William Reuben Aibinder

Artificial intelligence (AI) is permeating many aspects of daily life. Medicine, and specifically orthopaedic surgery, is being influenced by the rapid growth and success of AI in serving as an adjunct to medical expertise. AI includes many nested subfields and applications such as computer vision, language models, and predictive analytics. Using standard machine learning methods, clinical decision support tools are being developed for clinicians to provide evidence-based, individualized care. Some models predict clinical and functional outcomes after a procedure, whereas others may incorporate administrative metrics to predict factors, such as length of stay, return to the emergency department, and complications. The application of AI to medical imaging focuses on the extraction of a large amount of quantitative data in the form of predefined features from clinical images that humans cannot perceive, a field known as radiomics. There has been recent interest, particularly in oncology, to apply radiomics to aid in diagnosis and prognostication of outcomes. Given the robust growth of radiomics literature, and the anticipated application in orthopaedic surgery, the purpose of this review article is to educate the orthopaedic surgery community on the practical importance and the basics of radiomics.

Purpose: Shoulder arthroplasty (SA) utilization has increased substantially, with a corresponding rise in revision procedures, many driven by periprosthetic joint infection (PJI). Herpes zoster (HZ), or shingles, is increasingly prevalent among older and immunocompromised patients and shares overlapping risk factors with PJI. However, the association between shingles and postoperative complications following SA is unknown. This study evaluated the relationship between peri-operative shingles and postoperative complications after SA. Methods: Patient-level claims from adults ≥ 18 years undergoing SA between 2016 and 2022 from a 20% Medicare fee-for-service sample and the IBM ^® MarketScan ^® Commercial Database were retrospectively collected. Shingles diagnoses from one year preoperatively through one year postoperatively were identified. For postoperative shingles, the diagnosis date was required to precede the earliest censoring event (outcome occurrence, loss to follow-up, end of follow-up) to maintain appropriate temporality of events. Outcomes within one year included revision arthroplasty, irrigation and debridement (I&D), implant removal, or shoulder PJI. Incidence rates (IRs) per 100 person-years and incidence rate ratios (IRRs) were calculated for a composite primary outcome. Associations between shingles and outcomes were evaluated using three Cox regression approaches with covariate adjustment. Results: Among 26,543 patients, 1.4% had shingles preoperatively and 1.0% developed shingles on or after surgery. The IR of the primary outcome was lower in patients with (IR 3.78; 95% CI 2.20–5.36) than without shingles (IR 5.06; 95% CI 4.77–5.35), but without statistical significance (IRR 0.75; 95% CI 0.49–1.14). No analytic approach showed a significant association, and no effect modification was observed. Conclusions: Shingles diagnosed within one year before or after shoulder arthroplasty was not associated with increased risk of revision surgery, I&D, implant removal, or PJI within one year postoperatively.

Background: Shoulder arthroplasty is increasingly common, and the humeral head is routinely sent for pathologic evaluation post-operatively. Prior work has questioned the cost-effectiveness of this practice. The purpose of this study was to assess whether routine gross pathology of the humeral head after shoulder arthroplasty is cost-effective and clinically impactful. Methods: Primary total shoulder arthroplasty and hemiarthroplasty cases performed at a single tertiary institution from 2013 to 2024 were reviewed. Surgeon pre-operative diagnoses were compared with gross pathology findings to assess diagnostic agreement and any resulting change in patient care. Pathology costs were estimated using the 2024B Centers for Medicare and Medicaid Services fee schedule. Results: A total of 1274 cases met the inclusion criteria. Agreement between surgeon and pathology diagnoses was 95.6%, with 4.4% discordance. No discordant findings resulted in altered clinical management. The total pathology cost was $20,778.94, averaging $371.05 per discordant diagnosis. Discussion: Surgeon and pathologist diagnoses of humeral head pathology after shoulder arthroplasty show an exceedingly high agreement rate of 95.6%. Importantly, none of the 56 diagnoses in disagreement led to a change in patient care. Routine gross pathologic evaluation of the humeral head after shoulder arthroplasty shows limited clinical utility and is not cost-effective. Level of Evidence: Level IV, Economic analysis.

Background: The integration of artificial intelligence (AI) in medical research is becoming increasingly common. While AI has the potential to assist in various research tasks, its effectiveness in accurately identifying citations and compiling bibliographies is hampered by existing human errors linked to citation inaccuracies. This study aims to: 1) determine the prevalence of citation inaccuracies in orthopaedic surgery literature, and 2) propose a pragmatic classification system to clearly define citation accuracy to determine the validity of AI application in citation collection. Methods: We reviewed existing literature on citation accuracy to develop a novel citation accuracy classification system to determine if information is correctly cited. This scoring system distinguishing between major and minor errors, with subcategories that include secondary article citations, oversimplifications, statistical inaccuracies, missing facts, and contradictory conclusions, rated on a scale of 1 to 5. We analyzed 229 references from seven scientific articles and applied our novel classification system to assess accuracy. We also evaluated two artificial intelligence (AI) platforms for their potential to automate citation accuracy checks. Results: Improper citations were identified in 27.5% of the 229 references analyzed. Twenty references (8.3%) cited a secondary article. Minor errors were present in 25 references (10.9%), while major errors were found in 21 references (9.2%). Both tested AI platforms failed to complete a thorough review due to an inability to apply the classification system or access full-text sources. Conclusions: We present a novel citation accuracy classification system that can be applied across scientific disciplines. Citation inaccuracies are prevalent in orthopaedic surgery literature, with nearly 10% of citations containing major errors. These findings highlight the high prevalence of citation inaccuracies in orthopaedic surgery literature and raise concern about the unchecked use of AI tools that may inadvertently propagate such errors. Combining rigorous human oversight with carefully integrated AI support may help improve the accuracy and reliability of scholarly references in the future. Statement of Clinical Significance: The proposed citation accuracy classification system provides a foundation for characterizing inaccuracies, and we emphasize concerns about the unchecked deployment of AI tools that may inadvertently perpetuate the citation errors already present in the literature.