J Arthroplasty. 2025 Jun 02. pii: S0883-5403(25)00612-6. [Epub ahead of print]
In recent years, there has been a rise in the use of artificial intelligence (AI) in medical research, including within the field of orthopaedic surgery[1-4]. The increased volume and availability of digital data due to the widespread adoption of electronic health record systems provide rich datasets for training AI models. In addition, advancements in AI methodology and computational hardware have allowed for more powerful and flexible models. Despite these promising developments, the inherently multidisciplinary nature of medical AI research presents a challenge for reporting findings. The algorithms used to train AI models are mathematically complex, and medical data is notoriously messy. These difficulties complicate the effective communication and publication of research findings to the broader medical community. Consequently, there is a growing need for standardized reporting guidelines that can bridge this communication gap and ensure the transparency, reproducibility, and clinical relevance of AI research in orthopaedic surgery. The need for such standardization is particularly urgent given the combination of the recency of AI-driven research with the rapid pace of adoption. The concepts are unfamiliar to many readers and journal reviewers; thus, guidelines are critical to ensure quality and transparency. Currently, there are 17 reporting guidelines registered on the Enhancing the Quality and Transparency of Health Research (EQUATOR) network relating to AI and machine learning[5]. Many of these guidelines are study-type specific and have been adapted from earlier guidelines, such as the TRIPOD[6], SPIRIT[7], CONSORT[8], CHEERS[9], and DECIDE[10] guidelines. Others have created more generic recommendations on AI reporting, such as the MINIMAR[11], CAIR[12], and MI-CLAIM[13] guidelines, but these often do not take the intricacies of medical research into account. Also, there have been subspecialty-specific guidelines on AI research proposed, such as STREAM-URO[14] (urology), PRIME[15] (cardiology), and CLAIM [16,17] (radiology)[18,19]. A list of these guidelines and their item domains has been provided in Table 1. The purpose of this paper was to provide general guidelines and important considerations when publishing AI-derived findings in orthopaedic surgery. For more general information regarding AI in orthopaedics, we refer to these papers as a primer [20-26].