Social media is widely used and constitutes an important aspect of everyday life, particularly for adolescents (1). The extensive use of social media among adolescents has caused concern about potentially negative effects on adolescents’ mental health and well-being (2). Although early research investigated the effect of overall amount of social media use on adolescents’ well-being, the current consensus is that social media use entails a wide range of activities and experiences that must be considered separately (3, 4).

One concern with social media platforms is that they might extend the areas where one can have negative experiences. The pervasiveness of smartphone ownership and social media use has transformed when and where relational aggression may occur (5), and cyberbullying has been extensively studied (6, 7). Bullying is a specific form of peer aggression that is repeated, intentional, and with the purpose of harming the victim, and where there is a power imbalance in favor of the perpetrator (8). Although social media can create new opportunities for bullying (2), traditional bullying seems to be more prevalent than cyberbullying (9). The association between bullying and negative mental health outcomes is well-established [e.g. (10–12),], and cyberbullying is also associated with negative mental health outcomes (5, 13–15). In fact, a meta-analysis showed that cyberbullying is more strongly related to suicidal ideation than traditional bullying (16). This might be because cyberbullying can include anonymity, a larger audience, and difficulty keeping distance from the bullying (17). In addition to experiences classified as bullying (i.e., meet the Olweus criteria), adolescents can have a range of other negative experiences. These include name-calling, threats, and exclusion, and are often termed cybervictimization. This term is less stringent than cyberbullying, and has been defined as “the broad range of intentional acts of aggression and victimization that occur through social media” [(18), p. 298].

Virtual communities that enable adolescents to build a sense of belonging can also exclude and ostracize many others. The phenomenon of cyberostracism (19, 20), where individuals are ignored or excluded during online interactions, leads to adolescents feeling angry and aggressive (21), and even victimized leading to feelings of unworthiness and depression (22). In a recent study of cyberostracism and well-being in a Chinese university context (23), cyberostracism undermined their well-being and their need satisfaction, which is the essential source for facilitating a fully functioning state (24). These deleterious effects of cyberostracism pose a threat both to the adolescent and to society by increasing negative emotional, behavioral, and anti-social responses (25).

In a study including almost 25,000 adolescents in Spain and Ecuador, Rodríguez-Hidalgo et al. (26) explored the prevalence of cybervictimization, and found that 9% experienced one form of cybervictimization at least once or twice a month. A study of lifetime prevalence among adolescents found that 59% had experienced at least one type of relational aggression on social media, such as the spreading of false rumors or name calling (27). In a Norwegian survey of adolescents, 31% reported that others had written nasty comments to them online, during gaming, or on social media in the past year (28). Furthermore, 26% reported that someone had been mean or bullied them, 24% that they had been excluded or prevented from participating in an online group, while 15% had been threatened, and 14% that someone had posted a photo of them that made them feel sad or angry. There were gender differences in these negative online experiences. Bullying, nasty comments, and threats were more common among 15–18 year old males than females (28). Conversely, more females than males had experienced being excluded from online groups or that someone had posted a photo of them that made them sad or angry.

Other negative interactions on social media can include unwanted attention or contact from others, as one becomes approachable to a wide range of people. A Norwegian report showed that three out of 10 13–18 year-olds have received sexual comments in the past year, increasing from 19% among 13–14 year-olds to 42% among 17–18 year-olds (28). While some found such comments exciting, over one third of females found it uncomfortable and half had blocked the person who had made the comment.

While cyberbullying has received extensive attention in research (2, 6, 7), there has been little focus on subjective negative experiences that less readily classify as bullying. Our own study found that low socioeconomic status was associated with more negative experiences on social media in a sample of adolescents (29). Another study using the same dataset found that both the amount of negative experiences and the number of different negative experiences was associated with increased alcohol use and indications of potential alcohol-related problems (30). A study by Rosenthal and colleagues found an association between different negative experiences on Facebook and symptoms of depression, which was consistent after controlling for socioeconomic status (31). Due to the pervasive use of social media among adolescents, investigating the relationship between these types of negative experiences on social media and mental health and well-being is warranted. In the present study, we assessed the association between mental health outcomes and two aspects of negative experiences on social media, namely unwanted attention and negative acts and exclusion. Specifically, we hypothesized positive associations between negative experiences on social media and symptoms of anxiety and depression, and a negative association with mental well-being. Based on previous research, we included age, gender, subjective socioeconomic status, and amount of social media use as covariates.

First, results from descriptive analyses of the included variables are presented in Table 2, using mean and standard deviation and median and interquartile range for continuous data and frequency and proportion for categorical data. In separate logistic regression models, the three outcomes (case-level depression, case-level anxiety, and the median split of mental well-being) were sequentially regressed on the variables “negative acts and exclusion,” “unwanted attention from others,” and “number of negative experiences.” The results are presented as odds ratios with 95% confidence intervals in Table 3. In addition to the results from the logistic regression models, we also present the E-values for each model in Table 3. The E-value gives an estimate of the amount of unmeasured confounding needed to be present in order to fully explain away the observed associations, conditional on the measured covariates (16). As such, the E-values gives a numerical value to the liminal additional confounding necessary to reach non-significance. The associations between negative experiences on social media and symptoms of depression, symptoms of anxiety and mental well-being score were estimated using linear regression models across the variables “negative acts and exclusion” and “unwanted attention from others” and the results are presented in Table 4. The three outcomes were converted to z-scores, to obtain y-standardized regression coefficients. Finally, the association between each of the separate variables related to negative experiences and case-level depression, case-level anxiety, and mental well-being (median split) was estimated using logistic regression models and presented in Figure 1. All regression models included age, gender, subjective socioeconomic status, and amount of social media use as covariates. Data handling and analyses were performed in Stata version 15, and statistical analyses was performed both in Stata version 15 (36) and in R (37). Tables were produced using the “gtsummary” package in R (38). E-values were calculated using the E-value command available for Stata (39).

Summary statistics of the included variables are presented in Table 2. The median age was 17 years, and 56% of the sample were female. Just over half of the sample reported medium socioeconomic status. Half of the sample reported using social media many times a day, and 70% reported using social media at least 2 h a day. Almost three out of four (74.5%) reported at least one of the negative experiences seldomly or more frequently, and the median number of different negative experiences was 3.

Reporting at least one of the negative experiences at least seldomly was associated with increased odds for case-level anxiety [Odds ratio (OR) 2.16 (95% CI: 1.74–2.69), p < 0.001] and case-level depression [OR 2.26 (1.67–3.04), p < 0.001], and lower odds for above median mental well-being [OR 0.57 (0.49–0.66), p < 0.001].

All of the three main measures of negative experiences on social media were positively associated with case-level depression (ORs ranging from 1.25 to 2.29) and case-level anxiety (ORs ranging from 1.19 to 1.91). For mental well-being, negative associations were observed (ORs ranging from 0.85 to 0.49). The E-values ranged from 1.39 to 4.01. The highest E-value was observed for the association between “negative acts and exclusion” and case-level depression, while the lowest E-value was observed for number of negative experiences and the median-split of mental well-being. See Table 3 for further details.

Both “negative acts and exclusion” and “unwanted attention from others” on social media were positively associated with symptoms of depression [0.46 standard deviations (SD) and 0.25 SD, respectively] and anxiety [0.36 and 0.24 SD, respectively], and negatively associated with score on mental well-being (−0.37 and −13SD, respectively). The strength of the observed association (i.e., “effect size”) ranged between small to medium. See Table 4 for further details.

Results from separate logistic regression models for all the specific negative experiences on social media are presented in Figure 1. Across all the variables gauging negative experiences on social media, positive association with case-level depression and anxiety were observed, while a negative association was observed with mental well-being (median split). All estimates were statistically significantly different from the reference category “never,” except for case-level depression and reporting “receive unwanted nude pictures/sexualized content” seldom, and case-level anxiety and reporting “unwanted attention from strangers” seldom. Results from crude logistic and linear regression models yielded similar results to the presented findings (data not shown).

In the present study, we examined the potential association between a range of common negative experiences on social media and mental health and well-being. Our results indicate consistent positive associations between both “negative acts and exclusion” and “unwanted attention from others” on social media and self-reported symptoms of depression and anxiety. There were also consistent negative associations between both types of negative experiences and mental well-being. The same pattern was seen for the number of different negative experiences on social media and mental health and well-being. All the investigated associations were robust to adjustments for age, gender, socioeconomic status, as well as amount of social media use. The estimates for what levels of unmeasured confounding were needed to explain away our observed associations varied from small to strong. For the association with the highest point estimate, between “negative acts and exclusion” and case-level depression, the observed odds ratio of 2.29 could be explained away by an unmeasured confounder that was associated with both the exposure and the outcome by an odds ratio of 4.01-fold each, above and beyond the measured confounders (40). Weaker confounding could not do so. Although the other E-values were lower, our overall consistent findings can be considered robust to any reasonably expected impact of unmeasured confounding. Overall, the presented results are indicative of an important relationship between experiencing negative events on social media, even presumably less severe events, and worse mental health and well-being.

Our findings are in line with a previous study investigating negative experiences on Facebook and association with depressive symptoms among young adults (31). The paper by Rosenthal and colleagues included similar experiences to what we studied, such as “meanness” and “unwanted contact.” They also found that all the included negative experiences were related to depressive symptoms. Interestingly, they found that past year negative experiences was less strongly associated with depressive symptoms compared to lifetime negative experiences, which the authors interpret as being an indication that experiencing these kinds of events in adolescence is more distressing than in adulthood. Similarly, our findings also support the finding from Niu and colleagues that there is a positive correlation between cyber-ostracism and symptoms of depression among adolescents aged 12–18 years (22). Taken together, the evidence available indicates that the day-to-day interactions and events on social media have a real-life impact on its users (22).

In the present study, we found consistent associations between a range of commonly reported negative experiences on social media and symptoms of anxiety and depression as well as reduced mental well-being. Taken together with previous findings, our findings point to a potentially important public health issue, which needs to be addressed. Health-promoting intervention programs aimed at strengthening social media literacy and prosocial interaction on social media may be one way to specifically prevent negative experiences. In addition, a strengthening of the adolescents’ resilience in the face of adversity would probably also be beneficial. Future research should try to extricate the potential causal relationship between negative experiences on social media and mental health, as well as exploring potential precipitating and intermediate factors.

The data analyzed in this study is subject to the following licenses/restrictions: explicit consent from the participants is required by the Norwegian Health research legislation and the Norwegian Ethics committees in order to transfer health research data outside of Norway. Ethics approval for this study was also dependent on storing the research data on secure storage facilities located at the Norwegian Institute of Public Health, which prevents the authors from providing the data as supplementary information. Requests to access these datasets should be directed to jens.christoffer.skogen@fhi.no.

This study involving human participants were reviewed and approved by Regional Ethics Committee (REK) in Norway (REK#65611). Written informed consent from the participants' legal guardian/next of kin was not required to participate in this study in accordance with the national legislation and the institutional requirements.

JS: conceptualization, formal analysis, and writing and editing all drafts. AA and TF: review and editing, feedback on analytical approach, and literature review. PR: review and editing and literature review. GB: review and editing and feedback on analytical approach. GH: conceptualization, review and editing, feedback on analytical approach, and literature review. All authors contributed to the article and approved the submitted version.

The work of GH was supported by Dam Foundation (grant number 2021/FO347287) while the work of JS, AA, and TF was supported by The Research Council of Norway (grant number 319845). PR and GB received no external funding for the research, authorship, and/or publication of this article.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.