Among adolescents with JIA or SLE who were members of the Childhood Arthritis and Rheumatology Research Alliance (CARRA) Registry [45,46] and enrolled in a prospective multisite study to clinically validate PRO measures [33], we investigated associations between social media use and willingness to share social media data for research. Survey reports were collected via a tablet computer by using the REDCap (Research Electronic Data Capture; Vanderbilt University) secure web application [47,48] at regularly scheduled clinics visits, during which PRO and clinical data were also collected.

A small stipend was provided to participants in the form of a US $20 gift card. Trained research assistants obtained in-person informed assent and assigned participants a unique study ID that was linked to the ID used for the clinical validation study and registry to ensure confidentiality. The study protocol was reviewed and approved by the Boston Children’s Hospital institutional review board (protocol number: IRB-P00025665).

Adolescents were eligible if they were members of the CARRA Registry, were diagnosed with JIA or SLE, enrolled in the parent prospective clinical validation study [33], and were at 1 of 3 validation study sites that participated in this substudy. Additional eligibility criteria were an age of 13 to 18 years, the ability to complete the survey in English on a tablet computer, and the reported use of at least 1 of 4 popular social media platforms (Facebook, Twitter, Instagram, or Snapchat) in the past 30 days. Patients were ineligible if they were medically or emotionally unstable or were otherwise unable to assent, as determined by a clinician or site research team member; were unable to speak or read English at an eighth-grade reading level; or did not attend the data collection visit (absent at recruitment).

Of the 145 patients approached, 123 consented (84.8%), of whom 6 were excluded because they were absent during data collection or reported Patient-Reported Outcomes Measurement Information System (PROMIS) measures at a time that did not overlap social media data collection. Of the remaining 117 patients, 5 did not use Facebook, Twitter, Instagram, or Snapchat and were excluded, leaving an analytic sample of 112 (91.1%; Figure 1).

Clinical and demographic data were drawn from the CARRA Registry. PRO and social media survey data were collected electronically by using wireless touch screen tablets during the clinic visit. Social demographic measures included race, Hispanic ethnicity, sex at birth, date of birth, insurance status, and the highest education attained by a parent. Clinical characteristics included the study recruitment and treatment site, the Physician Global Assessment (PGA) score [49], a 10-point visual analog scale score (a value of ≥1 represented active disease), BMI, and disease duration in months.

PROMIS Pediatric measures [50,51] included short-form measures of fatigue, mobility, pain interference, depressive symptoms, anxiety, and meaning and purpose, which were administered by using computer-assisted technology [10,33]. Higher PROMIS symptom T-scores reflect worse symptom levels, and higher functioning scores reflect better functioning. PROMIS measures are designed such that the mean score of the relevant reference population (ie, healthy youth) is 50, with an SD of 10 [52]. A 3-point difference in the PROMIS Pediatric T-score metric is considered a minimally important difference [53].

Willingness to share social media data via direct observation by the research team was assessed for Facebook, Instagram, Twitter, and Snapchat. Willingness measures are summarized in Table 1, along with related measures of motivation to share social media data, reasons for not sharing, and passive and active use patterns.

Summary statistics were computed to characterize the study sample overall and by willingness to share social media data for research. The differences in demographic characteristics based on willingness to share social media data were analyzed by using appropriate statistical tests, including the Kruskal-Wallis test, 2-tailed t test, 2-sided Fisher exact test, and chi-square test. For 2 participants with JIA and 1 patient with SLE, we imputed missing values on the PGA by using the median score for their disease group and similarly imputed missing BMI values for 2 participants with JIA. In separate multivariable logistic regressions, we assessed the associations between willingness to share social media data for research (the dependent variable) and each PROMIS measure. All models controlled for age, biological sex, study site, and PGA score. The analyses were conducted by using SAS 9.4 (SAS Institute) [54]. Statistical significance was considered at P<.05.

Of the 112 participants, 98 (87.5%) were persons with JIA and 14 (12.5%) were persons with SLE. Overall, participants were aged 13 to 18 (mean 16.1, SD 1.6) years, 72 (64.3%) were female, 86 (76.8%) were White, 105 (93.8%) were non-Hispanic, and 101 (90.2%) had private insurance. For all, the average BMI was 23.2 (SD 4.3) kg/m², and the average PGA score was 0.8 (SD 1.5), indicating inactive disease.

Willingness to share social media data was reported by a majority (83/112, 74.1%) of participants. In all, 43.8% (49/112) reported viewing or reading about others with rheumatic diseases on social media, and 25% (28/112) reported posting about rheumatic disease (Table 1).

Willingness to share social media data was associated with female sex (P=.04) and greater disease activity (P=.04), which was measured as a mean PGA score (Table 2). Compared to nonsharers, sharers reported lower mobility (T-score: mean 49.0, SD 9.4 vs mean 53.9, SD 8.9; P=.02), greater pain interference (T-score: mean 45.7, SD 8.8 vs mean 40.4, SD 8.0; P=.005), more fatigue (T-score: mean 49.1, SD 11.0 vs mean 39.7, SD 9.7; P<.001), more depression (T-score: mean 48.1, SD 8.9 vs mean 42.2, SD 8.4; P=.003), and greater anxiety (T-score: mean 45.2, SD 9.3 vs mean 38.5, SD 7.0; P<.001).

In logistic regression analyses that controlled for age, sex, study site, and PGA score, each 1-unit increase in symptoms was associated with greater odds of willingness to share social media data, for measures of pain interference (Adjusted Odds Ratio [AOR] 1.07, 95% CI 1.001-1.14), fatigue (AOR 1.08, 95% CI 1.03-1.13), depression (AOR 1.07, 95% CI 1.01-1.13), and anxiety (AOR 1.10, 95% CI 1.03-1.18; Table 3).

The use of Instagram, Snapchat, Facebook, and Twitter was reported by 94.6% (106/112), 83.9% (94/112), 42.9% (48/112), and 31.3% (35/112) of participants, respectively, and poly-platform use was reported by 84.8% (95/112) of participants. More than two-fifths of participants (49/112, 43.8%) reported passive social media use, that is, reading about others with rheumatic diseases, while one-quarter (28/112, 25%) reported active social media use, that is, posting about rheumatic disease. Passive and active use patterns did not differ by age or by diagnosis; however, larger percentages of female participants than male participants reported both passive and active rheumatic disease–related social media activity (Table 4).

Among 49 youths who reported passive disease-related use of social media, there were high levels of agreement that such use is helpful for observational learning from others and for alleviating feelings of isolation (Figure 2). Similarly, many reported that passive use increases their access to understandable information and helps them feel prepared in speaking with family and their care team about rheumatic disease. Of the 28 social media users who reported that they use social media actively to post about rheumatic disease, many reported doing so to feel understood, support and connect with others with rheumatic diseases, and share experiences with family and the rheumatic disease community (Figure 3). Among all social media users, a plurality endorsed the importance of not defining themselves by their condition and not wanting others to feel badly for them because of their condition, reporting these as reasons for not posting about their condition. Others refrained from posting about their condition to avoid public disclosure, retain privacy, and protect themselves from others’ attention or because they considered their condition insufficiently serious to merit attention (Figure 4). Finally, participants who shared their social media data reported they were motivated to do so because they were interested in research or technology, out of altruism toward other patients with the same conditions, to feel valued, and to receive a stipend (Figure 5).

In this multisite cohort study centered on adolescents with rheumatic diseases, we found high levels of willingness to share social media data for health research. Contrary to our original hypothesis that willingness to share would be associated with lower levels of symptoms and better psychosocial status, sharing was associated with higher levels of disease-related symptoms, specifically pain interference and fatigue, and higher levels of depression and anxiety. Both passive and active disease-related use of social media were reported, and larger percentages of female participants than male participants engaged in these activities. Passive and active social media use patterns were motivated by participants’ goals regarding observational learning about their condition and connection to others with similar conditions, as well as by the exchange of social support. Motivations to refrain from posting about rheumatic disease reflected participants’ goals of wanting to avoid having their condition define them, remaining private, and protecting against disclosure and ridicule. Among participants who agreed to share their social media data for research (n=83), almost all did so out of interest in research (83/83, 100%), to help others with a rheumatic condition (80/83, 96%), and because they felt it was of value personally (70/83, 84%) and financially (ie, for compensation; 59/83, 71%).

A growing body of research indicates the potential for improving adolescent and young adult health behaviors and outcomes (eg, health food consumption, reduced BMI, and reduced tobacco use) through engagement with social media and related features for peer groups and messaging [55-57]. Nevertheless, understanding of the potential for using social media as a source of health information and as a platform for research engagement is constrained by the lack of insight into the differences in health status between persons who are and persons who are not willing to share their data. For participatory surveillance models, some evidence shows that greater sharing and openness exist among early adopters of research apps and among persons whose disease is better controlled [58]. Other studies have found less reticence to share digital health data for care improvement among technology users who report lower incomes when compared to those who report higher incomes [59].

This study adds to what is known about willingness to share social media, with the added advantage of assessing reports from a clinically characterized cohort whose actual sharing was directly observed, differing from studies on hypothetical willingness, which are more common [60-62]. Few studies of social media use and social media data sharing provide access to linked clinical data [42] or structured PROs, and to our knowledge, none have been undertaken among youth with rheumatic diseases. Research with teenagers is vital, since this group is assuming control over their own health care, health information, and social media, and as a group, teenagers are both heavily engaged with social media [63] and uniquely vulnerable to social influences communicated through web-based channels [58,59,64,65].

Our findings of differences between sharing and nonsharing cohorts have implications for the use of personally generated data from web-based cohorts. In this study, sharing was associated with worse health. It is not clear what explains differences in health status between sharing and nonsharing youth. It may be that youth with rheumatic diseases who are unwilling to share their social media data are more socially engaged in ways that they consider unsuitable or too sensitive to allow sharing. Alternatively, they may be less desirous of research attention if they feel well and are able to satisfy their social needs through offline means. Future work may help elucidate reasons for observed differences. Similarly, future work with youth affected by other conditions is merited to understand whether differences between sharing and nonsharing groups hold.

The findings from this study have larger implications for social media–related research. First, investigators who use social media platforms to engage adolescents in health research should be explicit about the potential for biases related to inferences when denominators are poorly specified or are unknown [41,66]. The results from this study suggest that observations about the health of youth with rheumatic diseases drawn from social media–engaged cohorts may be skewed toward describing youth with greater symptoms and worse psychosocial health. This is a limitation when the goal of a study is to understand the entirety of a patient population but may be an advantage when the aim of a study is to engage youth who are struggling. The internet and social media serve important supportive functions for youth with a chronic illness, of whom many have been disproportionately adversely affected by social isolation and the hardships of the recent COVID-19 pandemic [67]. Second, findings regarding the high value placed by youth with rheumatic diseases on obtaining social and informational support related to their condition from others on the web is revealing of the seriousness of the gaps in support available from organized health care systems, as others have reported [68]. As youth turn to social media for support to fill these gaps, it remains important to consider the accuracy and safety of information offered by web-based peers, accessibility to youth across a range of health literacy and technology access levels, and the potential for harm from exposure to misinformation. Steps for detecting and addressing this during clinical visits might include asking young patients about their need for information and support and their ability to access web-based resources, as well as providing links and pointers to reliable, vetted sources of web-based guidance. These “low-tech” strategies can be implemented among even more computationally sophisticated systems for evaluating and improving the quality and accuracy of web-based information.

This report draws from a convenience sample of youth, and generalizability is limited by several factors, including the focus on youth with rheumatic diseases who have access to and use the internet and social media. The study cohort consisted of youth who were previously enrolled in research, and our findings may not generalize to youth who refrain from or have limited access to research opportunities. Sociodemographic diversity is also limited. However, the clinical confirmation of disease status and availability of validated health measures (eg, PROMIS measures) are strengths. Nevertheless, the results may not generalize to youth at other clinical sites, a broader sample of youth with rheumatic diseases, youth with other chronic conditions, or youth who do not use social media. This study did not assess participants’ understanding of health information or their ability to discern information quality or misinformation. The cross-sectional nature of this study precludes causal interpretation.

We found high willingness to share social media data for health research among a clinically characterized cohort of adolescents with rheumatic diseases and substantial use of social media for disease-related observational learning and social connection. Differences in health status between sharing and nonsharing youth (as well as between social media users and nonusers) underscore the importance of considering the potential for biases in research results that rely on social media data and the importance of identifying opportunities to engage and improve the health of youth who may be on the web and in need of support.