Abstract
Introduction:
Previous studies have demonstrated cross-sectional associations between social media use and depression, but their temporal and directional associations have not been reported.
Methods:
In 2018, participants aged 18–30 years were recruited in proportion to U.S. Census characteristics, including age, sex, race, education, household income, and geographic region. Participants self-reported social media use on the basis of a list of the top 10 social media networks, which represent >95% of social media use. Depression was assessed using the 9-Item Patient Health Questionnaire. A total of 9 relevant sociodemographic covariates were assessed. All measures were assessed at both baseline and 6-month follow-up.
Results:
Among 990 participants who were not depressed at baseline, 95 (9.6%) developed depression by follow-up. In multivariable analyses conducted in 2020 that controlled for all covariates and included survey weights, there was a significant linear association (p<0.001) between baseline social media use and the development of depression for each level of social media use. Compared with those in the lowest quartile, participants in the highest quartile of baseline social media use had significantly increased odds of developing depression (AOR=2.77, 95% CI=1.38, 5.56). However, there was no association between the presence of baseline depression and increasing social media use at follow-up (OR=1.04, 95% CI=0.78, 1.38). Results were robust to all sensitivity analyses.
Conclusions:
In a national sample of young adults, baseline social media use was independently associated with the development of depression by follow-up, but baseline depression was not associated with an increase in social media use at follow-up. This pattern suggests temporal associations between social media use and depression, an important criterion for causality.
INTRODUCTION
Depression is highly prevalent in the U.S., and its incidence is increasing.1,2 Depression accounts for more disability-adjusted life years than all other mental disorders3 and was recently declared to be the leading global cause of disability by the WHO.4 In the U.S. alone, the economic burden of depression exceeds $200 billion annually from reduced worker productivity, increased medical expenses, and suicide.5
Although multiple factors contribute to depression,1,6 there is growing interest in the association between social media use (SMU) and psychological well-being.7–10 Defined as the consumption of “a group of Internet-based applications that allow the creation and exchange of user-generated content,”11 SMU involving sites such as Facebook, Instagram, Twitter, and Reddit has become an integral method by which individuals connect with others, share personal content, and obtain news and entertainment.12 SMU has increased dramatically among young adults who are at critical junctures around identity and brain development.12–15 More than 90% of U.S. young adults use social media, on average from 2 to 4 hours per day.12,14
Epidemiologic studies have demonstrated significant associations between SMU and depression.7,10,16,17 However, the cross-sectional design of these studies precluded the understanding of the temporal associations between SMU and depression.8,12 SMU could represent a precursor to depression through the displacement of in-person activities17 or by increasing continuous exposure to highly curated, idealized portrayals of life.8,18 Alternatively, the development of depression could lead individuals to shun in-person activities owing to anhedonia, thereby increasing SMU over time.
Therefore, a longitudinal study was conducted of a national sample of U.S. adults aged 18–30 to assess the independent, temporal associations between SMU and depression, controlling for baseline differences in sociodemographic characteristics.
METHODS
Study Sample
Participants were recruited online by Qualtrics Sampling Services, which specializes in web-based survey data collection that partners with >20 web-based panel providers to identify diverse respondents.19 The panel providers used a balanced start sampling methodology, in which potential participants were recruited in proportion to U.S. Census characteristics, including age, sex, race, education, household income, and geographic region. Participants were required to be aged 18–30 years, able to read English, and able to use an online interface to record their responses.
With the assistance of Qualtrics, 3 strategies were used to assure high-quality data. First, pilot tests were conducted with 30 individuals who were not a part of the final sample to assess survey processes and data quality. Second, a soft launch was performed on 240 participants (10% of the study baseline sample) before full implementation. Finally, standard quality checks were employed, such as screening for participants who provided the same response for all items (straightliners) and participants who completed it in less than one third of the median completion time (speeders). Of the 2,502 completers, 94 (3.8%) were eliminated for failing these quality checks.
The median time for survey completion was 18 minutes. Survey respondents received point-based credits from the panel provider that could be redeemed for incentives such as gift cards or charitable donations. All participants provided informed consent, and the study was approved by the University of Pittsburgh IRB.
Measures
Baseline data on participant sociodemographic and personal characteristics, depression, and SMU were collected in March 2018, with data on the 2 latter data elements reassessed approximately 6 months later. The exact mean and SD for follow-up were 170 days (5.7 months) and 51 days (1.7 months).
At baseline, self-reported characteristics previously associated with depression and SMU were captured, including age, sex, race and ethnicity, educational level, household income, relationship status, living situation, and Adverse Childhood Experiences (ACEs).
Age was classified as a continuous variable in years, and birth sex was classified as male or female. Standard categories for race and ethnicity were provided by Qualtrics. Relationship status was divided into 3 categories (single, member of an unmarried couple, married), and each of the following was divided into 4 categories: educational level (high school or less, some college or technical school, college graduate, graduate school), household income (< $25,000, $25,000–$49,999, $50,000–$74,999, ≥$75,000), and living situation (alone, with parent or guardian, with significant other, other). ACEs were assessed using an adapted 6-item questionnaire that asks about experiences before age 18 years, such as Did you live with anyone who was depressed, mentally ill, or suicidal? For all the items, response options were yes (1) or no (0); all responses were summed creating a scale from 0 to 6.20
At baseline and follow-up, depression was measured using the 9-Item Patient Health Questionnaire,21 which asks about the frequency of depression-related feelings and experiences over the past 2 weeks, such as having little interest or pleasure doing things or feeling down, depressed, or hopeless. For all the items, response options were not at all (0), several days (1), more than half of the days (2), or nearly every day (3). All responses were summed, creating a scale ranging from 0 to 27. A previously validated cut point of 10 was used to categorize respondents as nondepressed (0–9) or depressed (10–27).21
Respondents’ SMU was also assessed at baseline and follow-up. Participants were first given a list of the top 10 social media networks, which accounted for >95% of all social media activity at the time of the survey. These platforms included Facebook, Twitter, Reddit, YouTube, LinkedIn, Instagram, Tumblr, Snapchat, Pinterest, and WhatsApp. Then, participants were asked to provide the average total daily time spent on social media each day in hours and minutes. Participants were instructed to include the time only spent on social media for personal and not for occupational reasons. For the purposes of analysis, SMU was divided into quartiles using the xtile command in Stata.
Statistical Analysis
Because the first major goal was to assess the association between baseline SMU and the development of depression during follow-up, analyses were restricted to nondepressed individuals at baseline. Associations were determined between the primary independent variable (baseline SMU) and the development of depression using bivariable and multivariable regression analyses. The primary model included all sociodemographic variables.
The second goal was to examine associations in the opposite direction—between baseline depression and subsequent changes in SMU. Therefore, for this set of analyses, primary models used baseline depression (yes versus no) as the predictor variable and increase in SMU (yes versus no) as the outcome variable. These analyses included all participants, not only those depressed at baseline. Again, the primary multivariable model included all measured sociodemographic variables.
All primary regression models incorporated design-specific survey weights provided by Qualtrics to estimate effects for the general U.S. population aged 18–30 years. These weights also adjusted for undersampling or oversampling and nonresponse in terms of key demographic factors on the basis of the most recent U.S. Census data.
Several methods were used to ensure quality of analyses. For all multivariable models, variance inflation factors were used to screen for multicollinearity. Pearson residual, deviance residual, and Pregibon leverage plots were used to search for outliers and cases of undue influence. Interaction effects were tested between our primary independent variables and each of the sociodemographic covariates.
All statistical analyses were performed using Stata, version 15.0. For all analyses, statistical significance was defined using a 2-tailed p-value of <0.05.
There was no evidence of multicollinearity in the primary multivariable models with average variance inflation factors of 1.19 and 1.21, respectively. Visual inspection of residual and leverage plots revealed 3 potential outliers or influential cases in the first model assessing the temporal association between SMU and incident depression. Removal of these cases did not affect the significance or magnitude of associations in the model; therefore, they were retained. In the second multivariable model, assessing the temporal association between depression and SMU increase, residual and leverage plots did not indicate any outliers or particularly influential cases. There were no significant interaction effects between the primary independent variable and each covariate in both multivariable models.
Of the 2,408 respondents who provided baseline data, 1,339 (55.6%) provided follow-up data. There were sociodemographic differences between individuals who did not provide follow-up data and those who did. For example, compared with nonrespondents, respondents were slightly older (26.9 years vs 26.6 years, p=0.04) and more commonly female (57% vs 43%, p<0.001). However, nonresponders and responders did not differ significantly in terms of depression status (p=0.31). From these 1,339 individuals, the 34 individuals (2.5%) who did not provide complete data on primary measures (depression and SMU) were removed. The 16 individuals (1.2%) who were determined to have poor data quality on the basis of unfeasible responses, such as claiming to use social media >18 hours per day, were also removed. Despite the small numbers of individuals removed for missing data or poor quality, data were examined for patterns of missingness, and minor differences were found. Specifically, compared with individuals with missing or poor-quality data, good responders were more commonly married (32.4% vs 16.0%, p=0.04) and living with a significant other (47.1% vs 34.0%, p=0.03), according to chi-square tests. Therefore, the case-wise deletion was used to isolate the final sample of 1,289 individuals.
RESULTS
Of the 1,289 individuals in the final sample, 299 were depressed at baseline. Of the 990 participants who were nondepressed at baseline, 95 (9.6%) met the Patient Health Questionnaire criteria for depression at follow-up. Table 1 compares those who did with those who did not develop depression in terms of baseline SMU and sociodemographic characteristics.
Table 1.
Sample Characteristics and Bivariable Associations With the Development of Depression During the Study Period (n=990)
| Characteristic | % | Developed depression during study perioda | ||
|---|---|---|---|---|
| No,895 (90.4%) | Yes,95 (9.6%) | p-valueb | ||
| SMUc | <0.001 | |||
| Quartile 1 (0–120) | 31.0 | 32.3 | 18.5 | |
| Quartile 2 (121–195) | 29.1 | 29.7 | 22.8 | |
| Quartile 3 (196–300) | 25.3 | 24.5 | 32.6 | |
| Quartile 4 (>301) | 14.7 | 13.5 | 26.1 | |
| Age, years, mean (SD) | 27.0 (2.7) | 27.1 (2.7) | 26.5 (3.0) | 0.06 |
| Sex | 0.53 | |||
| Female | 55.1 | 54.8 | 58.2 | |
| Race | 0.55 | |||
| White, non-Hispanic | 71.0 | 71.8 | 63.3 | |
| Black, non-Hispanic | 6.3 | 6.2 | 7.8 | |
| Hispanic/Latino | 12.1 | 11.9 | 14.4 | |
| Asian | 9.6 | 9.2 | 13.3 | |
| Otherd | 1.0 | 1.0 | 1.1 | |
| Education | 0.08 | |||
| At least some high school | 10.7 | 10.5 | 13.0 | |
| Some college or technical school | 26.0 | 25.1 | 34.8 | |
| College graduate | 35.9 | 36.1 | 33.7 | |
| Graduate school | 27.5 | 28.4 | 18.5 | |
| Annual household income, $ | 0.49 | |||
| <25,000 | 12.3 | 12.0 | 15.2 | |
| 25,000 to <50,000 | 26.7 | 26.6 | 27.2 | |
| 50,000 to <75,000 | 22.7 | 22.4 | 26.1 | |
| ≥75,000 | 38.3 | 39.0 | 31.5 | |
| Relationship status | 0.94 | |||
| Single | 41.2 | 41.3 | 40.2 | |
| Member of unmarried couple | 24.7 | 24.7 | 23.9 | |
| Married | 34.1 | 34.0 | 35.9 | |
| Living situation | 0.40 | |||
| By myself | 17.7 | 18.1 | 14.1 | |
| With parent or guardian | 20.0 | 20.2 | 18.5 | |
| With significant other | 49.0 | 49.1 | 48.9 | |
| With acquaintances, friends, or roommates | 13.3 | 12.7 | 18.5 | |
| Employment | 0.27 | |||
| Regular employment | 79.1 | 79.6 | 73.9 | |
| Homemaker | 5.6 | 5.2 | 8.7 | |
| Student | 10.6 | 10.7 | 9.8 | |
| Unemployed | 4.8 | 4.5 | 7.6 | |
| ACE,e mean (SD) | 1.0 (1.4) | 0.9 (1.3) | 1.3 (1.7) | 0.01 |
Note: Boldface indicates statistical significance (p<0.05).
Depression was measured using the 9-Item Patient Health Questionnaire using the standard cut off of 10 of 27. Because all participants were nondepressed at baseline, this variable indicated the development of depression by follow-up (3–6 months).
p-value derived from chi-square tests for categorical variables and t-tests for variables measured on a continuous scale.
Self-reported SMU at baseline in minutes.
Other race included American Indian/Native Alaskan, Native Hawaiian/Pacific Islander, and Multiracial. There were too few individuals in each of these categories to warrant separate categories.
ACE was measured using a modified scale ranging from 0 to 6.
ACE, Adverse Childhood Experience; SMU, social media use.
There was a strong bivariable association between baseline SMU and the development of depression during the study period (Table 2). Compared with those in the lowest quartile of SMU, those in the highest quartile had 3.41 times the odds (95% CI=1.76, 6.60) for the development of depression. In multivariable models that adjusted for all measured covariates, compared with those in the lowest quartile of SMU, those in the highest quartile had 2.77 times the odds of developing depression (95% CI=1.38, 5.56). In the multivariable model, there was a significant linear trend in the association between baseline SMU and depression with each increasing quartile of SMU (p<0.001).
Table 2.
Bivariable and Multivariable Associations Among SMU, Personal and Sociodemographic Covariates, and Depression
| Characteristic | Development of depression during the study perioda | |
|---|---|---|
| OR (95% CI) | AOR (95% CI)b | |
| SMUc | ||
| Quartile 1 (0–120) | ref | ref |
| Quartile 2 (121–195) | 1.36 (0.70, 2.63) | 1.27 (0.64, 2.51) |
| Quartile 3 (196–300) | 2.36 (1.27, 4.41) | 1.91 (0.99, 3.67) |
| Quartile 4 (>301) | 3.41 (1.76, 6.60) | 2.77 (1.38, 5.56) |
| Age, years | 0.93 (0.87, 1.00) | 0.93 (0.85—1.02) |
| Sex | ||
| Male | ref | ref |
| Female | 1.14 (0.74, 1.77) | 0.94 (0.57—1.56) |
| Race | ||
| White, non-Hispanic | ref | ref |
| Black, non-Hispanic | 1.41 (0.61, 3.24) | 1.14 (0.44,2.97) |
| Hispanic/Latino | 1.38 (0.73, 2.61) | 1.08 (0.56, 2.09) |
| Asian | 1.64 (0.84, 3.19) | 1.66 (0.80, 3.46) |
| Otherd | 1.19 (0.15, 9.56) | 0.99 (0.13, 7.64) |
| Education | ||
| Started or graduated high school | ref | ref |
| Some college or technical school | 1.09 (0.53, 2.21) | 1.27 (0.57, 2.81) |
| College graduate | 0.76 (0.37, 1.54) | 1.04 (0.45, 2.38) |
| Graduate school | 0.52 (0.24, 1.13) | 0.74 (0.29, 1.89) |
| Annual household income, $ | ||
| <25,000 | ref | ref |
| 25,000 to <50,000 | 0.82 (0.41, 1.64) | 0.74 (0.35, 1.58) |
| 50,000 to <75,000 | 0.93 (0.46, 1.87) | 1.01 (0.46, 2.22) |
| ≥75,000 | 0.63 (0.32, 1.24) | 0.77 (0.34, 1.78) |
| Relationship status | ||
| Single | ref | ref |
| Member of unmarried couple | 0.96 (0.55, 1.67) | 1.14 (0.49, 2.63) |
| Married | 1.08 (0.66, 1.77) | 1.37 (0.50, 3.74) |
| Living situation | ||
| By myself | ref | ref |
| With parent or guardian | 1.17 (0.55, 2.51) | 0.95 (0.40, 2.60) |
| With significant other | 1.29 (0.68, 2.46) | 1.21 (0.45, 3.27) |
| With acquaintances, friends, or roommates | 1.85 (0.86, 3.97) | 1.86 (0.81, 4.26) |
| Employment | ||
| Regular employment | ref | ref |
| Homemaker | 1.85 (0.84, 4.09) | 1.20 (0.47, 3.10) |
| Student | 0.97 (0.46, 2.01) | 0.70 (0.31, 1.57) |
| Unemployed | 1.84 (0.79, 4.30) | 1.92 (0.78, 4.76) |
| ACEe | 1.20 (1.05, 1.38) | 1.17 (1.004, 1.36) |
Note: Boldface indicates statistical significance (p<0.05).
Depression was measured using the 9-Item Patient Health Questionnaire using the standard cut off of 10 of 27. Because all participants were nondepressed at baseline, this variable indicated the development of depression by follow-up (3–6 months).
These analyses controlled for all measured covariates.
Self-reported SMU at baseline. Numbers in parentheses represent minutes per day for each quartile.
Other race included American Indian/Native Alaskan, Native Hawaiian/Pacific Islander, and Multiracial. There were too few individuals in each of these categories to warrant separate categories.
Associated odds for each 1-year increase in age and each 1-unit increase on the ACEs questionnaire, which ranges from 0 to 6.
ACE, Adverse Childhood Experience; SMU, social media use.
A total of 2 sets of sensitivity analyses were conducted to examine the robustness of these results. First, the multivariable model included only covariates that had an association with the outcome of p<0.10 to ensure similar results when models were parsimonious. Second, all analyses were performed without using survey weights. Because findings from all sensitivity analyses were similar to those of the main analyses, only results from the primary analyses are presented in this report.
Of the 1,289 individuals in the sample, 517 (40.1%) increased SMU between baseline and follow-up. Table 3 compares those who did with those who did not increase SMU according to baseline depression and other sociodemographic characteristics.
Table 3.
Sample Characteristics and Bivariable Associations With SMU Increase During the Study Period (N=1,289)
| Characteristic | % | SMU increase during study perioda | ||
|---|---|---|---|---|
| No,772 (59.9%) | Yes,517 (40.1%) | p-valueb | ||
| Depressionc | 0.90 | |||
| No | 76.8 | 76.7 | 77.0 | |
| Yes | 23.2 | 23.3 | 23.0 | |
| Age, years, mean (SD) | 26.9 (2.7) | 26.9 (2.7) | 27.0 (2.8) | 0.84 |
| Sex | 0.46 | |||
| Female | 57.1 | 58.0 | 55.9 | |
| Race | 0.18 | |||
| White, non-Hispanic | 69.8 | 71.8 | 66.0 | |
| Black, non-Hispanic | 6.4 | 6.2 | 7.5 | |
| Hispanic/Latino | 13.3 | 11.9 | 14.9 | |
| Asian | 9.7 | 9.2 | 10.8 | |
| Otherd | 0.9 | 1.0 | 0.9 | |
| Education | 0.99 | |||
| At least some high school | 13.0 | 12.8 | 13.2 | |
| Some college or technical school | 28.4 | 28.5 | 28.2 | |
| College graduate | 34.2 | 34.1 | 34.2 | |
| Graduate school | 24.5 | 24.5 | 24.4 | |
| Annual household income, $ | 0.17 | |||
| <25,000 | 15.4 | 14.9 | 16.1 | |
| 25,000 to <50,000 | 27.0 | 27.0 | 27.1 | |
| 50,000 to <75,000 | 22.1 | 24.2 | 19.2 | |
| ≥75,000 | 35.4 | 33.9 | 37.7 | |
| Relationship status | 0.23 | |||
| Single | 43.3 | 45.3 | 40.4 | |
| Member of unmarried couple | 24.3 | 23.4 | 25.7 | |
| Married | 32.4 | 31.4 | 33.9 | |
| Living situation | 0.70 | |||
| By myself | 18.2 | 18.3 | 18.0 | |
| With parent or guardian | 24.4 | 22.0 | 20.4 | |
| With significant other | 47.1 | 45.9 | 49.0 | |
| With acquaintances, friends, or roommates | 13.4 | 13.9 | 12.6 | |
| Employment | 0.21 | |||
| Regular employment | 75.8 | 73.8 | 78.7 | |
| Homemaker | 6.1 | 6.2 | 5.8 | |
| Student | 11.7 | 12.8 | 10.1 | |
| Unemployed | 6.4 | 7.1 | 5.4 | |
| ACE,e mean (SD) | 1.2 (1.6) | 1.2 (1.6) | 1.2 (1.5) | 0.89 |
Based on self-reported SMU at baseline and follow-up.
p-value derived from chi-square tests for categorical variables and t-tests for variables measured on a continuous scale.
Baseline depression was measured using the 9-Item Patient Health Questionnaire using the standard cut off of 10 of 27.
Other race included American Indian/Native Alaskan, Native Hawaiian/Pacific Islander, and Multiracial. There were too few individuals in each of these categories to warrant separate categories.
ACE was measured using a scale ranging from 0 to 6.
ACE, Adverse Childhood Experience; SMU, social media use.
In bivariable logistic regression models (Table 4), there was no association between being depressed at baseline and increased SMU (yes versus no) during the study period (OR=0.97, 95% CI=0.75, 1.27). In the multivariable logistic regression model that included all covariates, there was still no association between being depressed at baseline and increasing SMU during the study period (OR=1.04, 95% CI=0.78, 1.38).
Table 4.
Bivariable and Multivariable Associations Among Depression, Personal and Sociodemographic Covariates, and SMU Increase
| Characteristic | SMU increase during the study perioda | |
|---|---|---|
| OR (95% CI) | AOR (95% CI)b | |
| Depressionc | ||
| No | ref | ref |
| Yes | 0.97 (0.75, 1.27) | 1.04 (0.78, 1.38) |
| Age, years | 1.00 (0.96, 1.04) | 0.98 (0.93, 1.03) |
| Sex | ||
| Male | ref | ref |
| Female | 0.92 (0.73, 1.15) | 0.92 (0.72, 1.17) |
| Race | ||
| White, non-Hispanic | ref | ref |
| Black, non-Hispanic | 1.47 (0.93, 2.33) | 1.62 (1.00, 2.63) |
| Hispanic/Latino | 1.33 (0.95, 1.85) | 1.40 (0.99, 1.98) |
| Asian | 1.33 (0.91, 1.95) | 1.48 (0.99, 2.20) |
| Otherd | 0.93 (0.27, 3.20) | 1.03 (0.29, 3.68) |
| Education | ||
| Started or graduated high school | ref | ref |
| Some college or technical school | 0.96 (0.66, 1.39) | 1.01 (0.67, 1.51) |
| College graduate | 0.98 (0.68, 1.41) | 0.96 (0.63, 1.46) |
| Graduate school | 0.95 (0.65, 1.40) | 0.93 (0.61, 1.42) |
| Annual household income, $ | ||
| <25,000 | ref | ref |
| 25,000 to <50,000 | 0.96 (0.67, 1.37) | 0.86 (0.58, 1.26) |
| 50,000 to <75,000 | 0.73 (0.50, 1.07) | 0.66 (0.44, 1.00) |
| ≥75,000 | 1.03 (0.73, 1.45) | 0.90 (0.60, 1.36) |
| Relationship status | ||
| Single | ref | ref |
| Member of unmarried couple | 1.22 (0.92, 1.62) | 1.29 (0.89, 1.87) |
| Married | 1.19 (0.92, 1.54) | 1.33 (0.85, 2.06) |
| Living situation | ||
| By myself | ref | ref |
| With parent or guardian | 0.97 (0.68, 1.39) | 0.95 (0.65, 1.41) |
| With significant other | 1.10 (0.81, 1.50) | 0.94 (0.60, 1.46) |
| With acquaintances, friends, or roommates | 0.96 (0.64, 1.44) | 0.93 (0.61, 1.42) |
| Employment | ||
| Regular employment | ref | ref |
| Homemaker | 0.86 (0.53, 1.38) | 0.85 (0.50, 1.46) |
| Student | 0.72 (0.50, 1.03) | 0.63 (0.41, 0.97) |
| Unemployed | 0.71 (0.44, 1.14) | 0.75 (0.44, 1.27) |
| ACE,e mean (SD) | 0.99 (0.92, 1.06) | 0.97 (0.90, 1.05) |
Based on self-reported SMU at baseline and follow-up.
These analyses controlled for all measured covariates.
Depression at baseline was measured using the 9-Item Patient Health Questionnaire using the standard cut off of 10 of 27.
Other race included American Indian/Native Alaskan, Native Hawaiian/Pacific Islander, and Multiracial. There were too few individuals in each of these categories to warrant separate categories.
Associated odds for each 1-year increase in age and each 1-unit increase on the ACEs scale, which ranges from 0 to 6.
ACE, Adverse Childhood Experience; SMU, social media use.
For these analyses, both sensitivity analyses described for Analysis 1 were also conducted. In addition, additional sensitivity analyses modeled baseline depression as continuous and increase in SMU as continuous. Because results from all these analyses were similar to those of the primary analyses, only primary results are presented in this report.
DISCUSSION
In a large national cohort of young adults aged 18–30 years, among individuals who were initially not depressed, baseline SMU was strongly and independently associated with the development of depression during the subsequent 6 months. However, there was no association between the presence of depression at baseline and an increase in SMU over the following 6 months.
Although no single study can determine causality, these results provide an important step forward in the assessment of directionality related to previously noted associations between SMU and depression.22 A temporal association is not sufficient to determine causality. For example, it may be that participants with high baseline SMU may also have had other unmeasured characteristics that led to the development of depression. However, these analyses did attempt to measure and control for such characteristics.
The degree of the association found between SMU and subsequent depression is noteworthy. A 3-fold increase in the odds of developing depression is extremely high. This is especially true considering that the outcome was the development of depression in a relatively brief period. By comparison, ACEs are well known to be predictors of the development of depression.20 However, in the multivariable model, each point on the ACE scale was associated with only a 17% increase in odds of the development of depression. This makes the potential role of SMU in the development of depression even more important. The magnitude of this finding supports previous work suggesting similar associations7,10,23 while contradicting work suggesting lower associations.24
It has been suggested that the association between SMU and depression may be explained by the fact that people with depression may feel anhedonic, shun in-person relationships, and subsequently turn to social media by default. However, this explanation is not supported by these empirical data, which find no longitudinal association between the initial presence of depression and later change in SMU.
There are 3 major conceptual reasons why SMU may be related to the development of depression. One is that SMU takes up a lot of time. In this sample, the average participant used about 3 hours of social media per day, consistent with national estimates.13 Therefore, it may be that this large amount of time displaces activities that may be more useful to the individual, such as forming more important in-person relationships, achieving true goals, or even simply having moments of valuable reflection.14 A second reason why SMU may be related to the development of depression relates to social comparison.25,26 Especially for young adults, who are at critical junctures related to the development of identity, exposure to unattainable images on social media sites may facilitate depressive cognitions.27
A third reason is that constant exposure to social media portrayals may interfere with normal developmental neurocognitive processes.28–30 For example, traditional path-ways related to social relationship development, such as social cognition, self-referential cognition, and social reward processing, involve complex interplay among multiple brain areas such as the dorsomedial prefrontal cortex, medial prefrontal cortex, and ventral striatum.28 Although research in this area is preliminary, it is possible that contextual features of SMU, such as the rapid cycling of these reward and cognitive processes, may interfere with normal development, which may in turn facilitate the development of conditions such as depression. More research needs to be done in this area to evaluate these possible mechanisms.
It is noteworthy that the association between SMU and depression increases in a linear fashion for each of the quartiles of SMU. The linear character of the association presents difficulties when offering clinical recommendations around SMU because there does not seem to be an ideal dose. Therefore, it will be important to conduct a further study that will help determine whether there are more fine-grained amounts of SMU or types of SMU that may minimize the risk of poor emotional health outcomes.
It should be noted that there are important contextual features of SMU that have been related to depression. For example, in cross-sectional studies, exposure to more negative content has been linked to depression.7 Similarly, passive exposure to social media, as opposed to more active engagement, seems to be more strongly associated with depression.31 Therefore, future studies should also examine associations such as these in longitudinal studies.
Given the strength of the association between SMU and subsequent depression as well as the lack of associations in the other direction, clinicians who work with patients who are depressed should be aware that SMU may be an important emerging risk factor for depression. These issues should also be carefully considered by public health practitioners emphasizing prevention of what is now the leading cause of disability worldwide.
Limitations
A key limitation of this study is that SMU was measured by self-report. Although this was a necessary limitation given the large scope of this study, it would be valuable for future studies to leverage more advanced methods of assessing SMU, such as ecologic momentary assessment.32 This concern is somewhat mitigated by the fact that self-report of SMU has been validated against more intensive assessments, especially in relative terms.33,34 For example, although some studies demonstrate that users overall tend to overestimate time spent on social media,35,36 the odds for depression according to relative quartile of SMU should still be accurate.
Another important limitation of this study is that it assessed overall SMU instead of more nuanced types of SMU. SMU is, of course, not the same experience for everyone: experiences vary widely according to the number of individuals interacted with, different platforms used, and topics discussed. Therefore, it will be critical to perform follow-up studies that address different types of SMU and their relative associations with the development of depression.
A final limitation relates to generalizability and selection bias. Follow-up rates were consistent with expected responses for longitudinal Internet-based assessments. However, consistent with previous literature, those who followed up were more likely to be female. This concern may be buffered by the fact that the association between SMU and the development of depression did not differ for female and male participants. In addition, survey weights were used that corrected for both coverage and follow-up.
CONCLUSIONS
This study provides the first large-scale data investigating the directionality of SMU and depression. It finds strong associations between initial SMU and subsequent development of depression but no increase in SMU after depression. This pattern suggests temporal associations between SMU and depression, an important criterion for causality. These results suggest that practitioners working with patients who are depressed should recognize SMU as a potentially important emerging risk factor for the development and possible worsening of depression.
ACKNOWLEDGMENTS
This study was funded by the Fine Foundation.
Fine Foundation had no role in study design; collection, analysis, and interpretation of data; writing of the report; or the decision to submit the report for publication.
No financial disclosures were reported by the authors of this paper.
REFERENCES
- 1.Malhi GS, Mann JJ. Depression. Lancet. 2018;392(10161):2299–2312. 10.1016/S0140-6736(18)31948-2. [DOI] [PubMed] [Google Scholar]
- 2.Wang J, Wu X, Lai W, et al. Prevalence of depression and depressive symptoms among outpatients: a systematic review and meta-analysis. BMJ Open. 2017;7(8):e017173. 10.1136/bmjopen-2017-017173. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Murray CJ, Vos T, Lozano R, et al. Disability-adjusted life years (DALYs) for 291 diseases and injuries in 21 regions, 1990–2010: a systematic analysis for the Global Burden of Disease Study 2010 [published correction appears in Lancet. 2013;381(9867):628]. Lancet. 2012;380(9859):2197–2223. 10.1016/S0140-6736(12)61689-4. [DOI] [PubMed] [Google Scholar]
- 4.Mathers CD, Loncar D. Projections of global mortality and burden of disease from 2002 to 2030. PLoS Med. 2006;3(11):e442. 10.1371/journal.pmed.0030442. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Greenberg PE, Fournier AA, Sisitsky T, Pike CT, Kessler RC. The economic burden of adults with major depressive disorder in the United States (2005 and 2010). J Clin Psychiatry. 2015;76(2):155–162. 10.4088/JCP.14m09298. [DOI] [PubMed] [Google Scholar]
- 6.Mendelson T, Tandon SD. Prevention of depression in childhood and adolescence. Child Adolesc Psychiatr Clin N Am. 2016;25(2):201–218. 10.1016/j.chc.2015.11.005. [DOI] [PubMed] [Google Scholar]
- 7.Primack BA, Bisbey MA, Shensa A, et al. The association between valence of social media experiences and depressive symptoms. Depress Anxiety. 2018;35(8):784–794. 10.1002/da.22779. [DOI] [PubMed] [Google Scholar]
- 8.Rosen LD, Whaling K, Rab S, Carrier LM, Cheever NA. Is Facebook creating “iDisorders”? The link between clinical symptoms of psychiatric disorders and technology use, attitudes and anxiety. Comput Hum Behav. 2013;29(3):1243–1254. 10.1016/j.chb.2012.11.012. [DOI] [Google Scholar]
- 9.Block M, Stern DB, Raman K, et al. The relationship between self-report of depression and media usage. Front Hum Neurosci. 2014;8:712. 10.3389/fnhum.2014.00712. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Primack BA, Shensa A, Escobar-Viera CG, et al. Use of multiple social media platforms and symptoms of depression and anxiety: a nationally-representative study among U.S. young adults. Comput Hum Behav. 2017;69:1–9. 10.1016/j.chb.2016.11.013. [DOI] [Google Scholar]
- 11.Kaplan AM, Haenlein M. Users of the world, unite! The challenges and opportunities of social media. Bus Horiz. 2010;53(1):59–68. 10.1016/j.bushor.2009.09.003. [DOI] [Google Scholar]
- 12.Uhls YT, Ellison NB, Subrahmanyam K. Benefits and costs of social media in adolescence. Pediatrics. 2017;140(suppl 2):S67–S70. 10.1542/peds.2016-1758E. [DOI] [PubMed] [Google Scholar]
- 13.Crone EA, Konijn EA. Media use and brain development during adolescence. Nat Commun. 2018;9(1):588. 10.1038/s41467-018-03126-x. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Primack BA, Escobar-Viera CG. Social media as it interfaces with psychosocial development and mental illness in transitional age youth. Child Adolesc Psychiatr Clin N Am. 2017;26(2):217–233. 10.1016/j.chc.2016.12.007. [DOI] [PubMed] [Google Scholar]
- 15.Institute of Medicine National Research Council. Investing in the Health and Well-Being of Young Adults. Washington, DC: The National Academies Press, 2015. 10.17226/18869. [DOI] [PubMed] [Google Scholar]
- 16.Shensa A, Escobar-Viera CG, Sidani JE, Bowman ND, Marshal MP, Primack BA. Problematic social media use and depressive symptoms among U.S. young adults: a nationally representative study. Soc Sci Med. 2017;182:150–157. 10.1016/j.socscimed.2017.03.061. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 17.Lin LY, Sidani JE, Shensa A, et al. Association between social media use and depression among U.S. young adults. Depress Anxiety. 2016;33(4):323–331. 10.1002/da.22466. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Moreno MA, Kota R, Schoohs S, Whitehill JM. The Facebook influence model: a concept mapping approach. Cyberpsychol Behav Soc Netw. 2013;16(7):504–511. 10.1089/cyber.2013.0025. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Ibarra JL, Agas JM, Lee M, Pan JL, Buttenheim AM. Comparison of online survey recruitment platforms for hard-to-reach pregnant smoking populations: feasibility study. JMIR Res Protoc. 2018;7(4): e101. 10.2196/resprot.8071. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.Marie-Mitchell A, Kostolansky R. A systematic review of trials to improve child outcomes associated with Adverse Childhood Experiences. Am J Prev Med. 2019;56(5):756–764. 10.1016/j.amepre.2018.11.030. [DOI] [PubMed] [Google Scholar]
- 21.Choi SW, Schalet B, Cook KF, Cella D. Establishing a common metric for depressive symptoms: linking the BDI-II, CES-D, and PHQ-9 to PROMIS depression. Psychol Assess. 2014;26(2):513–527. 10.1037/a0035768. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 22.Hill AB. The environment and disease: association or causation? Proc R Soc Med. 1965;58(5):295–300. 10.1177/003591576505800503. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 23.Twenge JM, Joiner TE, Rogers ML, Martin GN. Increases in depressive symptoms, suicide-related outcomes, and suicide rates among U. S. adolescents after 2010 and links to increased new media screen time. Clin Psychol Sci. 2018;6(1):3–17. 10.1177/2167702617723376. [DOI] [Google Scholar]
- 24.Heffer T, Good M, Daly O, MacDonell E, Willoughby T. The longitudinal association between social-media use and depressive symptoms among adolescents and young adults: an empirical reply to Twenge et al. (2018). Clin Psychol Sci. 2019;7(3):462–470. 10.1177/2167702618812727. [DOI] [Google Scholar]
- 25.Yoon S, Kleinman M, Mertz J, Brannick M. Is social network site usage related to depression? A meta-analysis of Facebook–depression relations. J Affect Disord. 2019;248:65–72. 10.1016/j.jad.2019.01.026. [DOI] [PubMed] [Google Scholar]
- 26.Hoge E, Bickham D, Cantor J. Digital media, anxiety, and depression in children. Pediatrics. 2017;140(suppl 2):S76–S80. 10.1542/peds.2016-1758G. [DOI] [PubMed] [Google Scholar]
- 27.Seabrook EM, Kern ML, Rickard NS. Social networking sites, depression, and anxiety: a systematic review. JMIR Ment Health. 2016;3(4): e50. 10.2196/mental.5842. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Meshi D, Tamir DI, Heekeren HR. The emerging neuroscience of social media. Trends Cogn Sci. 2015;19(12):771–782. 10.1016/j.tics.2015.09.004. [DOI] [PubMed] [Google Scholar]
- 29.Lamblin M, Murawski C, Whittle S, Fornito A. Social connectedness, mental health and the adolescent brain. Neurosci Biobehav Rev. 2017;80:57–68. 10.1016/j.neubiorev.2017.05.010. [DOI] [PubMed] [Google Scholar]
- 30.Choudhury S, McKinney KA. Digital media, the developing brain and the interpretive plasticity of neuroplasticity. Transcult Psychiatry. 2013;50(2):192–215. 10.1177/1363461512474623. [DOI] [PubMed] [Google Scholar]
- 31.Escobar-Viera CG, Shensa A, Bowman ND, et al. Passive and active social media use and depressive symptoms among United States adults. Cyberpsychol Behav Soc Netw. 2018;21(7):437–443. 10.1089/cyber.2017.0668. [DOI] [PubMed] [Google Scholar]
- 32.Schembre SM, Liao Y, O’Connor SG, et al. Mobile ecological momentary diet assessment methods for behavioral research: systematic review. JMIR Mhealth Uhealth. 2018;6(11):e11170. 10.2196/11170. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 33.Bennett BL, Whisenhunt BL, Hudson DL, et al. Examining the impact of social media on mood and body dissatisfaction using ecological momentary assessment. J Am Coll Health. 2020;68(5):502–508. 10.1080/07448481.2019.1583236. [DOI] [PubMed] [Google Scholar]
- 34.Stieger S, Lewetz D. A week without using social media: results from an ecological momentary intervention study using smartphones. Cyberpsychol Behav Soc Netw. 2018;21(10):618–624. 10.1089/cyber.2018.0070. [DOI] [PubMed] [Google Scholar]
- 35.Ernala SK, Burke M, Leavitt A, Ellison NB. How well do people report time spent on Facebook? An evaluation of established survey questions with recommendations. In: Proceedings of the 2020 CHI Conference on Human Factors in Computing Systems; 2020. April 25–30; Honolulu, HI. New York: ACM, 2020. 10.1145/3313831.3376435. [DOI] [Google Scholar]
- 36.Scharkow M The accuracy of self-reported internet use—a validation study using client log data. Commun Methods Meas. 2016;10(1):13–27. 10.1080/19312458.2015.1118446. [DOI] [Google Scholar]