Social media platforms have become increasingly popular forums for people who share similar experiences to connect. TikTok has gained traction since its creation in 2016 and has become one of the most popular platforms used by people to share their stories, request advice, and connect with others [1]. Medical content, including videos that discuss symptoms, diagnoses, and treatment of various conditions, is widespread on TikTok, including content pertaining to skin health and diseases.

Atopic dermatitis (AD) is a common chronic skin condition with a strong patient presence in social media communities on various platforms, where patients often post about their experiences and concerns related to the disease [2]. AD affects 15% to 30% of children and 2% to 10% of adults in industrialized countries [3,4]. Due to its itchy, painful, and visible nature, AD can often impact self-image and quality of life. Patients with AD have reported lower satisfaction with life scores and lower mental and physical health scores [5]. Previous studies have found that online posts from patients with dermatoses, including AD, often discuss issues such as self-image, self-confidence, and therapeutic options [6].

There are many systemic barriers to care that patients with AD may experience, including financial and geographic barriers as well as patient knowledge and beliefs [7]. In light of these potential access issues, the public nature of videos makes TikTok an easy and accessible way for patients to find information related to their disease. However, misinformation can reach patients just as easily. Studies have found that TikTok content produced by board-certified physicians is more accurate and reliable than content produced by others [8,9], but they do not produce much of the content that reaches patients. One study found that most skin care–related TikTok content was produced by laypeople, and of the most liked and most viewed skin care posts, only 2.5% were produced by board-certified dermatologists [10].

By combining accessibility and accuracy, videos produced by board-certified physicians may be a helpful avenue for patients to obtain accurate information regarding AD. Understanding how people respond to these videos by analyzing the comments they leave may be helpful to determine the utility of such content and assess whether patients are able to find supportive communities on TikTok.

Prior studies have analyzed illness-related content on other diseases on social media to better understand its purpose and utility for patients. Many of these studies have found that most illness-related content is produced by patients or loved ones, and producing content on social media can serve as an outlet for these individuals to discuss their experiences with a disease or treatment [11].

Some studies have attempted to characterize patients’ perceptions of such illness-related content and have found mixed reactions. Concern about misinformation has been cited as a primary reason for disliking health-related social media content, whereas observing empathy and shared experiences has been reported as a primary benefit [12]. This suggests that while patients have varying responses to illness-related content on social media, many appreciate the feelings of relatability and support that this content can produce.

Prior studies have found that public reactions on social media can be reasonably representative of public opinion gathered via conventional surveys [13]. Additionally, multiple studies have used comments on social media to assess and analyze public audience reactions in the context of illness-related content (eg, videos; [14,15]). Naslund et al [16] found that interactions among commenters on social media conveyed themes of hopefulness, minimizing isolation, peer support, and learning from shared experiences. This suggests that commenters may feel support from the initial video or post itself but also from interactions with other users or the video creator in the comment section.

Our study aimed to (1) characterize the audience’s online response to board-certified physicians’ TikTok content related to AD according to established protocols [14,15,17] and (2) better understand the interactions that happen among members of the audience in the comment section of this content. In doing this, we aimed to understand whether the supportive environments that have been observed as a result of patient-driven illness-related content on social media are still observed with physician-driven content related to AD.

Three unique TikTok accounts were used to search 2 separate terms (“atopic dermatitis” and “eczema”) in December 2023. Two accounts were previously established personal accounts, whereas 1 account was newly created for the purpose of the study in an attempt to mitigate the effects of a personalized algorithm. Using the “Top” video categorization on TikTok, the first 30 videos to appear under each term for each account (for a total of 180 videos) were screened for inclusion. From this initial list of 180 videos, duplicate videos (75/180, 41.7%), videos that were not in English (12/180, 6.7%), and videos that did not appear to be posted by a board-certified physician (57/180, 31.7%) were excluded, resulting in 36 videos. To confirm board certification status, we did a preliminary search of the creator’s publicly available full name to find an associated hospital, university, or clinic. On the basis of the location of their associated institution, we cross-checked their certification status with national online databases. This resulted in the use of the American Board of Medical Specialties (United States), General Medical Council (United Kingdom), or Royal College of Physicians and Surgeons of Canada (Canada) databases. Of the remaining 36 videos, 7 were further excluded through this process, and 1 additional video was excluded as the content was not related to AD, leading to 28 final videos to be analyzed. All analyzed videos were posted to the social media platform between 2020 and 2023.

All analyses were performed in Prism (version 10.2.3; GraphPad Software), and P values of less than .05 were considered statistically significant. Frequency distributions were performed on categorical video, original comment, and secondary comment data separately. Engagement rate analyses were performed using Kruskal-Wallis one-way ANOVAs. For ease of analysis, certain related comment categories were consolidated (Figure 2). “Personal experiences” and “giving advice” were combined into “sharing information.” “Tagging another user,” “off-topic,” and “humor” were combined into a single “off-topic” category.

This study was exempt from institutional review board approval as there was no involvement with human participants and only publicly available data were used.

The 28 videos analyzed (Multimedia Appendix 1) had a total view count of 98,900,200, with an average of 3,532,150 views per video (SD 9,756,600; range 21,200-51,400,000). Video length ranged from 4 to 89 seconds, with an average length of video of 40.25 seconds. Among the total videos, 7.1% (n=2) were tagged for being inclusivity focused, 7.1% (n=2) were tagged for containing sponsored content, 35.7% (n=10) were tagged for having a “stitched” format, and 53.6% (n=15) were tagged for use of a trending sound or song (Figure 3).

Most videos categorized fell into the “recommendation” category (21/28, 75%), followed by “explanation of disease” (16/28, 57.1%), “debunking misinformation” (5/28, 17.9%), and “informal or anecdotal” (5/28, 17.9%). Most videos fell under 2 categories (Table 1).

Video engagement rate ranged from 0.38 to 7.54 (Multimedia Appendix 2). Video category did not have a significant impact on engagement rate (Figure 4).

The 28 videos analyzed had a total of 30,019 comments, with comment counts on individual videos ranging from 10 to 14,700. For videos with more than 50 comments, the top 50 comments were analyzed, resulting in a total of 1149 comments included in the analysis. The length of time that a video continued to generate top replies varied substantially, and our analysis was limited by the fact that some videos were posted only weeks before the date of analysis. Nevertheless, among the videos we analyzed, the median interval between the earliest and most recent top comment was 268 (IQR 113-477) days.

Of note, 37.2% (427/1149) of the analyzed comments were made by users who self-identified as patients with AD in their comments, and 6.8% (78/1149) were made by users who self-identified as relatives or loved ones of patients with AD (Table 2).

Across all video categories, “sharing information” and “requesting advice” comments made up a larger percentage than those that were critical, off-topic, or appreciative (P<.001; Figure 5). There was no significant difference between the percentage of comments involving sharing information and requesting advice (P=.95). Comments that were categorized as “learning” made up only 4.3% (49/1149) of the total.

Of the comments related to personal experience, the percentage of negative personal experience comments trended higher than that of neutral or positive personal experience comments across all video categories. The percentage of negative personal experience comments was significantly higher than that of positive personal experience comments (P=.001; Figure 6); however, there was no significant difference between the percentage of neutral and negative personal experience comments across video categories or between the percentage of positive and neutral personal experience comments.

Video category did not have a significant impact on the distribution of secondary comment categories. Across video categories, the percentage of “recommendation” secondary comments was significantly higher than the percentage of “argumentative” (P<.001), “educational” (P<.001), and “other” (P<.001) secondary comments (Figure 7). Across video categories, the percentage of “emotional support” secondary comments was also significantly higher than the percentage of “argumentative” (P<.001), “educational” (P<.001), and “other” (P<.001) secondary comments (Figure 7). There was no significant difference between the percentage of “recommendation” and “emotional support” secondary comments (P=.96). There were also no significant differences between the percentage of “argumentative,” “educational,” and “other” secondary comments (Figure 7). The number of creator replies did not have a significant impact on the distribution of secondary comment categories (data not shown).

Our study analyzed TikTok videos retrieved through searches conducted in December 2023. Given the constantly evolving nature of social media, and TikTok in particular, our findings are limited to the period in which the videos were posted (June 2020 to December 2023) and may not reflect content from other time frames. Additionally, studies have critiqued systematic analysis of TikTok content as limited by variability in the content presented to each user based on the app’s algorithm [23]. We attempted to improve replicability by using 3 separate accounts, including 1 newly created account with no established user preferences, to search for AD-related videos. Among the “Top” videos collected by the new account, 60% of the videos (18/30) and 75% (21/28) of the video creators overlapped with the “Top” videos collected by the other 2 accounts, suggesting some similarity among video presentation across users with different personalized algorithms. However, a larger number of varying accounts would need to be studied to confirm this. It is also important to note that the “Top” videos under a search term may not accurately reflect the videos that are most likely to appear in a user’s “For You Page” or personalized feed. The “Top” videos may also not be representative of all AD content on TikTok.

Another limitation is that comment sections represent only a subset of all viewers and may not accurately reflect the general population’s perceptions of these videos. The categorizations in this study were based on interpretations of online comments, and because we did not directly interview individuals, some comments may have been misinterpreted. This is particularly challenging due to the text form nature of comments, which makes tone and intention difficult to judge. Therefore, future research could incorporate qualitative interviews with patients with AD and their loved ones to gain a deeper understanding of how they are affected by these videos and what value they derive from them.

It is also important to note that our methodology excluded all creators whose board certification status was not publicly verifiable, resulting in mostly US-based creators, which may have disproportionately excluded international creators. Additionally, the formula used to calculate the engagement rate may have resulted in inflated engagement rates, particularly for videos with many spam comments or creator replies. Finally, the subset used for interrater reliability only represented 1.7% (20/1149) of the total comment dataset and may constitute a limitation in the reliability of ratings across all samples.

Our study suggests that videos created by dermatologists on TikTok are generally well received. The comment sections appear to provide transient supportive environments where users connect over shared challenges and exchange personal experiences and recommendations. While previous studies have identified virtual support spaces, they have primarily focused on patient-generated content. Our finding that supportive interactions also occur under physician-produced videos suggests that increasing physician-created content on social media could help expand access to accurate, evidence-based information while maintaining supportive communities for those affected by an emotionally and socially distressing illness such as AD.