• PURPOSE
    • To use advanced topic modeling, specifically the Bidirectional Encoder Representations from Transformers Topic (BERTopic) Model, to analyze research topics in Arthroscopy: The Journal of Arthroscopic and Related Surgery (Arthroscopy).
  • METHODS
    • Text data from the titles and abstracts of 7,304 original articles and reviews published in Arthroscopy between 1985 and 2023 were included to train the BERTopic artificial intelligence (AI) model for topic generation. BERTopic, an advanced natural language processing tool implemented in Python via Jupyter Notebook, uses contextual embeddings and clustering algorithms to efficiently group large datasets into topics based on semantic similarity. The AI-generated topics were then analyzed by frequency (i.e., the number of studies classified under each topic from 1985 to 2023) and popularity (i.e., "hot" and "cold" topic patterns based on linear regression models of topic frequency from 2020 to 2023).
  • RESULTS
    • The BERTopic model categorized 6,901 articles into 35 topics. The most common topics from 1985 to 2023 were anterior cruciate ligament reconstruction, hip arthroscopy and femoroacetabular impingement (FAI), and shoulder instability. From 2020 to 2023, hip arthroscopy and femoroacetabular impingement, superior capsular reconstruction, and anterior cruciate ligament reconstruction were identified as "hot" or popular topics, whereas suture anchor biomechanics, platelet-rich plasma, and arthroscopic irrigation were identified as "cold" topics, indicating a decline in popularity.
  • CONCLUSIONS
    • Using BERTopic, the study showed an efficient way to analyze large amounts of data to establish patterns within orthopaedic sports medicine literature. This study shows the capacity of the BERTopic model to synthesize thousands of articles within Arthroscopy: The Journal of Arthroscopic and Related Surgery into 35 key topics. The ability to process large amounts of data with accuracy and efficiency provides a powerful tool for establishing and defining the current landscape and potential future directions of orthopaedic literature.
  • CLINICAL RELEVANCE
    • Using AI to investigate topics a journal has published will allow us to recognize patterns, identifying common topics, emerging topics, and shifts in focus over time. It will also allow us to identify research gaps that may need to be addressed.