U.S. cohort · Baseline 2016–2018, follow-up 2018–2020
How was social-media use related to later weight concerns?
In a U.S. childhood cohort, more reported social-media time at ages 9–10 was associated with higher odds of weight-related concerns at the two-year follow-up.
For each additional typical daily hour at baseline, the adjusted odds ratio was 1.55 for “fear of obesity” (95% CI 1.21–1.98) and 1.30 for self-worth tied to weight (95% CI 1.03–1.63). These are associations, not percentages harmed or proof that social media caused the concerns.
ABCD cohort · Youth-reported social-media hours at baseline · Youth-reported symptoms at year two · Observational study
Source: Chu et al. (2024), Table 2, social-media column, published September 4, 2024. Original study: Eating and Weight Disorders 29, 57. The data describe 2016–2020, not current platform use.
Reported associations, with their uncertainty
Linear axis 0.8–2.1. The dashed line at 1.0 marks equal modeled odds per additional hour. Intervals describe uncertainty in each association; they do not test differences between outcomes.
Source: Chu et al., Table 2, CC BY 4.0. SocialUsage selects the social-media column and redraws reported estimates. Third row label shortened here; exact labels and values in the accessible table.
The first two outcomes concern weight-related fear and self-worth. The other three concern eating-related symptoms or behaviors. Keeping them separate matters: this table does not measure a single “negative body image” score, nor can its overlapping outcomes be added into an overall harm rate.
Read all five estimates, intervals and source p values
| Year-two symptom | Adjusted OR | 95% CI | Source p |
|---|---|---|---|
| Fear of obesity | 1.55 | 1.21–1.98 | 0.001 |
| Self-worth tied to weight | 1.30 | 1.03–1.63 | 0.025 |
| Inappropriate compensatory behaviors to prevent weight gain | 1.18 | 1.01–1.40 | 0.039 |
| Binge eating | 1.28 | 1.10–1.49 | 0.002 |
| Distress with binge eating | 1.31 | 1.06–1.61 | 0.012 |
The authors highlight these p values as below .05. The paper does not describe a multiple-comparison adjustment for this selection. We report the original values and do not perform additional significance tests or rank the associations.
An odds ratio is not a percentage of children
An adjusted odds ratio compares modeled odds while accounting for the listed covariates. Odds are the probability of an outcome divided by the probability of its absence. A value of 1.0 represents equal odds; a value above 1.0 represents higher odds.
The 1.55 estimate does not mean 1.55% of children had the concern, and it is not a percentage-point increase in prevalence. Recovering absolute probabilities would require additional model information and a defined starting probability. This page supplies neither an absolute-risk prediction nor an extrapolation across several extra hours.
Exposure was measured before the follow-up assessment, which helps establish timing. It does not remove all alternative explanations. The authors describe residual confounding and reporting bias, and explicitly state that their observational study cannot establish causality.
Symptoms were reported by parents at baseline but by the young participants at follow-up. Adjusting for the parent report cannot establish that every later report was a newly developed symptom. The results also do not diagnose an eating disorder or predict an individual child's experience.
Interpretation follows the original statistical methods and discussion limitations. It is an explanation of study measures, not clinical advice.
Social-media hours were a separate exposure
At baseline, children reported typical weekday and weekend daily hours for several activities. Social media, with examples such as Facebook, Instagram and Twitter, was separate from television, online videos, gaming, texting and video calls. The study combined weekday and weekend answers as:
(weekday hours * 5 + weekend hours * 2) / 7
That is the authors' exposure construction, not a new SocialUsage calculation from participant data. It measures reported typical use, not a device log, a specific platform's effects or the content children saw.
The paper also studies total screen time and a later problematic-use questionnaire. Those are different exposures. This page selects only the baseline social-media-hours column of Table 2, so the other models cannot be substituted into this chart.
The follow-up assessed symptoms, not appearance comparison
Year-two symptoms came from youth responses to the KSADS-5 assessment. The study uses “fear of weight gain” in its narrative and “fear of obesity” as the table label; we preserve the table label for traceability. Self-worth tied to weight is a separate reported concern.
The article does not establish one common recall window for all five selected outcomes. “At two-year follow-up” identifies when they were assessed, not a verified two-year incidence window. This selection does not directly measure comparing one's appearance with others, satisfaction with a particular body feature, or exposure to edited images.
Source: Methods: exposures and year-two eating disorder symptoms.
Methods, dates and reuse
Sample, adjustment and missing information
The paper reports a final analytical sample of 10,246 participants from the U.S. ABCD cohort, recruited largely through schools at 21 sites. Baseline ages were 9–10 in 2016–2018, with year-two assessments in 2018–2020. That overall sample is not a verified denominator for every model. The main-text recruitment and exclusion figures do not fully reconcile with the stated final sample; we retain the reported final sample without reconstructing participant flow.
The authors applied ABCD propensity weights aligned with American Community Survey age, sex and race/ethnicity distributions. Missing exposure or outcome records were excluded. Missing baseline sociodemographic values were imputed using Gaussian normal regression. Neither weighting nor adjustment makes this a current all-teen prevalence survey.
Table 2 footnote a lists these covariates: race/ethnicity; sex; household income; parent education; site; baseline parent-reported eating disorder symptom; baseline BMI percentile; baseline anxiety symptoms; baseline BAS reward responsiveness. We preserve that table-specific list. The models used multiple logistic regression in Stata 18. Model-specific n, standard errors and a common outcome recall window are not supplied in this selection.
Source dates and attributed reuse
- Data
- Baseline 2016–2018; year-two assessment 2018–2020
- Article published
- September 4, 2024
- Source access checked
- 2026-10-08
- Assisted review
- 2026-10-08
Credit: Jonathan Chu, Kyle T. Ganson, Alexander Testa, Abubakr A. A. Al-shoaibi, Dylan B. Jackson, Rachel F. Rodgers, Jinbo He, Fiona C. Baker and Jason M. Nagata. Screen time, problematic screen use, and eating disorder symptoms among early adolescents: findings from the Adolescent Brain Cognitive Development (ABCD) Study, Eating and Weight Disorders 29, 57 (2024). Original article and rights notice; Creative Commons Attribution 4.0 International.
SocialUsage selects five aggregate results, redraws their intervals and adds original interpretation; the reported numbers are unchanged. The article's licence allows attributed reuse and adaptation, including commercial use, subject to its terms and any third-party exceptions. No source graphics, questionnaire pages or participant records are reproduced. Export and embed utilities are not provided for this guide; that product choice does not narrow the source licence. Assisted review is not independent human review, and the authors do not endorse this page.