Editorial Commentary: Artificial Intelligence Analysis of Biomedical, Large, Clinical Registry Data Using Machine Learning Requires Tens of Thousands of Subjects and a Focus on Substantial Clinical Benefit: Minimal Clinically Important Difference Is too Low a Bar

As the field of arthroscopic hip preservation surgery grows, large high-quality registries represent a foundational study design for establishing whether hip arthroscopy is effective for patients with femoroacetabular impingement syndrome (FAIS). Original research publications from experienced high-volume surgeons tell us “Can it work”. A registry tells us “Does it work?”. The ability of preservation to truly preserve the joint, delay the arthritis process, and reduce the risk of arthroplasty requires long-term follow-up. A geographic registry can follow this. The registry represents the “real world”, a heterogeneous set of variables pertaining to the doctor, patient, intervention, and outcome. The vast array of factors that can be analyzed before, during, and after surgery makes machine learning an ideal technique for analysis of large quantities of data. A global hip preservation surgery registry is a desirable and achievable goal. In order to optimally predict outcome of hip arthroscopy, given the known large number of patient- and hip-specific factors that influence outcomes, a deep learning model with tens of thousands of subjects for this medium-scale task would be needed. Measures of clinical relevance need to include more than just MCID (minimal clinically important difference), which is the lowest bar minimal threshold. Patient expectations often far exceed MCID—requiring other metrics like SCB (substantial clinical benefit), PASS (patient acceptable symptom state), and MOI (maximal outcome improvement). Registries should include validated, reliable, and responsive patient-reported outcome scores (e.g., International Hip Outcome Tool [iHOT-12]) with measures of clinical relevance and expectations assessed routinely. The United Kingdom's NAHR (Non-Arthroplasty Hip Registry) and Denmark's DHAR (Danish Hip Arthroscopy Registry) are the two largest geography-based registries in current hip preservation research both with 11 years of patient enrollment.