Abstract
Introduction
Creator-side analytics can characterize viewing behavior around surgical videos but cannot establish learning or educational quality. We compared advanced procedural and patient-facing videos from one orthopedic YouTube channel.
Methods
We performed a retrospective cross-sectional analysis of cumulative video-level YouTube Studio metrics extracted on February 5, 2026, for videos published from June 2018 through February 4, 2026. One author classified videos using predefined title and metadata criteria. Co-primary outcomes were average view duration and video-attributed net subscriber change per 1,000 engaged views. Mann–Whitney U tests and rank-biserial correlations were used.
Results
Among 262 included videos (175 advanced procedural; 87 patient-facing), advanced procedural videos were longer (median, 357 vs 60 seconds), had longer average view duration (107 vs 32 seconds), and had higher net subscriber change per 1,000 engaged views (8.04 vs 2.57; both p< .001). Patient-facing videos had higher average percentage viewed (61.16% vs 30.30%; p< .001). Post hoc sensitivity analyses preserved the direction and statistical significance of the comparative findings.
Conclusion
Within this single channel, content type corresponded to distinct and consistent within-channel engagement patterns: advanced procedural videos sustained longer absolute viewing and greater net subscriber change per 1,000 engaged views, whereas patient-facing videos were watched more completely. Because absolute and proportional engagement diverged, these dimensions should be interpreted separately. These creator-side metrics show how different orthopedic video formats are watched and how video-attributed net subscriber change varies relative to engagement across formats, providing a single-channel behavioral reference rather than a measure of educational effectiveness or learning.
Keywords: education, medical, social media, video recording, orthopedics, internet, YouTube, learning analytics
1. Introduction
The traditional “see one, do one, teach one” model of surgical education is increasingly challenged by restricted resident working hours and reduced operative exposure.1-3
Digital video platforms have consequently become embedded in surgical training, with YouTube commonly used for just-in-time procedural review and preoperative preparation.4,5 Social-media resources are also increasingly discussed within orthopedic and broader surgical education.6,7
The educational utility of open-access medical videos remains debated. Most prior investigations have evaluated public-facing popularity metrics or structured quality instruments such as DISCERN and the Global Quality Score.8-10 Highly visible procedural videos may nevertheless score poorly against validated video-assessment instruments, including the LAP-VEGaS guidelines and, for operative performance, the GOALS score.11-14
Creator-side analytics provide video-level measures of viewing duration, retention, and subscription behavior that are unavailable on public video pages.15-17 Learning-analytics and social-media studies support the use of platform-derived measures to describe behavior, while emphasizing that such measures do not directly establish learning.18,19
We therefore compared Advanced Procedural Content (APC) and Patient-Facing Content (PF) on one orthopedic YouTube channel. The objectives were to compare average view duration and video-attributed net subscriber change per 1,000 engaged views, describe proportional viewing through average percentage viewed, and characterize the channel’s audience and referral pathways. All outcomes were interpreted as behavioral platform measures rather than validated indicators of learning, educational quality, or viewer profession.
2. Materials and Methods
2.1. Study Design and Data Source
We conducted a retrospective cross-sectional analysis of cumulative creator-side analytics from one orthopedic YouTube channel. The analytic unit was an individual video. The prespecified publication window included videos published from June 2018 through February 4, 2026; cumulative video-level metrics were extracted from YouTube Studio Advanced Mode on February 5, 2026. The study was reported in accordance with the STROBE recommendations for cross-sectional studies, and a completed checklist is provided as supplementary material. 20
The raw content export contained 499 videos. Five videos published on or after February 5, 2026, were excluded before cohort formation, leaving a 494-video publication cohort. Channel-level contextual exports for temporal growth, geography, traffic source, age, and gender were obtained separately from the video-level publication cohort and were used only for descriptive context. Cohort comparisons were restricted to the 494-video publication cohort. Percentages were calculated within each source export, and totals from separate exports were not forced to match. In the geography export, reported country rows accounted for 2,766,073 of 4,132,914 full-channel views (66.9%); country shares used the export total as the denominator, and the unassigned remainder was not redistributed. In the traffic-source export, detailed source rows exceeded the export total by 1,652 views; therefore, “Other sources” was calculated as the residual after subtracting YouTube Search, External, and Suggested videos from the export total rather than by summing the remaining rows.
The raw exports contained video duration, engaged views, views, watch time, average view duration, average percentage viewed, the exported “Subscribers” field, publication date, traffic source, geography, viewer age, and viewer gender. Engaged views were retained exactly as reported by YouTube Studio and were not interpreted as unique viewers. Because the exported “Subscribers” field included a negative value in the full publication cohort, it was analyzed as video-attributed net subscriber change rather than subscribers gained. Average view duration and average percentage viewed were treated as separately exported platform metrics; video-level subscription activity is attributed to the specified video’s watch page. 21
2.2. Operational Video Classification
All 494 eligible videos were evaluated using a predefined title- and metadata-based rubric. APC comprised operative or anatomically demanding material, including surgical approaches, fixation techniques, amputations, specialist instrument use, and advanced fracture or dislocation reduction. PF comprised bedside or immediately patient-facing procedures, including simple joint reduction, injection, aspiration, and drainage procedures. Videos that were nonprocedural, nonclinical, ambiguous, or outside either definition were excluded from the primary comparison.
A predefined rule-based screen assigned 482 videos without case-specific adjudication. One author reviewed the rule outputs and manually adjudicated the remaining 12 videos because generic reduction or injection terminology did not reflect the context available in the accompanying metadata: 11 advanced fracture/dislocation or specialist-instrument videos were assigned to APC, and one complication-focused injection video was excluded. No independent second rater or interrater-reliability statistic was available. A deidentified 499-video audit, including the five post-window uploads, is supplied as Supplementary Data File 1; the operational rubric and adjudicated cases appear in Supplementary Tables S1 and S2.
2.3. Outcomes
The co-primary outcomes were (1) average view duration in seconds, obtained directly from the creator-side export, and (2) video-attributed net subscriber change per 1,000 engaged views. This metric was selected as a co-primary outcome because it captures video-attributed net subscription activity relative to engaged-view volume, complementing viewing-duration measures. Unlike playback metrics, subscription represents an explicit channel-level action. Expressing net subscriber change per 1,000 engaged views standardizes subscription activity to engagement volume and reduces dependence on absolute audience size; it does not standardize for video duration. The underlying fields are exported per video in YouTube Studio and can be extracted using the platform’s native export functions without third-party analytics tools. Average percentage viewed was a secondary proportional-viewing metric. Video duration was analyzed as a content characteristic. None of these variables was treated as a measure of educational quality, knowledge acquisition, skill transfer, clinical competence, or repeat use.
Average percentage viewed values were retained exactly as exported, including values exceeding 100% in some short videos; no capping, interpolation, or synthetic replacement was performed. All 262 APC/PF videos had complete data for duration, views, engaged views, watch time, average view duration, average percentage viewed, and net subscriber change. No imputation was required.
2.4. Statistical Analysis
No a priori sample-size calculation was performed. The primary analytic sample was a census of all publication-window videos meeting the predefined APC or PF criteria. Continuous variables were summarized as median and interquartile range (IQR). APC and PF groups were compared with two-sided Mann–Whitney U tests. Rank-biserial correlation (r_rb) quantified effect size, with positive values indicating higher ranks in APC. A Bonferroni-adjusted alpha of .025 was used for the two co-primary outcomes; secondary analyses were exploratory and used p< .05.
Three post hoc sensitivity analyses examined robustness. First, net subscriber change was recalculated per 1,000 total views rather than per 1,000 engaged views. Second, all three comparative outcomes were reanalyzed after excluding videos 60 seconds or shorter. Third, percentage viewed was reconstructed as 100 × displayed whole-second average view duration divided by current video duration to assess whether the proportional-viewing result was sensitive to video-level discrepancies between the two separately exported platform metrics. A descriptive quadrant analysis used the overall medians of average view duration and net subscriber change per 1,000 engaged views across the 262-video analytic sample; values equal to a median were assigned to the high category. Analyses were performed in Python 3.13.5 with pandas 2.2.3, NumPy 2.3.5, and SciPy 1.17.0.
Tables 1-3 summarize aggregate performance, unadjusted comparisons, and cohort accounting, respectively. Figures 1-3 were generated with Matplotlib 3.10.8 from the deidentified video-level audit. Contextual Figures 4-6 were generated from separate full-channel exports; panel B of Figure 5 used Natural Earth 1:110m Admin 0 Countries public-domain boundary data.
Table 1.
Aggregate Performance of the Included Analytic Sample Based on Video-Level Creator-Side Exports (n = 262)
| Metric | APC | PF | Total |
|---|---|---|---|
| Number of videos, n | 175 | 87 | 262 |
| Engaged views | 1,782,407 | 1,473,972 | 3,256,379 |
| Views | 1,785,392 | 1,719,722 | 3,505,114 |
| Watch time, hours | 57,569.3 | 13,190.7 | 70,760.0 |
| Video-attributed net subscriber change | 15,897 | 2,882 | 18,779 |
APC indicates Advanced Procedural Content; PF, Patient-Facing Content. Values are cumulative through the February 5, 2026 extraction for videos published through February 4, 2026.
Table 3.
Video-Level Accounting From the Full Raw Export to the Primary Analytic Sample
| Cohort stage | Videos, n | Views | Engaged views | Watch time, h | Video-attributed net subscriber change |
|---|---|---|---|---|---|
| Full raw export | 499 | 4,120,793 | 3,857,408 | 85,871.9 | 22,995 |
| Post-window uploads | 5 | 3,010 | 2,106 | 41.3 | 12 |
| Publication cohort | 494 | 4,117,783 | 3,855,302 | 85,830.6 | 22,983 |
| Included APC + PF | 262 | 3,505,114 | 3,256,379 | 70,760.0 | 18,779 |
| Other/excluded | 232 | 612,669 | 598,923 | 15,070.5 | 4,204 |
Post-window uploads were five videos published on or after February 5, 2026. “Other/excluded” refers to publication-cohort videos that did not meet the APC or PF operational definitions. Subscriber totals represent the exported video-attributed net subscriber change and may differ from channel-level totals that include subscription activity from other sources. Component rows may not sum exactly to the publication-cohort total because of rounding.
Figure 1.
Average view duration by content group. Box-and-jitter plot comparing Advanced Procedural Content (APC; n = 175) and Patient-Facing Content (PF; n = 87). Boxes show the interquartile range, central lines show medians, whiskers extend to the most extreme observation lying within 1.5 times the interquartile range, and points represent individual videos. Two-sided Mann–Whitney U test, p< .001
Figure 3.
Quadrant analysis of average view duration and video-attributed net subscriber change. Dashed lines indicate the overall medians (78 seconds and 6.10 net subscriber change per 1,000 engaged views). Values equal to a median were assigned to the high category. The analysis is descriptive and unadjusted
Figure 4.
Cumulative full-channel engaged views through February 4, 2026. The curve shows the cumulative total reported in the full-channel daily export. This contextual figure includes all videos present in the channel-level export and was not used in APC/PF inferential comparisons
Figure 6.
Full-channel traffic-source distribution. “Other sources” was calculated as the residual obtained by subtracting YouTube Search, External, and Suggested videos from the export total. This residual approach was used because the detailed traffic-source rows exceeded the export total by 1,652 views; the component rows were not forced to reconcile. Displayed percentages may not total 100% because of rounding. This contextual figure was not used in APC/PF inferential comparisons
Figure 5.
Geographic distribution of full-channel views. (A) Top 10 countries by share of total views in the geography export. (B) Reported country-level views shown on a logarithmic color scale. Reported country rows accounted for 2,766,073 of 4,132,914 full-channel views (66.9%); the unassigned remainder was not redistributed. Gray areas indicate locations without a reported geography row or without a mappable polygon in the boundary dataset. The choropleth was generated de novo in Python from creator-side YouTube Studio geography data using Natural Earth 1:110m Admin 0 Countries public-domain boundary data
All observations originated from one channel and accumulated over unequal exposure periods. Videos shared channel-level, temporal, and platform influences and therefore could not be assumed to be strictly independent. No multivariable, mixed-effects, upload-age-adjusted, or formally time-stratified model was performed. P values and effect sizes were interpreted as exploratory within-channel comparisons.
3. Results
3.1. Cohort Accounting and Channel Scale
The raw video-level export included 499 videos. After excluding five videos uploaded outside the study window, the publication cohort comprised 494 videos, totaling 4,117,783 views, 3,855,302 engaged views, 85,830.6 watch hours, and a cumulative video-attributed net subscriber change of 22,983. Of these videos, 175 were classified as APC and 87 as PF, giving an analytic sample of 262 videos; the remaining 232 videos met neither operational definition and were excluded from the group comparisons (Table 3).
The full-channel daily export recorded 3,868,061 engaged views from June 1, 2018, through February 4, 2026 (Figure 4).
This channel-level contextual time series was used only for descriptive purposes and was not used in the APC/PF inferential comparisons.
3.2. Audience and Referral Profile
In full-channel contextual exports, 79.3% of views were attributed to male viewers and 66.5% to viewers aged 25–44 years. These platform-reported categories did not contain occupation, credentials, or training status and therefore could not establish a professional audience.
India accounted for 24.3% of total views in the geography export, followed by the United States (15.8%), Türkiye (3.0%), and the United Kingdom (2.5%); reported country rows collectively accounted for 66.9% of the export total, and the unassigned remainder was not attributed to a country (Figures 5A and B). YouTube Search accounted for 58.8% of traffic-source views, followed by residual Other sources (15.3%), External sources (14.7%), and Suggested videos (11.3%) (Figure 6).
3.3. Aggregate Performance of the Analytic Sample
The 175 APC videos generated 1,782,407 engaged views, 1,785,392 views, 57,569.3 watch hours, and a cumulative video-attributed net subscriber change of 15,897. The 87 PF videos generated 1,473,972 engaged views, 1,719,722 views, 13,190.7 watch hours, and a cumulative video-attributed net subscriber change of 2,882 (Table 1). These cumulative totals were unadjusted for video count, duration, upload date, or exposure time.
3.4. Co-Primary Outcomes
APC videos had longer average view duration than PF videos (median, 107 [IQR, 73-139.5] vs 32 [IQR, 20-51] seconds; U = 13,959.5; p< .001; r_rb = .834; Figure 1 and Table 2). The net subscriber change rate was also higher for APC videos (8.04 [IQR, 5.43-11.63] vs 2.57 [IQR, 1.41-4.88] per 1,000 engaged views; U = 13,006.5; p< .001; r_rb = .709; Figure 2 and Table 2). Both co-primary comparisons remained below the Bonferroni-adjusted alpha of .025.
Table 2.
Unadjusted Comparison of Video Characteristics and Behavioral Analytics
| Metric | APC median [IQR] | PF median [IQR] | U | p | r_rb |
|---|---|---|---|---|---|
| Video duration, seconds | 357 [199.5-622.5] | 60 [16.0-120.5] | 13,897.0 | < .001 | .826 |
| Average view duration, seconds | 107 [73-139.5] | 32 [20-51] | 13,959.5 | < .001 | .834 |
| Net subscriber change per 1,000 engaged views | 8.04 [5.43-11.63] | 2.57 [1.41-4.88] | 13,006.5 | < .001 | .709 |
| Average percentage viewed, % | 30.30 [20.98-39.50] | 61.16 [45.29-95.05] | 1,962.0 | < .001 | −.742 |
All comparisons used two-sided Mann–Whitney U tests. A Bonferroni-adjusted alpha of .025 was used for the two co-primary outcomes (average view duration and net subscriber change per 1,000 engaged views). Positive r_rb values indicate higher ranks in APC.
Figure 2.
Video-attributed net subscriber change by content group. Box-and-jitter plot of net subscriber change per 1,000 engaged views for APC (n = 175) and PF (n = 87). Boxes show the interquartile range, central lines show medians, whiskers extend to the most extreme observation lying within 1.5 times the interquartile range, and points represent individual videos. Two-sided Mann–Whitney U test, p< .001
3.5. Video Duration and Proportional Viewing
APC videos were longer than PF videos (357 [IQR, 199.5-622.5] vs 60 [IQR, 16.0-120.5] seconds; U = 13,897.0; p< .001; r_rb = .826). In contrast, PF videos had higher average percentage viewed (61.16% [IQR, 45.29%-95.05%] vs 30.30% [IQR, 20.98%-39.50%]; U = 1,962.0; p< .001; r_rb = −.742) (Table 2). Thus, APC videos had longer absolute playback duration, whereas PF videos had greater proportional viewing.
3.6. Sensitivity and Quadrant Analyses
Using total views rather than engaged views as the denominator produced essentially unchanged group separation for net subscriber change (APC, 8.04 [IQR, 5.43–11.63] vs PF, 2.57 [IQR, 1.06–4.88] per 1,000 views; U = 12,994.5; p< .001; r_rb = .707). After excluding videos 60 seconds or shorter, APC retained longer average view duration (107.5 vs 52.0 seconds; U = 6,234.5; p< .001; r_rb = .746) and higher net subscriber change (8.11 vs 4.34 per 1,000 engaged views; U = 5,665.5; p< .001; r_rb = .587), whereas PF retained higher average percentage viewed (43.80% vs 30.10%; U = 1,557.0; p< .001; r_rb = −.564). Reconstructing percentage viewed from displayed whole-second average view duration and current duration also preserved higher proportional viewing in PF (61.67% vs 30.12%; U = 2,040.5; p< .001; r_rb = −.732) (Supplementary Table S3).
Using overall medians of 78 seconds and 6.10 net subscriber change per 1,000 engaged views, 95 APC and 1 PF video were in the high-duration/high-change quadrant; 69 PF and 26 APC videos were in the low-duration/low-change quadrant. The remaining videos occupied discordant quadrants (Figure 3).
4. Discussion
4.1. Principal Findings
Within this channel, APC videos had longer average view duration and higher video-attributed net subscriber change per 1,000 engaged views than PF videos, whereas PF videos had higher average percentage viewed. For educational content creators, net subscriber change per 1,000 engaged views provides a standardized description of video-attributed net subscription activity relative to engaged-view volume rather than absolute reach alone. Within a channel, this metric may help compare how subscription activity varies across content formats, without implying causal conversion or educational effectiveness. The alternative-denominator, short-video-exclusion, and reconstructed-percentage analyses preserved the direction of the principal findings. Because APC and PF videos also differed in duration, production style, and intended audience, these unadjusted contrasts should be read as composite category-level differences rather than the isolated effect of procedural complexity; these confounders are examined in Section 4.6. These results describe distinct dimensions of platform behavior only; their educational interpretation is bounded as set out in Section 4.3.
4.2. Absolute and Proportional Viewing Measures
Absolute and proportional viewing measures pointed in different directions. APC videos were substantially longer and had longer average playback, whereas PF videos had a larger proportion of their duration viewed. Average view duration should therefore not be interpreted as duration-standardized retention. Conversely, a high average percentage viewed in very short videos does not establish greater educational value. The sensitivity analysis excluding videos 60 seconds or shorter reduced but did not eliminate the between-group differences.
4.3. Engagement Is Not Educational Quality
Platform engagement is not equivalent to educational quality. Studies using validated instruments have repeatedly shown that visible or frequently viewed surgical videos may score poorly on pedagogical criteria.11-14 The present measures characterize playback and subscription behavior only. They cannot verify factual accuracy, instructional design, learner comprehension, skill acquisition, or clinical performance.
4.4. Audience and International Reach
The full-channel descriptors showed international reach. However, geography, age, and gender do not identify occupation or educational purpose. Open-access digital material may broaden access where cadaveric laboratories, simulation resources, and formal fellowships are limited, but barriers involving bandwidth, language, device access, and contextual relevance remain.22-24
4.5. Role as an Adjunct
Technique-focused videos may offer asynchronous exposure to operative anatomy and workflow, and short-form or modular resources may support preparation and review.25-28 Such material should be considered an adjunct rather than a substitute for supervised practice, structured feedback, deliberate practice, and competency-based assessment.
4.6. Limitations
This study has several limitations. First, the single-channel design limits external validity: the engagement patterns described here reflect the output, audience, and algorithmic context of a single orthopedic channel and should not be generalized to other creators, to other surgical specialties, or to other educational platforms, whose audiences, recommendation systems, and content ecosystems differ. Second, the cumulative metrics were measured at one extraction date and were affected by unequal time since upload, channel growth, recommendation-algorithm changes, and other temporal factors. Third, all videos shared channel-level influences and were not statistically independent; no multivariable, mixed-effects, upload-age-adjusted, or time-stratified model was performed. Fourth, APC and PF videos differed substantially in duration, topic complexity, and production characteristics and may also have differed in unmeasured search demand. These differences are potential confounders rather than incidental features. Duration is mechanically coupled to both absolute and proportional viewing metrics. Production style—for example, operating-room recording versus bedside demonstration—may independently influence retention and subscription behavior. These production characteristics were not coded or measured as separate variables in the present dataset. Intended audiences may also have differed, with APC material more likely to be relevant to surgeons and trainees and PF material more likely to be relevant to patients; however, viewer profession or role was not available in the platform exports. Because these factors are intertwined with the content categories themselves, the unadjusted between-group comparisons should be read as composite category-level differences rather than the isolated effect of procedural complexity. The large duration imbalance may have favored APC on absolute viewing duration, whereas PF showed the opposing proportional-viewing pattern. Excluding videos 60 seconds or shorter attenuated but did not eliminate the group differences. This sensitivity analysis does not eliminate residual confounding by duration. Fifth, average percentage viewed and average view duration were separately exported platform metrics and were not algebraically identical for every video. The reason for individual discrepancies could not be determined from the deidentified audit. However, reconstruction from the displayed whole-second average view duration and current duration preserved the direction and statistical significance of the proportional-viewing result. Sixth, the direction of residual bias from upload age, algorithm changes, and production characteristics cannot be determined from these data. Seventh, classification relied on one author’s review of titles and metadata, and no second rater or content-level validation was available. Any misclassification would most plausibly involve boundary videos, such as complex reduction procedures, and could have either attenuated or exaggerated the observed between-group differences; the direction and magnitude of such bias cannot be determined from these data. Only 12 of 494 eligible videos (2.4%) required case-specific adjudication beyond the predefined rule-based screen; however, this does not exclude misclassification among videos assigned by the rule-based screen. The complete deidentified audit (Supplementary Data File 1) permits independent inspection of the recorded classifications. Future studies should nonetheless employ at least two independent raters with a prespecified adjudication protocol and report interrater reliability, for example Cohen’s κ. Eighth, platform-reported audience categories could not verify professional status. Finally, behavioral analytics cannot establish content accuracy, learning, skill transfer, or clinical benefit. The findings are exploratory, associative, and channel-specific.
5. Conclusion
In this seven-year, single-channel cross-sectional analysis, content type was associated with distinct and consistent within-channel engagement patterns. APC videos sustained longer average view duration and showed greater video-attributed net subscriber change per 1,000 engaged views, whereas PF videos achieved higher proportional viewing; these patterns persisted across all post hoc sensitivity analyses. Together, these creator-side measures quantify how distinct orthopedic video formats are watched and how video-attributed net subscriber change varies relative to engagement across formats. They provide a single-channel behavioral reference for this orthopedic channel and generate hypotheses for controlled educational research. Because absolute and proportional engagement diverged, these two dimensions should be interpreted separately rather than merged into a single engagement claim. These behavioral findings characterize audience interaction rather than educational effectiveness and should be confirmed in multi-channel studies incorporating standardized content classification, validated quality measures, learner outcomes, upload-time adjustment, and channel-level modeling. Until such replication is available, no inference should be extended beyond this single orthopedic channel to other surgical specialties or educational platforms.
Supplemental Material
Supplemental Material for Viewer Engagement Patterns in Advanced Procedural and Patient-Facing Orthopedic YouTube Videos: A Seven-Year Single-Channel Cross-Sectional Analysis by Ahmet Aybar, Kemal Gokkus, Abdullah Sukun, Yakup Yildirim in Inquiry: The Journal of Health Care Organization, Provision, and Financing.
Supplemental Material for Viewer Engagement Patterns in Advanced Procedural and Patient-Facing Orthopedic YouTube Videos: A Seven-Year Single-Channel Cross-Sectional Analysis by Ahmet Aybar, Kemal Gokkus, Abdullah Sukun, Yakup Yildirim in Inquiry: The Journal of Health Care Organization, Provision, and Financing.
Appendix.
Abbreviations
- APC
Advanced Procedural Content
- GOALS
Global Operative Assessment of Laparoscopic Skills
- GQS
Global Quality Score
- IQR
interquartile range
- LAP-VEGaS
Laparoscopic Surgery Video Educational Guidelines
- LMICs
low- and middle-income countries
- PF
Patient-Facing Content
- STROBE
Strengthening the Reporting of Observational Studies in Epidemiology
Author Contributions: A.A.: Conceptualization, investigation, data curation, writing—original draft. K.G.: Conceptualization, methodology, formal analysis, data curation, visualization, writing—original draft, writing—review & editing, project administration. A.S.: Methodology, validation, writing—review & editing. Y.Y.: Supervision, writing—review & editing. All authors have read and approved the final version of the manuscript, and all authors meet the ICMJE criteria for authorship.
Funding: The authors received no financial support for the research, authorship, and/or publication of this article.
The authors declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: Kemal Gokkus owns and operates the orthopedic YouTube channel analyzed in this study, produced the analyzed content, and had privileged access to nonpublic creator-side analytics. The channel was monetized during the study period; he therefore has personal, professional, and financial interests in its performance. Ahmet Aybar, Abdullah Sukun and Yakup Yildirim declared no competing interests. No external party influenced the study design, analysis, interpretation, or reporting.
Declaration of Generative AI and AI-Assisted Technologies: During revision, the authors used ChatGPT (OpenAI; GPT-5.5 Thinking; accessed June–August 2026) for language editing, structural organization, consistency checking, and assistance in drafting and reviewing Python code used to reproduce analyses and figures from the authors’ supplied raw YouTube Studio exports, and Claude (Anthropic; Claude Fable 5; accessed August 2026) for consistency checking and language editing during the minor-revision stage. The authors reviewed the code, independently checked the outputs against the raw exports, verified all statistical results and references, made all scientific decisions, and accept full responsibility for the manuscript. No synthetic or altered research data were generated.
Supplemental Material: Supplemental material for this article is available online.
ORCID iD
Kemal Gokkus https://orcid.org/0000-0002-4916-3471
Ethical Considerations
Ethical approval was not required because the study used aggregated, non-identifiable platform analytics and involved no intervention or access to identifiable user-level data.
Data Availability Statement
Creator-side analytics are restricted by platform access controls and cannot be deposited publicly in full. Aggregated data supporting the reported tables and a deidentified video-level classification audit are provided as supplementary material. The separate full-channel daily, geography, and traffic-source exports underlying contextual Figures 4-6 are not included in the public supplementary dataset; additional aggregated data may be available from the corresponding author on reasonable request, subject to platform terms and applicable privacy constraints. The Figure 5B choropleth was generated de novo in Python using Natural Earth 1:110m Admin 0 Countries public-domain boundary data; no copyrighted third-party map image, tile service, or proprietary cartographic product was reproduced.
References
- 1.Reznick RK, MacRae H. Teaching surgical skills — changes in the wind. N Engl J Med. 2006;355:2664-2669. doi: 10.1056/NEJMra054785 [DOI] [PubMed] [Google Scholar]
- 2.Ahmed N, Devitt KS, Keshet I, et al. A systematic review of the effects of resident duty hour restrictions in surgery. Ann Surg. 2014;259:1041-1053. doi: 10.1097/SLA.0000000000000595. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 3.Bilimoria KY, Chung JW, Hedges LV, et al. National cluster-randomized trial of duty-hour flexibility in surgical training. N Engl J Med. 2016;374:713-727. doi: 10.1056/NEJMoa1515724. [DOI] [PubMed] [Google Scholar]
- 4.Rapp AK, Healy MG, Charlton ME, Keith JN, Rosenbaum ME, Kapadia MR. YouTube is the most frequently used educational video source for surgical preparation. J Surg Educ. 2016;73:1072-1076. doi: 10.1016/j.jsurg.2016.04.024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Sterling M, Leung P, Wright D, Bishop TF. The use of social media in graduate medical education: a systematic review. Acad Med. 2017;92:1043-1056. doi: 10.1097/ACM.0000000000001617. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Feroe AG, Only AJ, Murray JC, et al. Use of social media in orthopaedic surgery training and practice: a systematic review. JBJS Open Access. 2024;9(1):e23.00098. doi: 10.2106/JBJS.OA.23.00098 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Lima DL, Viscarret V, Velasco J, Lima RNCL, Malcher F. Social media as a tool for surgical education: a qualitative systematic review. Surg Endosc. 2022;36(7):4674-4684. doi: 10.1007/s00464-022-09150-9. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Kunze KN, Krivicich LM, Verma NN, Chahla J. Quality of online video resources concerning patient education for the meniscus: a YouTube-based quality-control study. Arthroscopy. 2020;36(1):233-238. doi: 10.1016/j.arthro.2019.07.033. [DOI] [PubMed] [Google Scholar]
- 9.Curran V, Simmons K, Matthews L, et al. YouTube as an educational resource in medical education: a scoping review. Med Sci Educ. 2020;30:1775-1782. doi: 10.1007/s40670-020-01016-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 10.Mohammadzadeh Azarabadi J, Gökkuş K, Shalalfa N. Evaluating the quality of osteonecrosis and avascular necrosis content on YouTube: a multidimensional analysis using DISCERN, JAMA, and GQS tools. J Orthop Surg Res. 2025;20:695. doi: 10.1186/s13018-025-06116-0. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 11.Gupta T, Haidery TH, Sharma R, Sharma S, Kumar A. How reliable are YouTube videos for general surgery residents learning? Cureus. 2023;15:e34718. doi: 10.7759/cureus.34718. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 12.de’Angelis N, Gavriilidis P, Martínez-Pérez A, et al. Educational value of surgical videos on YouTube: quality assessment of laparoscopic appendectomy videos by senior surgeons vs. novice trainees. World J Emerg Surg. 2019;14:22. doi: 10.1186/s13017-019-0241-6. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Ghersi A, Mansour J, Marchand P, Al Rubaie A, Kouyoumdjian P, Coulomb R. Surgical videos on the internet: is this a reliable pedagogical tool in residency training? SICOT J. 2022;8:39. doi: 10.1051/sicotj/2022039. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Çakıcı K, Köseoğlu S. Assessment of the quality of endoscopic tympanoplasty and endoscopic myringoplasty videos on YouTube. BMC Med Educ. 2025;25:1181. doi: 10.1186/s12909-025-07775-7. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Yee A, Padovano WM, Fox IK, et al. Video-based learning in surgery: establishing surgeon engagement and utilization of variable-duration videos. Ann Surg. 2020;272(6):1012-1019. doi: 10.1097/SLA.0000000000003306. [DOI] [PubMed] [Google Scholar]
- 16.Wever GS, Elliott I, McCaul J, et al. Viewership footprint for a low-resource, student-centred collaborative video platform to teach orthopaedics in southern Africa. S Afr Med J. 2020;110:532-536. doi: 10.7196/SAMJ.2020.v110i6.14348. [DOI] [PubMed] [Google Scholar]
- 17.Sui W, Sui A, Rhodes RE. What to watch: practical considerations and strategies for using YouTube for research. Digit Health. 2022;8:20552076221123707. doi: 10.1177/20552076221123707. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 18.Toofaninejad E, Dawson S, Sohrabi S, Kalantarion M. Exploring the transformative potential of learning analytics in medical education: a systematic review. J Adv Med Educ Prof. 2025;13:12-24. doi: 10.30476/jamp.2024.103973.2034. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 19.Kemppainen A, Nuutinen H, Koljonen V, Kulmala P. Instagram video engagement in medical education: cross-sectional study. JMIR Med Educ. 2026;12:e79124. doi: 10.2196/79124. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 20.von Elm E, Altman DG, Egger M, et al. The Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) statement: guidelines for reporting observational studies. Lancet. 2007;370:1453-1457. doi: 10.1016/S0140-6736(07)61602-X [DOI] [PubMed] [Google Scholar]
- 21.Google. YouTube Analytics and Reporting APIs: metrics. Google for Developers. Accessed June 15, 2026. https://developers.google.com/youtube/analytics/metrics
- 22.Meara JG, Leather AJM, Hagander L, et al. Global Surgery 2030: evidence and solutions for achieving health, welfare, and economic development. Lancet. 2015;386:569-624. doi: 10.1016/S0140-6736(15)60160-X. [DOI] [PubMed] [Google Scholar]
- 23.Brown KE, Flores MJ, MacKechnie MC, et al. Novel e-learning platform for orthopaedic training in LMICs: a descriptive review of the IGOT portal. Surg Open Sci. 2023;13:24-26. doi: 10.1016/j.sopen.2023.04.003. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 24.O’Flynn E, Ahmed A, Biswas A, Bempong-Ahun N, Perić I, Puyana JC. E-learning supporting surgical training in low-resource settings. Curr Surg Rep. 2024;12:151-159. doi: 10.1007/s40137-024-00399-8. [DOI] [Google Scholar]
- 25.Palmon I, Brown CS, Highet A, et al. Microlearning and social media: a novel approach to video-based learning and surgical education. J Grad Med Educ. 2021;13:323-326. doi: 10.4300/JGME-D-20-01562.1. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Maertens H, Madani A, Landry T, Vermassen F, Van Herzeele I, Aggarwal R. Systematic review of e-learning for surgical training. Br J Surg. 2016;103:1428-1437. doi: 10.1002/bjs.10236. [DOI] [PubMed] [Google Scholar]
- 27.Selten JW, Nazari T, Andriessen EH, Konings S, Wiggers T, de Jonge J. Standardized videos in addition to the surgical curriculum in medical education for surgical clerkships: a cohort study. BMC Med Educ. 2022;22:384. doi: 10.1186/s12909-022-03314-w. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 28.Parker AS, Steffes BC, Hill K, et al. An online, modular curriculum enhances surgical education and improves learning outcomes in East, Central, and Southern Africa: a mixed-methods study. Ann Surg Open. 2022;3:e140. doi: 10.1097/AS9.0000000000000140. [DOI] [PMC free article] [PubMed] [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
Supplemental Material for Viewer Engagement Patterns in Advanced Procedural and Patient-Facing Orthopedic YouTube Videos: A Seven-Year Single-Channel Cross-Sectional Analysis by Ahmet Aybar, Kemal Gokkus, Abdullah Sukun, Yakup Yildirim in Inquiry: The Journal of Health Care Organization, Provision, and Financing.
Supplemental Material for Viewer Engagement Patterns in Advanced Procedural and Patient-Facing Orthopedic YouTube Videos: A Seven-Year Single-Channel Cross-Sectional Analysis by Ahmet Aybar, Kemal Gokkus, Abdullah Sukun, Yakup Yildirim in Inquiry: The Journal of Health Care Organization, Provision, and Financing.
Data Availability Statement
Creator-side analytics are restricted by platform access controls and cannot be deposited publicly in full. Aggregated data supporting the reported tables and a deidentified video-level classification audit are provided as supplementary material. The separate full-channel daily, geography, and traffic-source exports underlying contextual Figures 4-6 are not included in the public supplementary dataset; additional aggregated data may be available from the corresponding author on reasonable request, subject to platform terms and applicable privacy constraints. The Figure 5B choropleth was generated de novo in Python using Natural Earth 1:110m Admin 0 Countries public-domain boundary data; no copyrighted third-party map image, tile service, or proprietary cartographic product was reproduced.