Optimizing the management of periprosthetic infections with artificial intelligence: current evidence and future directions

Abstract

Periprosthetic joint infections (PJIs) are a serious complication in both primary arthroplasty and revision arthroplasty, posing a major challenge in orthopaedic surgery and creating a substantial financial burden. This literature review examines the current knowledge on PJI prediction, diagnosis and prognosis, with a focus on scoring systems and machine learning (ML) tools developed to improve their management. We conducted a narrative literature review by searching Medline, CENTRAL and Embase up to October 1, 2024, with independent dual-reviewer screening. Nine non-randomized studies were included, covering 297,981 prostheses and 7,190 PJIs. Two studies assessed prediction based on patient history but highlighted the need for refinement and multi-centre prospective validation. Five studies evaluated ML in diagnosis, showing promising accuracy, yet requiring broader validation in larger, more diverse clinical contexts. Two studies addressed prognosis, but models remain limited in providing individualized, treatment-specific insights. The development of ML models represents an interesting approach, given the rising prevalence of PJIs and the need for better management. However, available studies face important limitations, including small sample sizes, lack of external validation and limited transparency regarding parameters and models. To make ML tools clinically relevant, future research should prioritize external validation, larger multi-centre prospective studies and transparent reporting. Ultimately, robust ML models have the potential to enhance PJI management, improve patient outcomes and reduce healthcare costs.

Ähnliche Arbeiten