JMIR Publications

ABSTRACT

Background:

Recognizing adolescence as a critical window for establishing lifelong health and well-being underscores the necessity of access to accurate, comprehensive, and age-appropriate sexuality education. Concurrently, the ascent of video platforms as principal information sources for adolescents introduces a significant variable: the quality of content hosted on these platforms exerts considerable influence on their subsequent physical and mental health outcomes.

Objective:

This study aimed to evaluate the quality, reliability, understandability, and actionability of adolescent sexuality education videos on major Chinese platforms (Bilibili, TikTok/Douyin, Kwai), analyze associated user comment sentiment and topics, identi[*Corresponding author. E-mail address:78742728@qq.com(Duo Zhao)]fy predictors of quality/reliability, and provide recommendations.

Methods:

A cross-sectional analysis was conducted (April 2025) on the top 100 comprehensively ranked CSE videos (N=300 total) retrieved from each platform using the keyword "青春期性教育". Videos were assessed using the Global Quality Scale (GQS), modified DISCERN (mDISCERN), and Patient Education Materials Assessment Tool (PEMAT-U/A), with inter-rater reliability assessed via Cohen's kappa. A corpus of ~49,000 user comments (processed from 86,647 scraped) underwent sentiment analysis (fine-tuned RoBERTa) and topic modeling (BERTopic, yielding 29 topics grouped into 6 themes). Statistical analyses included Kruskal-Wallis H tests, Spearman correlations, and stepwise linear regressions (SPSS 27.0, P<.05).

Results:

Video quality and reliability were moderate on Bilibili and TikTok but generally poor on Kwai. Content from verified sources (Doctors, Educators, Institutional Media) demonstrated superior quality and stability compared to highly variable content from Individual Media (the predominant source type, especially on Kwai: 87%). Paradoxically, Kwai exhibited the highest user engagement despite the lowest quality scores. Understandability(PEMAT-U) was consistently the strongest positive predictor for both quality (GQS, final model Adj R²=0.383, β=0.485) and reliability (mDISCERN, final model Adj R²=0.209, β=0.319). Actionability(PEMAT-A) and video duration were also significant positive predictors. Understandability scores (PEMAT-U) were generally high (~69%), while actionability scores (PEMAT-A) were moderate-to-low (33-50%). Sentiment analysis revealed comments were predominantly Neutral (71.2%), with Negative (18.4%) significantly outweighing Positive (10.4%). Key discussion themes identified included Sources of Knowledge Acquisition, Sexual Safety/Prevention, Physiology, and Sexual Health/Practices.

Conclusions:

While online video platforms offer accessible channels for adolescent sexuality education in China, the current content landscape is characterized by moderate-to-poor quality, high variability, questionable reliability, and limited actionability, with significant platform differences often linked to verification policies and engagement dynamics. Understandability is paramount, yet high engagement does not necessarily correlate with high quality or reliability, potentially amplifying misinformation. Enhancing content quality through adherence to evidence-based frameworks like ITGSE, strengthening platform accountability (verification, algorithms), and promoting user media literacy are critical steps to effectively leverage these platforms for empowering youth, contributing to a healthier future, and supporting global goals such as the SDGs and ending the AIDS epidemic by 2030.