A scoping review of social media in child, adolescents and young adults: research findings in depression, anxiety and other clinical challenges

Approach

The literature search was conducted from January 2000 to December 2020. The philosophical approach to the search was done according to the original six-stage process¹⁵⁰ and updated modifications¹⁵¹ (purposeful research question, identifying relevant studies, selecting studies based on an iterative process, charting the data, analysis of findings and consultation from stakeholders). The Preferred Reporting Extension for Systematic Reviews and Meta-Analyses (PRISMA) for scoping reviews¹⁵² has additional suggestions for sources of information, the search and appraising data.

Research question

This review addresses the overarching question: ‘What is the nature of the relationship between social media use, psychopathology and mental and/or behavioural health conditions or problems?’ The population of interest is children, adolescents and young adults (aged ≤25 years). Secondary questions are as follows.

1. What social media is commonly used, in what ways and for what purpose(s) (i.e. approach, interest, motivation)?

2. In what ways is social media helpful, neutral or negative related to clinical populations for depression and anxiety, and specific problems like cyberbullying, sexting and suicide?

3. What is the relationship (i.e. association, mediation, causation and/or other) between social media (e.g. Facebook, Twitter, Instagram) and behavioural health?

4. What methods of assessment, triage and approaches, interventions and professional development can help providers, parents, teachers and others in the community to help?

Identifying relevant studies

Eleven databases were queried: PubMed/Medline, APA PsycNET, Cochrane Database of Systematic Reviews, EMBASE, PsycINFO, Web of Science and Scopus, Social Sciences Citation Index (SSCI), Centre for Reviews and Dissemination, Cochrane Central Register of Controlled Trials, Cumulative Index to Nursing and Allied Health Literature (CINAHL) and Google Scholar.

The search focused on youth (adolescent, child, children, high, junior, juvenile, middle, minor, secondary, teenager, youth) and social media use in five concept areas (Fig. 1): clinical depression and anxiety and problematic challenges (e.g. suicidal ideation, cyberbullying); quantitative data; social media mode; engagement and qualitative dimensions; and health and psychological well-being. Definitions were used based on consensus literature: bullying is a subset of aggressive behaviour that involves repeated and intentional attempts to damage/distress a weaker victim by a more powerful perpetrator;¹⁵³ and sexting is sending or receiving of sexually explicit pictures, videos, or text messages via smartphone, digital camera or computer.⁹⁶ Exclusion criteria included studies focusing on anorexia, attention-deficit hyperactivity disorder, physical or intellectual disabilities, genetics, substance use, gambling, sleep/insomnia, cognitive disorders and aggression/violence beyond cyberbullying and suicide) (Fig. 1).

Study selection

One author (D.M.H.) screened titles and abstracts of potential references, excluding duplicates and those that did not meet the search criteria. Two authors (D.M.H., D.S.) reviewed the full text of remaining abstracts to find those meeting inclusion criteria; additional studies that met inclusion criteria were added from references.

Data charting

A data-charting form was used to extract data, and notes were organised with a descriptive analytical method. The reviewers (D.M.H., D.S.) compared and consolidated information by using a modified content analysis with thematic components;¹⁵⁴ a third author (A.J.M.) moderated any disagreement and a fourth author (S.-T.T.L.) analysed consistency of the approach. The information was shared with selected experts, their input summarised and themes extracted.

Analysis, reporting and the meaning of findings

Results were organised into tables, with key concepts and components outlined and described, partially based on excerpts from published topics. The studies varied considerably, and therefore were challenging to compare. Qualitative steps to analyse disparate populations, methods and data of studies were used (Fig. 2).¹⁵⁴ Content, discourse and framework qualitative analysis techniques were to analyse findings from papers and classify, summarise and tabulate the behavioural data; discourse and thematic analyses were used to search for themes and patterns; and framework analysis was used to sift through, chart and sort data in accordance with key issues and themes a series of steps (e.g. indexing, charting, mapping and interpretation).¹⁵⁴ Data in Supplementary Tables 1–3 are organised by study, sample size, population (e.g. country), objective and design, methods and measures, outcomes and clinical implications/challenges and training/research foci.

Expert opinions and feedback

Expert opinions were solicited to review preliminary findings and suggest additional steps for improvement. A list of relevant experts was compiled from (a) behavioural health organisations across professions internationally; (b) technology-related special interest groups of organisations (e.g. American Telemedicine; Medical, Nursing and Informatics Associations); (c) educational and professional development organisations (e.g. Accreditation Council of Graduate Medical Education, American Academy of Child and Adolescent Psychiatry, American Academy of Pediatrics); (d) academic institutions and (e) researchers, authors, editors and editorial board members of journals related to social media.

Experts were invited by email (N = 24) and attended a live expert feedback session for discussion and feedback; completed a qualitative and quantitative five-item Likert scale survey (n = 20; 83.3%) and/or provided qualitative feedback via email (n = 4; 16.7%). The data charting and the search criteria plan were reviewed; their input did not suggest a search with additional terminology or otherwise change the scope. Input was summarised and themes were extracted to guide the organisation (e.g. headings in rows) and content (e.g. in the columns) of Table 1 and Supplementary Tables 1–3, based on previous work using consensus and modified Delphi processes.¹⁵⁴ Results showed that the majority agreed or strongly agreed that the search strategy was effective using the research question (n = 21; 87.5%); it was systematic/thorough (20; 83.3%); and adequately scientific in methodology (n = 18; 75%); and ‘[The tables] are organised in a practical way to summarise social media study findings for providers, teachers and systems’ (n = 20; 83.3%), once more specific outcomes were entered in the final column of each.

Literature overview

Out of 2820 potential literature references, 112 duplicates and 2284 studies that were outside of the scope of this review were excluded (Fig. 1). Full-text review of 424 articles revealed that 128 met full inclusion criteria; 12 additional studies were found within those, for a total of 140 studies.^{12,17–95,97–149,155} The studies focused on clinical populations for depression and anxiety,⁷¹ clinical challenges (e.g. suicidal ideation, cyberbullying)²⁷ or psychological well-being.²¹ Studies were of children aged 12 years and younger (n = 1; 0.01%), adolescents (13–18 years) (n = 72; 54.1%) and young adults (19–25 years) (n = 48; 34.2%); the rest were aggregates of the above (n = 18; 12.9%). The overall mean age was 18.78 years. The most common social media studied were Facebook (n = 62), Twitter (n = 20), Instagram (n = 11), YouTube (n = 6) and MySpace (n = 5). Studies varied in identifying gender identity (n = 63; 45%), ethnicity and race (n = 42; 30%) or neither (n = 35; 25%).

Most studies were cross-sectional cohort studies using self-report questionnaires. Few studies were longitudinal in design,¹⁹ had comparison groups²⁰ or were randomised controlled trials.³ Few studies used clinician/provider-administered instruments^2,20,30 or structured assessments.^{4,17,51,82,131} Timing or temporal dimensions are generally quite limited and studies span across acute disorders, subacute symptoms and trait/personality factors among a wide variety of ethnic, clinical and non-clinical populations. Broadly speaking, the studies looked at associations (n = 120; 85.7%),^{17–23,25–29,31–33,36–41,43–53,55–61,63–73,75,76,78–84,86–95,97–100,102,103,105,112,114–127,129–131,133–142,144,146,148,149,154,155} and mediating (n = 16; 11.4%)^{24,30,34,35,62,74,77,85,101,104,113,128,132,143,145} and causal (n = 4; 2.9%)^{42,54,104,147} relationships between social media and behavioural health issues.

Clinical populations, depression and anxiety

There were 78 studies of social media with outcomes in clinical populations and disorders (Supplementary Table 1). The mean age was 18.4 years (median 18 years) and included adolescents⁴² and young adults.²¹ The study populations were diverse in terms of ethnicity, but were predominately White, and 46 studies were ≥50% female. The mean sample size was 8332.4 (median 310). The most common social media sites studied were Facebook (n = 37), Twitter (n = 10), Instagram (n = 5), MySpace (n = 2) and YouTube (n = 2); two were on screen time.^13,33

Cross-sectional and longitudinal studies^{12,19,22,23,25,27,28,31,35,41,42,47,48,51,65,67,76,79,87,89} of social media use and depression found that shorter periods of social media use (<3 h), particularly with purposeful or active engagement, are associated with better mood and psychological well-being, whereas longer periods of social media use predict depression (and often anxiety) or poorer psychological function^{22,25,27,28,31,35,48,59,65,87} (particularly browsing^30,89), partly because of sleep disruptions.⁴⁷ Cross-sectional and longitudinal studies are consistent with one prospective study that suggests a threshold effect around 3 h that has negative impact for many, but not all, users: low use-stable (80% at 3–4 h/day/item), high use-decreasing (12.3% at 4–5 h/day/item) and low use-increasing (7.3% at 3 to nearly 5 h/day).³³ Two studies found that depression predicts social media use^33,67 and reduces perception of support.⁵¹ Specifically, Twitter use may be associated with depressive thoughts and symptoms, but only for people with low initial levels of in-person social support, and conveying positive sentiment helped to reduce depressive thoughts and feelings irrespective of people's level of in-person social support.²³ Depressive signals observed in Tweets may predict future depression.⁷⁶ Instagram browsing was associated with increases in depressed mood in adolescents.⁴²

Type of media use is important, since hours spent on social media and internet use were more strongly associated with self-harm behaviours, depressive symptoms, low life satisfaction and low self-esteem than hours spent electronic gaming and watching television.¹² In addition, girls generally demonstrated stronger associations between screen media time and mental health indicators than boys (e.g. heavy internet users were 166% more likely to have clinically relevant levels of depressive symptoms than low users for girls, compared with 75% more likely for boys). A cross-sectional study showed that cortisol systemic output was positively associated with Facebook network size and negatively associated with Facebook peer interactions.⁵⁰

Studies of anxiety disorders are similar to findings in depression studies, with social anxiety symptoms mediated by spending more time on Facebook and passively using Facebook (i.e. viewing other's profiles without interacting).⁶² In a study with three focus groups of those with anxiety disorders, six themes emerged: seeking approval, fearing judgement, escalating interpersonal issues, wanting privacy, negotiating self and social identity and connecting and disconnecting.⁴¹ A qualitative study revealed three types of negative use, including ‘oversharing’ (frequent updates or too much personal information), ‘stressed posting’ (sharing negative updates) and encountering ‘triggering posts’.⁴⁶ Both social anxiety and need for social assurance had a significant positive association with problematic use of Facebook^41,54 or ‘fear of missing out’ (FOMO).^24,39

Clinical challenges like suicide, cyberbullying, sexting and other behaviours

The review found 34 studies on clinical challenges such as cyberbullying, sexting and posts on suicide (Supplementary Table 2). The primary populations were children (n = 1), adolescent (n = 15) and young adults (n = 14, with 3 for college students), with a mean age of 18 (median 17.9) years. The study populations were diverse in terms of ethnicity, but were predominately White and 15 studies were ≥50% female. The mean sample size was 34934.5 (median 524). The most common social media types studied were Facebook (n = 10), Twitter (n = 8), Instagram (n = 5), YouTube (n = 3) and MySpace (n = 2).

Excessive social media use, depression, suicide and school burn-out appear strongly related.^{103,107,109,115} One longitudinal study found that, compared with matched non-suicide-related Twitter posts, suicide-related posts were characterised by a higher word count, increased use of first-person pronouns and more references to death.¹⁰³ In this study, emotional engagement, school burn-out and depression contributed to excessive social media use. Similarly, students with burn-out are at higher risk for depression and excessive social media use. Excessive social media use leads to school burn-out and school burn-out leads to excessive social media use. Individuals who were suicidal felt significantly less belongingness and significantly higher burdensomeness; they also use a higher proportion of achievement-related words and appear protective. Studies have compared artificial intelligence/machine learning to self-report measures to evaluate risk of suicide,¹⁰⁷ para-suicidal events,¹⁰⁹ suicide-related Tweets¹¹² and other behaviors.¹¹⁵ Machine learning can easily differentiate people who are at high suicidal risk from those who are not (linguistic inquiry and word count, decision tree and cross-validation analyses).¹⁰⁷ Machine-learning algorithms accurately identify the clinically significant suicidal rate in 92% of cases (sensitivity: 53%, specificity: 97%, positive predictive value: 75%, negative predictive value: 93%); a higher proportion of achievement-related words appears protective. For a single point of performance for comparison, artificial intelligence/machine learning had roughly 10% false alarms, but correctly identified about 70% of those who will attempt suicide.¹⁰⁹

The relationship of depression, self-esteem and cyberbullying has been evaluated. A study of 8- to 13-year-olds evaluated whether cybervictimisation is prospectively related to negative self-cognitions and depressive symptoms beyond other types of victimisation.¹¹⁰ The majority of participants reported experiencing at least some degree of peer victimisation at either wave 1 or wave 2 (physical: 68.1%, relational: 89.8%, verbal: 87.9%, property related: 65.8%, cyber: 63.1%). Of note, 16.1% of participants obtained raw scores >75 on the Reynolds Adolescent Depression Scale – Version 2 (RADS-2), and 8.1% obtained scores >82 (signifying mild and moderate depression, respectively). Victimisation was correlated with negative cognition and depressive symptoms; it predicted depressive symptoms; age and gender were not predictors of cybervictimisation or depression. Depression is associated with problematic social media use and indirectly predicted cyberbullying perpetration (associations were weak). Another study found that problematic social media use is weakly correlated with depression (r = 0.22), gender (r = −0.15), age (r = −0.13) and self-esteem (r = −0.11).⁹⁵ Experiences of LGBTQ participants included both help for coping and cyberbullying leading to depression, stress and suicidal ideation.⁹⁷

Bystander responses to suicidal behaviour and cyberbullying are in sharp contrast. Only 33.6% of participants left a positive, supportive comment on at least one of two suicide posts. Content severity, experience with a loved one's suicide attempts and use of Facebook to meet people were predictive of providing positive comments.⁹⁴ Positive bystander responses (PBRs) were higher in cyberbullying than traditional bullying incidents.¹⁵⁴ Females exhibited more PBRs across both types of bullying. Bullying severity affected PBRs, in that PBRs increased across mild, moderate and severe incidents, consistent across traditional bullying and cyberbullying. PBRs related to cyberbullying included (a) seek help from a teacher or parent, (b) seek help from a peer or friend, (3) direct intervention and (d) providing comfort or emotional support.

Provider access to a patient's social media could assist in identifying suicidal ideation and/or acts, since patients fail to disclose risk factors to physicians; however, there are ethical and privacy concerns about searching a patient's social media platforms.¹⁰⁰

Health behaviour and well-being topics

There were 28 studies on health behaviour and well-being (Supplementary Table 3). The primary populations were adolescents,⁸ college students¹⁴ and young adults.⁶ The study populations were diverse in terms of ethnicity, but were predominately White and 19 studies were ≥50% female. The mean sample size was 1558.8 (median 15.8). The most common social media types studied were Facebook,¹⁵ Twitter (n = 2), Instagram (n = 1), YouTube (n = 1) and MySpace (n = 1). The most study population or disorder was depression (8) or anxiety (6).

Of the longitudinal studies, one found that a group deactivated from Facebook for 4 weeks showed small increases in well-being, but no changes in loneliness, compared with a usual use group.¹²⁴ Another study over 2 months examined internalising symptoms (e.g. depression, anxiety and loneliness) related to the content of their Facebook communication and the responses they received from peers.¹³⁵ The mean number of posts was 60.2 overall (88 for girls and 37 for boys). For girls, internalising symptoms predicted negative affect, somatic complaints and eliciting support; they also predicted receiving more peer comments expressing negative affect and peer responses offering support. A study over 9 months evaluated how social media activity affected individual social communication skill and self-esteem.¹⁴⁶ Active social media use (i.e. directed, person-to-person exchanges) increases bonding and bridging social capital and decreases loneliness; passive use does not.

Cross-sectional studies of teenagers examined psychological well-being and differences between girls and boys in use of technologies,¹² screen time ^27,125,126 and social networking services (SNS).¹⁴² The study found that adolescent girls spent more time on smartphones, social media, texting, general computer use and online, and boys spent more time gaming and on electronic devices in general.¹² Associations between moderate or heavy digital media use and low psychological well-being/mental health issues were generally larger for girls than for boys. For both girls and boys, heavy users (≥5 h) often rated twice as likely to experience well-being and mental health issues (e.g. risk factors for suicide) as low users. Also important was that the time 12th graders spent online doubled between 2006 and 2016; girls tend to spend more time in friendship dyads and boys in groups, and girls focus more on social relationships and popularity. A study of SNS and social self-concept, self-esteem and depressed mood found that the association between having an SNS and these negative indicators is more common with female youth; overall, frequency of SNS use is a positive predictor of social self-concept.¹⁴²

With regard to college students, studies examined the relationship of social medial with well-being,¹²⁸ FOMO,¹³⁰ attachment, social capital¹³⁰ and social closeness based on activity.¹³⁹ Social media use is not associated with mental health problems, nor is emotional regulation; however, emotional regulation is associated with perceived stress and perceived stress is associated with mental health problems.¹²⁸ Social media use does not indirectly predict mental health problems as mediated by perceived stress or emotional regulation. Social media use may indicate challenges with mental health issues or be a way of dealing with difficult emotions. When attachment theory was used to explore individuals’ attachment orientations towards Facebook use related to online and offline social capital, a secure attachment was positively associated with online bonding, bridging and all capital, and offline bridging capital; an avoidant attachment was negatively associated with online bonding capital.¹³⁸ Anxious–ambivalent attachment had a direct association with online bonding capital and an indirect effect on all capital through Facebook. Users in the study on social closeness spent 7.82 min consuming content and 3.13 min on participation.¹³⁹ Interacting with others on social media (e.g. commenting on updates) helps users feel closer to other people and this predicts positive emotional states after Facebook use. A study on FOMO involved two groups (10 min/day versus usual use), and both showed decreases in anxiety and FOMO; only the experimental group showed additional decreases in loneliness and depression.¹³⁰ Moderation helps with mood and loneliness, and reduces anxiety and FOMO.

In a study on giving up Facebook, pre- and post-evaluation of perceived stress and well-being was measured by salivary cortisol between 14.00 and 17.00 h; those using Facebook had lower cortisol levels, less perceived stress, decreased life satisfaction and lower social loneliness on the Social and Emotional Loneliness Scale for Adults.¹³¹ One study examined that a user's activities on Twitter estimate a depressive tendency, based on a medium positive correlation (r = 0.45) between the Zung Self-Rating Depression Scale and the model estimations of potentially meaningful words (≤20).¹⁴⁶ Although a total of 99 words had absolute values of correlation coefficients with Zung scores >0.4, the highest scores were associated with the following words: even if, very, workplace, hopeless, disappear, too much, sickness, bad and hospital.

Implications for clinicians and researchers across clinical populations, problems and well-being

Findings of this scoping review inform approaches by providers, families and teachers when working with social media in children, adolescents and young adults (Table 1). To understand how technology affects the lives of adolescents and emergent adults, it is necessary to engage them in a conversation, share ideas and be available to help with problems. As many young people (and adults) may consider the internet their ‘lifeline’ to social engagement, consideration of the problematic aspects of internet use may be met with reluctance.^6,12,96,156 Exploring beliefs, norms, values, cultural and language factors, and the meaning of technology to the individual, is integral to understanding and meeting the needs of each patient.^16,23,24,132 For providers, the value of forming and maintaining a trusting, therapeutic alliance with youth cannot be overstated, as quality care depends on patient–provider engagement, open and honest communication and shared decision-making for treatment.^11,96,157

An accurate assessment or history is needed of online activities and associated health and risk factors. Internet use may be healthy or problematic, and this continuum may be explored with youth and parents via non-judgemental questioning to clarify the types and extent of technology used (Table 1).^4,5,17,156 Assessment is enhanced with multiple informants: parents, significant others, schools, primary care providers and/or others that know the youth well.^156,157 How they use their time, what they enjoy, how they want others to view them, awareness/use of privacy settings and proneness to risky behaviours is a snapshot of esteem and quality of relationships.^157–159

Providers, families and others need an approach to promote healthy use of social media and prevent problematic social media behaviours. Data on the relationship of social media use and its impact on behaviour – association, mediation or causation – and clinical interventions are limited.^4,5,9,14,158 Nonetheless, positive family/home life, good engagement, supervision and other approaches may reduce risk of risky or dangerous behaviour.^4,24,38,156 A shared understanding is needed about healthy versus problematic use, how to monitor use and blending social media with alternative activities to meet emotional needs. Individual, peer/group and family education and therapy is often helpful. Motivational interviewing techniques may help co-construct a plan that meshes with values, with parent and provider input.^3,24,156