Artificial intelligence in organ transplantation within European healthcare systems: A scoping review of model...

Artificial intelligence (AI) and machine learning (ML) are increasingly developed and evaluated for the organ transplantation pathway, yet the breadth of empirical evidence generated within European healthcare systems, and its actual translation into routine clinical use, has not been mapped. This scoping review charts the nature, clinical targets, modelling approaches and reported performance of AI applications in organ transplantation in Europe. Following Arksey and O'Malley's framework and PRISMA-ScR reporting, MEDLINE and Embase were searched using terms covering organ transplantation, AI, ML, deep learning, allocation, decision support, immunosuppression and graft survival. Empirical primary studies conducted wholly or partly in Europe were prioritised for charting, while systematic, narrative and position-paper literature informed the background and comparative discussion. Empirical European studies spanned kidney, liver, lung, heart and pancreas transplantation and addressed graft and patient survival prediction, donor-recipient matching, allocation, graft and biopsy assessment, rejection detection, and decision support. Tree-based ensembles and neural networks predominated, with discrimination commonly between 0.70 and 0.95 and frequently exceeding conventional scores such as MELD, KDRI and SOFT. Registry-based national cohorts from the United Kingdom, Spain, France, Germany, Italy and multinational European consortia were prominent. Few tools reached prospective or randomised evaluation, and benefits in shared decision-making were not consistently demonstrated. Crucially, almost all identified studies described model development or validation rather than adoption into routine care, and evidence of sustained clinical implementation within European transplant units was scarce. AI shows substantial promise across the European transplantation pathway, particularly for risk stratification, organ assessment and rejection detection, but most evidence remains retrospective and developmental rather than clinically deployed, and is concentrated in a small number of countries with mature registries. Prospective and externally validated studies, attention to equity and generalisability across diverse populations, workflow integration, and alignment with European regulation are required before routine clinical adoption.