JMIR Publications

ABSTRACT

Background:

Social media platforms, particularly short-form videos on Instagram (known as “Reels”), have become widely used sources of health information for the general public and cancer patients. However, a widening gap exists between evidence-based clinical guidance and popular wellness narratives, often impacting patient health literacy and treatment decisions.

Objective:

This study aimed to quantitatively compare the readability and emotional sentiment of cancer-related content produced by institutional/medical accounts, wellness influencers, and patients on Instagram Reels.

Methods:

Using a newly created Instagram account to avoid preformed content algorithms affecting search results, a cross-sectional content analysis was conducted on 50 publicly available Reels (25 from board-certified clinicians/academic institutions and 25 from holistic/wellness influencers and patients). Visual text was manually transcribed, and audio was transcribed using the Microsoft Word Dictate tool, with accuracy being confirmed by the authors. Linguistic complexity was assessed via the Flesch-Kincaid Grade Level formula and Flesch Reading Ease Score, and emotional sentiment was quantified using the VADER (Valence Aware Dictionary and sEntiment Reasoner) sentiment lexicon to calculate a normalized compound score (-1.0 to +1.0). Differences between account types were evaluated using Fisher’s exact, pooled, and Satterthwaite tests, while engagement metrics (views, likes, and likes per 1,000 views) were compared using Wilcoxon rank-sum tests. All analyses were performed using SAS 9.4 with a two-sided significance threshold of p < 0.05.

Results:

Medical/Institutional Reels demonstrated a higher average reading complexity (Mean Grade Level = 8.91 ± 4.03 vs. 6.45 ± 2.47 for wellness accounts; p = 0.013), indicating approximately 9th-grade versus 6th-grade complexity. Mean VADER compound sentiment scores did not differ significantly between account types (Mean 0.10 ± 0.40 vs. 0.14 ± 0.57, p = 0.77). Wellness/Influencer accounts commanded substantially higher median like counts (5,483 vs. 360 likes; p = 0.010); however, engagement efficiency was equivalent between groups, yielding similar median like rates (21.3 vs. 20.1 likes per 1,000 views; p = 0.48).

Conclusions:

Well-meaning academic institutions and oncology clinicians lock evidence-based data behind dense clinical jargon that often exceeds the average reading level of half the U.S. population. Conversely, wellness and influencer accounts utilize highly accessible, easy-to-read plain language. This can result in the dissemination of medical misinformation that can range from moderately incorrect to dangerous. To effectively increase the spread of evidence-based cancer information and audience engagement, medical professionals should adapt short-form video strategies to prioritize easy-to-understand language, use reading-level calculators, and adopt user-trust techniques such as validation badging.