Introduction
Decades of research indicate that social integration – the extent to which one has meaningful, positive social ties and with many other people or a large peer group – may “get under the skin” to shape trajectories of health and disease (Holt-Lunstad et al.,
2010). For example, individuals with a greater number of friendships may have improved general health outcomes (Ho,
2016), while those who lack social support may experience elevated rates of cardiovascular disease (Czajkowski et al.,
2022). But the rise of social media has ushered in a new era of expanded social experiences, the consequences of which have yet to be fully explored (Nesi et al.,
2018a). Scholars have theorized that social experiences that take place online might not simply mirror those that take place offline; rather, they might be “transformed”— rendered uniquely intense and impactful by the social media context (Nesi et al.,
2018b). That is, the perpetual accessibility, permanence, and publicity afforded by social media platforms may amplify the psychological and physiological sequelae of the social experiences that unfold upon them. It is critical to understand the specific ways in which online social integration may predict or shape physical health outcomes, particularly for youth, among whom the use of social media is now near ubiquitous (Auxier & Anderson,
2021). Therefore, the aim of the present study is to examine how key aspects of social integration represented in posts on late adolescents’ social media may be predictors of their subsequent health outcomes as young adults.
Studies that have been conducted “offline” suggest that social integration is a particularly potent predictor of physical wellbeing for people of all ages. Strong social ties, one aspect of social integration, is associated with reduced inflammation (Uchino et al.,
2018), improved sleep (de Grey et al.,
2018), more active antiviral responses (Leschak & Eisenberger,
2019), and lower rates of all-cause mortality (Holt-Lundstad et al.,
2010). Conversely, the health risks posed by lacking social ties are comparable to the effects of smoking a pack of cigarettes a day (Holt-Lundstad et al.,
2010). Another aspect of social integration is indicated by the individual fitting in with the social norms of the broader peer group. In line with findings on the benefits of strong social ties, an emerging literature also suggests that fitting in with the peer group may predict physical health outcomes (Allen et al.,
2015). Adolescents and adults who adhere to social norms have been found to report better global health outcomes up to a decade later (Ferguson et al.,
2020), and individuals who value maintaining social harmony have been found to have lower rates of hypertension (Merz et al.,
2016). In a study of patients with ulcerative colitis, those who demonstrated an above-average level of social conformity reported experiencing less symptom severity (Boye et al.,
2008). Finally, another study found that teens who maintained unusually time-consuming romantic relationships in adolescence—in other words, those teens whose relationships were out of step with developmental norms— experienced heightened blood pressure 14 years later (Allen et al.,
2021).
The achievement of social integration may be so influential on physical health because of the evolutionary necessity of group acceptance and cooperation for survival. Social baseline theory (SBT; Beckes & Coan,
2011) posits that the human brain innately expects access to supportive others who can assist in necessary survival goals. In the actual or anticipated absence of enough supportive others, the brain perceives a heightened level of threat and acts accordingly, e.g., increasing vigilance and preparing the body for conservation of resources (Gross & Medina-DeVilliers,
2020). Sustained neural perceptions of threat have been associated with metabolic shifts such as increased basal glucose levels and alterations in immune functioning such as heightened inflammatory responses (Slavich,
2020). This may explain why having a friendship with a close other is beneficial, but if the individual remains more broadly disconnected or out of step with their peer group (i.e., not socially integrated overall), they remain at risk for poor health outcomes as has been seen in other peer contexts (Mrug et al.,
2012).
Remarkably few studies have examined the relationship between specific online social experiences and physical health; even fewer have focused specifically on markers of social integration or adherence to social norms. One issue is that social media use may involve many distinct activities, some of which tap into social integration (e.g., “liking” a friend’s picture or receiving a comment from a friend), and some of which do not tap into social integration (e.g., “liking” or clicking on links to articles that have been placed into one’s Facebook “news feed” by an algorithm). Studies that fail to distinguish between such activities may ultimately be measuring something more akin to “time spent on the internet” than social integration. These studies typically find social media use to be related to worse health outcomes (Lee et al.,
2022). However, spending more time looking at screens and spending more time online have both been independently linked to poorer health outcomes (Do et al.,
2013; Domingues-Montanari,
2017), so these studies may not provide much insight into the true effects of online social integration.
A handful of studies have examined number of Facebook friends in relation to physical health; this measure has some potential to serve as a proxy for social integration. These studies are mixed in their findings: two have found number of Facebook friends to be unrelated to physical health (Lima et al.,
2017; Shakya & Christakis,
2017), while two have found number of Facebook friends to be positively associated with measures of physical health (Hobbs et al.,
2016; Nabi et al.,
2013). The mixed findings may be attributable to the limited utility of number of Facebook friends as a stand-alone measure of social support or social integration. For example, some individuals have a large objective number of Facebook “friends” but do not have strong social ties with nor fit in with these people, and in some cases, may not even know them.
Some research has been able to examine more fine-grained data about social media friendships (Hobbs et al.
(2016)). This research was identified how many Facebook friendships a given participant had initiated (i.e., how many people the participant “friended”) and how many Facebook friendships other parties had initiated with the participant (i.e., how many people “friended” the participant). Participants who were friended more by others had lower mortality rates 2 years later. The authors hypothesized that receiving many friend requests was reflective of higher of social integration. The strong design of this study, including its longitudinal design and use of a coded, objective measure of online social experiences, underscore the study’s credibility. It seems that measures of online social integration may indeed hold important implications for physical health outcomes, but that no additional studies have sought to explore this possibility.
Similarly, there appears to be a lack of research that has sought to look at fitting in with social norms in a social media context in relation to physical health outcomes. This is somewhat surprising, as a growing literature explores norms of social media use. Researchers have been particularly interested in understanding the parameters of what is and is not appropriate to post on social media, and the social consequences of violating such norms of propriety. To this end, the term “faux pas” has been used by researchers to describe a certain type of inappropriate social media post: “items that might affect employer hiring decisions,” including photos and comments depicting or referencing nudity, firearms, illegal drugs, alcohol use, and sexual activities (Karl et al. (
2010). In a study of adults of all ages, most participants reported being “very unlikely” to post such content (mean of 1.35 on a 1–5 likelihood scale; Roulin,
2014). In a study of college students, most participants reported being “very unlikely” or “unlikely” to post such content (mean of 1.51 on a 1–5 likelihood scale; Karl et al.,
2010). Another study of college students found that participants were significantly more likely to engage in faux pas posting on Twitter than on Facebook and Instagram (Miller,
2020). However, most participants still reported being either “unlikely” or “neither likely nor unlikely” to post such content on Twitter (mean of 2.415 on a 1–5 likelihood scale). Thus, faux pas posting would indeed appear to be a non-normative activity among adults of all ages, particularly on Facebook.
Violating social norms via inappropriate online posting could hold negative social consequences for social integration. For example, one study that asked college students about the appropriateness of posting about romantic relationship drama on social media found that 74% of the participants considered such content inappropriate, and 28.9% of those participants would either block or unfriend the inappropriate posters (Roche et al.,
2015). This translates to approximately 20% of the whole sample – meaning that potentially, 20% of all college students who come across such a post would remove the poster from their social network. Another study found that approximately 59% of college students would find posts about getting drunk to be inappropriate, and 88% find posting about vomiting due to drinking to be inappropriate (Wolfer,
2017). Moreover, 22.1% percent of participants reported that they would unfriend someone who posted about vomiting due to drinking. Crucially, these studies did not ask about participants’ judgments of people who have relationship drama or who get drunk in their personal lives – they focused on judgments of people who
post about these things on social media. Collectively, this literature suggests that young adults who do not fit in with the broader peer group’s social norms about appropriate posting are at risk for low social integration. It therefore seems possible that the implications of faux pas posting would extend to the domain of physical health.
Discussion
Although offline social experiences have been linked to markers of physical health, social experiences taking place online during adolescence have yet to be considered as predictors future physical health outcomes, particularly after controlling for other likely explanatory factors. This study examined social media posts received from peers as predictors of participants’ future physical health outcomes (e.g., Interleukin-6 (inflammation), sleep problems, problems with physical functioning, and BMI) at age 28. The results of this study found that several indicators of social integration observed in posts on participants’ social media pages in their early 20 s were strong predictors of their long-term physical health across a range of outcomes 7 years later. Importantly, these results remained after accounting for potential social, psychological, and physical confounds. The results indicated two key features of the types of posts that were most predictive of future physical health. Predictive posts contained clear markers of either social ties with peers or potential deviance from peer norms about posting. Second, it was the number of different individuals making such posts on the participant’s, rather than the sheer frequency of these posts, that were most often predictive. In contrast, more general markers social media activity such as friend network size, overall number of posts received, or even the number of different individuals posting (without regard to what they were posting) were not ultimately predictive of future physical health outcomes after accounting for controls. Each of these findings is discussed in turn below.
Generally speaking, social media posts from friends that displayed evidence of a social tie or personal connection between the participants and friends, either through their verbal content or via pictures, were correlated with lower levels of IL-6, better sleep quality, and lower BMI. The current study also found that faux pas posts - featuring content deviating from peer norms about posting - were associated with participants’ future higher levels of IL-6, worse sleep quality, and lower physical functioning. This mirrors findings from a large body of research underscoring the importance of social integration for physical health. while extending these indicators of social integration to the online domain.
Indeed, social media use has become ubiquitous for young people and social integration in this domain has likely become a social competency unto itself (Politte-Corn et al.,
2023). Youth who are successfully integrated in the online domain (as shown by having strong social ties on social media) are likely to reap benefits for their physical health, because so much of their social interaction occurs online. Conversely, a failure to integrate socially in this context (as shown by having content that deviates from peer norms about social media posts) may put individuals at risk of alienation from their peers, with potentially significant consequences for their health and well-being. Young adults may also use social media as a context in which to enact the same relationship patterns (healthy or unhealthy) that they display in offline settings (Mikami & Szwedo,
2011). Research has suggested that offline and online social skills are related, though not identical (
r = 0.18; Resnik & Bellmore,
2019); this research also found positive correlations between online social skills and a sociometric measure of social media acceptance, and additional positive correlations between social media acceptance and broader peer acceptance.
However, when entered into regression models including possible alternative explanatory variables, a particular feature of the most predictive types of social media posts emerged. Although it might have been expected that simply receiving more posts indicating social ties, or conversely, more posts that deviated from peer norms, would strengthen predictions to outcomes, nearly all correlations between such “frequency” variables and outcomes were diminished in regression analyses. Instead, the most predictive variables were often those which captured the number of different individuals making a particular type of post.
When considering posts that indicate social ties (connection posts), these results are perhaps suggestive of the greater importance of broader social integration/acceptance in the online domain relative to ties with one or two particular individuals for predicting adaptative outcomes. This is an interesting possibility given recent evidence that a stronger friendship with just one individual seems generally to serve as a more robust predictor of long-term adjustment relative to self-perceived broader social acceptance (Narr et al.,
2019). Because online social media essentially maintains a historical record of recent and previous social experiences with peers that is easily reviewable, it may be that in this domain social ties are more likely to be considered broadly than individually. Perhaps benefits are gained when individuals are able to look through a social media feed and recognize examples of social ties with many others among their posts rather than a few. Indeed, research has found that while a measure of close friendship quality with another person contributed indirectly to later physical health quality, it was teens’ ability to understand, adapt to, and fit in with the norms of their larger peer social group that directly predicted future health quality (Allen et al.,
2015).
Notably, however, receiving posts that deviated from peer norms from a greater number of individuals did
not similarly predict enhanced physical outcomes, but rather predicted the opposite: poorer physical health. Interacting with a greater number of individuals online thus seems to cut two ways: positively if done in a way that demonstrates or strengths social ties, or negatively if done in a way that promotes social deviancy relative to expected peer norms for posting. Although it is unclear if interacting with more individuals in a socially deviant way online has any psychological benefits for teens (some research has suggested youth with a deviant friend have lower friendship quality overall (Poulin, Dishion, & Haas,
1999), though some has shown that while such youth have similar levels of depression as friendless youth, they are less lonely (Brendgen, Vitaro, & Bukowoski,
2000), the data point to long-term problems in the physical domain, at least for young adults. Even if there may be some short term social or psychological benefits to such interactions, the nature of these interactions is unlikely to be accepted more broadly by the peer group (Roche et al.,
2015), which could create difficulties in forming and maintaining appropriate relationships with others in the future, such as with romantic partners or co-workers. This may be particularly problematic when considering physical health outcomes, as the benefits of sustained romantic relationships for physical health are well-established, and interpersonal difficulties in the workplace might make it more difficult to sustain employment necessary to help pay for physical health care if needed (Johnson, Backlund, Sorlie, & Loveless,
2000).
Although these data help identify the types of social media posts that may be predictive of future physical health outcomes, the potential mechanisms of these predictions are somewhat less clear. The findings related to posts showing social ties (connection posts) are consistent with the idea of how building a system of positive social support may confer health benefits. Ample research has identified both loneliness (Holt-Lundstad et al.,
2010) and stress (Glaser & Kiecolt-Glaser,
2005) as predictive of physical health problems. Indicators of social ties online were often suggestive of real-world interactions between friends, suggesting that such posts may be indicators that individuals are spending time with others, thereby potentially reducing feelings of loneliness. As suggested above, the fact that social ties are indicated and available online also means that individuals may be able to access these ties even when they are alone (indeed, that may be when they are most likely to). Thus, individuals may not necessarily have to feel likely they are alone very often, as they can both be in touch with friends at almost any time via social media and can also review their messages from friends and be reminded of the ties they have with them (Allen et al.,
2014).
Taken further, to the extent that connection posts do reflect more intimate friendships, individuals who receive such posts may truly be able to call upon those friends as sources of support when stressed. These possibilities suggest that there may be psychological or social features that help account for the associations between connection posts and physical health outcomes. Interestingly, individuals receiving connection posts from a greater number of friends neither reported being less troubled by internalizing symptoms, nor having greater social competence, as compared to participants receiving these posts from fewer friends online, and their predictive effects held after considering these variables in regression models. This suggests there may be alternative processes to consider with regard to how receiving such posts might be predictive of future health. Indeed, they may be more specific, in terms of actually being able to call on others for emotional or practical support, or feeling positive about the state of one’s friendships in a way that promotes physical health through stress reduction. On this point, it should be noted that having more pictures posted to their social media page that featured same-age peers was predictive of a lower future BMI, even after controlling for baseline levels of physical attractiveness that were highly negatively correlated with BMI. This suggests the possibility that friendships indicating social ties may also feature more in-person interactions, which has the potential to include behaviors that could promote positive physical health via in-person activity.
As suggested above, although having friends who post content that deviates from peer norms may have some immediate psychosocial benefits (e.g. less loneliness), there may be limitations to it that contribute to negative physical health outcomes. Notable in the data was that receiving a greater number of posts that deviated from peer norms from others was negatively correlated with online friendship network size and showed a trend toward negative correlations with number of connection posts received from others and pictures featuring same-age peers. This suggests that although such individuals may have others with whom to communicate, they may have a smaller social network to draw from overall and be less likely to share more social ties with their friends online. In the long term, this type of online social experience may play out oppositely from that experienced by individuals with many social ties; their friends may be less reliable sources of support, and they may lack the skills needed to even request such support in the first place. Individuals who engage in more deviant conversations online may also be more likely to engage in riskier behaviors that could contribute to negative health consequences (Dishion et al.,
1996). For example, previous research has identified deviant behavior in photos online as directly predictive of individuals’ problematic alcohol use (Szwedo et al.,
2012).
Importantly, more general markers of social media activity such as friend network size, overall number of posts received, or number of different individuals posting overall, were ultimately not predictive of future physical health outcomes in regression models. Initial correlations suggested some promise for these variables, with greater friendship network size correlated with better sleep quality, and the number of different individuals posting overall correlated with every outcome in a way as to indicate more positive physical health. However, it is suspected that these associations may have been driven by receiving a larger number of connection posts in particular from among the larger number overall. Such results indicate the importance of looking beyond general indicators of online activity and instead toward specific types of interactions that youth are having online.
Beyond limitations noted above, several others merit consideration. First, this study was correlational in nature, precluding causal conclusions from being drawn from any of its findings. It is possible that other processes unaccounted for in the present study might help improve understanding of the observed associations between online social media posts and future health outcomes. Moreover, there are likely other relevant qualities of posts that could be predictive of future health outcomes (e.g., emotional support received, aggressive behavior). Although the coding system captured observed posts on the participants’ social media page (from peers and from the participant), at the time online data collection occurred there was not an easily accessible way to see what participants were posting on friends’ pages. It would be helpful to have a sense of whether or not there is evidence of a social tie on both ends (reciprocation), and if this is a significant predictor of health outcomes. Similarly, it would be useful to see if participants themselves are making posts on friends’ pages that deviated from peer norms for posting. Individuals’ own behaviors online may be the strongest predictors of their future health. More recent iterations of social media also allow for private and backchannel messaging, which undoubtedly contributes to individuals’ overall well-being, but is not typically available for others to see.
Another limitation was that we did not have baseline physical health information to include in analyses. As a result, we selected a range of social, psychological, and physical variables that have been linked to physical health in previous studies as controls. We also further probed analyses by including a post-hoc variable of historical prior illness, which did not significantly alter the pattern of results (nor did externalizing behavior or alcohol use). Still, baseline health assessments would help strengthen the study’s ability to suggest that predictors were indicative of change over time in physical health. It is also possible that participants’ actual offline behaviors (e.g., poor sleep hygiene for sleep problems) might mediate associations between online predictors and health outcomes. Future studies would benefit from greater consideration and inclusion of variables representing such behaviors to better understand the contribution of online behaviors relative to offline behaviors. The relative youth of the sample is also relevant, given that many participants may not begin to develop more significant health problems until later in their lives. As such, we included broad markers of physical health that might be more relevant to a broad age-range, including stress (IL-6), sleep, overall physical functioning, and BMI. Future studies would benefit by exploring additional health outcomes that might include more specific health behaviors (e.g., cigarette smoking, alcohol use), or other health factors (e.g. blood pressure, cholesterol), and possibly at older ages. Finally, though this study aimed to use more objective (i.e. coded) measures when possible, many control and outcome variables were self-reported and potentially subject to biases (i.e. social desirability).