Social Media and Depressive Symptoms in Childhood and Adolescence: A Systematic Review

The internet is a ubiquitous medium for business, information and entertainment, but arguably it has had most profound impact as a means of interpersonal communication. Use of social networking sites grew exponentially after the launch of MySpace and Facebook in 2004. Within a few years, Facebook was being used by four-fifths of internet users aged 13–16 in the UK (Livingstone et al. 2011). Twitter, allowing short messages to be sent to unlimited recipients, was introduced in 2006 and rapidly gained global popularity. Since the launch of internet-connected smartphones, which overtook sales of cell phones in 2013, instant messaging sites such as Snapchat and WhatsApp have become standard tools of communication. In the USA, recent data show that 93% of people aged 15–17 have mobile internet access through a phone or tablet; while Facebook remains highly popular, Instagram and Twitter are more valued by teenagers (Statista 2016). Whether on conventional computer screen or mobile device, young people use social media in every waking hour, in countries rich and poor.

The internet may be a double-edged sword. Neuroscientist Susan Greenfield (2014) argued in her book Mind Change that digital technology has deleterious effects on personality, identity and relationships. Applying Prensky’s (2001) distinction between “digital immigrants” and “digital natives”, Greenfield explained that whereas the former were schooled in a pre-digital era, now using the internet alongside older media, the latter have known nothing else. According to this conceptualization, digital immigrants confer higher value on face-to-face interaction, sometimes criticising younger people for contravening traditional social norms when focusing on their mobile devices, while digital natives dismiss this as older people’s fear of change.

As described in Turkle’s book Alone Together: Why We Expect More from Technology and Less from Each Other (2011), family relationships are weakened when proximate reality is neglected in favor of online contact. Digital technology may be changing conceptualizations and language of human relationships. Defined as a “dyadic, co-constructed phenomenon characterized by reciprocity, closeness and intimacy” (Amichai-Hamburger et al. 2013: 34), friendship is vital for the well-being of children and adults, but the advent of social media appears to have modified its meaning. For “Generation Wired” (a term used by Tam and Walter 2013), such relationships are increasingly generated and maintained online. Data from the Pew Research Center (Madden et al. 2013) show an average of 300 Facebook friends for teenagers in the United States, but the quality of such relationships is more important than the quantity. With a much larger social circle than in traditional friendships, inevitably much communication is shallow (Grieve et al. 2013). Virtual reality may become the real world for some users, to the extent that friends known only through cyberspace become their closest confiding relationships (McKenna et al. 2002).

Evidence suggests that while people with strong social skills and technological abilities accrue benefit from online interaction, those who are less adept do not fare so well. This exacerbation of differences was portrayed by Kraut and colleagues (2002) as “the rich get richer”. By contrast, the social compensation hypothesis postulates that socially-awkward people derive benefit from online contact that they do not find with face-to-face interaction (Valkenburg and Peter 2007; Amichai-Hamburger and Schneider 2014). However, while Dolev-Cohen and Barak (2013) suggested that online communication is supportive for shy, anxious or depressive young people, it may compound their difficulties by reinforcing poor self-esteem (Staksrud et al. 2013).

Concerns have arisen about the mental health impact of internet activity on the young, with frequent coverage of this topic in the mass media. Early evidence of adverse psychological impact was presented by Kraut and colleagues (1998) and by Young and Rodgers (1998), who found that frequent internet use raised the risk of depressive symptoms. Since then, studies have shown correlations of online activity by younger people with low self-esteem (Caplan 2002), loneliness (Clayton et al. 2013), self-harm (Lam et al. 2009) and autistic traits (Finkenauer et al. 2012). However, other studies have indicated higher self-esteem and satisfaction with life, and reduced risk of mental health problems (Valkenburg et al. 2006; Bessièrre et al. 2008; Grieve et al. 2013; Best et al. 2014). Development of supportive social bonds and belongingness can protect against adversities such as loneliness and bullying (Wu et al. 2016).

A high proportion of serious mental health problems in adulthood emerges during adolescence (Kessler et al. 2005; Children Young People’s Health Outcomes Forum 2012). Epidemiological data predating mass use of online social media showed a high risk of depression in this age group, with estimates of 2–5% prevalence of major depressive disorder (Costello et al. 2003), but recent reports show an alarming increase in depressed younger people (Office for National Statistics 2014). The internet, and related social trends, may be a major factor in the rise of psychological morbidity in the young.

Various theories have been proposed for the putative link between social media use and psychological problems in younger people. Socialization is crucial to the progression from adolescence to adulthood, and use of social media may have profound influence on this adjustment (Wood et al. 2016). Applying John Bowlby’s psychanalytic theory, Oldmeadow and colleagues (2013) found that people with attachment anxiety were more likely to turn to Facebook for emotional support. However, reduced face-to-face contact detracts from a traditional supportive environment that can help young people to manage the challenges of adolescence. Development of self-awareness may be inhibited in young people who lack engagement in reflective interactions with family and friends (Siegel 2014). Empathy is honed through social relationships, which may not be as close and meaningful online, where superficial behavior such as virtue-signalling prevails.

The internet may be changing the process of identity formation, which psychoanalyst Erik Erikson (1980) emphasized in the adolescent stage of his model of eight stages of the human lifespan. Each stage presents conflict, which must be resolved to advance successfully to the next stage. Most young people overcome the stress and difficulties of adolescence, but some flounder. Successful progress from adolescence to adulthood entails acquisition of social skills, confirmed and rewarded by social acceptance. Self-presentation is tested through display and response. As friendships become more complex in adolescence, with the emergence of romantic intimacy and sexual interest, there is need for privacy, yet social media encourage openness and divulgence of personal information. A review of studies of online identity development by Wängqvist and Frisén (2016) showed that aspects of identity constrained in offline contact are freely expressed online, and that anonymity in internet communication has implications for cohesive identity formation.

As argued in an influential book Generation Me by Jean Twenge (2006), narcissism may be increasing in Western society. Twenge and colleagues (2008) found considerably higher scores in this trait in students compared to a group of 20 years earlier. Narcissistic behavior has always been more evident in younger people, who have relatively little responsibility to others while tending to be self-absorbed. The extent is probably exaggerated in the media, as the term is used for people merely taking “selfies”: such behavior may be vain, but is socially normative and should not be pathologized if it does not pervasively disrupt daily functioning (Webber 2016). However, the internet has provided more opportunity for expression of narcissistic aspects of personality. People with narcissistic traits are prone to low mood when their high expectations are not fulfilled (Webber 2016). Huprich (2014) described a narcissistic personality pattern including depressive and masochistic tendencies as “malignant self-regard”. A depressive reaction to setbacks is a prominent feature of the DSM-V condition of narcissistic personality disorder.

Several socio-cultural theories have emerged on the effect of digital media on mental health. The internet can be a harsh environment for young people, who are heavily influenced by peer pressure. A review by Wu and colleagues (2016) of research on use of the internet for social purposes showed that a major motive for young people is positive reinforcement of their social connectedness. Social media are the forum for the setting and reinforcing of norms. Conformity is rewarded, while a careless remark might result in a person being ostracized. Unrealistic expectations arise as users see the relative popularity of others, as indicated by their number of friends and “likes”. A study of college students (Feinstein et al. 2013) showed that negative comparisons with peers on Facebook leads to rumination, which increases the risk of depression. Online self-disclosure may relieve stress, generates supportive messages and raises a person’s profile (Tamir and Mitchell 2012), but control of sensitive information is lost. Depressed or anxious young people do not always make sensible decisions about privacy, sometimes revealing personal details in a way that they later regret (McKenna et al. 2002).

Young people are expected to be in perpetual contact, and to project themselves visually as well as verbally. Attractiveness is a major criterion of status and popularity. Young female internet users are particularly keen to choose the most favorable image of themselves on Facebook (Pempek et al. 2009). “Selfies” may be uploaded to seek approval, but an adverse remark may be distressing for someone of delicate self-esteem. Young people are increasingly transmitting sexualized messages or images (“sexting”), with little concern for consent or for exploitation by abusive peers or strangers (Staksrud et al. 2013). Impulsive behavior online may jeopardize future careers, and in some instances children have been criminalized for disseminating sexual images. Aggressive behavior or “trolling” is a common problem in internet use by young people (Ko et al. 2012; Hinduja and Patchin 2013). A review of 113 studies by Kowalski and colleagues (2014) found that cyberbullying correlates with mental health problems in adolescence; in some cases it has led to suicide (Hinduja and Patchin 2010). Bullying may be worse online than in physical proximity, factors being the anonymity of the bully and the inescapable public embarrassment and shame (Slonje et al. 2013). The three most frequent problems arising in counselling sessions provided by ChildLine (a British helpline for children) in 2016 were low self-esteem or unhappiness, family relationships and bullying (online and offline); the latter was the most common reason for counselling in children aged 11 and under, and third in the 12 to 15 years age group (National Society for the Prevention of Cruelty to Children 2016).

Gender differences are an important consideration. Rodgers and colleagues (2013) found that body image concerns correlate with social media use by young female but not male users; such perceptions may lead to eating problems and poorer outcomes of adolescent adjustment. A review of 67 studies of internet use and body image concerns in adolescence by Rodgers and Melioli (2016) described various theoretical perspectives on this link. One theory is self-objectification, which is based on the feminist argument that women are seen as sexual objects under a male gaze. Self-objectification is a form of consciousness manifesting in habitual monitoring of physical appearance, with tendencies for anxiety and shame. Work by Tiggemann and Slater (2013) suggests that self-objectification is a significant cause of psychological problems in adolescence. The combination of media and peer pressure on girls to be thin and sexually attractive is compounded by the internet. Being constantly connected turns a young person into a commodity, to be compared with others (Gonzales and Hancock 2011). However, a limitation of socio-cultural theories is their emphasis on structure over agency, which reduces internet users to passivity.

Research has repeatedly shown that girls use the internet mostly for relational purposes (thus being highest users of social media), while boys tend to use it more for instrumental activities such as video-gaming (Park 2009; Kuss and Griffiths 2012). In a study of Canadian teenagers by Sampasa-Kanyinga and Lewis (2015), 71% of girls used social networks for more than 2 h daily, compared to 29% of boys, which partly explained their finding of a marked gender imbalance in psychological distress. A recent UK government survey of children aged 14–15 (Department of Health and NHS England 2015) showed that girls were more than twice as likely as boys to suffer from symptoms of common mental disorder (37–15%), with the proportion of girls with anxiety or depression rising by 10% since the previous survey 10 years earlier. Among various factors discussed by experts was excessive use of social media (Times 22nd August 2016). Gender-related differences in case ascertainment for mental health problems also need consideration, with girls possibly more likely to seek help.

Addictive personality traits may be stimulated by the internet. Soon after the emergence of the worldwide web, psychiatrist Ivan Goldberg (1995) proposed internet addiction as a specific disorder; manifestations are similar to other compulsive behaviors such as gambling, including preoccupation, mood problems, functional impairment and withdrawal effects (Leung 2004). Young (1999) devised the Internet Addiction Scale, which has been used widely in research on internet use. Another instrument is the Facebook Addiction Scale, which measures mood and withdrawal symptoms (Andreassen et al. 2012). Although such addiction was not included by the American Psychiatric Association (2013) in the latest Diagnostic & Statistical Manual of Mental Disorders, internet gaming disorder was entered as a condition for further investigation. However, such diagnostic expansion has been criticized as medicalisation of everyday life (Moynihan and Henry 2006).

Disruption of the body clock may also be a factor in the psychological consequences of internet use. Blue light emitted by digital screens inhibits secretion of melatonin, which is necessary for sleep. A meta-analysis by Carter and colleagues (2016) showed that sleeping beside mobile devices stimulates the brain. Disturbed sleep patterns may cause obesity, weakened immunity and stunted growth. Daytime tiredness and irritability may have social and educational consequences. Change to normal circadian rhythms has known influence on mood (Lewy et al. 1998).

Thus it can be seen that research on the relationship between social media and mental health in young people is multifaceted. Much of the discussion of this topic has occurred in lay media including the popular press, possibly leading to uncritical acceptance of untested ideas and assumptions. Studies indicate benefits as well as harmful effects of internet use, but evidence is complicated by the lack of either causal mechanisms or a certain direction of relationship between variables. Results in one study setting may not be generalizable to young people elsewhere. Whether any increase of psychological distress in adolescence is attributable to online social media activity is not yet understood. A review of relevant peer-reviewed studies was therefore indicated.

This review examined empirical research on the relationship between social media use and depressive symptoms in the child and adolescent population, with three objectives. First, we aimed to produce a critique of the design and conclusions of relevant studies. It is apparent that research findings, particularly as reported in the popular media, may lack inferential validity in measuring the impact of social media on mental health. Secondly, the review was to analyze correlations between social media use and depressive symptoms, including a “dose” effect, taking account of limitations considered above. Thirdly, the review investigated the role of gender, as differences between male and female behavior and response to social media use have been highlighted in research.

A systematic literature search was conducted, seeking relevant articles in peer-reviewed journals. Eligible studies had a generic child or adolescent sample, rather than selecting groups by mental health morbidity or vulnerability. Social media were defined as websites used primarily for social interaction: these include social networking sites such as Facebook, instant messaging (e.g., WhatsApp) and image-sharing applications (e.g., Instagram). Excluded were studies measuring depressive symptoms in relation to use of the internet rather than social media specifically. Although the internet is a rapidly changing phenomenon, no time period was applied, or geographical restriction; such limits would be arbitrary and a risk of selection bias (McCrae and Purssell 2015). As there was no resource for translation, only studies in English language were included. The databases Medline, PsychInfo, and Embase were used, with the following search strategy:

Articles were screened first by title, then by abstract. At the final stage of screening, full articles were read and assessed independently by all three authors, with any differences in verdict followed by discussion to gain consensus. Eligible articles were summarized and assessed for risk of bias, using the Cochrane Collaboration Methods Group Tool to assess risk of bias in cohort studies.

For statistical analysis, the most relevant result was extracted from each study on the relationship between social media use and depressive symptoms. Most studies had a correlational design, but for studies presenting other statistics, these were converted to correlations using the compute.es package (Re 2014). These were analyzed and forest plots generated using the meta (Schwarzer 2015) and metafor (Viechtbauer 2015) packages in R (R Core Team 2016). The metacor command was used with Fisher’s z transformation for the correlation and the DerSimonian-Laird estimator for τ². Because a-priori heterogeneity was assumed, a random-effects model was used for the primary analysis; although the fixed-effect equivalent was also calculated as comparison of their differing inferential assumptions can be instructive. The random-effects model provides an unconditional inference about a broader set of studies of which the studies included in the meta-analysis are assumed to be a random sample, while the fixed-effects approach makes an inference based only on and about those studies actually included in the meta-analysis (Viechtbauer 2015).

The “rich get richer” and social compensation hypotheses were tested by Selfhout and colleagues (2009), who compared incidence of depression and anxiety between use of internet for communication and other uses in high school pupils. Van der Eijnden and colleagues (2008) studied the relationship between compulsive online communication and psychological well-being, based on previous research indicating that unlike instrumental activities online, use of the internet for social purposes raises the risk of loneliness and depression. The researchers tested several hypotheses to investigate a possible bidirectional relationship.

Self-identity was the theoretical basis for three studies. Social comparison and feedback-seeking are important means of forming a self-identity in adolescence, but Nesi and Prinstein (2015) were interested in how young people may engage in such behaviors in maladaptive ways on the internet. They hypothesized that a high frequency of social comparison and feedback-seeking online would predict depression, although this would be moderated by peer popularity. Neira and Barber (2014) applied self-concept theory in their study of adolescent use of social media. Social comparison and peer feedback are integral to the younger person’s self-evaluation, and these are dramatically increased by internet use, thus making a plausible link between social media, self-concept and depressed mood. Dumitrache and colleagues (2012) studied self-image and depressive tendencies in teenage Facebook users. Referring to the contrasting hypotheses of enhancement and compensation, they examined the relationship between positive or negative self-image and the quantity and type of information posted online.

In a female sample, Tiggemann and Slater (2015) investigated correlations between self-objectification, body shame, dieting and use of various types of media. Internet use was considered as a predictor of self-objectification and its adverse consequences, including depression. Gámez-Gaudix (2014) applied the cognitive-behavioral model to investigate problematic internet use and depressive symptoms in teenagers. In this model, online social communication is less threatening than face-to-face interaction, but this readily available source of emotional support can lead to excessive and dysfunctional use. Online activity may be a maladaptive response to depressive tendencies.

Stress was the focus of two studies. Frison and Eggermont (2015) noted that stress increases in adolescence, due to pressures at school and in family relationships. Coping mechanisms include actively seeking social support and avoidance, which are respectively adaptive and maladaptive. While the internet facilitates social support, friendship on Facebook is often weaker, to the effect that support may not be received, with potentially adverse psychological consequences. The researchers studied relationships between daily stress, seeking and receiving of social support, and depressed mood. A psycho-physiological study by Morin-Major and colleagues (2016) investigated Facebook activity (frequency of use, network size, self-presentation and peer interaction) with basal cortisol level (a measure of stress) and depressive symptoms. It is known that social support is a buffer to biological response to acute stressors.

Hwang and colleagues (2009) considered the internet as a means of social support in the rapidly changing context of Taiwan. In contrast to the individualism of Western societies, Taiwan has a collectivist culture, and young people are subjected to high social pressures in academic performance, sometimes to the detriment of their well-being. Whereas depression has been normalized in American life, it remains stigmatized in Oriental countries, causing double jeopardy for sufferers. The study by Hwang and colleagues was primarily concerned with the online and offline activity of younger people with depressed mood, but it recruited a general adolescent sample and measured the behavioral and psychological variables in a regression model.

The study by Ybarra and colleagues (2005) had no stated theoretical rationale, but they referred to previous research showing differences in use of the internet by young people in relation to depressive symptoms. Through the Youth Internet Safety Survey they measured online communication, self-disclosure and exposure to sexual content and harassment. Depressive symptoms were assessed using DSM categories of minor and major depressive disorder.

Statistical Analysis

The overall random effects pooled estimate was 0.13 (0.05, 0.2), p = 0.001; Q = 131.47, df = 10, p = < 0.0001, I² = 92.4% (Fig. 2), suggesting a clinically and statistically significant relationship between social media use and depressive symptoms (Table 2). There was little evidence of publication bias: the linear regression test of funnel plot asymmetry showing no evidence to reject the null hypothesis of funnel plot symmetry (t = 0.3, df = 9, p = 0.77); although this measures small study effects rather than bias directly. The funnel plot for this is shown in Fig. 3. Additionally the trim and fill analysis showed two outlying studies (Dumitrache et al. 2012; Nesi and Prinstein 2015), trimming and filling of which had some effect in reducing the random effects estimate, r = 0.09 (95% CI 0.01, 0.16), p = 0.03; Q = 180.67, df = 12, p < 0.0001, I² = 93.4% (see Supplementary File 2).

Open image in new window

Table 2

Selected study results

Authors (year)	Included sample	Correlates	Correlation result
Ybarra et al. (2005)	1061	Internet used most frequently for chatroom versus other purposes/depressive symptoms	0.14**
Van der Eijnden et al. (2008)	663	Instant messaging (time 1)/depression (time 2)	0.17**
Hwang et al. (2009)	6341	Online communication/depressed mood	0.13*
Selfhout et al. (2009)	307	Instant messaging (time 1)/depression (time 2)	−0.02
Neira and Barber (2014)	1819	Social media use/depressive symptoms	−0.09*
Gámez-Gaudix (2014)	699	Preference for online social interaction (time 1)/depression (time 2)	0.13*
Frison and Eggermont (2015)	910	Social support seeking on Facebook/depressed mood	0.13**
Nesi and Prinstein (2015)	619	Technology-based social comparison and feedback-seeking (time 1)/depressive symptoms (time 2)	0.34**
Tiggemann and Slater (2015)	204	Social media use/depressive symptoms	0.19*
Morin-Major et al. (2016)	88	Facebook use/depressive symptoms	−0.097
Dumitrache et al. (2012)	76	Items of identity-related information on Facebook/depression	0.355*

*p < 0.01

**p < 0.001

Open image in new window

To assess any possible impact from publication type, sub-group analyses were conducted for the two types of study design. The pooled estimate for cross-sectional studies (n = 6) was r = 0.12, (95% CI 0.02, 0.22), Q = 81.01, I² = 93.8%; while that for longitudinal studies (n = 5) was r = 0.12, (95% -0.01, 0.25), Q = 38.98, I² = 89.7%; suggesting little effect, although the difference between study types was not always marked and the time over which the longitudinal studies were conducted varied widely. Although there was a difference between the two estimates, this is not clinically significant and the test for subgroup difference was not statistically significant (Q = 0.0, df = 1, p = 0.98). In order to try to understand the results Euclidean cluster analysis was undertaken based on the results alone. The agglomerative coefficient was strong (0.94) and showed three main clusters (Fig. 4), suggesting that study outcomes could be broadly put into three groups. Examination of the clusters revealed two groups of outlying studies; one showing small negative correlations and the other large positive effects. However, the third and largest cluster, accounting for the majority of studies, had a limited range of outcomes. Thus although we cannot account for the clusters methodologically, this distribution of results is supportive of our pooled estimate being an accurate reflection of the underlying phenomenon.

Open image in new window