Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2026 Oct 6.
Published before final editing as: Int J Hum Comput Interact. 2026 Sep 25:10.1080/10447318.2026.2734696. doi: 10.1080/10447318.2026.2734696

Modeling Daily Digital Experiences in Adolescents: A Multilevel Latent Class Analysis of Social Media Engagement Patterns and Exposure to Alcohol-Related Content

Xiangyu Tao 1,2, María E Contreras-Pérez 1,2, Alex C Clement 3, Tim Janssen 4, Joy Gabrielli 3, Suzanne M Colby 4, Tyler Wray 4, Matthew K Meisel 4, Stephanie T Lanza 5, Kristina Jackson 1,2
PMCID: PMC13637484  NIHMSID: NIHMS2210282  PMID: 42835697

Abstract

This study used multilevel Latent Class Analysis (mLCA), a day-level, person-centered approach to examine adolescents’ social media engagement, and associated it with exposure to alcohol-related content. Participants were 302 adolescents (Mage = 16.21) who completed 21 days of ecological momentary assessment and weekly surveys. Using multilevel latent class analysis, we identified six distinct engagement types: No Use, Passive Use, Liked/Reacted, Moderate Engagement, Discussion, and High Engagement. Sixty-one percent of adolescents spent most days in one dominant class. Exposure to alcohol-related content from peers or followed accounts predicted more active engagement days, consistent with social learning processes. Exposure from peers is related to temporary social media disengagement, whereas exposure from followed accounts predicted reactive use. Reliance on self-report and the absence of private-messaging measures are noted limitations. Findings highlight dynamic daily processes linking online engagement with substance-use risk contexts.

Keywords: social media, alcohol-related content, multilevel latent class analysis, ecological momentary assessment

1. Introduction

Social media plays a central role in contemporary adolescent development, serving not only as a source of entertainment but also as an important space for identity exploration, social comparison, and behavioral modeling (Balcha & Chen, 2025; Engel et al., 2024). The majority of U.S. adolescents report daily use of platforms like YouTube (73%) and TikTok (57%), with nearly one in six saying they are online “almost constantly” (Sidoti, 2024). As adolescents engage in these digital spaces, they routinely encounter a wide range of content, including sensitive or risk-related material such as depictions of alcohol use. Prior work shows that such content is common and that greater social media engagement is associated with higher exposure (Erevik et al., 2017; Kurten et al., 2022; Russell et al., 2022; Zhang & Esser, 2024). Meta-analytic evidence further demonstrates that exposure to alcohol-related posts relates to shifts in beliefs, willingness, and behavioral risk (Cheng, Lim, Rutherford, et al., 2024; Curtis et al., 2018; K. M. Jackson et al., 2021; Kurten et al., 2022, 2022; Steers et al., 2025).

While this body of research provides strong evidence of importance of social media exposure in adolescent development, far less is known about how adolescents’ daily social media behaviors organize into meaningful patterns that shape their exposure environments from day to day. Adolescents do not use social media uniformly: on some days they may browse passively, on others they may interact with peers, and on still others they may post, share, or engage with trending content (Ehrenreich et al., 2021). These day-to-day configurations form distinct micro-contexts that differ in the kinds of interactions adolescents experience and the opportunities they have to encounter specific types of content.

Guided by Social Cognitive Theory (SCT; Bandura, 1986, 2001), we conceptualize these daily social media behaviors as part of the antecedent component in the ABC behavioral model (Antecedent-Behavior-Consequence). SCT highlights that learning occurs through observation and reinforcement in the moment, as adolescents attend to and interpret the content they encounter (Helbig et al., 2019). Social media posts that portray alcohol use in positive contexts (e.g., parties) act as symbolic models, reinforcing perceptions that drinking is normative or desirable (Nagata et al., 2023), which may be more likely to occur among heavy social media users. These dynamics may be particularly pronounced during late adolescence (ages 15–18 years), a developmental period characterized by increased autonomy, intensified peer influence, and heightened sensitivity to social rewards (Crone & Dahl, 2012). During this age period, the likelihood of alcohol use also rises significantly (WHO, 2018).

From an SCT perspective, such patterns reflect reciprocal determinism, defined as the dynamic interaction between cognitive factors (e.g., anticipated outcomes), environmental influences (e.g., peer content), and behavioral inputs (e.g., social media platform use)(Bandura, 2001). Existing research has primarily focused on between-person differences in overall social media use (e.g., heavier vs. lighter users) in relation to alcohol content exposure or associated alcohol use outcomes (Cheng, Lim, Rutherford, et al., 2024; Erevik et al., 2018; Nagata et al., 2023; Riehm et al., 2019). Such work has been foundational for identifying population-level associations; however, SCT also highlights the importance of proximal, moment-to-moment behavioral patterns in shaping the learning opportunities adolescents encounter. Indeed, a growing body of work argues that understanding social media's role in adolescent development requires moving beyond global frequency- or time-based metrics. which capture only how much adolescents use social media, toward more specific dimensions of use, including particular behaviors, content types, and contexts of engagement (Orben et al., 2024; Schønning et al., 2020). Specifically, day-level engagement patterns can link to increased attention to specific content and may also guide algorithmic curation, influencing future exposure, consistent with SCT’s emphasis on dynamic and bidirectional influence that could be unfolded using a person-centered approach.

Prior studies have begun differentiating content characteristics among adult populations, building on typologies developed to classify alcohol-related posts (Hendriks et al., 2018). For example, exposure to posts emphasizing positive alcohol consequences (e.g., fun, bonding) is associated with higher alcohol use risk (Erevik et al., 2017), whereas exposure to posts describing negative consequences or recovery experiences has been linked to stronger treatment-seeking intentions (Russell et al., 2022). Moreover, alcohol-related posts from close friends or peers tend to carry more influence than those from influencers or people they don’t do not know (Corcoran et al., 2024; Cristello et al., 2024). This is consistent with SCT, which highlights that the learning model is most influential when content features proximal others, such as friends and peers, whose behaviors are more likely to be seen as relevant and achievable (Hendriks et al., 2020; Vanherle et al., 2024). However, no work has examined how adolescents’ daily social media engagement patterns correspond to their day-level exposure to content, including whether certain types of days make encounters with alcohol-related material more likely.

1.1. The Current Study

The present study extends prior work by shifting attention from overall levels of social media engagement to the day-level patterns through which adolescents navigate digital environments. Rather than examining alcohol use behavior itself, we use alcohol-related content as a case domain to illustrate how daily engagement patterns function as antecedent processes within the ABC model. Using 21 consecutive days of intensive longitudinal data, we applied multilevel latent class analysis (mLCA) to identify recurring profiles of daily social media use nested within individuals. We then examined whether person-level exposure to different types of alcohol-related content (positive vs. negative; friends vs. other accounts) predicted the likelihood of adolescents experiencing each day type. Although mLCA is exploratory and does not specify directional hypotheses about classes, SCT broadly suggests that adolescents who encounter more alcohol-related content, particularly from close friends, may be more likely to experience days marked by active engagement, the behavioral channels through which observational learning and reinforcement would plausibly operate. We therefore treated this as a broad conceptual expectation rather than a confirmatory hypothesis and interpreted the results as overall patterns rather than specific class-to-class contrasts. Overall by adopting this day-level, person-centered approach, the current study offers new insight into the dynamic micro-contexts adolescents encounter online and demonstrates how mLCA can enrich SCT-informed research by capturing the proximal, unfolding nature of digital exposure environments (see Figure 1 for our conceptual model).

Figure 1. The Conceptual Model.

Figure 1.

Open in a new tab

2. Methods

2.1. Participants and Procedures

Data were gathered from 302 adolescents (Mage=16.21, SD=0.50, 50.66% women/transgender women) enrolled in a longitudinal study examining youth in-vivo exposure to alcohol content in the media. All study procedures were approved by the Brown University Institutional Review Board (see (K. Jackson et al., 2024) for full study procedures). Participants were recruited nationally via social media and enrolled in four seasonal cohorts between March 2022 and December 2022. Inclusion criteria were: (1) age between 15 and 18 years, (2) enrollment in a U.S. high school, (3) English literacy, (4) owning a smartphone, and (5) either lifetime alcohol use or having at least one close friend who used alcohol. Written informed consent was obtained from participants (over 18 years) or their parent/guardian (under 18 years). Study staff also obtained assent from participants under 18 years of age.

Following a group orientation conducted over Zoom and completion of an online baseline survey including demographics and measures related to media use and substance use, participants completed a 21-day ecological momentary assessment (EMA) data collection. Each day, they were prompted at random times during four “blocks” (morning, afternoon, early evening, and bedtime) to report on media use including any exposure to alcohol-related content; if no alcohol content was encountered, participants completed a set of general media questions about platform use and digital interactions unrelated to alcohol (EMA data on alcohol-related media content is not included in the present study). This was done to minimize non-response due to “learning” that an alcohol report was more burdensome and ensured that all participants had the same opportunities to earn compensation based on survey compliance regardless of their daily alcohol content exposure. Participants also completed weekly surveys that were administered after each 7-day period across the 21-day study. These weekly surveys assessed adolescents’ exposure to social media posts depicting alcohol-related content more broadly, including exposure to positive (e.g., fun, bonding) and negative (e.g., regret, harm) consequences, and whether such posts came from friends or other followed accounts. Three weekly surveys were completed in total.

The level of compensation was commensurate with study adherence, with participants compensated $5 per day along with bonuses for high compliance (6 of 7 complete days per week) along with $35 for the orientation session and baseline survey, and $10 for each weekly assessment completed.

2.2. Measures

2.2.1. Daily Social Media Measures/MLCA Indicators

Eleven binary items related to engagement with social media content served as indicators for the multilevel latent class analysis (mLCA) that was the focus of this study. For this project, “social media” was defined as interactive, networked platforms that allow users to generate content and interact with peers (e.g., Instagram, TikTok, Snapchat, Twitter, Facebook). Streaming services (e.g., Netflix, Hulu), YouTube, music platforms (e.g., Spotify), and television or movie viewing were not included, as they do not constitute interactive social networking. At each EMA report, participants reported if they engaged with those social media content in each of the following ways: liking or reacting (e.g., emoji), commenting, sharing publicly, sharing privately, saving, following the user, blocking the user, clicking on an affiliated link, discussing with someone else, or nothing/kept scrolling (this option being mutually exclusive with other options). A final binary indicator was coded to indicate whether the participant endorsed not using any social media platforms that day. Endorsed reactions were aggregated across up to four EMA reports for each day and binarized to indicate whether they occurred (or not) during that day. Overall, participants reported social media activities on 97% of days (See Results section for details).

2.2.2. Alcohol Content Exposure and Other Covariates

The primary covariates were exposure to any alcohol content in social media and exposure to depictions of positive and negative alcohol-related consequences in social media over three weeks, as assessed in each week’s end-of-week survey. In these surveys, participants responded to three prompts: “This past week, did you see any text, pictures, or videos related to alcohol, drinking, being drunk, or hung-over in the media?” and “This past week, did you see any text, pictures, or videos in the media either publicly or privately that referred to [positive consequences of alcohol use (e.g., having fun, socializing/fitting in, relaxing)?]/[negative consequences of alcohol use (e.g., hangovers, loss of control)?]. For each prompt, participants indicated whether the content was (1) posted by close friends on social media, (2) posted by other people or accounts that they followed on social media, (3) seen from other media sources, or (4) not seen. Participants could select more than one response option for each question. Although the item stem referred broadly to alcohol-related content “in the media,” response options distinguished social media exposure from exposure through other media sources. Because the present analyses focused on social media, only exposure attributed to close friends or other followed accounts on social media was included in the social media exposure variables; exposure attributed to other media sources was not included. Participant reports of the type and source of alcohol content/consequence to which they were exposed were aggregated to the person level (i.e., the 3-week EMA period) and binarized to have occurred (1) or not occurred (0) during the 21-day period.

To account for school holidays and breaks, we examined multiple high school academic calendars from diverse U.S. regions (East Coast, West Coast, and Midwest) and identified the widest overlapping periods to approximate a standardized holiday schedule across participants. Days were additionally coded as weekdays (Monday–Friday) or weekends (Saturday–Sunday) to capture weekday–weekend variability in social media use.

2.3. Analytic Plan

We used multilevel latent class analysis methods to attempt to identify classes of social media engagement using MPlus version 8.11 (Muthén & Muthén, 2017). We explored solutions with as few as two, and as many as eight classes, using the 11 binary, daily-level engagement indicators. The most parsimonious model was identified through comparison of several fit indices across these solutions including Akaike information criterion (AIC), Bayesian information criterion (BIC), and sample size adjusted Bayesian information criterion (SABIC), while also ensuring sufficient entropy and an adequate minimum number of days represented in the smallest latent class to support reliable interpretation of class-specific patterns (Asparouhov & Muthén, 2012; Celeux & Soromenho, 1996; Ramaswamy et al., 1993; Weller et al., 2020). Lower values of AIC, BIC, and SABIC indicate a more optimal balance between model fit and parsimony relative to competing models with higher values on these indices. Classes comprised of a very small (i.e., < 5% of expected days) proportion were avoided, unless the class was distinct and conceptually interpretable (Weller et al., 2020). The bootstrap likelihood ratio test is not available when LCA is applied to clustered data, such as daily data.

We adjusted the standard errors of LCA parameter estimates to account for the non-independence of person-days clustered within individuals (Skinner et al., 1989). The model was specified at the day-level, treating participant ID as a clustering variable rather than explicitly modeling between-person heterogeneity. This approach assumes that all days are drawn from a shared population of latent classes and produces population-averaged parameter estimates. Model identification was addressed by specifying 1,000 random sets of starting values in Mplus, with the best loglikelihood solution required to be replicated.

The manual Bolck-Croon-Hagenaars (BCH) method was conducted in Mplus to estimate unadjusted associations between day-level social media engagement class membership and person-level alcohol content exposure (Bolck et al., 2004). Following recommended procedures, class-specific BCH weights were first saved from the latent class model, then incorporated into subsequent models to regress class membership on covariates (Asparouhov & Muthén, 2014). This approach accounts for classification error while providing bias-adjusted estimates when examining covariate–class relationships. As sensitivity analyses addressing the potential for inflated Type I error across the many pairwise contrasts, we conducted an overall Wald test (df = 5) of each covariate's association with class membership and applied the Benjamini–Hochberg false discovery rate (FDR) correction to the pairwise contrasts within each covariate.

3. Results

3.1. Descriptive Statistics

Participants (n = 302) contributed a total of 21,820 survey-level observations across the 21-day ecological momentary assessment (EMA) period, including 17,116 (78.44%) answered questions related to social media use. The complete EMA dataset was aggregated into 6,013 day-level records, including those with missing responses on social media use (n day = 173, i.e., all alcohol exposure surveys across the day, so did not answer any general social media use questions). For latent class analysis, such days were excluded, resulting in 5,840 valid day-level records with non-missing social media reports. After excluding six participants who contributed fewer than five valid days, the final analytic sample included 5,830 (96.95%) day-level observations from 296 participants. Logistic regression predicting complete exclusion from the analytic sample based on age, sex, ethnicity, and race was not significant, χ2(9) = 4.21, p = .90. A separate logistic regression predicting missingness >20% EMA days indicated that demographic characteristics were generally unrelated to missing data, χ2(9) = 17.92, p = .037. The only significant predictor was age, with older adolescents less likely to be in the high-missing group (OR = 0.60, 95% CI [0.36, 0.99], p = .049). On average, participants completed 19.70 days (SD = 2.86; range = 6–21), with 95.27% completing a minimum of one EMA report at least 14 out of 21 days.

Participants in the analytical sample (n = 296, Mean Age = 16.20, SD = 0.71; 51.69% women or transgender women) were predominantly in 10th or 11th grade (80.65%), with around 25% identified as Hispanic/Latine. Less than half identified as White (45.27%), followed by Black or African American (18.24%), and 48.31% reported past-year alcohol use (See Table 1). Endorsement of the individual social media engagement indicators was common: for example, 94.3% of participants liked or reacted to content and 80.7% commented on at least one day (Supplementary Table S1). At study enrollment, nearly all participants held Instagram (99.3%) and YouTube (97.3%) accounts, and most held TikTok (85.8%) and Snapchat (82.6%) accounts; daily use exceeding one hour was most common for TikTok (89.7%), Instagram (83.4%), and YouTube (63.5%). Although 68.1% held Facebook accounts, daily Facebook use was minimal (5.7% reported an hour or more; M = 13 minutes; K. M. Jackson et al., 2026).

Table 1.

Participant Demographic Information (n=296)

Variable n (%)
Age
15 years 60 (20.27%)
16 years 116 (39.19%)
17 years 118 (39.86%)
18 years 2 (0.68%)
Grade
9th 43 (14.53%)
10th 126 (42.57%)
11th 113 (38.18%)
12th 14 (4.73%)
Gender
Woman/Transgender Woman 153 (51.69%)
Man/Transgender Man 93 (31.42%)
Non-binary/Gender Fluid/Not listed 50 (16.89%)
Race
American Indian/Alaska Native 3 (1.01%)
Asian 68 (22.97%)
Black or African American 54 (18.24%)
White 134 (45.27%)
Multiracial 23 (7.77%)
Not listed 14 (4.73%)
Ethnicity
Hispanic/Latine 76 (25.68%)
Past-Year Drinking Status
No Use 153 (51.69%)
Any Use 143 (48.31%)
Exposure to Alcohol-related Social Media Content (n = 294)
Alcohol-related posts — Close friends 131 (44.56%)
Alcohol-related posts — People followed 218 (74.15%)
Positive consequences — Close friends 101 (34.35%)
Positive consequences — People followed 175 (59.52%)
Negative consequences — Close friends 62 (21.09%)
Negative consequences — People followed 112 (38.10%)
Both positive & negative consequences — Close friends 51 (17.35%)
Both positive & negative consequences — People followed 97 (32.99%)
Open in a new tab

Across the study period, 294 adolescents (99.32%) provided responses regarding alcohol-related content exposure for at least one week. Among them, 44.56% reported seeing posts showing alcohol-related content from close friends, and 74.15% from other people they follow (not including close friends). In addition, 34.35% encountered posts portraying positive consequences of alcohol use from close friends, 59.52% from people they follow. In contrast, 21.09% reported negative portrayals from close friends, 38.10% from other followed accounts. Finally, 17.35% were exposed to both positive and negative consequences from close friends, and 32.99% from people they follow.

3.2. Day-Level Latent Class Solution

We estimated multilevel latent class models specifying two- through eight-class solutions to identify distinct patterns of daily social media use behaviors. Fit indices (AIC, BIC, and adjusted BIC) improved consistently as more classes were added, with the 8-class model showing the lowest values across most indices, except that the 6-class model showed the lowest BIC value (see Table 2). However, entropy decreased in the 8-class solution (0.72) relative to the 6-class model (0.83), suggesting reduced classification certainty. Moreover, the 8-class solution split the discussion-oriented profile into a discussion-only class and a discussion-plus-liking class; the latter largely overlapped with liking- and engagement-based profiles already represented in more parsimonious solutions, such that the additional classes did not yield conceptually distinct day types (For details, see Supplementary Table S2). In contrast, the 6-class solution provided coherent and qualitatively distinct profiles. Thus, we retained the 6-class solution as the final model to balance model fit with interpretability and parsimony. Item-response probabilities for each class, along with class labels, are presented in Figure 2.

Table 2.

Fit Indices for 2 through 8 Class Latent Class/Profile Solutions

Solution AIC BIC SABIC Entropy Model- Implied Class Sizes (%)
2 47767.34 47920.77 47847.68 0.75 43.7%, 56.3%
3 45084.16 45317.64 45206.42 0.93 13.1%, 28.5%, 58.4%
4 44074.85 44388.38 44239.02 0.83 11.5%, 13.1%, 28.3%, 47.1%
5 43907.04 44300.62 44113.13 0.85 3.4%, 12.0%, 13.1%, 20.9%, 50.6%
6 43720.31 44193.94 43968.32 0.83 2.6%, 3.2%, 13.1%, 20.9%, 27.3%, 32.9%
7 43658.69 44212.37 43948.62 0.78 2.9%, 3.3%, 6.8%, 13.1%, 19.8%, 20.9%, 33.2%
8 43619.06 44252.79 43950.90 0.77 3.2%, 3.4%, 8.8%, 10.0%, 10.3%, 13.1%, 20.9%, 30.3%
Open in a new tab

Note. AIC: Akaike information criteria, BIC: Bayesian information criteria, SABIC: sample size-adjusted Bayesian information criteria. Fit indices are reported to two decimal places, and model-implied class sizes are rounded to one decimal place for readability.

Figure 2. What Types of Social Media Use Days are Youth Reporting? Prevalence of endorsing each type of social media use given types of day.

Figure 2.

Open in a new tab

Liked/Reacted Days (C1, 32.9% of days) represented the most prevalent social media use type of day, marked by near-universal endorsement of liking or reacting to posts, with minimal engagement in other social media behaviors. Moderate Engagement Days (C2, 27.3%) reflected a mix of actions, including liking, sharing, following, and moderate levels of commenting and saving posts. Passive Use Days (C3, 20.9%) were defined by very low engagement across all behaviors except for passive scrolling, as indicated by high endorsement of “did nothing.” No Use Days (C4, 13.1%) captured days on which no social media use was reported. Discussion Days (C5, 3.2%) reflected low platform-based activity but elevated rates of discussing social media content with others. Finally, High Engagement Days (C6, 2.6%) represented a small subset of highly active days, characterized by consistently high endorsement across all social media behaviors, including posting, sharing, following, and clicking affiliated links.

We also examined whether day types varied by calendar context. On school holidays and breaks, adolescents were significantly more likely to have Passive Use Days, High Engagement Days, or Moderate Engagement Days than No Use Days (ps ≤ .027). In contrast, weekend effects emerged only between No Use and Moderate Engagement Days: adolescents had lower odds of Moderate Engagement relative to No Use Days on weekends (p = .041), suggesting a tendency for some youth to disengage more fully at the end of the week. No significant differences by holiday or weekend status were observed for Liked/Reacted Days or Discussion Days.

For descriptive purposes, we also calculated the proportion of days each participant spent in each class based on model class assignment using the highest posterior probability on that day as shown in Figure 3(A-B). Participants engaged in a median of 4 types of social media behavior days over the course of the 21 days (M = 3.59; SD = 1.04). Around 53.0% of youth engaged with four or more distinct latent classes across the study period, while only 0.7% remained in a single class throughout. However, despite this variability, 61.8% of youth spent more than half of their days in one dominant class, suggesting some level of consistent daily social media use behavior.

Figure 3. Person-level Patterns of Daily Social Media Use Class Membership.

Figure 3.

Open in a new tab

Note. Figure shows burst-level patterns of latent class engagement across the 21-day EMA period. (A) Pie chart shows the distribution of participants by the number of distinct day-level classes they experienced across the 21-day period, based on each day's modal class assignment. (B) Histogram shows the distribution of participants by the proportion of days spent in their most frequent class; the dashed line marks the 50% threshold, with values above it indicating participants who had a dominant day-level class.

3.3. Associations between Alcohol Content Exposure and Social Media Day Types

Table 3 presents the pairwise comparison results using person-level alcohol content exposure aggregated across the three weekly surveys in the 21-day study period, to predict social media day types. Adolescents reporting exposure to content depicting negative alcohol consequences posted by close friends during this period were more likely to have High Engagement Days compared with Passive Use, Discussion, and Liked/Reacted Days, while seeing positive alcohol consequences posted by close friends was associated with higher odds of having Discussion Days compared with High Engagement Days, Passive Days, and Liked/Reacted Days. Meanwhile, adolescents reporting general alcohol-content exposure from close friends reported higher odds of Moderate Engagement Days and No Use Days compared to Passive Use Days or Discussion Days.

Table 3.

Pairwise Comparison for Person-Level Alcohol Content Exposure Predicting Day-Level Latent Class Membership

Contrast Exposure: close friends Exposure: followed accounts Positive consequences: close friends Positive consequences: followed accounts Negative consequences: close friends Negative consequences: followed accounts
Overall Wald(5), p 26.46, p < .001 43.62, p < .001 10.53, p = .062 8.98, p = .110 14.98, p = .011 9.99, p = .076
C2 vs C1 1.45 [1.08, 1.93] † 0.45 [0.32, 0.62] † 1.29 [0.91, 1.83] 1.29 [0.95, 1.74] 1.39 [0.92, 2.08] 1.04 [0.77, 1.40]
C3 vs C1 0.88 [0.67, 1.14] 0.41 [0.31, 0.54] † 1.15 [0.84, 1.56] 1.23 [0.96, 1.58] 1.35 [0.95, 1.94] 1.24 [0.97, 1.60]
C4 vs C1 1.31 [1.00, 1.70] 0.66 [0.49, 0.89] † 1.36 [0.99, 1.85] 0.92 [0.70, 1.21] 1.52 [1.05, 2.20] 1.12 [0.86, 1.47]
C5 vs C1 0.67 [0.41, 1.09] 0.51 [0.31, 0.83] † 2.00 [1.25, 3.20] 1.34 [0.86, 2.09] 1.01 [0.56, 1.83] 1.37 [0.89, 2.11]
C6 vs C1 0.58 [0.29, 1.19] 1.04 [0.41, 2.64] 0.91 [0.46, 1.79] 1.11 [0.50, 2.46] 3.11 [1.64, 5.90] † 2.19 [1.17, 4.10]
C3 vs C2 0.61 [0.48, 0.77] † 0.91 [0.73, 1.15] 0.89 [0.69, 1.15] 0.96 [0.76, 1.20] 0.98 [0.74, 1.29] 1.20 [0.96, 1.50]
C4 vs C2 0.90 [0.71, 1.15] 1.46 [1.13, 1.90] † 1.05 [0.80, 1.38] 0.71 [0.55, 0.92] 1.10 [0.81, 1.48] 1.08 [0.84, 1.39]
C5 vs C2 0.46 [0.28, 0.75] † 1.12 [0.70, 1.81] 1.55 [0.98, 2.45] 1.04 [0.67, 1.62] 0.73 [0.41, 1.28] 1.32 [0.86, 2.03]
C6 vs C2 0.40 [0.19, 0.86] † 2.31 [0.87, 6.10] 0.71 [0.35, 1.43] 0.86 [0.37, 2.00] 2.24 [1.16, 4.35] 2.11 [1.08, 4.12]
C4 vs C3 1.49 [1.16, 1.92] † 1.60 [1.25, 2.06] † 1.18 [0.90, 1.56] 0.74 [0.58, 0.96] 1.12 [0.83, 1.52] 0.90 [0.71, 1.15]
C5 vs C3 0.76 [0.46, 1.25] 1.23 [0.77, 1.96] 1.74 [1.10, 2.75] 1.09 [0.71, 1.67] 0.74 [0.42, 1.31] 1.10 [0.72, 1.68]
C6 vs C3 0.67 [0.32, 1.37] 2.53 [0.99, 6.44] 0.79 [0.41, 1.55] 0.90 [0.40, 2.01] 2.30 [1.23, 4.29] † 1.76 [0.94, 3.31]
C5 vs C4 0.51 [0.31, 0.84] † 0.77 [0.47, 1.25] 1.47 [0.93, 2.34] 1.47 [0.94, 2.30] 0.66 [0.37, 1.18] 1.22 [0.79, 1.89]
C6 vs C4 0.45 [0.22, 0.92] 1.58 [0.61, 4.07] 0.67 [0.34, 1.32] 1.21 [0.54, 2.73] 2.05 [1.08, 3.87] 1.95 [1.03, 3.70]
C6 vs C5 0.88 [0.38, 2.03] 2.06 [0.74, 5.71] 0.46 [0.21, 0.98] 0.82 [0.34, 2.00] 3.09 [1.40, 6.80] † 1.60 [0.77, 3.31]
Open in a new tab

Note. N = 4,777 days. C1 = Liked/Reacted Days; C2 = Moderate Engagement Days; C3 = Passive Use Days; C4 = No Use Days; C5 = Discussion Days; C6 = High Engagement Days. Cells present odds ratios (OR) with unadjusted 95% confidence intervals from the manual BCH three-step model; each contrast gives the odds of membership in the first-listed class relative to the second per unit increase in the covariate. Overall Wald tests (df = 5) jointly test each covariate’s five multinomial logit coefficients as sensitivity analyses. Bolded estimates are significant at p < .05 (unadjusted).

†

Contrast remains significant after Benjamini–Hochberg false discovery rate correction applied within covariate (15 contrasts per family).

Exposure = general alcohol-content exposure; positive/negative consequences = exposure to content depicting positive or negative alcohol consequences.

Consistent with patterns with close friends’ content, adolescents exposed to negative consequences from other followed accounts were more likely to have High Engagement Days compared with Liked/Reacted, Moderate Engagement, or No Use Days, and those exposed to positive consequences were more likely to report Passive or Moderate Engagement Days than No Use Days. Adolescents who reported general alcohol-content exposure from followed accounts were more likely to have Liked/Reacted Days compared with all other day types except High Engagement Days, and were also more likely to have No Use Days compared with Moderate Engagement or Passive Use Days.

In sensitivity analyses, overall Wald tests were significant for general alcohol-content exposure from followed accounts, Wald(5) = 43.62, p < .001, general exposure from close friends, Wald(5) = 26.46, p < .001, and negative-consequence content from close friends, Wald(5) = 14.98, p = .011, but not for the two positive-consequence covariates or negative-consequence content from followed accounts (ps = .062–.110). Pairwise contrasts involving the first three covariates were largely robust to FDR correction, whereas contrasts involving positive-consequence content and negative-consequence content from followed accounts were not (Table 3); the latter associations should therefore be considered preliminary.

4. Discussion

This study used intensive longitudinal data from a consecutive 21-day EMA period to examine how adolescents’ day-level social media behavior patterns relate to their overall exposure to alcohol-related content online. Unlike prior research focused on between-person differences in media use or alcohol-content engagement (Cheng, Lim, Teixeira, et al., 2024; Erevik et al., 2018; Nagata et al., 2023; Riehm et al., 2019), our approach identifies which types of social media use days are more likely to occur among adolescents who report such exposure, offering a dynamic, in-vivo view of the associated conditions that may shape adolescents’ future behavior without assuming immediate behavioral outcomes.

Through multilevel latent class analysis, we identified six distinct types of social media use days, labeled No Use, Passive Use, Liked/Reacted, Moderate Engagement, Discussion, and High Engagement. These classes reflect both variability and structure in adolescents’ daily digital behavior, capturing a range of interaction intensity. We also examined the temporal context of the interaction and found that, on holidays and school breaks, no-use days were less likely to occur, with adolescents instead showing higher odds of passive, moderate, or high engagement. This suggests that structural breaks increase overall activity across the spectrum of use, which is consistent with prior work showing that adolescents’ social media and broader screen engagement increase during school holidays, likely reflecting less-structured days that afford more discretionary screen time (Olds et al., 2019). By contrast, weekends were characterized by a relative shift from moderate engagement toward complete disengagement, indicating that some adolescents may “log off” more fully at the end of the week. While the study’s analytic focus was on day-level associations, person-level patterns also revealed meaningful trends: over 50% of adolescents shifted between four or more day-level classes across the study period, yet the majority spent more than half their days in a single dominant class, suggesting individualized but somewhat stable engagement routines.

Despite this heterogeneity in daily engagement, exposure to alcohol-related content on social media platforms was common, which is consistent with prior research (Erevik et al., 2018; Russell et al., 2022; Zhang & Esser, 2024). Across the study period, 74.15% of adolescents reported encountering alcohol-related posts from accounts they followed, and 44.56% from close friends. Positive depictions of alcohol consequences were more prevalent than negative ones from both sources (positive: 59.52% followed, 34.35% friends; negative: 38.1% followed, 21.09% friends), and a substantial minority encountered both valences (32.99% followed, 17.35% friends). Across all content types, followed accounts were consistently more frequent sources of exposure than close friends. These findings underscore the pervasiveness of alcohol content in adolescents’ social media environments, even among those with fluctuating or lower-intensity use.

The pairwise contrasts were exploratory; accordingly, we focus on broader patterns across exposure sources and interpret them in light of the sensitivity analyses. We found that days characterized by more active interaction, such as liking, commenting, sharing, and clicking on affiliate links, were especially likely to occur among adolescents who reported exposure to alcohol-related consequences from close friends. For negative consequences, High Engagement Days occurred more often than Passive Use, Discussion, and Liked/Reacted Days, a pattern that was robust in sensitivity analyses, whereas for positive consequences, greater odds of Discussion Days were suggested compared with High Engagement, Passive, and Liked/Reacted Days. One possible explanation is that negative content may draw adolescents to engage more actively online (e.g., liking, commenting, or sharing) either to signal disapproval, process risk-related messages, or simply because the content is emotionally salient and attention-grabbing. By contrast, positive content from close friends may be more likely to spark conversation and discussion, consistent with prior work showing that adolescents are especially attuned to peers’ posts that depict alcohol as socially rewarding (Erevik et al., 2017). From a Social Cognitive Theory perspective, exposure to content from close friends may be especially influential, as adolescents are more likely to observe, attend to, and learn from models who are socially similar and personally meaningful (Cristello et al., 2024).

A similar, though less robust, pattern was suggested for followed accounts. Adolescents exposed to content depicting negative alcohol consequences from accounts they followed, but were not close friends with, were more likely to have High Engagement Days. Such negative content may be especially emotionally salient, prompting adolescents to respond more actively through liking, commenting, or sharing. In contrast, those exposed to positive consequences from followed accounts were more likely to report Passive or Moderate Engagement Days than No Use Days, suggesting that positive portrayals may be integrated into routine online interactions and acknowledged without eliciting intensive engagement. However, apart from the association between close friends' negative-consequence content and High Engagement Days, these valence-specific associations did not withstand the sensitivity analyses correcting for multiple comparisons and should be considered preliminary and hypothesis-generating. In addition, adolescents reporting general alcohol-content exposure from followed accounts were more likely to have Liked/Reacted Days than all other day types except High Engagement Days. This suggests that algorithmic or platform-driven exposure may be particularly tied to quick, visible forms of interaction that keep users minimally but consistently engaged. Unlike passive scrolling, liking or reacting generates a traceable signal that can reinforce content circulation and increase the likelihood of seeing similar posts in the future.

In contrast, overall exposure to alcohol-related content from friends showed a less intuitive pattern: both Moderate Engagement days and No Use days occurred more frequently during the 21-day period in which alcohol-related content was reported, compared to Passive Use or Discussion days. Likewise, adolescents reporting general exposure from followed accounts were also more likely to have No Use Days compared with Moderate Engagement or Passive Use Days. Given all participants are active social media users, No Use days reflect occasional breaks in engagement rather than abstinence from social media. This pattern suggests that temporary disengagement does not reduce adolescents’ likelihood of encountering such content over time. Exposure may instead be driven by platform-level factors, such as prior engagement history, interest-based algorithms, or feed structures that resurface content across multiple days, rather than direct behavioral activity on a given day.

4.1. Strengths, Limitations, and Future Directions

By characterizing distinct patterns of daily engagement rather than overall time or frequency of use, this approach responds to growing calls to move beyond global frequency-based metrics and toward more specific behaviors and contexts of social media use (Orben et al., 2024; Schønning et al., 2020). This study provides novel insights into how adolescents’ daily social media behaviors are associated with exposure to alcohol-related content. The use of intensive ecological momentary assessment across 21 consecutive days is a key strength, as it allowed us to capture recently occurring behaviors in adolescents’ everyday environments while reducing reliance on global retrospective recall. This design also enabled examination of day-to-day variability in social media engagement, and the application of multilevel latent class analysis provided a novel analytic approach to characterize daily variability in social media engagement.

Of course, several limitations should be noted. First, although EMA offers advantages over traditional retrospective surveys by shortening the recall window and assessing behaviors in adolescents’ daily environments, social media engagement and alcohol-content exposure were still participant-reported. EMA does not eliminate recall error, missed exposures, or social-desirability bias, and adolescents may have under-disclosed alcohol-related content or behaviors because of stigma or perceived consequences (Stone et al., 2023). Such under-reporting could lead to conservative estimates of exposure prevalence and could also bias associations if under-reporting differed across patterns of social media engagement. The relatively high prevalence of reported alcohol-content exposure (e.g., 73.91% reported exposure from followed accounts) suggests that participants were willing to disclose some alcohol-related experiences, although this does not rule out selective or missed reporting. Notably, nearly half of the sample (48.31%) reported past-year alcohol use, and endorsement of the daily engagement indicators was common (Supplementary Table S1), which is difficult to reconcile with systematic under-reporting of either alcohol-related experiences or social media activity. Thus, future studies combining EMA with passive sensing, platform logs, device-level data, or other digital-trace approaches would further strengthen measurement.

Second, although participants completed multiple EMA prompts per day, we aggregated data to the day level for analysis, and platform identity was not captured for general social media use; we therefore could not determine which specific platforms were used or which content types were encountered at particular times. Future studies should consider platform-specific tracking to enhance the ecological precision of the data. Future work drawing on the richer momentary data collected in the parent study could enable more detailed characterization of the alcohol-related media content adolescents encounter. Meanwhile, at the survey level, participants who reported alcohol-related content exposure did not complete items about their social media activity on that same survey, resulting in approximately 3% of day-level data missing social media behavior reports. This limits our ability to fully characterize social media use patterns on exposure days and may bias estimates of co-occurrence between activity and exposure. However, given the relatively small number of alcohol-related surveys compared with non-alcohol surveys, the likelihood of substantial bias is low. Moreover, the six participants excluded from the analytic sample did not differ demographically from retained participants, and EMA missingness was generally unrelated to demographics apart from a slight age effect, suggesting that systematic bias at the participant level is unlikely. When aggregating to create person-level alcohol media exposure variables across the three study weeks, missing weeks were treated as ‘no exposure,’ which may slightly bias estimates but provided a conservative estimate across participants.

Third, our exposure measures did not capture all possible pathways through which adolescents may encounter alcohol-related content. We did not code exposure to alcohol content from users that the adolescents did not follow, such as recommended or algorithmically surfaced posts. Considering the growing role of non-friend and non-followed content on many platforms, future work should explicitly assess these sources to better understand the full spectrum of exposure pathways. Relatedly, our daily indicators did not capture private peer-to-peer messaging, such as reading direct messages or participating in group chats. Although items assessing private sharing and discussion partially reflected private interaction, they were not designed to capture the receipt or reading of direct messages, which are common forms of adolescent social media use and peer communication (Ali et al., 2023; Garrett et al., 2025). Consequently, our engagement classes may underrepresent private modes of use, and alcohol-related content encountered through direct messages may not be fully reflected in our exposure measures. Future work should incorporate private messaging behaviors to more fully characterize daily engagement patterns and exposure pathways. In addition, studies designed a priori to test valence-specific effects with adequate power for source-by-valence comparisons are needed to confirm the preliminary patterns observed here.

5. Conclusion

This study advances our understanding of how adolescents’ daily social media behavior patterns (e.g., liking, commenting, and sharing) are associated with exposure to alcohol-related content in their digital environments. By applying a day-level person-centered approach, we showed that alcohol content exposure is not uniformly linked to intensity of use, and that different content sources, such as close friends versus followed accounts, may relate to distinct patterns of interaction. These findings underscore the importance of moving beyond frequency-based metrics of social media use and toward more nuanced, behaviorally anchored models that capture dynamic risk contexts. Future research and prevention efforts should consider how momentary social media behaviors, rather than static user characteristics, shape adolescents’ opportunities to encounter health-relevant content in everyday life. By focusing on alcohol-related content exposure as a behavioral antecedent, this line of research can highlight an upstream point of intervention in the adolescent risk trajectory, which is distinct from, but foundational to, future substance use behavior. These findings also highlight the need for greater social media algorithmic accountability from platforms, such as transparency in recommendation systems and safeguards that prevent adolescents from being disproportionately targeted with alcohol-related content.

Supplementary Material

Supplementary_Table_S2
Supplementary_Table_S1

Acknowledgements

We thank the study participants, research assistants, and data collection team for their time and contributions.

Funding Statement

Research reported in this publication was supported by the National Institute on Alcohol Abuse and Alcoholism of the National Institutes of Health under award R01AA027968 (PI: Jackson) and by the National Institute on Drug Abuse under awards R01DA057588 (PI: Lanza) and K99DA064782 (PI: Tao). The content is solely the responsibility of the authors and does not necessarily represent the official views of the National Institutes of Health.

Biographies

Xiangyu Tao, Ph.D., is a postdoctoral fellow in the Department of Psychiatry at Rutgers Robert Wood Johnson Medical School and the Rutgers Addiction Research Center. Her research uses intensive longitudinal, computational, and community-engaged methods to examine how digital and social environments shape substance use, recovery, and health disparities.

María Eugenia Contreras-Pérez, Ph.D., is a postdoctoral associate at Rutgers University whose research focuses on substance use, motivational interventions, and health disparities among adolescents and young adults. Her work integrates quantitative and qualitative methods to understand treatment mechanisms and develop more responsive, equitable interventions for clinically underserved populations.

Alex Clement is a PhD candidate in the Department of Clinical and Health Psychology at the University of Florida and is completing his clinical residency at Brown University. His research examines media and technology use as contributors to adolescent risk behavior and develops novel digital tools for research and intervention.

Tim Janssen, Ph.D., investigates the development of alcohol use initiation and escalation in adolescents and young adults. Specifically, his research examines momentary changes in individual characteristics as well as context effects on the decision to start drinking and to keep drinking.

Joy Gabrielli, Ph.D., is a clinical child psychologist and Associate Professor at the University of Florida’s College of Public Health and Health Professions. Her research examines how media and technology shape cognitive, social, and emotional development in youth, informing evidence-based practices and public policy to support resilient digital futures.

Suzanne M. Colby, Ph.D., is a Professor of Behavioral and Social Sciences at Brown University’s Center for Alcohol & Addiction Studies. She is interested in factors associated with early substance use and progression to substance use disorders in youth and young adults, and in developing and testing interventions for youth.

Tyler Wray, Ph.D., is an Associate Professor of Behavioral and Social Sciences at the Brown University School of Public Health. His work explores how technology influences addictive behaviors and their determinants, how technology use supports or undermines recovery, and how technology can be designed to improve outcomes in addiction treatment.

Matthew Meisel, Ph.D., is an Assistant Professor in the Department of Behavioral and Social Sciences at the Center for Alcohol and Addiction Studies within the Brown University School of Public Health. His research examines how social factors relate to individuals’ substance use, particularly alcohol use.

Stephanie Lanza, Ph.D., is a Professor of Biobehavioral Health at The Pennsylvania State University. She holds a PhD in Human Development and Family Studies and a master’s degree in Applied Statistics from Penn State. Her research advances and disseminates quantitative methods for addressing substance use and addiction.

Kristina M. Jackson, Ph.D., studies the developmental course of substance use among adolescents and young adults and the individual- and contextual-level risk factors for initiation and progression. Her work examines in vivo exposure to alcohol-related digital media content and the social and cognitive mechanisms linking this exposure with youth drinking.

Footnotes

Declaration of Interest

The authors declare no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

Ethical Approval Statement

All study procedures were reviewed and approved by the Institutional Review Board at Brown University (Protocol #: 2009002813). The study was conducted in accordance with the ethical standards of the Declaration of Helsinki and relevant institutional guidelines.

Participant Consent Statement

Informed consent was obtained from all participants prior to participation. For participants under the age of 18, parental or guardian consent and adolescent assent were obtained in accordance with Institutional Review Board requirements.

Data Availability Statement

The data that support the findings of this study are deposited in the National Institute on Alcohol Abuse and Alcoholism Data Archive (NIAAADA), held within the National Institute of Mental Health Data Archive (NDA Collection #3698: Jackson, K. M., Understanding Adolescent In-Vivo Exposure to Alcohol Content in the Media; https://nda.nih.gov/edit_collection.html?id=3698). Data will be made available to qualified researchers in accordance with NDA data access policies and the collection’s data sharing schedule. The data are not publicly available without restriction because they contain information that could compromise the privacy of adolescent research participants. Mplus syntax for the multilevel latent class analyses is available at https://www.latentclassanalysis.com/.

Reference

  1. Ali S, Razi A, Kim S, Alsoubai A, Ling C, Choudhury M, Wisniewski P, & Stringhini G (2023). Getting Meta: A Multimodal Approach for Detecting Unsafe Conversations within Instagram Direct Messages of Youth. Proceedings of the ACM on Human-Computer Interaction, 7, 1–30. 10.1145/3579608 [DOI] [Google Scholar]
  2. Asparouhov T, & Muthén B (2012). Using Mplus TECH11 and TECH14 to test the number of latent classes. Mplus Web Notes, 14(22), 1–17. [Google Scholar]
  3. Asparouhov T, & Muthén B (2014). Auxiliary variables in mixture modeling: Using the BCH method in Mplus to estimate a distal outcome model and an arbitrary secondary model. Mplus Web Notes, 21(2), 1–22. [Google Scholar]
  4. Balcha AR, & Chen H-L (2025). Child-Computer Interaction Dynamics: A Systematic Review of Social Media Embedded Interactive Tools and Their Psychological Impacts. International Journal of Human–Computer Interaction, 41(23), 14990–15006. 10.1080/10447318.2025.2492297 [DOI] [Google Scholar]
  5. Bandura A (1986). Social foundations of thought and action. Englewood Cliffs, NJ, 1986(23–28), 2. [Google Scholar]
  6. Bandura A (2001). Social Cognitive Theory of Mass Communication. Media Psychology, 3(3), 265–299. 10.1207/S1532785XMEP0303_03 [DOI] [Google Scholar]
  7. Bolck A, Croon M, & Hagenaars J (2004). Estimating latent structure models with categorical variables: One-step versus three-step estimators. Political Analysis, 12(1), 3–27. [Google Scholar]
  8. Celeux G, & Soromenho G (1996). An entropy criterion for assessing the number of clusters in a mixture model. Journal of Classification, 13(2), 195–212. [Google Scholar]
  9. Cheng B, Lim CCW, Rutherford BN, Huang S, Ashley DP, Johnson B, Chung J, Chan GCK, Coates JM, Gullo MJ, & Connor JP (2024). A systematic review and meta-analysis of the relationship between youth drinking, self-posting of alcohol use and other social media engagement (2012–21). Addiction (Abingdon, England), 119(1), 28–46. 10.1111/add.16304 [DOI] [PubMed] [Google Scholar]
  10. Cheng B, Lim CCW, Teixeira JP, Gullo MJ, Chan GCK, & Connor JP (2024). The relationship between young people, social media use and alcohol use: A prospective cohort study. Drug and Alcohol Dependence, 265, 112478. 10.1016/j.drugalcdep.2024.112478 [DOI] [PubMed] [Google Scholar]
  11. Corcoran E, Doucette H, Merrill JE, Pielech M, López G, Egbert A, Nelapati S, Gabrielli J, Colby SM, & Jackson KM (2024). A qualitative analysis of adolescents’ perspectives on peer and influencer alcohol-related posts on social media. Drug and Alcohol Review, 43(1), 13–27. 10.1111/dar.13720 [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Cristello JV, Litt DM, Sutherland MT, & Trucco EM (2024). Subjective norms as a mediator between exposure to online alcohol and marijuana content and offline use among adolescents. Drug and Alcohol Review, 43(1), 111–121. 10.1111/dar.13620 [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Crone EA, & Dahl RE (2012). Understanding adolescence as a period of social–affective engagement and goal flexibility. Nature Reviews Neuroscience, 13(9), 636–650. 10.1038/nrn3313 [DOI] [PubMed] [Google Scholar]
  14. Curtis BL, Lookatch SJ, Ramo DE, McKay JR, Feinn RS, & Kranzler HR (2018). Meta‐analysis of the association of alcohol‐related social media use with alcohol consumption and alcohol‐related problems in adolescents and young adults. Alcoholism: Clinical and Experimental Research, 42(6), 978–986. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Ehrenreich SE, George MJ, Burnell K, & Underwood MK (2021). Importance of Digital Communication in Adolescents’ Development: Theoretical and Empirical Advancements in the Last Decade. Journal of Research on Adolescence, 31(4), 928–943. 10.1111/jora.12643 [DOI] [PubMed] [Google Scholar]
  16. Engel E, Gell S, Heiss R, & Karsay K (2024). Social media influencers and adolescents’ health: A scoping review of the research field. Social Science & Medicine, 340, 116387. 10.1016/j.socscimed.2023.116387 [DOI] [PubMed] [Google Scholar]
  17. Erevik EK, Pallesen S, Andreassen CS, Vedaa Ø, & Torsheim T (2018). Who is watching user-generated alcohol posts on social media? Addictive Behaviors, 78, 131–137. 10.1016/j.addbeh.2017.11.023 [DOI] [PubMed] [Google Scholar]
  18. Erevik EK, Torsheim T, Andreassen CS, Vedaa Ø, & Pallesen S (2017). Disclosure and Exposure of Alcohol on Social Media and Later Alcohol Use: A Large-Scale Longitudinal Study. Frontiers in Psychology, 8. 10.3389/fpsyg.2017.01934 [DOI] [PMC free article] [PubMed] [Google Scholar]
  19. Garrett SL, Shipkova M, Prinstein MJ, Telzer EH, & Lindquist KA (2025). Positive and Negative Sentiment in Social Media Direct Messages Predicts Negative Emotion Differentiation Among Adolescents. Affective Science, 6(2), 272–279. 10.1007/s42761-025-00302-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Helbig KA, Wright SJ, Derieux JR, Schrieber SR, & Radley KC (2019). Behavioral interventions. In Handbook of behavioral interventions in schools: Multi-tiered systems of support (pp. 14–28). Oxford University Press. 10.1093/med-psych/9780190843229.003.0002 [DOI] [Google Scholar]
  21. Hendriks H, van den Putte B, & Gebhardt WA (2018). Alcoholposts on Social Networking Sites: The Alcoholpost-Typology. Cyberpsychology, Behavior, and Social Networking, 21(7), 463–467. 10.1089/cyber.2017.0729 [DOI] [PubMed] [Google Scholar]
  22. Hendriks H, Wilmsen D, Van Dalen W, & Gebhardt WA (2020). Picture me drinking: Alcohol-related posts by Instagram influencers popular among adolescents and young adults. Frontiers in Psychology, 10, 2991. [DOI] [PMC free article] [PubMed] [Google Scholar]
  23. Jackson K, Gabrielli J, Colby S, Wray T, Janssen T, Rogers M, & Delapaix C (2024). Real-Time Exposure to Alcohol Content in Digital Media in Adolescents: Protocol for a Multi-burst Ecological Momentary Assessment Study. JMIR Research Protocols, (forthcoming/in press). 10.2196/50649 [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Jackson KM, Gabrielli J, Colby SM, Wray TB, Rogers ML, Janssen T, Corcoran E, Nelapati S, Clement A, & Meisel M (2026). Characterizing in vivo exposure to alcohol content in the media: Media platform, engagement, and associations with attitudes. Alcohol: Clinical and Experimental Research, 50, e70311. 10.1111/acer.70311 [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Jackson KM, Janssen T, Cox MJ, Colby SM, Barnett NP, & Sargent J (2021). Mechanisms Underlying Associations between Media Alcohol Exposure, Parenting, and Early Adolescent Drinking: A Moderated Sequential Mediation Model. Journal of Youth and Adolescence, 50(9), 1896–1910. 10.1007/s10964-020-01373-0 [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Kurten S, Vanherle R, Beullens K, Gebhardt WA, van den Putte B, & Hendriks H (2022). Like to drink: Dynamics of liking alcohol posts and effects on alcohol use. Computers in Human Behavior, 129, 107145. 10.1016/j.chb.2021.107145 [DOI] [Google Scholar]
  27. Muthén LK, & Muthén BO (2017). Mplus (Version 8) [Computer software]. [Google Scholar]
  28. Nagata JM, Smith N, Zamora G, Sajjad OM, Ganson KT, Testa A, & Jackson DB (2023). Problematic social media use and alcohol expectancies in early adolescents. BMC Public Health, 23(1), 430. 10.1186/s12889-023-15298-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Olds T, Maher C, & Dumuid D (2019). Life on holidays: Differences in activity composition between school and holiday periods in Australian children. BMC Public Health, 19(2), 450. 10.1186/s12889-019-6765-6 [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Orben A, Meier A, Dalgleish T, & Blakemore S-J (2024). Mechanisms linking social media use to adolescent mental health vulnerability. Nature Reviews Psychology, 3(6), 407–423. 10.1038/s44159-024-00307-y [DOI] [Google Scholar]
  31. Ramaswamy V, DeSarbo WS, Reibstein DJ, & Robinson WT (1993). An empirical pooling approach for estimating marketing mix elasticities with PIMS data. Marketing Science, 12(1), 103–124. [Google Scholar]
  32. Riehm KE, Feder KA, Tormohlen KN, Crum RM, Young AS, Green KM, Pacek LR, La Flair LN, & Mojtabai R (2019). Associations Between Time Spent Using Social Media and Internalizing and Externalizing Problems Among US Youth. JAMA Psychiatry, 76(12), 1266–1273. 10.1001/jamapsychiatry.2019.2325 [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Russell AM, Ou T-S, Bergman BG, Massey PM, Barry AE, & Lin H-C (2022). Associations between heavy drinker’s alcohol-related social media exposures and personal beliefs and attitudes regarding alcohol treatment. Addictive Behaviors Reports, 15, 100434. 10.1016/j.abrep.2022.100434 [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Schønning V, Hjetland GJ, Aarø LE, & Skogen JC (2020). Social Media Use and Mental Health and Well-Being Among Adolescents – A Scoping Review. Frontiers in Psychology, 11. 10.3389/fpsyg.2020.01949 [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Sidoti MF and O. (2024, December 12). Teens, Social Media and Technology 2024. Pew Research Center https://www.pewresearch.org/internet/2024/12/12/teens-social-media-and-technology-2024/ [Google Scholar]
  36. Skinner CJ, Holt D, & Smith TF (1989). Analysis of complex surveys. John Wiley & Sons Ltd. [Google Scholar]
  37. Steers M-LN, Strowger M, Tanygin AB, Ward RM, & Nolfi DA (2025). The Relationship Between Alcohol-Related Content on Social Media and Alcohol Outcomes in Young Adults: A Scoping Review. Alcohol Research : Current Reviews, 45(1), 04. 10.35946/arcr.v45.1.04 [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Stone AA, Schneider S, & Smyth JM (2023). Evaluation of Pressing Issues in Ecological Momentary Assessment. Annual Review of Clinical Psychology, 19(Volume 19, 2023), 107–131. 10.1146/annurev-clinpsy-080921-083128 [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Vanherle R, Beyens I, & Beullens K (2024). Exposure to Specific Types of Alcohol-Related SNS Content and Adolescents’ (Underage) Drinking: A Two-Wave Daily Diary Study. Media Psychology, 27(1), 50–75. 10.1080/15213269.2023.2228684 [DOI] [Google Scholar]
  40. Weller BE, Bowen NK, & Faubert SJ (2020). Latent Class Analysis: A Guide to Best Practice. Journal of Black Psychology, 46(4), 287–311. 10.1177/0095798420930932 [DOI] [Google Scholar]
  41. WHO. (2018). Global status on alcohol and health 2018. Geneva: World Health Organization 2018. [Google Scholar]
  42. Zhang L, & Esser MB (2024). U.S. Adolescents’ Exposure to Alcohol Marketing: Self-Reported Exposure on the Internet and Traditional Media. AJPM Focus, 3(5), 100243. 10.1016/j.focus.2024.100243 [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary_Table_S2
NIHMS2210282-supplement-Supplementary_Table_S2.docx (26.9KB, docx)
Supplementary_Table_S1
NIHMS2210282-supplement-Supplementary_Table_S1.docx (24.8KB, docx)

Data Availability Statement

The data that support the findings of this study are deposited in the National Institute on Alcohol Abuse and Alcoholism Data Archive (NIAAADA), held within the National Institute of Mental Health Data Archive (NDA Collection #3698: Jackson, K. M., Understanding Adolescent In-Vivo Exposure to Alcohol Content in the Media; https://nda.nih.gov/edit_collection.html?id=3698). Data will be made available to qualified researchers in accordance with NDA data access policies and the collection’s data sharing schedule. The data are not publicly available without restriction because they contain information that could compromise the privacy of adolescent research participants. Mplus syntax for the multilevel latent class analyses is available at https://www.latentclassanalysis.com/.

RESOURCES