Background <p>Quality assessment in arthroplasty is gaining increasing importance in the context of growing transparency requirements in healthcare systems. With the introduction of the German Hospital Transparency Act and the Federal Hospital Atlas, the comparable presentation of treatment outcomes has become a&#xa0;central focus. Reliable risk adjustment models are a&#xa0;prerequisite for this, as they must adequately account for differences in patient case mix.</p> Objective <p>This article aims to describe the current state of risk adjustment in arthroplasty, highlight existing limitations, and discuss potential future developments based on the authors’ own research.</p> Methods <p>A&#xa0;narrative review of national and international approaches to risk adjustment in registries and certification systems was conducted. The focus is on the observed-to-expected (O/E) ratios used in the German Arthroplasty Registry (EPRD) and their comparison with model-based approaches applied in international registries. In addition, results from the authors’ own studies on the integration of joint-specific risk factors and machine learning methods are presented.</p> Results <p>Current risk adjustment models are largely based on routine data and patient-specific risk factors and capture clinically relevant joint-specific risk factors insufficiently. The authors’ own studies demonstrate that systematic inclusion of such parameters improves the prediction of postoperative complications following total knee arthroplasty.</p> Conclusion <p>Further development of risk adjustment is essential for fair and valid quality assessment in arthroplasty. In particular, joint- and procedure-specific risk factors should be more strongly integrated in future models to avoid bias in quality comparisons.</p>