The mental health and well-being profile of young adults using social media

The study examines who uses major social media platforms and how their mental health and well-being profiles differ, addressing gaps in population-representative data that hinder interpretation of associations between social media use and mental health. Prior work often assumes harm from social media, yet evidence suggests effects are mixed, reciprocal, and heterogeneous across individuals, contexts, and platforms. The authors aim to provide descriptive, cross-sectional profiles of users of Facebook, Instagram, Twitter, Snapchat, and YouTube in a UK cohort, across multiple mental health (depression, self-harm, suicidal thoughts, disordered eating) and positive well-being constructs (happiness, mental well-being, psychological needs, life satisfaction, optimism, meaning in life). Research questions: (1) demographic differences in social media use frequency; (2) demographic differences across platforms; (3) differences in mental health and well-being by use frequency; (4) differences by platform user groups.

The paper situates its contribution within literature using social media data for mental health prediction and research on social media’s influence on mental health. It notes assumptions of harm are common, but meta-analytic and large-scale studies (e.g., Orben & Przybylski; Appel et al.) indicate small or mixed effects, likely reciprocal and context-specific, with within-person variability and individual differences. Benefits of social media (e.g., peer support) and positive well-being are under-studied relative to pathology. A central challenge is the lack of high-quality, population-representative user profiles; existing demographic estimates (e.g., Pew) help, but many studies rely on algorithmic inference, and platform user bases differ from the general population, introducing selection bias that can confound outcomes and complicate benchmarking of predictive models. The paper addresses this by providing descriptive profiles by platform and frequency within a representative cohort.

Design and sample: Cross-sectional descriptive analyses using the Avon Longitudinal Study of Parents and Children (ALSPAC). Eligible pregnancies: 14,541; children alive at 1 year: 14,901. Analyses used a main sample of N=4083 respondents (complete data on self-harm, suicidal thoughts, disordered eating, social media use) from a questionnaire at mean age 24 (2016/17). Of 9211 invited, 4345 returned; 4083 had complete relevant data. A sub-sample of N=2991 linked responses from age 23 (well-being, MFQ) to age 24 social media data. Demographics included sex (assigned at birth), ethnicity (White, Ethnic Minority Groups, Unknown), education (A Levels by age 20: Yes/No), and parental occupation (Registrar General’s Social Class; grouped manual vs non-manual; highest of parents; measured pre-birth). Measures: Social media use—(1) account presence (Yes/No/Don't know); (2) platform-specific frequency for Facebook, Twitter, Instagram, Snapchat, YouTube (Daily, Weekly, Monthly, Less than monthly, Never). For analyses, Weekly/Monthly/Less than monthly were combined as Less than daily; (3) overall frequency of visiting any social media (More than 10/day; 2–10/day; Once/day or less). Mental health: Depressive symptoms via Short Mood and Feelings Questionnaire (MFQ; 13 items; score 0–26; ≥12 indicates depression). Suicidal thoughts: endorsement within past year to “ever thought of killing yourself…”. Self-harm: any intentional self-harm within past year. Disordered eating: composite—self-reported professional diagnosis (anorexia, bulimia, binge eating disorder, other) and/or monthly past-year weight-control behaviors (fasting, vomiting, laxatives/medication), following Micali et al. Well-being: seven instruments—Warwick-Edinburgh Mental Well-being Scale (WEMWBS) total 14–70; Satisfaction with Life Scale (5–35); Subjective Happiness Scale (mean 1–7); Gratitude Questionnaire GQ-6 (6–42); Life Orientation Test-Revised (optimism; 0–24); Meaning in Life Questionnaire subscales: Presence and Search (each 5–35); Basic Psychological Needs (BPN) in General: Autonomy, Competence, Relatedness (each mean 1–7). For multi-item scales, missing items were imputed with person-level mean/mode where ≥50% items answered; all well-being scales scored positively (higher=more of construct). Analysis: Descriptive statistics and visualizations in R (v4.0.1), RStudio, tidyverse, ggplot2. Chi-squared tests for demographic differences in frequency (Table 2). Results stratified by sex due to imbalance and known differences. Ethics: ALSPAC Ethics and Law Committee approvals; informed consent obtained.

Use frequency: Among N=4083, 39% used social media more than 10 times/day (n=1576), 53% used 2–10 times/day (n=2144), and 8.7% once/day or less (n=356). Sex was significantly associated with frequency (p<0.001): females 40% >10/day, 53% 2–10/day, 7.1% ≤1/day; males 35% >10/day, 53% 2–10/day, 12% ≤1/day. Other demographics showed no significant association. Platform use (any frequency; Table 3): Facebook 97%; Twitter 56%; Instagram 69%; Snapchat 70%; YouTube 73%. Sex differences: Instagram (female 76% vs male 54%), Snapchat (73% vs 64%), YouTube (68% vs 83%), Facebook (98% vs 97%), Twitter (56% vs 57%). Twitter users were more often A Level educated and from non-manual parental backgrounds; Snapchat showed the reverse. Ethnic minority participants were more likely to use Twitter than White participants. Mental health by frequency: Patterns were not uniformly linear. Among women, the lowest use group (≤1/day) had the highest proportions of disordered eating, self-harm, and suicidal thoughts. In men, only the highest use group (>10/day) showed higher depressive symptoms (MFQ≥12). Findings partially align with a non-linear (Goldilocks) pattern but differ by outcome and sex. Well-being by frequency: Subjective happiness and optimism (LOT-R) were relatively stable across frequencies. Relatedness was higher among women in the two most frequent use groups. Different well-being constructs yielded different patterns, underscoring construct-specific relationships. By platform (daily users): YouTube daily users—especially women—had the highest prevalence of disordered eating, self-harm, suicidal thoughts, and depression. In well-being, YouTube users showed lower life satisfaction, relatedness, and (notably for women) competence; WEMWBS was only marginally lower. Instagram (and in some cases Snapchat) daily users showed the highest subjective well-being across several measures, particularly relatedness, gratitude, and happiness. Overall, platform user groups are heterogeneous, with sex being a primary differentiator, and platform-specific profiles evident.

The findings address the research questions by demonstrating that demographic composition (notably sex) differs across both use frequency and platform, that platform user groups are not homogeneous, and that mental health and well-being vary by both frequency and platform in outcome-specific ways. These differences imply potential selection bias when sampling from particular platforms and highlight the need for benchmarking and stratification in studies of social media and mental health. The non-linear associations between frequency and outcomes support considering dose-response shapes beyond linear models. The heterogeneity across well-being measures (e.g., stability for happiness/optimism versus variability for relatedness) illustrates the jingle-jangle issue; conflating distinct constructs under “well-being” can lead to conflicting findings and poor replicability. Platform-specific results—poorer mental health among YouTube daily users, higher well-being among Instagram/Snapchat daily users—challenge treating “social media” as a single entity and suggest that content type and interaction modes (e.g., passive vs interactive) may relate differently to mental health. The profiles provided can contextualize predictive model performance and help interpret prior mixed findings.

This study provides population-based descriptive profiles of young adult social media users across five platforms and three use-frequency categories, reporting prevalence of four mental health outcomes and means for seven well-being constructs. Users differ primarily by sex, and platform-specific profiles are evident: YouTube users are more likely to show poorer mental health and lower well-being on several indicators, whereas Instagram and Snapchat users generally exhibit higher positive well-being. Associations between frequency and outcomes are often non-linear and depend on the specific well-being construct. Implications include the need to stratify analyses by sex, avoid conflating platforms and well-being constructs, and to model potential non-linear relationships. Future research should extend profiling across age groups, employ longitudinal designs to probe reciprocal effects, incorporate objective use measures and granular activity types (passive vs interactive), and examine how platform content and communication modes may differentially relate to mental health.

- Age- and cohort-specific sample (early 20s in a UK cohort), limiting generalizability across ages and generations; two data collection points were one year apart. - Limited ethnicity variable (White vs Ethnic Minority Groups) and cohort area predominantly White; unable to analyze specific ethnic groups. - Differential attrition led to underrepresentation of men and overrepresentation of higher SES participants, potentially biasing results. - Self-reported use frequency may be imprecise versus objective measures; definitions of “use” vary, recall bias present; frequency is distinct from screen-time, and time spent per visit varies by platform. - Lack of specificity about activity types (passive consumption vs active posting/communication) which likely differ by platform (e.g., YouTube more passive). - Cross-sectional focus for most analyses precludes causal inference; potential for unmeasured confounding and selection effects.