews. This ensured that we were able to assess the relevant and available literature showing the effect of (i) mobile device screen time/use on sleep (RQ1), (ii) causal mechanisms through which mobile device screen time/use can affect young people's sleep outcomes (RQ2), and (iii) impact on mental health (RQ3,4). RQ5 was limited to qualitative research from the searches for RQ3 and RQ4.

Key terms and subject areas within each component were combined with the Boolean operator 'OR' and all three components were combined with the Boolean operator 'AND'. We constructed and adapted search strategies for each database.

Inclusion and exclusion criteria

Only studies that met the inclusion criteria were included. All literature searches were limited to publications in English language from 2007 as this was when the first commercial smartphone was released.

Population: Young people aged 10-19 years. Studies including children <10 years or adults >19 years alongside young people were only included when results were reported for young people separately and/or if at least one third of the sample was target age group or if mean age was 10-19 years. Only Western countries classified as high-income countries (HICs) by the World Bank (https://data.worldbank.org/income-level/high-income) were included and all low or middle income countries (LMICs) were excluded. Literature from non-Western HICs (e.g. Japan) was collected, but not included in data extraction and synthesis as the focus for this report was restricted to countries that were culturally more similar to Scotland. Studies from the general population and findings from subsets of that sample with specific medical conditions (autism, ADHD, etc.) were included, but clinical populations, i.e. samples based on having a specific condition/disorder/disease, were excluded.

Exposures: Engagement with digital/electronic mobile devices (e.g. smartphones, tablets, laptops) and software accessible through mobile devices (e.g. social media, games, websites, messaging applications), including studies which assessed the effectiveness of mobile applications or websites designed to improve sleep or mental health outcomes. Studies which referred to screen time in general without specifying the device and/or specific use (e.g. an app) were not considered for inclusion. This was to ensure that we captured evidence on contemporary screen technology rather than older screen technology such as televisions.

Outcomes: 1) Sleep - objectively assessed or subjectively reported indicators of sleep health, and diagnosed sleep disorders, 2) Mental health and wellbeing - indicators of absence or presence of emotional, psychological, and social well-being assessed using validated psychometric questionnaires, and/or diagnosed mental disorder.

Study types: To ensure a rapid synthesis of the evidence, we applied a hierarchical approach to study selection. First, we selected reviews from 2007 to 2019 and searched them for primary experimental and longitudinal studies that met criteria for our Research Questions. We also selected primary experimental and longitudinal studies between 2017 and 2019 because the most recent reviews conducted literature searches in 2017. We did not consider cross-sectional evidence (or reviews of cross-sectional studies) for RQ1-4 because of the limitation related to identifying temporality and causal relationships between mobile device screen time/use and sleep or mental health outcomes. Cross-sectional qualitative studies were selected to answer RQ5 in the absence of literature reviews. Studies which included open questions in questionnaires, as well as interview or focus group methodologies, were eligible for answering RQ5.

Study selection

Five authors on this report screened titles, abstracts and full texts of all potentially relevant studies (AM, JP, KS, CM, JR). To allow a rapid evidence synthesis while reducing the risk of study selection bias, a randomly selected 50% of title and abstracts (AM, CM) and 20% of full-text articles were independently screened in duplicate (AM, JP, KS, CM). The decision regarding study eligibility was cross-checked and disagreements resolved through discussion or involvement of a third reviewer.

Data extraction

We developed and tested an electronic data extraction form on two included quantitative studies and one qualitative study. We extracted the study design, sample size, population characteristics (e.g. age, gender), type of exposure and outcomes, study results (direction, effect size, statistical significance, and topics on young people's views). Data extraction for all included studies was cross-checked by a second reviewer.

Quality assessment

We assessed the quality of primary quantitative studies using The Cochrane Collaboration tool for assessing risk of bias in intervention and longitudinal observational studies. Five quality domains were assessed: (i) population selection bias, (ii) performance bias (i.e. bias in assessment of the exposure), (iii) detection bias (bias in assessment of the outcome), (iv) attrition bias (bias of missing outcome data), and (v) selective reporting bias (incomplete reporting of all outcomes). Studies were judged to be of high, unclear or low risk of bias. One researcher appraised the quality of all included reviews/studies (JR) and another researcher (AM) cross-checked the quality assessment for all studies. JR's assessment of the study quality was known to AM when cross-checking.

Qualitative research studies were appraised using the National Institute for Health and Care Excellence (NICE) appraisal checklists (14). The checklist includes 14 items which are shown in Appendix Table C. Studies were rated as being of high, unclear or low quality. One researcher appraised the quality of all included reviews/studies (JR) and another researcher cross-checked the quality appraisal for all studies (CM). We resolved disagreement in quality appraisal through discussion and, if needed, sought the opinion of a third reviewer.

Data synthesis

For data synthesis we mapped each mobile device activity (including time spent using the device) identified from the literature against its findings on sleep outcomes (RQ1-2) and mental health and wellbeing outcomes (RQ3-5). This involved providing a summary of the number of studies, characteristics of studies, and findings on sleep, mental health and wellbeing outcomes grouped by type of mobile device screen activity (e.g. social media use, gaming, mobile app use). We intended to summarise the findings across the body of evidence for each Research Question as consistent, inconsistent, fragmented or incomplete, or no evidence. However, assessment of consistency/ inconsistency was not possible due to the low number of included studies for each unique relationship between type of mobile device exposure and sleep outcome (see Results). Judging consistency or inconsistency based on one or two studies is not meaningful. We indicated incompleteness and where there was no evidence of an association. In addition to narratively describing the findings addressing each Research Question, we compiled a summary of findings in tables (Appendix Tables E and F), providing details of the studies included in the review.