This study was a cross-sectional observational analysis of publicly available social media posts, conducted in accordance with the ethical guidelines of the Association of Internet Researchers. According to these guidelines, publicly accessible social media data that do not involve direct participant interaction or contain sensitive personal information do not require institutional review board approval.

To ensure a comprehensive dataset, 4 major social media platforms—Instagram, Facebook, Reddit, and X (formerly Twitter)—were selected based on their widespread use and the presence of dedicated community spaces where individuals discuss orthopaedic procedures and rehabilitation experiences. TikTok (ByteDance, Beijing, China) and YouTube (Alphabet, Mountain View, CA) were excluded because of their primarily video-based format, restrictive data-access policies, and lack of standardized searchable written content, which limited feasibility for consistent text-based analysis.

An advanced keyword search was performed across platforms using a predefined set of hashtags and keyword phrases relevant to ACLR with QTG and BTB autograft. Hashtags (e.g., #ACLR, #QuadricepsGraft, #ACLRecovery, #PatellarGraft, #BTBGraft, and #SportsSurgery) were prioritized for platforms on which they are algorithmically integral, such as Instagram, Facebook, and X. Meanwhile, Reddit was queried using Boolean keyword combinations [e.g., (‘ACL reconstruction’ OR ‘ACL surgery’ or ACLR) AND (‘quadriceps graft’ OR ‘quad tendon’ OR ‘BTB graft’ OR ‘patellar graft’)] to account for its limited hashtag infrastructure. To minimize algorithmic bias, searches were conducted using non-personal anonymized accounts in a cleared browser state (with no cache or cookies).

Only original, public-facing posts were included in the dataset. Comments, replies, and re-shares were excluded to maintain consistency and reduce variability in tone and content structure. Posts authored by caregivers, family members, or clinicians were also excluded. Inclusion criteria were any post made from 2012 to 2024 specifically mentioning ACLR with quadriceps or patellar tendon graft. The year 2012 was chosen as the start year to ensure that all platforms were publicly available and widely used by that point to provide a consistent and comparable dataset. Exclusion criteria were as follows: (1) posts from health care professionals or promotional content, (2) posts not directly related to patient experience (e.g., general information or advice about ACL injuries), and (3) posts that were not written in English. No scraping tools or application programming interfaces were used; all data were reviewed and logged manually in compliance with platform policies.

Sex and postoperative time point were inferred based on explicit textual references within the posts (e.g., “as a female athlete” or “I’m 6 months post-op”). If such information was not clearly stated, it was coded as “unspecified.” Age was inconsistently reported across posts and was therefore not included in the analysis. Only posts clearly written from the first-person perspective of the patient were included to preserve the authenticity of experiential reporting.

Two raters (S.S.T., G.B.), who were trained in qualitative content analysis, conducted a review of the collected social media posts. Self-training consisted of a comprehensive review of thematic analysis methodologies, practice coding sessions using a predefined coding framework, and consensus discussions to resolve discrepancies in scoring and thematic assignments.

A predefined thematic coding framework, developed based on existing literature and input from the senior authors (W.F.P, E.S.C), guided the classification of posts into sentiment categories (positive, neutral, negative) and thematic groupings. Core themes such as “functional outcomes,” “return to sports,” “pain management,” and “donor-site concerns” were established a priori. Raters remained open to emergent patterns, and additional themes such as “psychological effects” and “recovery milestones” were added inductively based on recurring content during the review process. This hybrid approach allowed for iterative refinement to better reflect the discourse within the dataset. Because individual posts often discussed multiple aspects of the recovery experience, each post could be assigned to more than 1 theme or sub-theme. This overlapping thematic coding allowed for a more nuanced analysis of patient discourse across categories. The final thematic schema is defined in Table 1 .

Each post was assigned a sentiment score based on the overall emotional tone expressed (positive = 1, neutral = 0, negative =–1). Sentiment assignment was conducted independently by both raters, with disagreements resolved through discussion and consensus.

Descriptive statistics were used to calculate frequencies of sentiment, first-order themes, and sub-themes. One-way analysis of variance was used to determine significant differences for mean sentiment scores between posts about QTG and posts about BTB. A χ ² test of proportions was implemented to detect significant differences in theme and sub-theme frequencies across the 2 graft-option cohorts.

To ensure reliability, both raters independently coded the first 200 posts, after which an intercoder reliability was assessed using the Cohen κ statistic (κ = 0.72). This level of agreement was deemed acceptable for qualitative research. Intrarater reliability was evaluated by having each rater re-code a random sample of 50 posts after a 2-week interval, with agreement assessed using the Cohen κ statistic (κ = 0.78) to account for consistency over time.

To assess temporal differences in themes between graft groups, posts were stratified into the following postoperative time intervals: 0 to 2 weeks, 2 to 6 weeks, 6 to 10 weeks, 2.5 to 6 months, 6 months to 1 year, 1 to 2 years, and more than 2 years. χ ² Analyses were used to compare theme frequency proportions within each interval between the QTG and BTB cohorts. A total of 492 QTG and 469 BTB posts included identifiable postoperative timing data and were eligible for this temporal analysis. The threshold for significance was set at P <.05. All statistics were performed using Stata, version 19 (StataCorp, College Station, TX).