• PURPOSE
    • To utilize advanced topic modelling through the Bidirectional Encoder Representations from Transformers Topic (BERTopic) model to investigate research topics in the journal Knee Surgery, Sports Traumatology, Arthroscopy (KSSTA).
  • METHODS
    • Titles and abstracts from 7886 original research articles and reviews published in KSSTA between 1993 and 2023 were examined using the BERTopic artificial intelligence (AI) model. BERTopic applies contextual embeddings and clustering algorithms to group large textual data efficiently sets into topics based on semantic similarity. The generated AI topics were assessed by frequency (the total number of articles per topic from 1993 to 2023) and popularity trends ('hot' or increasing and 'cold' or decreasing trends determined by linear regression analyses of topic frequency from 2020 to 2023).
  • RESULTS
    • The BERTopic model organized 7410 publications into 33 distinct topics. From 1993 to 2023, the most frequently reported topics included arthroscopic shoulder surgery, meniscus injury and treatment, and total knee arthroplasty (TKA): design biomechanics. Between 2020 and 2023, arthroscopic shoulder surgery, TKA: design biomechanics, and TKA: alignment & kinematics were identified as increasingly popular ('hot') topics. Conversely, ankle instability, non-anterior cruciate ligament (ACL) knee ligament injuries, and ACL reconstruction: bone tunnels showed declining popularity ('cold' topics).
  • CONCLUSIONS
    • This study demonstrates the efficacy of the BERTopic model in analyzing large textual data sets to identify relevant research patterns within orthopaedic literature. The results highlight BERTopic's ability to summarize thousands of articles from KSSTA into 33 central topics, underscoring its utility in accurately and efficiently capturing current trends and future directions in orthopaedic sports medicine research.
  • LEVEL OF EVIDENCE
    • Level IV, systematic review.