You are on page 1of 13

Research on Child and Adolescent Psychopathology (2021) 49:241–253

https://doi.org/10.1007/s10802-020-00715-7

Social Media and Depression Symptoms: a Meta‑Analysis


Simone Cunningham1 · Chloe C. Hudson1 · Kate Harkness1

Accepted: 29 September 2020 / Published online: 6 January 2021


© Springer Science+Business Media, LLC, part of Springer Nature 2021

Abstract
Social Networking Sites (SNS) have close to 3 billion users worldwide. Recently, however, SNS have come under media
scrutiny for their potential association with depression. Two previous meta-analyses failed to find evidence for a robust
concurrent association between SNS use and depression symptoms. However, these analyses focused primarily on the time
spent using SNS. The current meta-analysis is the first to consider the multi-dimensional nature of SNS use, and examines
separately the quantitative associations of depression symptoms to SNS use in three types of SNS studies examining three
distinct constructs of SNS use: time spent using SNS, intensity of SNS use, and problematic SNS use. Sixty-two studies
(N = 451, 229) met inclusion criteria. Depression symptoms were significantly, but weakly, associated with time spent using
SNS (r = 0.11) and intensity of SNS use (r = 0.09). However, the association of depression symptoms to problematic SNS
use was moderate (r = 0.29), was significantly higher than for time spent using SNS (Qbetween = 35.85, p < 0.001) or intensity
of SNS use (Qbetween = 13.95, p < 0.001), and was not significantly moderated by age, gender, year of study publication, or
mode of recruitment. These results suggest that future research examining causal models of the relation of SNS use and
depression, as well as research on intervention and prevention, should focus in more detail on individuals who are engaging
in a pattern of problematic SNS use.

Keywords Depression symptoms · Social networking sites · Problematic social media use

Social Networking Sites (SNSs) are a worldwide platform negative social comparison and cyber-bullying (e.g.,
for communication (Kaplan and Haenlein 2010). In March Kırcaburun et al. 2019). A growing body of research sug-
2020, Facebook was estimated to have over 2.6 billion users gests that SNS use is associated with symptoms of depres-
worldwide, and Instagram recently surpassed 1 billion users sion, anxiety, low self-esteem, and loneliness (Saiphoo et al.
(Clement 2020). Because of the rise of SNSs we now have 2019; Vannucci et al. 2017), with a recent high-impact report
unparalleled opportunities for information exchange and suggesting that this association in adolescents may be causal
discovery, as well as easy and instantaneous connection (Twenge et al. 2018).
with people all over the world. Some research has touted The potential relation between SNS use and depression is
the advantages of SNS use at the individual level, includ- particularly concerning. Depression affects over 300 million
ing benefits on mood, cognitive function, and social support people worldwide and is the leading global cause of dis-
(Ellison et al. 2007; Myhre et al. 2017; Park et al. 2009; Patel ability (World Health Organization 2018). It typically first
et al. 2015; Primack and Escobar-Viera 2017). At the same onsets in adolescence or young adulthood (Weissman et al.
time, SNSs have come under intense scrutiny in recent years 2006), which is both the developmental period associated
for their negative aspects, including the potential to promote with the highest SNS use (Smith and Anderson 2018), and
the age cohort in which rates of depression have been stead-
ily rising (Twenge 2015; Twenge et al. 2018). Very few pro-
Electronic supplementary material The online version of this
article (https​://doi.org/10.1007/s1080​2-020-00715​-7) contains spective or experimental studies exist testing causal models
supplementary material, which is available to authorized users. of the relation between SNS use and depression. However,
the studies that do exist appear to point to the possibility of
* Simone Cunningham a bi-directional relation; that is, high levels of SNS use have
0scc6@queensu.ca
been found to prospectively predict increases in depression
1
Department of Psychology, Queen’s University, 62 Arch (e.g., Vernon, Modecki, & Barber, 2016), and individuals
Street (Humphrey Hall), Kingston, ON K7L 3L3, Canada

13
Vol.:(0123456789)
242 Research on Child and Adolescent Psychopathology (2021) 49:241–253

with depression appear to be more vulnerable to the nega- separate meta-analytic investigation. Another, larger, set of
tive impacts of SNS use (Vernon et al., 2016). Qualitative studies has focused more broadly on the emotional invest-
reviews of the much larger number of studies examining the ment, or “intensity,” of use, as well as the extent to which
cross-sectional association between depression and SNS use SNS use has become addictive or problematic. While these
reveal that this relation has been inconsistently replicated latter studies may not address the very specific features of
and is complex, likely varying as a function of a number of SNSs, they do address the general ways in which individuals
methodological and sample moderators (Baker and Algorta use SNSs that may be associated with negative outcomes. In
2016; Blease 2015; Frost and Rickwood 2017; Seabrook this quantitative review we focus on three distinct operation-
et al. 2016). Therefore, the time is ripe for a rigorous meta- alizations of SNS use that have received the most attention
analysis to quantify the strength of the concurrent relation in the literature and, thus, are frequent enough to permit
between SNS use and depression symptoms, and to identify meta-analytic investigation: (a) time spent using SNSs; (b)
key moderators that can be used to inform much-needed the “intensity” of SNS use; and (c) problematic SNS use.
studies designed to test causation. It is important to note that there is no agreed-upon con-
sensus operationalization of the above three constructs in the
literature, and studies vary, sometimes widely, in the ways
Prior Research on the Relation Between SNS in which these constructs are operationalized and assessed.
Use and Depression Symptoms In general, however, time spent using SNSs is typically
assessed with a small number of items querying about the
To the best of our knowledge, only two meta-analytic studies amount of time (e.g., minutes, hours) per day or per week
have quantitatively examined the average effect sizes of the spent on social media (e.g., Baker et al. 2016; Lin et al.
concurrent relation between SNS use and depression symp- 2016; Tandoc et al. 2015). Examples include, “Indicate in
toms (Huang 2017; Yoon et al. 2019). Both meta-analyses hours and minutes (e.g., 3 h and 6 min) how much time you
included studies conducted in North America, Europe, Asia, spend social networking each day” (Baker et al. 2016) and
and Australia. First, in separate meta-analyses, Yoon and asking participants to estimate their “total time per day on
colleagues (Yoon et al. 2019) reported small positive cor- social media for personal use” (Lin et al. 2016).
relations between depression symptoms and time spent on Intensity of SNS use is a heterogeneous construct that
SNSs (e.g., minutes using SNSs per day, hours using SNSs is typically assessed with SNS intensity scales, such as the
per week; 33 articles; r = 0.11) and SNS checking frequency Facebook Intensity Scale (FIS; Ellison, Seinfeld, & Lampe,
(e.g., number of times SNSs were checked per day or per 2007). These scales are designed to measure users’ emo-
week; 12 articles; r = 0.10). Age (adults versus adolescents/ tional connectedness to SNSs, as well as the extent to which
children) and gender (percent of the sample who identified SNS use is integrated into their day-to-day life (Avery 2018;
as female) were not significant moderators in either case. Daniels 2014; Ellison et al., 2007). These scales generally
Second, in a meta-analysis of 61 studies, Huang (2017) yield a total score based on multiple features of SNS use,
reported that the relations of time spent using SNSs (e.g., and may include items assessing time spent on SNSs (e.g.,
minutes per day, hours per week) to overall psychological minutes or hours per day or week), number of followers and/
wellbeing was negative, but small (r = −0.07). When the or friends, number of posts (e.g., number of posts per day or
negative indicators of psychological well-being (i.e., lone- per week), feelings of connectedness with SNS communities
liness and depression) were considered separately, they were (e.g., “I feel I am part of the Facebook community”; Ellison
also weak (rs = −0.08, −0.11, respectively). Age and gender et al., 2007), and the extent to which SNSs have become
were again not significant moderators of these relations. The integrated into individuals’ daily routine (e.g., “Facebook
meta-analytic knowledge to date, then, suggests that there has become a part of my daily routine”; Ellison et al., 2007).
is a statistically significant, but potentially not a clinically Therefore, the ‘intensity’ construct includes time spent on
meaningful, concurrent association between depression SNSs in its definition and assessment, but also goes beyond
symptoms and SNS use when use is defined as time spent this to include other distinct features of SNS use.
using SNSs or checking frequency. The construct of problematic SNS use has also been
There are many different ways to operationalize and variously defined in the literature, but, in general, reflects
assess SNS use beyond time spent using SNS and frequency a pattern of use that is characterized by behavioral and
of checking SNS. For example, some studies have specifi- psychological features of addiction (e.g., Andreassen et al.
cally compared active versus passive use (e.g., Escobar- 2017; Bányai et al. 2017; Kırcaburun et al. 2019). Most
Viera et al. 2018), or the extent to which SNSs involve social recently, problematic SNS use has been situated in the
comparison (e.g., Avery 2018). However, research on these biopsychosocial model of addiction and, like other behav-
more specific constructs is still at an early stage, and, thus, ioral addictions, has been theorized to involve a compul-
for many of them there are not enough studies to permit sion to engage in a non-substance related behavior despite

13
Research on Child and Adolescent Psychopathology (2021) 49:241–253 243

the potential negative consequences (Kırcaburun et al. Moderators of the Relation Between
2019). Thus, problematic SNS use is often determined by Depression Symptoms and SNS Use
assessing whether there is high frequency of SNS use in
combination with a variety of addiction symptoms, such Age Adolescents are higher users of SNSs than adults (Per-
as dependence, tolerance, and withdrawal (Bányai et al. rin 2015), and many of the negative aspects of SNS use,
2017). Problematic SNS use has been assessed with a vari- including cyber-bullying, negative social comparison, and
ety of self-report questionnaires (e.g., Andreassen et al., ‘fear of missing out’ (FOMO) have been examined primar-
2012; Andreassen et al. 2017; Elphinston and Noller 2011; ily in adolescent samples (e.g., Kırcaburun et al. 2019).
Jung and Kim 2014). One of the most commonly used Because adolescence is a critical period of transition in
scales, the Bergen Social Media Addiction Scale (BMAS; terms of identify formation, peer and romantic relationship
Andreassen et al. 2017), includes items such as “You have building, and neuro-emotional development, generally, ado-
tried to cut down on the use of social media without suc- lescents have been theorized to be especially vulnerable to
cess” and “You use social media so much that it has a the depressogenic aspects of SNS use (Shapiro and Mar-
negative impact on your job/studies,” rated on a scale from golin 2014). Indeed, in a recent large-scale study, Twenge
1 (very rarely) to 5 (very often). and colleagues (Twenge et al. 2018) suggest that increases
As noted above, the meta-analyses by Huang (2017) in rates of depression and suicidality in teens observed in
and Yoon and colleagues (Yoon et al. 2019) found sta- the period since 2010 are directly attributable to increases
tistically significant, but very small, concurrent associa- in SNS use by this age group over this period. As noted
tions between depression symptoms and both time spent above, age was not a significant moderator of the concur-
on SNSs and frequency of checking SNSs. However, to rent relation between time spent using SNSs and depression
date, there does not exist a meta-analytic summary of the symptoms in the two previous meta-analyses (Huang 2017;
relation between depression symptoms and the extent to Yoon et al. 2019). However, it is not known whether age
which SNS use is emotionally consuming (i.e., intensity moderates the relation of depression symptoms to potentially
of SNS use) and potentially addictive. Theoretically, an more depression-relevant indices, including the intensity and
intense or problematic pattern of SNS use may be more problematic use of SNSs.
strongly associated with depression than non-problematic
use, regardless of total time spent using SNSs, because Gender Adolescent and adult women have been reported
of the potential for addictive behavior, in general, to fuel to spend more time on SNS than men, and are more likely
emotions of guilt, shame, and loneliness (e.g., Bilevicius to use SNSs for relational purposes, whereas men are more
et al. 2018). Similarly, already depressed individuals may likely to use SNSs for general information-seeking (Brooks
be more likely than non-depressed individuals to become and Longstreet 2015; Krasnova et al. 2017). Further, the
emotionally invested in SNSs, and to engage in prob- prevalence of depression is at least twice as high in women
lematic SNS use, given the strong association between compared to men, and this gender gap is even more pro-
depression and addictive behavior, in general (Edgerton nounced in adolescents (Salk et al. 2017). Some studies
et al. 2018; Swendsen and Merikangas 2000). Therefore, have, indeed, found that the relation between depression
examination of the intensity and addictive nature of SNS symptoms and SNS use is stronger in women than men (e.g.,
use is a crucial step in better understanding the potential Brooks and Longstreet 2015). Further, a recent longitudi-
relation between SNS use and depression symptoms. Fur- nal study with a large sample found that SNS use predicted
ther, given the significant costs associated with disorders depression symptoms only in adolescent girls, not boys (Hef-
like depression (World Health Organization 2018), under- fer et al. 2019). However, other studies have failed to find
standing this relation is crucial from a prevention and pub- evidence of a gender difference in SNS use (Baker et al.
lic health perspective as it may highlight more clinically 2016; Barry et al. 2017). Further, as noted above, gender
meaningful intervention targets. was not a significant moderator of the concurrent relation
Overall, the inconsistent findings in the qualitative litera- between depression symptoms and time spent using SNSs
ture regarding the concurrent relation between depression or frequency of checking SNSs in the two previous meta-
symptoms and SNS use, and the weak meta-analytic correla- analyses (Huang 2017; Yoon et al. 2019). However, it is not
tions between depression and time spent using SNSs, specifi- known whether gender moderates the relation of depression
cally, highlight the need to broaden the operationalization to potentially more depression-relevant constructs of SNS
of SNS use. The current meta-analysis represents the first to use, including the intensity and problematic use of SNSs.
our knowledge to quantitatively examine and compare the
concurrent relations between depression symptoms and SNS Year of Publication Over the past several years, the num-
use in studies that have assessed (a) time spent using SNSs; ber of individuals who use SNSs on a regular basis has
(b) intensity of SNS use, and (c) problematic SNS use.

13
244 Research on Child and Adolescent Psychopathology (2021) 49:241–253

drastically increased. For example, in March 2012 Face- (e.g., Karsten et al. 2010); and subthreshold symptoms are
book had approximately 901 million active users. By the strongest prospective risk factor for the future develop-
March 2020, Facebook had an estimated 2.6 billion users ment of a full-blown depressive disorder and, thus, are an
(Clement 2020). Further, increased use of handheld, port- important indicator of risk (Shankman et al. 2009).
able electronic devices such as smartphones has likely In terms of moderating variables, we hypothesized that
improved the accessibility of SNSs and the way that SNS the relation between depression symptoms and SNS use
is used and integrated into individuals’ daily routines. would be stronger (a) in studies that included participants
Because of these changes and their potential impact on under the age of 18 than in studies that only included adults;
SNS research data, we examined studies’ year of publica- (b) in studies that included a higher proportion of female
tion as a moderator in the relation between SNS use and participants; (c) in studies that excluded versus included
depression symptoms. non-SNS users; and (d) in studies that recruited participants
exclusively through SNSs relative to using more diverse
Design Characteristics We examined two important design recruitment strategies.
features that may moderate the focal association of SNS
use and depression. First, we examined whether the study
included versus excluded non-SNS users, as we reasoned Methods
that the exclusion of non-users may fail to capture an impor-
tant part of the SNS use distribution. Second, we examined Literature Search Strategy
whether or not a study’s participants were recruited solely
through advertisements placed on SNSs, as we reasoned that We used the following databases to identify articles: PSY-
participants who respond to SNS advertisements may use CInfo, PubMed, Web of Science, and ProQuest Dissertations
SNSs more frequently than those who do not, potentially and Theses. For all databases, we searched for the follow-
resulting in a less representative sample. Studies also differ ing string in titles, keywords, and abstracts: “(depress* or
widely in terms of the specific SNS examined (e.g., Face- dysthym*) and (social media or social network* or Face-
book vs. Instagram vs. a combination, etc.). Unfortunately, book or Twitter or Instagram or Snapchat or LinkedIn)”.
given the vast heterogeneity across studies on this variable, Searches were conducted in January 2019. In addition to our
it was not possible for us to include site as a moderator, main search, we also performed backward searches (exam-
although descriptions of the SNSs included in each study ining the references section of eligible studies) and forward
are provided. searches (examining articles that had cited eligible studies).
The searches initially yielded 4118 articles (see Fig. 1). Of
these, 1013 duplicates were detected and removed, leaving
The Current Study 3105 articles for initial examination.
During the first stage of the search examination, we
The goal of this meta-analysis is to quantitatively summa- screened the titles and abstracts of all articles and excluded
rize and compare three types of studies that assessed the those that were (1) not quantitative; (2) not written in Eng-
concurrent relations between depression symptoms and: (a) lish; (3) did not assess depression symptoms, or (4) did not
time spent using SNSs, (b) intensity of SNS use, and (c) assess any kind of SNS use. Studies were deemed eligible for
problematic SNS use. We hypothesized that all three groups inclusion as long as they included a measure of depression
of studies would show significant effect sizes in association symptoms, whether self-report or clinician-rated.1 Studies
with depression symptoms. However, we expected the aver- that included participants who met diagnostic criteria for a
age effect size to be significantly greater among studies that depressive disorder were not excluded as long as they also
assessed intensity of use and problematic SNS use compared included a measure of depression symptoms.
to studies that assessed time spent using SNSs. A total of 288 articles were identified as potentially eli-
This meta-analysis focuses on depression symptoms, and gible for inclusion. During the second stage of examina-
not the diagnosis of a depressive disorder, for two main rea- tion, the full texts of articles were fully reviewed by the first
sons. Most pragmatically, there were not sufficient studies to and second authors. During this second stage, articles were
permit comparison of effect sizes between studies of symp- excluded if (1) it became apparent that they did not meet
toms versus diagnosis. More conceptually, however, depres- the inclusion criteria from the first stage; (2) they did not
sion symptoms are an important correlate in and of them-
selves: Depression is a dimensional phenomenon in nature
(Kotov et al. 2017); subthreshold symptoms are much more 1
The vast majority of studies utilized self-report measures of depres-
common in the population than a depression diagnosis and sion. Therefore, it was not possible to include mode of assessment
are associated with a similar level of functional impairment (i.e., self- vs. clinician-report) as a moderator.

13
Research on Child and Adolescent Psychopathology (2021) 49:241–253 245

Articles identified through database Articles identified through forward


searches (k = 4118) and backward searches (k = 135)

Articles screened after


duplicates removed (k = 3105)

Articles excluded based on


title and abstract (k = 2817)

Full-text articles assessed for


reliability (k = 288)

Articles excluded based on


full text (k = 225)

Articles included in meta-analysis


(k = 62)

Studies that reported Studies that reported SNS Studies that reported
time spent using SNS (k intensity (k = 6)1 problematic use (k = 20)1
= 44)1

Fig. 1  Study flowchart. 1Note that these numbers do not add up to 62 articles included analyses for depression and more than one construct
because they account for a instances where there were separate stud- of SNS use
ies with different participants within an article and b because some

measure at least one of time spent using SNSs, intensity of only one author was able to provide the relevant statistics.
SNS use, or problematic SNS use, and (3) they assessed only The second stage of examination that included a full text
nighttime use, in-bed use, active use, or passive use. These review resulted in 62 articles that were eligible for inclusion.
studies were not included as they assessed only a specific Appendix C provides a summary of all 62 studies included
subset of individuals’ total SNS use, and, thus, may not be in the meta-analyses.
representative of participants’ overall SNS use.
If more than one study used participants that were part of Study Coding and Inter‑Rater Reliability
the same larger study, the study with the larger sample was
included in the meta-analysis. In instances in which a paper All articles were coded by the first and second authors and
included more than one study with different samples, the inter-rater reliability was assessed separately for initial
two studies were coded independently. Finally, when stud- screening, eligibility screening, and for data extraction.
ies appeared to meet inclusion criteria but did not report the Inter-rater reliability was high for the initial screening, with
relevant statistics required for the meta-analysis, the corre- consensus on the status of 2981 of the 3105 articles screened
sponding author was contacted and unpublished data were (96% agreement). Similarly, inter-rater reliability was high
requested. If the author did not respond or could not provide for study inclusion versus exclusion, with consensus on the
the information, the study was excluded. In total, there were status of 282 of the 288 potential studies (98% agreement).
10 authors contacted. Of these, three responded. However, Finally, inter-rater was high for data extraction. Eleven data

13
246 Research on Child and Adolescent Psychopathology (2021) 49:241–253

points were extracted for each of the 62 studies included regression of effect sizes onto the continuous moderator
in the meta-analysis (see Appendix C), resulting in a total (Card 2012). In this case, a significant meta-regression
of 682 data points extracted. There was initial agreement indicates that the moderator changes the relation between
on 646 data points (95% agreement). Disagreements for the depression symptoms and SNS use.
initial screening, eligibility screening, and data extraction Categorical moderators (age category of participants,
were discussed and a consensus reached. exclusion of non-SNS users, and recruitment through SNSs)
The moderator variables were coded as follows: Age were analyzed using the Qbetween statistic. The Qbetween sta-
was assessed as mean age and age category. Age category tistic assesses whether between-group heterogeneity was
categorized studies as including 0-only individuals under greater than expected (Card 2012). In this case, a signifi-
age 18 (children and adolescents), 1-only individuals age cant Qbetween statistic indicates that the effect size for the
18 or older (adults), or 2-both. Gender was coded as the relation between depression symptoms and SNS use varies
proportion of the sample that was female.2 Year of publi- significantly for different levels of the categorical moderator.
cation was included as a continuous variable ranging from In cases where there was a significant Qbetween statistic for
2012 to 2018. Exclusion of non-SNS users was coded as a moderator with more than two levels, follow-up analyses
0-including participants who did not use SNSs at all or were conducted comparing across levels of the moderator.
1-excluding non-SNS users. Finally, study recruitment was Publication bias was assessed by examining the fun-
coded as 0-recruited participants through SNSs or 1-did nel plots for each of the three meta-analyses (time spent
not recruit participants through SNSs. using SNSs, intensity of SNS use, and problematic SNS
use). If significant funnel plot asymmetry was detected,
Data Analysis suggesting publication bias, the “trim and fill method”
(Duval 2005; Duval and Tweedie 2000) was used to
The meta-analyses were performed using the Compre- determine the number of missing studies due to publica-
hensive Meta-Analysis software (Borenstein et al. 2005). tion bias. Specifically, studies that contributed to fun-
Three sets of analyses were performed to examine studies nel plot asymmetry were temporarily “trimmed” until
that reported on the relation between depression symptoms the funnel plot was symmetrical. Next, trimmed studies
and (a) time spent using SNSs, (b) intensity of SNS use, were reinstated and studies that fell within the under-
and (c) problematic SNS use. Random effects models were represented section were imputed, which allowed for
used for all analyses because they better manage outliers more accurate estimation of the mean effect size. Funnel
in effect sizes and sample sizes compared to fixed-effect plots for the present paper are included in electronic sup-
models and, thus, results are more generalizable (Card plementary material. Publication bias was also assessed
2012). To test for heterogeneity in the variation in effect quantitatively using Egger’s weighted regression analysis
sizes across studies, we calculated Q and I2 statistics. For- (Egger et al. 1997).
est plots were used to visualize the extent of heterogeneity
across studies. Further, a fourth set of analyses were con-
ducted to statistically compare the effect sizes across the Results
three types of studies. Several studies included in the pre-
sent meta-analysis reported the relation between depres- Depression and Time Spent Using SNSs
sion symptoms and more than one measurement of SNS
use (e.g., both time spent using SNSs and problematic Across the 44 studies that assessed the relation between
SNS use; see Appendix C). Although all of these studies depression symptoms and time spent using SNSs, the
were included within the initial three meta-analyses, they pooled effect size expressed as a correlation was 0.11
were excluded from final analyses comparing the effect (95% CI = 0.086 – 0.13), p < 0.001. The results were het-
sizes between the three types of studies in order to retain erogeneous, I2 = 96.65, Q = 2091.52, p < 0.001. A forest
independent samples. plot for the meta-analysis of depression symptoms and
Continuous moderators (mean age of sample, proportion time spent using SNSs is depicted in Fig. 2.
of sample that was female, and year of publication) were The relation between depression symptoms and
analyzed using a meta-regression, which involves a weighted time spent using SNSs was not significantly moder-
ated by mean age (regression coefficient = 0.002, 95%
CI = −0.002 – 0.007, p = 0.321) or gender (regres-
sion coefficient = 0.001, 95% CI = −0.003 – 0.002,
2
Many studies included in the present meta-analysis assessed gen- p = 0.173). When age was modelled as a categorical
der through self-report items and did not appear to specifically query
about biological sex. Thus, we use the term “gender” throughout the variable, however, the pooled effect size between depres-
present paper as opposed to sex. sion symptoms and time spent using SNSs was greater

13
Research on Child and Adolescent Psychopathology (2021) 49:241–253 247

Fig. 2  Forest plot of depression


symptoms and time spent using
SNS

as a trend in studies that used only adults (mean effect significantly higher in studies that recruited participants
size = 0.11, k = 24) compared to studies that used only through SNSs (pooled effect size = 0.21, k = 6) compared
children or adolescents (mean effect size = 0.02, k = 6), to studies that did not (pooled effect size = 0.10, k = 38),
Q between(1) = 3.40, p = 0.065. Year of publication did not Qbetween(1) = 4.94, p = 0.027.
significantly moderate the relation between depression The funnel plot was asymmetrical, suggesting that publica-
symptoms and time spent using SNSs (regression coef- tion of studies of the relation between depression symptoms
ficient = 0.01, 95% CI = −0.005 – 0.02, p = 0.203). and time spent using SNSs is biased towards reporting higher
There was a trend for the effect size between depression effect sizes. The quantitative measure of publication bias also
symptoms and time spent using SNSs to be higher in stud- was significant (Egger’s weight regression analysis, p < 0.001).
ies that only included SNS users in the analyses (pooled The trim and fill method imputed that approximately 14 stud-
effect size = 0.14, k = 15) compared to studies that also ies might be missing due to publication bias. After adjustment
included non-SNS users (pooled effect size = 0.09, k = 29), for publication bias, the effect size of the relation between
Q between (1) = 3.34, p = 0.067. The effect size between depression symptoms and time spent using SNSs was 0.06
depression symptoms and time spent using SNSs was (95% CI = 0.04 – 0.08).

13
248 Research on Child and Adolescent Psychopathology (2021) 49:241–253

Fig. 3  Forest plot of depres-


sion symptoms and intensity of
SNS use

Fig. 4  Forest plot of depres-


sion symptoms and problematic
SNS use

Depression and Intensity of SNS Use Depression and Problematic SNS Use

Across the six studies that assessed the relation between Across the 20 studies that assessed the relation between
depression symptoms and intensity of SNS use, the depression symptoms and problematic SNS use, the
pooled effect size expressed as a correlation was 0.09 pooled effect size expressed as a correlation was 0.29 (95%
(95% CI = −0.003 – 0.18), p = 0.058. The results were CI = 0.23 – 0.35), p < 0.001. The results were heterogeneous,
heterogeneous, I 2 = 75.06, Q = 20.05, p < 0.01. A forest I2 = 95.49, Q = 421.61, p < 0.001. A forest plot is depicted
plot for the meta-analysis of depression symptoms and in Fig. 4.
SNS use intensity is depicted in Fig. 3. Moderator analy- Of the studies that assessed problematic SNS use, only
ses were not conducted for intensity of SNS use as, with one had a sample comprised of individuals under age
only 6 studies total, there was not adequate coverage of 18. However, the relation between depression symptoms
the varying areas or levels of the moderator variables and problematic SNS use was not systematically related
(see Card 2012). to age when examined continuously (regression coeffi-
The funnel plot was symmetrical, suggesting no pub- cient = −0.002, 95% CI = −0.011 – 0.008, p = 0.715). Fur-
lication bias. Similarly, the quantitative assessment of ther, the relation between depression symptoms and prob-
publication bias was not significant (Egger’s weighted lematic SNS use was not significantly moderated by gender
regression analysis, p = 0.154). (regression coefficient = 0.001, 95% CI = −0.004 – 0.004,

13
Research on Child and Adolescent Psychopathology (2021) 49:241–253 249

p = 0.871). Year of publication did not significantly mod- SNSs (r = 0.11) and SNS intensity (r = 0.09) were statisti-
erate the relation between depression symptoms and time cally significant or approaching statistical significance, the
spent using SNSs (regression coefficient = −0.01, 95% effect sizes were small, suggesting that these may not be
CI = −0.05 – 0.03, p = 0.524). clinically meaningful relations. Although this was expected
The effect size of the relation between depression symp- for time spent using SNSs, we had hypothesized that inten-
toms and problematic SNS use was not significantly different sity of SNS use, similar to problematic SNS use, would have
in studies that exclusively included SNS users in the analy- a stronger association with depression symptoms than time
ses (pooled effect size = 0.32, k = 8) compared to studies spent using SNSs.
that also included non-SNS users (pooled effect size = 0.27, The current meta-analytic results suggest that problem-
k = 12), Qbetween(1) = 1.18, p = 0.277. Only two studies exclu- atic SNS use may be the primary construct driving the rela-
sively recruited participants through SNSs. Thus, no mod- tion between SNS use and depression, more so than time
erator analyses were conducted on this variable. spent using SNSs or intensity of SNS use more generally.
The funnel plot was relatively symmetrical, suggesting If so, this finding has important implications for identify-
no publication bias. Similarly, the quantitative assessment ing individuals who are most at risk for negative outcomes,
of publication bias was not significant (Egger’s weighted regardless of the directionality of this relation, and provides
regression analysis, p = 0.102). a research target for future longitudinal studies examining
the transactional relations between depression symptoms
Comparing Time, Intensity, and Problematic Use and SNS use, as well as mediators of this relation, over time.
of SNSs For example, if problematic SNS use prospectively predicts
depression, then future research could investigate inter-
The omnibus comparison of effect sizes across the total time, ventions for the addictive behaviors associated with SNS
intensity, and problematic nature of SNS use was significant, use as a means of preventing the emergence of depression
Qbetween(2) = 31.61, p < 0.001. As hypothesized, the mean symptoms and disorder. Similarly, if depression symptoms
effect size of studies that assessed problematic SNS use prospectively increase the likelihood of problematic SNS
(pooled effect size = 0.29, k = 15) was significantly greater use, then existing prevention and intervention strategies for
than that of studies assessing time spent using SNSs (pooled depression could be amended to include modules that spe-
effect size = 0.11, k = 38; Qbetween(1) = 35.88, p < 0.001) cifically target teaching more adaptive ways to use SNSs.
and intensity of SNS use (pooled effect size = 0.09, k = 5; A possible alternative explanation for the above associa-
Qbetween(1) = 13.92, p < 0.001), while the latter two effect tion between problematic SNS use and depression symptoms
sizes did not differ significantly, Qbetween(1) = 0.18, p = 0.666. is construct overlap in these cross-sectional studies. As noted
It is noteworthy that the pooled effect sizes in these analy- previously, problematic SNS use is characterized by symp-
ses, which excluded several studies that reported more than toms of behavioral addiction and, by definition, negatively
one type of SNS use measurement, did not differ from the impacts functioning. Thus, it is not surprising that individ-
pooled effect sizes for each meta-analysis that included these uals who endorse elevated levels of depression symptoms
studies. would also endorse items on scales of measures of prob-
lematic SNS use (e.g., “used [SNSs] to reduce feelings of
guilt, anxiety, helplessness, and depression,” “used [SNSs]
Discussion to forget about personal problems” from the BSMAS). This
possibility points, again, to the importance of prospective,
The goal of this meta-analysis was to statistically synthesize longitudinal research that attempts to determine the causal
and compare effect sizes in the existing literature for the con- and transactional associations between SNS use and symp-
current relation between depression symptoms and SNS use toms of psychopathology, including depression.
across three types of studies examining distinct constructs The observation of a small pooled effect size for the rela-
of SNS use: (a) time spent using SNSs; (b) intensity of SNS tion between SNS use intensity and depression symptoms
use; and (c) problematic use of SNS. The findings suggest was unexpected. One possible explanation for this finding
that, in general, greater SNS use is significantly associated is that although some items included in intensity scales
with elevated depression symptoms. Further, partially con- may overlap with features associated with problematic use
sistent with hypotheses, the effect size for studies that specif- (e.g., frequent use; high integration into one’s daily routine),
ically examined a pattern of problematic SNS use indicated some items may also be assessing more positive or protec-
a moderate strength of effect (r = 0.29) that was significantly tive aspects of use. For example, individuals who endorse
higher than the effect for studies that assessed either time the FIS item “I feel I am part of the Facebook community”
spent using SNSs or intensity of SNS use. Although the (Ellison et al., 2007) may also experience more social sup-
pooled effect sizes for studies examining time spent using port and more positive interactions on SNSs, which has in

13
250 Research on Child and Adolescent Psychopathology (2021) 49:241–253

turn been associated with positive outcomes (e.g., Davila recruited through SNS advertisements (r = 0.21) was similar
et al., 2012). Therefore, it is possible that the heterogeneity in strength to the overall effect size for problematic SNS
in items on SNS intensity scales may have resulted in a small use. This finding suggests that there may be important dif-
pooled effect size in the relation between intensity of SNS ferences between individuals who respond to research study
use and depression symptoms. advertisements on SNSs versus individuals recruited from
In addition to showing the largest effect size, the relation other sources and signals a need to be aware of sample bias
of depression and problematic SNS use was not significantly in future studies. It is possible that individuals who respond
moderated by age, gender, year of study publication, or to advertisements on SNSs are also high-frequency users,
design characteristics. It should be noted that only one study thus resulting in a restricted range on the variable of time
assessed problematic SNS use in adolescents. As a result, spent using SNSs. Further, to the extent that time spent
age was examined as a continuous moderator in what were using SNSs is correlated with problematic use, it is possi-
primarily adult samples. Therefore, our review highlights ble that samples recruited solely from SNS advertisements
the need for more research in adolescents examining the may include a disproportionate number of problematic SNS
specific construct of problematic SNS use. Nevertheless, the users, which could spuriously inflate the effect sizes for time
moderation findings indicate that the relation of problematic spent using SNSs. Therefore, the current results highlight a
SNS use to depression is no stronger in younger adults (e.g., need for future studies to ensure that samples are representa-
college-age, who are the next highest users after adoles- tive of the population to which they would like to general-
cents) than older adults. Further, problematic SNS use was ize, and that studies examine the intersection of total time,
not significantly moderated by gender. These null findings intensity, and problematic SNS use.
are important as they suggest that despite different use pat- Finally, there was a trend for the pooled effect sizes of the
terns and rates of depression across ages and genders, if use relation of time spent using SNSs and depression symptoms
rises to the level of being problematic, its association with to be significantly larger in studies of adults than in stud-
depression is consistent. An intriguing and important area ies of children or adolescents. It should be noted, however,
for future research in this area is to examine in a more fine- that both effect sizes were small (r = 0.11, r = 0.02, respec-
grained manner the specific ways in which younger versus tively), thus is it unlikely that this difference is clinically
older individuals’ or men’s versus women’s problematic use meaningful. Further, the small number of studies examining
of SNSs differ (e.g., addictive on-line gaming versus addic- the younger age group (k = 6) collapsed across studies of
tive scrolling through posts promoting social comparison) children and adolescents. Therefore, this highlights again
in order to determine whether the content of intervention the importance of more research on the relation of SNS use
and prevention strategies should differ across these impor- and depression in adolescents specifically.
tant demographic groups. Finally, although the prevalence The results of this meta-analysis should be interpreted
and modes of SNS use (i.e., increased smartphone use) have in light of the following limitations. First, although there
undergone drastic changes over the last several years, the were 62 studies included overall, there were smaller sam-
year of study publication was not a significant moderator ple sizes for each of the three SNS use groups, particularly
in the relation between depression symptoms and SNS use. the intensity of SNS use group (k = 6). This small sample
These null findings may suggest that changes in SNS use size prevented us from examining any of the moderators
over time are not a relevant factor in the potential association within the intensity of SNS use group, and in the other two
between SNS use and depression; however, year of publi- groups of studies we were not able to examine interactions
cation is ultimately a proxy for changes in the way SNSs among our moderator variables. Further, there are a number
are used over time. Future research that more specifically of potentially important mechanisms may moderate the rela-
examines these changes and their association with depres- tion of SNS use and depression that we could not examine
sion symptoms warrants further attention. Overall, however, due to an insufficient number of studies. For example, some
our results suggest that problematic SNS use may have a studies have found that passive social media use (e.g., scroll-
more stable and consistent relation to depression symptoms ing through other individual’s posts) is more strongly associ-
than time spent using SNSs or intensity of SNS use, one that ated with depression symptoms than active social media use
is not altered by, or dependent upon, other factors. (e.g., posting photos or status updates, commenting on other
In contrast, the relation between depression symptoms posts, etc.; e.g., Escobar-Viera et al. 2018). Other research
and time spent using SNSs was significantly moderated has found that time of day may be relevant, with nighttime/
by mode of recruitment. Specifically, studies that exclu- in-bed SNS use associated with depression symptoms more
sively recruited participants through SNSs had a signifi- strongly than daytime use (Becker and Lienesch 2018;
cantly larger pooled effect size than studies that relied on Woods and Scott 2016). These potential moderators may
other recruitment methods. Indeed, the effect size for the also be of relevance to the statistically significant hetero-
relation of time spent using SNSs to depression in studies geneity (i.e., significant I2 values) across studies examining

13
Research on Child and Adolescent Psychopathology (2021) 49:241–253 251

all three constructs of SNS use. As more research exam- use and depression symptoms was moderate in size, and was
ining these specific factors of SNS use becomes available, not significantly moderated by age, gender, or design char-
meta-analytic research will be useful in further specifying acteristics, suggesting that this relation may be more stable
the nature of the relation between depression and SNS use. and universal. An important implication of these results is
Second, the current meta-analyses only included studies that future work with the goal of establishing the directional-
that reported bivariate correlational statistics for the con- ity of the association between depression and SNS use, and
current relation between depression symptoms and SNS mediators and moderators of this association, may want to
use. The purpose of this inclusion criterion was to create focus in more detail on problematic SNS use. Further, we
more homogeneity within our groups of studies, as we rea- suggest that prevention and intervention strategies to lessen
soned that including studies that dichotomized continuous the functional impact of SNS use may be most cost-effective
variables (e.g., created depression groups based on symptom if they target individuals who are engaging in problematic
scale scores) would alter the nature of the research question. or addictive use.
However, our inclusion criterion resulted in the exclusion
of several studies relevant to the topic of the present meta-
analysis. Assessing the relation of a categorical depression
variable and SNS use thus represents an important area of References
investigation.
Finally, as with all meta-analyses, the results of the cur- Andreassen, C. S., Pallesen, S., & Griffiths, M. D. (2017). The rela-
tionship between addictive use of social media, narcissism, and
rent study were limited by what was available in the existing self-esteem: Findings from a large national survey. Addictive
published literature. In this meta-analysis we detected evi- Behaviours, 64, 287–293.
dence for a “file drawer” effect for studies that assessed time Andreassen, C.S., Brunborg, G. S., Pallesen, S., & Torsheim, T. (2012).
spent using SNSs, such that the published literature may Development of a Facebook Addiction Scale. Psychological
Reports, 110(2), 501-517.
overestimate the strength of the relation between depres- Avery, P. M. (2018). The interactive effects of social support, envy, and
sion symptoms and SNS use. However, we were only able supportive interactions on the relationship between Facebook use
to include published studies in the meta-analysis (includ- and depression among emerging adults (Doctoral Dissertation).
ing theses published on ProQuest), and did not include any Retrieved from https:​ //search​ .proque​ st.com/docvie​ w/213244​ 3358?​
accou​ntid=6180
unpublished data or data from the grey literature (e.g., con- Baker, D. A., & Algorta, G. P. (2016). The relationship between online
ference presentations, theses not available on ProQuest). social networking and depression: A systematic review of quanti-
Therefore, the current findings of publication bias may be tative studies. Cyberpsychology, Behaviour, and Social Network-
an underestimate of the true publication bias present in this ing, 19(11), 638–648.
Baker, Z. G., Krieger, H., & LeRoy, A. S. (2016). Fear of missing
field. Further, and importantly, of the studies included in this out: Relationships with depression, mindfulness, and physical
meta-analysis did not include information on participants’ symptoms. Translational Issues in Psychological Science, 2(3),
ethnicity or cultural background. As a result, we were not 275–282.
able to examine the potential moderating role of ethnicity Bányai, F., Zsila, A., Király, O., Maraz, A., Elekes, Z., … Demetrovics,
Z. (2017). Problematic social media use: Results from a nationally
on the relation between depression symptoms and SNS use. representative adolescent sample. PLoS ONE, 12(1).
Further, most studies did not clarify whether they assessed Barry, C. T., Sidoti, C. L., Briggs, S. M., Reiter, S. R., & Lindsey, R.
gender or biological sex, and few studies reported the inclu- A. (2017). Adolescent social media use and mental health from
sion of participants who identified as anything other than adolescent and parent perspectives. Journal of Adolescence, 61,
1–11. https​://doi.org/10.1016/j.adole​scenc​e.2017.08.005
male or female. Future research that investigates sex, gen- Becker, S. P., & Lienesch, J. A. (2018). Nighttime media use in ado-
der, ethnicity, and their intersection are necessary to further lescents with ADHD: Links to sleep problems and internalizing
clarify the relation between SNS use and depression. symptoms. Sleep Medicine, 51, 171–178.
The results of the current meta-analysis indicate a sig- Bilevicius, E., Single, A., Bristow, L. A., Foot, M., Ellery, M., Keough,
M. T., & Johnson, E. A. (2018). Shame mediates the relationship
nificant association of depression symptoms and all three between depression and addictive behaviours. Addictive Behav-
constructs of SNS use examined (time spent using SNSs, iors, 82, 94–100.
intensity of SNS use, or problematic SNS use). Nevertheless, Blease, C. (2015). Too many ‘friends,’ too few ‘likes’? Evolutionary
associations with time spent using SNSs and intensity of use psychology and ‘Facebook depression’. Review of General Psy-
chology, 19(1).
were weak, suggesting that they may not be clinical mean- Borenstein, M, Rothstein, D & Cohen, J (2005) Comprehensive meta-
ingful. Further, for time spent using SNSs, this association analysis: A computer program for research synthesis [Computer
rose to moderate size only when studies included partici- software]. Englewood: Biostat.
pants recruited exclusively from SNSs, raising the potential Brooks, S & Longstreet, P (2015). Social networking’s peril: Cognitive
absorption, social networking usage, and depression. Cyberpsy-
of a confounding between time spent using SNSs and prob- chology-Journal of Psychosocial Research on Cyberspace,
lematic use. In contrast, the association of problematic SNS 9(4). https​://doi.org/10.5817/CP201​5-4-5

13
252 Research on Child and Adolescent Psychopathology (2021) 49:241–253

Card, N. A. (2012). Applied meta-analysis for social science research. adolescents and emerging adults: Associations between cyberbul-
Ney York: Guilford. lying perpetration, problematic social media use, and psychosocial
Clement, J. (2020). Global social networks ranked by number of users in factors. International Journal of Mental Health Addiction, 17,
2020. Retrieved from: https:​ //www.statis​ ta.com/statis​ tics/​ 272014​ / 891–908. https​://doi.org/10.1007/s1146​9-018-9894-8
globa​l-socia​l-netwo​rks-ranke​d-by-numbe​r-of-users​/ Kotov, R., Krueger, R. F., Watson, D., Achenbach, T. M., Althoff, R.
Daniels, M. (2014). Facebook and depression in late adolescence: R., Bagby, R. M., … Zimmerman, M. (2017). The Hierarchical
intensity of use, quality of interactions, and the role of self-defini- Taxonomy of Psychopathology (HiTOP): A dimensional alterna-
tion and identity (Doctoral Dissertation). Retrieved from https​:// tive to traditional nosologies. Journal of Abnormal Psychology,
searc​h.proqu​est.com/docvi​ew/17795​43659​?accou​ntid=6180 126, 454–477. https​://doi.org/10.1037/abn00​00258​
Davila, J., Hershenberg, R., Feinstein, B. A., Gorman, K., Bhatia, V., Krasnova, H., Veltri, N., Eling, N., & Buxmann, P. (2017). Why men
& Starr, L. R. (2012). Psychology of Popular Media Culture, 1(2), and women continue to use social networking sites: The role of
72-86. gender differences. Journal of Strategic Information Systems, 26,
Duval, S. (2005). The trim and fill method. In H. R. Rothstein, A. 261–284.
J. Sutton, & M. Borenstein (Eds.), Publication bias in meta- Lin, L. L., Sidani, J. E., Shensa, A., Radovic, A., Miller, E., … Pri-
analysis: prevention, assessment and adjustments (pp. 127–144). mack, B. A. (2016). Association between social media use and
Hoboken: Wiley. depression among U.S. young adults. Depression and Anxiety,
Duval, S., & Tweedie, R. (2000). Trim and fill: A simple funnel-plot- 33, 323–331.
based method of testing and adjusting for publication bias in meta- Myhre, J. W., Mehl, M. R., & Glisky, E. L. (2017). Cognitive benefits
analysis. Biometrics, 56(2), 455–463. of online social networking for healthy older adults. The Journal
Edgerton, J. D., Keough, M. T., & Roberts, L. W. (2018). Co-develop- of Gerontology, Series B: Psychological Sciences and Social Sci-
ment of problem gambling and depression symptoms in emerging ences, 72, 752–760.
adults: A parallel-process latent class growth model. Journal of Park, N., Kee, K. F., & Valenzuela, S. (2009). Is there social capital in
Gambling Studies, 34, 949–968. a social network site?: Facebook use and college students’ life sat-
Egger, M., Smith, G. D., Schneider, M., & Minder, C. (1997). Bias in isfaction, trust, and participation. Journal of Computer-Mediated
meta-analysis detected by a simple, graphical test. British Medical Communication, 14(4), 875–901.
Journal, 315, 629–634. Patel, R., Chang, T., Greysen, S. R., & Chopra, V. (2015). Social media
Ellison, N. B., Steinfield, C., & Lampe, C. (2007). Connection strate- use in chronic disease: A systematic review and novel taxonomy.
gies: Social capital implications of Facebook-enabled communica- The American Journal of Medicine, 128, 1335–1350. https​://doi.
tion practices. New Media and Society, 13(6), 873–892. org/10.1016/j.amjme​d.2015.06.015
Elphinston, R., & Noller, P. (2011). Time to face it! Facebook intru- Perrin, A. (2015). Social media usage: 2005–2015. Washington, D.C.:
sion and the implications for romantic jealousy and relationship Pew Internet & American Life Project.
satisfaction. Cyberpsychology, Behavior, and Social Networking, Primack, B. A., & Escobar-Viera, C. G. (2017). Social media as it
14(11), 631–635. interfaces with psychosocial development and mental illness in
Escobar-Viera, C. G., Shensa, A., Bowman, N., Sidani, J. E., Knight, J., transitional age youth. Child and Adolescent Psychiatric Clinics
… Primack, B. A. (2018). Passive and active social media use and of North America, 26, 217–233.
depressive symptoms among United States adults. Cyberpsychol- Saiphoo, A. N., DahoahHalevi, L., & Vahedi, Z. (2019). Social net-
ogy, Behaviour, and Social Networking, 21(7), 437–443. working site use and self-esteem: A meta-analytic review. Person-
Frost, R. L., & Rickwood, D. J. (2017). A systematic review of the ality and Individual Differences, 153.
mental health outcomes associated with Facebook use. Computers Salk, R. H., Hyde, J. S., & Abramson, L. Y. (2017). Gender differences
and Human Behaviour, 76, 576–600. in depression in representative national samples: Meta-analysis
Heffer, T., Good, M., Daly, O., MacDonell, E., & Willoughby, T. of diagnoses and symptoms. Psychological Bulletin, 143(8),
(2019). The longitudinal association between social-media use 783–822.
and depressive symptoms among adolescents and young adults: Seabrook, E. M., Kern, M. L., & Rickard, N. S. (2016). Social net-
An empirical reply to Twenge et al. (2018). Clinical Psychological working sites, depression, and anxiety: A systematic review. JMIR
Science, 7(3), 462–470. Mental Health, 3.
Huang, C. (2017). Time spent on social network sites and psychologi- Shankman, S. A., Lewinsohn, P. M., Klein, D. N., Small, J. W., Seeley,
cal wellbeing: A meta-analysis. Cyberpsychology, Behaviour, and J. R., & Altman, S. E. (2009). Subthreshold conditions as precur-
Social Networking, 20(6), 346–354. sors for full syndrome disorders: A 15-year longitudinal study
Instagram: About us (2019). Retrieved June 20, 2019 from Instagram: of multiple diagnostic classes. Journal of Child Psychology and
https​://www.insta​gram.com/about​/us/ Psychiatry, and Allied Disciplines, 50, 1485–1494. https:​ //doi.org/
Jung, S. Y., & Kim, J. N. (2014). Development and validation of SNS 10.1111/j.1469-7610.2009.02117​.x
addiction proneness scale for college students. The Korean Jour- Shapiro, L. A., & Margolin, G. (2014). Growing up wired: Social net-
nal of Health Psychology, 19(1), 147–166. working sites and adolescent psychosocial development. Clinical
Kaplan, A. M., & Haenlein, M. (2010). Users of the world, unite! The Child and Family Psychology Review, 17, 1–18.
challenges and opportunities of social media. Business Horizons, Smith, A., & Anderson, M. (2018). Social Media Use in
53, 59–68. 2018. Retrieved from: https ​ : //www.pewre ​ s earc ​ h .org/
Karsten, J., Hartman, C. A., Ormel, J., Nolen, W. A., & Penninx, B. W. inter​net/2018/03/01/socia​l-media​-use-in-2018/
J. H. (2010). Subthreshold depression based on functional impair- Snap Inc. (2019). Retrieved June 20, 2019 from Snapchat: https:​ //www.
ment better defined by symptom severity than by number of DSM- snapc​hat.com/en-US/
IV symptoms. Journal of Affective Disorders, 123, 230–237. https​:// Swendsen, J. D., & Merikangas, K. R. (2000). The comorbidity of
doi.org/10.1016/j.jad.2009.10.013 depression and substance use disorders. Clinical Psychology
Kendler, K. S., & Gardner, C. O. (1998). Boundaries of major depres- Review, 20, 173–189.
sion: An evaluation of DSM-IV criteria. American Journal of Tandoc, E. C., Ferrucci, P., & Duffy, M. (2015). Facebook use, envy,
Psychiatry, 155, 172–177. and depression among college students: Is Facebook depressing?
Kırcaburun, K., Kokkinos, C. M., Demetrovics, Z., Király, O., Griffiths, Computers in Human Behaviour, 43, 139–146.
M. D., & Çolak, T. S. (2019). Problematic online behaviors among

13
Research on Child and Adolescent Psychopathology (2021) 49:241–253 253

Twenge, J. M. (2015). Time period and birth cohort differences in Woods, H. C., & Scott, H. (2016). #Sleepyteens: Social media use in
depressive symptoms in the U.S., 1982-2013. Social Indicators adolescence is associated with poor sleep quality, anxiety, depres-
Research, 121(12), 437–454. sion, and low self-esteem. Journal of Adolescence, 51, 41–49.
Twenge, J. M., Joiner, T. E., Rogers, M. L., & Martin, G. N. (2018). World Health Organization. (2018). Depression. Retrieved from https:​ //
Increases in depressive symptoms, suicide-related outcomes, www.who.int/menta​l_healt​h/manag​ement​/depre​ssion​/defin​ition​/
and suicide rates among U.S. adolescents after 2010 and links to en/
increased new media screen time. Clinical Psychological Science, Yoon, S., Kleinman, M., Mertz, J., & Brannick, M. (2019). Is social
6(1), 3–17. https​://doi.org/10.1177/21677​02617​72337​6 network site usage related to depression? A meta-analysis of
Vannucci, A., Flannery, K. M., & McCauley Ohannessian, C. (2017). Facebook-depression relations. Journal of Affective Disorders,
Social media use and anxiety in emerging adults. Journal of Affec- 248, 65–72.
tive Disorders, 207, 163–166.
Vernon, L., Modecki, K. L., & Barber, B. L. (2016). Tracking effects Publisher’s Note Springer Nature remains neutral with regard to
of problematic social networking on adolescent psychopathology: jurisdictional claims in published maps and institutional affiliations.
The mediating role of sleep disruptions. Journal of Clinical Child
and Adolescent Psychology, 46(2), 1-15.
Weissman, M. M., Pilowsky, D. J., Wickramaratne, P. J., Talati, A.,
Wisniewski, S. R., … Rush, A. J. (2006). Remissions in maternal
depression and child psychopathology: A Star*D-child report.
Journal of the American Medical Association, 295, 1389–1398.

13

You might also like