Determinants of Viewer Engagement in Health and Sports Videos: A Quantile Regression Forest Machine Learning Approach Applied to Reformer Pilates Content

This study assessed the quality, reliability, and popularity of Reformer Pilates videos on YouTube, evaluating educational value, credibility, and viewer engagement using standardized scoring while exploring factors affecting popularity. On November 1, 2024, a YouTube search for “Reformer Pilates“ identified the top 70 most-viewed videos. Videos were excluded if irrelevant, promotional, or under 25 seconds. Video quality and reliability were assessed using the Global Quality Scale (GQS), Modified DISCERN (mDISCERN), and JAMA benchmarks. Viewer engagement and popularity were measured with the Video Power Index (VPI). Factors associated with engagement were analyzed using Quantile Regression Forest modeling. Most videos (87.1%) were uploaded by individual channels, 12.9% by Pilates studios. Average GQS (3.06), mDISCERN (3.09), and JAMA (2.61) scores indicated moderate quality and reliability. Videos claiming instructor expertise had lower mDISCERN scores (p = 0.005). Studio-recorded videos had higher GQS scores than home-recorded videos, though not statistically significant. Quantile Regression Forest analysis indicated that mentioning safety information, shorter video duration, and higher GQS scores were among the factors associated with higher viewer engagement (VPI). Older videos tended to exhibit lower engagement levels, reflecting a preference for newer content. Findings highlight the importance of high-quality, concise, and safety-focused Reformer Pilates videos, as these characteristics are associated with higher engagement and popularity. Content creators may benefit from prioritizing these factors to attract and retain audiences, underscoring YouTube’s role in delivering potentially reliable Pilates content.