This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR Formative Research, is properly cited. The complete bibliographic information, a link to the original publication on https://formative.jmir.org, as well as this copyright and license information must be included.
Increasing youth mental health problems over time correlate with increasing rates of social media use (SMU); however, a proposed contributory relationship remains unproven. To better understand how SMU impacts mental health requires a more nuanced understanding of the relationship between different patterns of SMU and specific individual factors. Studies suggest that more
Using ecological momentary assessment (EMA), this study aims to investigate potential differences in affective states during
We recruited patients seeking care at a large urban adolescent medicine clinic who exhibited at least mild depressive symptoms based on Patient Health Questionnaire-9 (PHQ-9) scores. Participants completed an enrollment survey and a 7-day EMA protocol, receiving 5 EMA questionnaires per day, which assessed real time SMU behaviors and affective states using the Positive and Negative Affect Schedule–Expanded form subscales. To correct for nonindependent data with EMA responses clustered within individuals, data were analyzed using mixed-effects modeling, allowing for a random intercept at the individual level to examine associations between EMA-reported SMU and affective states while adjusting results for age, gender, race and ethnicity, PHQ-9 score, and EMA response rate.
A racially and ethnically diverse group of 55 adolescents aged 14 to 19 years provided a total of 976 EMA responses, averaging 17.76 (SD 8.76) responses per participant, with a response rate of 51.15%. Participants reported higher mean levels of negative affect during
Although in aggregate, adolescents with depressive symptoms experienced more negative affect during
Social media use (SMU), the use of “internet-based networks that enable users to interact with others, verbally and visually,” [
Although some evidence suggests that children and adolescents with high-frequency SMU experience increased rates of mental health problems and greater depressive symptoms, the overall data are inconclusive, with some studies demonstrating no difference or even suggesting minor psychological benefits [
Variations in the relationships between SMU and mental health outcomes may have resulted from how SMU was defined and measured, what mental health symptoms were considered, and the potential influence of confounders. When SMU is considered as a binary or duration-of-use variable, without discriminating among user characteristics or their specific SMU behaviors, a more granular and useful understanding of how SMU can affect mental health may be lost.
Most SMU studies thus far have relied on retrospective self-reports, which may experience recall bias. Studies that incorporate EMA, which allows for
Some recent studies have taken a more nuanced look at different patterns of SMU and their relationship with specific mental health outcomes [
Regarding individual patterns of use, SMU that is
On the other hand,
At present, it remains unclear how the characteristics of individuals using social media may affect their frequency or type of SMU or lead to different mental health outcomes. Although there are reasons to hypothesize a contributory link between certain types of SMU and declining mental health (eg, SMU that displaces face-to-face social interactions and leads to increased social isolation and loneliness), underlying mental or behavioral health issues may actually drive high-frequency SMU. In fact, young people with certain behavioral health conditions—especially attention-deficit/hyperactivity disorder, social anxiety disorder, autism spectrum disorder, and depression—use screen media more and are predisposed to
Multiple studies have suggested that race/ethnicity may be an important moderator of the impact of SMU on mental health. Different experiences using social media networks may confer different levels of risk by race/ethnicity for mental health outcomes. Although adolescents of different racial/ethnic groups do not differ in their total number of
Conversely, minority groups face potential exposure to racist and/or discriminatory content during SMU. Among Black young adults, SMU on certain platforms is associated with increased anticipatory race-related stress, bodily alarm responses, and anger expression, in part mediated by individual experiences of perceived racism and everyday discrimination [
This study has 3 aims. The first is to investigate, within a sample of youths with depression, momentary associations between SMU and three types of affective states: positive affect, negative affect, and sadness. The second aim is to investigate whether adolescents’
This study used EMA survey data from a clinical sample of patients aged 14 to 19 years recruited from a large urban adolescent medicine clinic during the period from August 2016 to March 2018. Patients seeking well care were asked to complete the Patient Health Questionnaire-9 (PHQ-9), and those with a score of ≥5, indicating at least
The institutional review board of Boston Children’s Hospital approved this study (Institutional Review Board protocol number IRB-P00019244), and participants were assured of their confidentiality. All participants or parents of minor participants provided written informed consent. Once enrolled, participants completed surveys at the time of enrollment and EMA questionnaires multiple times a day for the next 7 days.
After completing the PHQ-9 as a screener for inclusion, qualifying participants completed an enrollment survey that assessed key demographic information. Consistent with the most recent US national census questionnaire, race/ethnicity/ancestry (hereafter referred to as
For the week following enrollment and based on standard EMA protocol, participants received 5 daily EMA assessments administered at random intervals sent to their personal smartphones (or 1 provided to them) using MetricWire [
SMU behaviors reported in EMA surveys were characterized with respect to frequency and type of use. For the purpose of this study, SMU was defined as any behavior for which the participant indicated using a web-based platform that allows for interaction and communication. This included behaviors such as
Each EMA survey assessed emotional affect using subscales of the Positive and Negative Affect Schedule–Expanded form, which measured participants’ positive affect, negative affect, and sadness [
We used the EMA responses to compare reported levels of positive affect, negative affect, and sadness during SMU moments and non-SMU moments. To correct for nonindependence of the data with EMA responses clustered within an individual, we used mixed-effects modeling, allowing for a random intercept at the individual level to examine associations between SMU and EMA-reported affective scores while adjusting the results for age, gender, race/ethnicity, PHQ-9 score, and EMA response rate. We used similar procedures to examine differences in affect during
To investigate the differences in the associations between SMU and affect by race/ethnicity, the abovementioned mixed-effects models
The sample was mostly female and comprised similar percentages of Black non-Hispanic, White non-Hispanic, and Hispanic participants (
Sample characteristics.
Characteristics | Values | ||||
|
|||||
|
Age (years), mean (SD; range) | 17.42 (1.50; 14-19) | |||
|
Gender (female), n (%) | 37 (67) | |||
|
|
||||
|
|
Black non-Hispanic | 16 (29) | ||
|
|
White non-Hispanic | 15 (27) | ||
|
|
Hispanic | 15 (27) | ||
|
|
Mixed race or other | 8 (15) | ||
|
PHQ-9a score, mean (SD) | 11.27 (5.26) | |||
|
|||||
|
|
||||
|
|
Active SMU | 95 (9.7) | ||
|
|
Passive SMU | 41 (4.2) | ||
|
Response rate (%) | 51.15 | |||
|
EMA responses per participant, mean (SD) | 17.76 (8.76) |
aPHQ-9: Patient Health Questionnaire-9.
bEMA: ecological momentary assessment.
cSMU: social media use.
When comparing moments with
Effects of SMUa type on mean affect scoresb.
SMU type and affect type | Coefficient | SE | Significance, |
95% CI | |||||||
|
|||||||||||
|
Negative affect | 0.08 | 0.04 | 2.30 (131) | .02 | 0.01 to 0.16 | |||||
|
Positive affect | –0.10 | 0.10 | –0.94 (55) | .35 | –0.30 to 0.11 | |||||
|
Sadness | 0.16 | 0.08 | 2.14 (39) | .04 | 0.01 to 0.32 | |||||
|
|||||||||||
|
Negative affect | 0.09 | 0.05 | 1.96 (215) | .05 | 0.00 to 0.18 | |||||
|
Positive affect | –0.04 | 0.13 | –0.30 (40) | .77 | –0.31 to 0.23 | |||||
|
Sadness | 0.17 | 0.10 | 1.74 (154) | .08 | –0.02 to 0.04 | |||||
|
|||||||||||
|
Negative affect | 0.03 | 0.06 | 0.46 (256) | .64 | –0.09 to 0.14 | |||||
|
Positive affect | –0.18 | 0.09 | –1.98 (369) | .049 | –0.35 to 0.001 | |||||
|
Sadness | 0.04 | 0.11 | 0.39 (17) | .70 | –0.20 to 0.28 |
aSMU: social media use.
bThe above results are from mixed-effects modeling with adjustments for age, sex, race/ethnicity, Patient Health Questionnaire-9 score, and ecological momentary assessment response rate.
c2-tailed.
Using generalized linear modeling with the addition of a race×SMU type interaction term, there were some observed differences in reported affective states by race/ethnicity during moments of
During the moments of
Group differences in mean affect during SMUa versus all other momentsb.
Affect and race/ethnicitya | Coefficient | Mean difference | SE | Significance, |
95% CI | ||||||||
|
|||||||||||||
|
Race and ethnicity×active SMU (1) | 0.125 | N/Ac | 0.07 | 1.89 (132) | .06 | –0.01 to 0.26 | ||||||
|
Race/ethnicity×active SMU (2) | 0.012 | N/A | 0.11 | 0.12 (132) | .91 | –0.20 to 0.22 | ||||||
|
Black non-Hispanic | N/A | 0.13 | 0.05 | 2.51 (81) | .01 | 0.03 to 0.24 | ||||||
|
White non-Hispanic | N/A | 0.01 | 0.04 | 0.18 (118) | .86 | –0.08 to 0.09 | ||||||
|
Hispanic | N/A | 0.12 | 0.09 | 1.33 (144) | .19 | –0.06 to 0.30 | ||||||
|
|||||||||||||
|
Race/ethnicity×passive SMU (1) | 0.405 | N/A | 0.18 | 2.24 (677) | .03 | 0.05 to 0.76 | ||||||
|
Race/ethnicity×passive SMU (2) | 0.047 | N/A | 0.24 | 0.20 (677) | .84 | –0.42 to 0.52 | ||||||
|
Black non-Hispanic | N/A | –0.03 | 0.12 | –0.22 (586) | .82 | –0.26 to 0.21 | ||||||
|
White non-Hispanic | N/A | –0.43 | 0.13 | –3.24 (295) | .001 | –0.69 to –0.17 | ||||||
|
Hispanic | N/A | –0.07 | 0.20 | –0.36 (868) | .72 | –0.48 to 0.33 |
aSMU: social media use.
bThe abovementioned results, including the mean differences, are adjusted for age, gender, race/ethnicity, Patient Health Questionnaire-9 score, and ecological momentary assessment response rate. Mean differences represent average affect during SMU type (
cN/A: not applicable.
In summary, during moments of
This study attempts to enrich our understanding of the complex interrelations between SMU and affective states. Previous research has revealed that the association between SMU and indicators of mental health varies considerably among individual youths [
Differences in media and SMU by ethnicities were well-documented during the emergence of mobile and social media, as minority youths (ie, Black, Hispanic, and Asian) were shown to consume 4.5 hours a day more media than their White peers and were more likely to be early users of social networking sites [
Our findings regarding differences in affect during
However, considering that our participants were determined to be at least mildly depressed before the study, this result may be better understood as indicating that young people who are depressed are more likely to engage in
For Black non-Hispanic participants in our study, interacting directly with other people while using social media (
Such a response by Black non-Hispanic adolescents may be explained by the dual nature of SMU for minority youths. The risk of experiencing racism-based cyberaggression, blatant acts of racism, microaggressions, and other forms of systemic oppression on social media, which can be further amplified by platform-based algorithms, may lead to different SMU experiences for minority-identifying individuals than for their White peers [
Even in web-based community building and activism, the positive side of the dual nature of SMU for minorities may hold experiences that could be more upsetting for Black non-Hispanic youths as they can include being exposed to, sharing experiences about, and building solidarity and organizing against racism and inequality. Previous research has demonstrated that Black and Hispanic participants report higher rates of
Again, it is important to remember that our study did not test the direction of effect. Black non-Hispanic youths may turn to
On the basis of these findings, researchers should investigate the possibility that SMU is not experienced uniformly. One’s racial/ethnic identity may be an important mediator of affective experiences during SMU. Although depression and anxiety have often been a major focus of research on the effects of SMU, there may be value in studying affect in greater detail, given the relationship between lower positive affect and depression and between higher negative affect driven by irritability/hostility and activist solidarity.
From a clinical standpoint, as associations between SMU and behavioral health conditions become more clear, clinicians will need to become comfortable with assessing and providing counseling on
If it is found that the design of social media platforms may contribute to inequitable SMU experiences or unhealthy behavioral outcomes, technology companies will have the opportunity to respond with design changes that promote greater equity and more positive mental health outcomes. In 2019, the social media platform Instagram started testing the potential value of hiding
The study’s limitations include the relatively small sample size, which reduces the generalizability of the results. However, the sample included a diverse group of participants drawn from a population of patients with evidence of at least mild depression, which is an at-risk group of great clinical interest. In addition, our analyses were performed at the moment level, thereby increasing the statistical power.
Although the overall EMA survey response rate (approximately 50%) was similar to that of other studies using this methodology with adolescents, it does represent a potential selection bias. Participants with higher response rates could provide qualitatively different responses than those with lower response rates. For instance, participants with higher levels of social anxiety exhibited higher EMA response rates. In addition, adolescents may have avoided answering surveys during SMU, especially when they were actively engaged in it. However, EMA measures essentially eliminate recall bias, which likely greatly enhances the reliability of patient reports.
Another limitation of this study was the inability to prove that SMU contributes to mental health problems. Given the dynamic nature of social media, our results may be sensitive to the specific period during which the study was performed. Study participants’ SMU characteristics and affective experiences may be highly dependent on news stories that happened to be trending at a given point, creating the potential for chronology bias.
Moving forward, it will be important to measure momentary affective responses to SMU in more detail, with larger populations, with different social media designs, and over longer periods to achieve greater granularity and demonstrate the stability of these findings. Given the significant differences in affective experiences among racial/ethnic groups, future research should broaden to include gender, culture, and other experience-influencing user characteristics.
At a time marked by increasing trends in mental health problems and unprecedented levels of interactive media use among children and adolescents, there is a real urgency to achieve a better understanding of the relationships between different media use behaviors and mental health problems. Identifying helpful and harmful digital design features and risk and protective factors in users will be of paramount importance, as will be the continued study of the role of racial/ethnic identities and other individual factors as potential moderators of media use experiences. As the nature of our social interactions continues to evolve in an increasingly digital landscape, so too must our understanding of the potential influence of these behaviors on the health and well-being of the children, adolescents, and young adults we serve.
ecological momentary assessment
Patient Health Questionnaire-9
social media use
The authors would like to thank Jill R Kavanaugh for her bibliographic assistance with the study.
This work was supported by the Digital Wellness Lab and the Division of Adolescent Medicine, Boston Children’s Hospital, Boston, Massachusetts, United States, with funding from the Leadership Education in Adolescent Health (grant T71 MC 00009) from the Maternal and Child Health Bureau, Health Resources and Services Administration, and the Boston Children’s Hospital Office of Faculty Development Award to DSB, PhD.
None declared.