Do engagement metrics reflect educational value in short-form diabetes videos?

To the Editor,

We read with great interest the study by Tao et al., which addresses an increasingly important question in digital diabetes education by evaluating type 1 diabetes mellitus (T1DM)-related short videos across Bilibili and TikTok. ¹ The platform-comparative design, inclusion of 200 videos, and use of multiple assessment instruments are important strengths. The study is also clinically relevant because individuals living with T1DM frequently rely on rapidly accessible information for day-to-day self-management decisions, particularly where specialist educational support is limited. Several aspects of the analysis, however, deserve closer consideration.

The interpretation of the Video Information and Quality Index (VIQI) findings is not straightforward because one VIQI subitem was partly derived from the number of likes, while likes were also used as the primary regression outcome. In the study, higher likes contributed directly to a higher VIQI-1 score, and the total VIQI score subsequently emerged as the strongest positive predictor of likes in the negative binomial model. This introduces overlap between predictor construction and outcome definition. Consequently, the association between VIQI and engagement may not fully reflect an independent relationship between audiovisual quality and audience response. ² This distinction matters because the inference that production quality drives dissemination may be overstated if part of the association is structurally embedded within the scoring approach. ³

The adjusted finding that higher Global Quality Scale (GQS) scores predicted fewer likes also requires cautious interpretation. In the correlation analysis, Global Quality Scale (GQS), Video Information and Quality Index (VIQI), and modified DISCERN (mDISCERN) were positively associated with engagement metrics. In contrast, the multivariable regression showed a strong inverse association for Global Quality Scale (GQS) after adjustment. The study’s sensitivity analyses identified substantial collinearity among these instruments, reflected by elevated Variance Inflation Factor (VIF) values and overlapping contributions within the Principal Component Analysis (PCA). Under such conditions, coefficient inversion may arise from instability in partitioning shared variance rather than a true negative relationship between educational quality and engagement. ⁴ The findings therefore support a mismatch between educational rigor and platform performance but may not establish a direct suppressive effect of quality on interaction.

The interpretation of transparency differences between platforms also appears more nuanced than the Journal of the American Medical Association (JAMA) score contrast suggests. The scoring framework captures authorship, attribution, disclosure, and currency as visible content elements. On short-video platforms, these elements are often distributed across captions, profiles, or linked metadata rather than embedded within the video. ⁵ Differences in platform interface may therefore influence observed scores, complicating direct comparison. ⁶

The thematic distribution further highlights interpretive considerations. Videos centered on psychological support and experience sharing were more prevalent on TikTok yet scored lower on conventional quality metrics. For T1DM, however, psychosocial coping and self-management confidence are integral to outcomes, ⁷ suggesting that educational value may not be fully captured by instruments emphasizing informational completeness. ⁸

We commend the authors for addressing an important aspect of digital diabetes communication. Future work should separate engagement from quality metrics and adopt platform-sensitive, short-video–specific evaluation tools.

ORCID iD

Neeraj Singh https://orcid.org/0009-0002-6917-5813

Data Availability Statement

Not applicable, as no data were generated or analyzed in this study*