Overview of the video selection process
A search on TikTok using the above keywords identified 251 relevant videos. After removing 67 irrelevant videos, 47 videos with no explanation and 14 duplicates, 123 eligible videos were included for further analysis (Fig. 1).
Search strategy and video screening procedure.
General information about the eligible videos
The general information of the eligible videos on TikTok is shown in Table 1. The included videos were all uploaded after 2019. The longest number of days the videos uploaded was 1350 days, while the newest videos were uploaded in the first two weeks of data collection, with a relative median of 376 days. Among the videos included in this study, the video duration was 54 (10, 1197) seconds. The maximum number of likes received for the included videos was 617,000, the minimum was 534, and the median was 2,066. And the number of comments received for the included videos was 219 (32, 50000). In addition, the number of collections and forwarding for these videos ranged from 71 to 34,000 and 50 to 93,000, and the median values were 507 and 588. The included schizophrenia-related videos were uploaded between November 23, 2019 and July 21, 2023. From 2019 to 2023, the uploaded videos included in this study on the social networking platform TikTok included 1 video (0.81%) uploaded in 2019, 2 videos (1.63%) in 2020, 24 videos (19.51%) in 2021, 61 videos (49.59%) in 2022, 35 videos (28.45%) in 2023. Depending on the origin of the uploader, the included videos were classified into two main categories: individual users and organizational users. Of these, individual users were subdivided into three categories: health professionals, science communicators and general users. Among organizational users, three types of sources were identified: news agencies, nonprofit organizations, and for-profit organizations. The results of the study showed that individual users published the majority of videos (n = 85, 69.11%). Among individual users, health professionals uploaded the largest percentage of videos (n = 45, 36.59%), followed by science communicators (n = 21, 17.07%), while general users uploaded the least number of videos (n = 19, 15.45%). In addition, only 38 videos were uploaded by organizational users, accounting for 30.89% of the total number of videos included. Among organizational users, news agency sources accounted for the largest number of videos (n = 16, 13.01%), followed by nonprofit organizations (n = 12, 9.76%%), while for-profit organizations uploaded the fewest number of videos (n = 10, 8.13%). Table 1 shows the detailed results of the JAMA score, GQS score and Modified DISCERN score for the videos. The results of all three scoring tools showed the largest percentage of 2’s with 57 (46.34%), 48 (39.02%) and 45 (36.59%), respectively.
Assessment of video content and source
The videos included in the study were classified into four categories based on their content: science introduction, treatment method, symptom and etiology, and causation. The completeness of the content of the videos was analyzed using radar charts and the results were as in Fig. 2A. Out of the four categories of videos with different content, the science introduction category had the largest number of 50, followed by symptoms at 30, treatment method at 28, and the least number of videos was etiology and causation at 15. In addition, the videos were categorized into six categories based on the origin of the uploader: health professionals, science communications, general users, news agencies, nonprofit organizations, and for-profit organizations. The completeness of the video sources was analyzed using a radar chart, and the results were as shown in Fig. 2B, among the six different sources, the for-profit organizations category has the lowest number of 10, the health professionals category has the highest number of 45, and the other sources are, in order, science communications with 21, general users with 19, news agencies with 16, nonprofit organizations with 16.
Analysis of the integrity of the video content and source.
The videos were divided into six categories according to the origin of the uploader: health professionals, science communications, general users, news agencies, nonprofit organizations, and for-profit organizations. The Kruskal-Wallis Wallis test was conducted to characterize the videos from these six different sources. The results showed a significant correlation between video source and duration only, at P = 0.014. Videos from for-profit organizations were 90(39–199) longer than other sources. These for-profit videos received 6266.5 (652–369000) likes, 982 (122–26000) comments, 879 (220–9398) favorites, and 879 (220–9398) shares.
However, videos from news organizations were shorter in length at 44.5 (10–141) and these news organizations received 1526.5 (648 − 66,000) likes, 197 (32–5866) comments, 463 (106–898) favorites and 289 (50–2544) shares. The video duration from other sources in order of magnitude was 76.5 (21–834) for non-profit organizations, 60 (10–1197) for general users, 54 (20–135) for science communicators, and 49 (16–181) for health professionals.
However, there was no statistically significant correlation between the video source and days of uploading (P = 0.053), indicating that there was no significant difference between the identity of the user who uploaded the video and the number of days the video was posted on the social network platform. In addition, there was no significant correlation between the source of the video and the number of likes, comments, collects, and forwarding volume, respectively, P = 0.089, P = 0.335, P = 0.052, and P = 0.144. Three scoring tools, JAMA, GQS, and Modified DISCERN, were used to assess the quality and reliability of the video information of the six sources, and after comparing the six groups of videos from different sources of the scoring variability, the test found that there was no correlation between video sources and JAMA, GQS and Modified DISCERN. This indicates that there is no significant difference in the quality of videos from different sources on TikTok (Table 2).
Assessment of video quality
The quality of the videos was evaluated using the correlation between the results of the three ratings, JAMA, Modified DISCERN, and GQS. The correlation test showed that the JAMA score was significantly positively correlated with the number of video likes (r = 0.721, P < 0.001), while it was negatively correlated with the number of video comments (r=−0.345, P < 0.001) and the GQS score (r=−0.366, P < 0.001). The Modified DISCERN score was positively correlated with the JAMA score (r = 0.273, P < 0.001) were positively correlated. In addition, there was a positive correlation between GQS scores and video duration (r = 0.378, P < 0.001), a significant positive correlation with JAMA scores (r = 0.721, P < 0.001), and a negative correlation with the number of video likes (r=−0.282, P = 0.002) and video comments (r=−0.278, P < 0.001) (Table 3). Videos with more likes and comments had higher JAMA scores and GQS scores, and longer-duration videos had higher GQS scores. However, there was no statistical correlation between GQS and Modified DISCERN scores. Also, there was no statistically significant correlation between all three GQS, JAMA, and Modified DISCERN scores and collects, forwarding, and time since the video upload date (Table 3).
