Artificial intelligence-driven kidney organ allocation: systematic review of clinical outcome prediction, ethical...

Kidney transplantation remains the optimal treatment for end-stage renal disease, yet persistent organ shortages and inequitable allocation necessitate innovative solutions. Artificial intelligence (AI) and machine learning (ML) have emerged as promising tools for improving clinical outcomes and optimizing donor-recipient matching. However, their integration into clinical practice remains limited, and significant challenges regarding validation and ethical implementation persist. This systematic review synthesizes current research on AI-driven kidney allocation, focusing on predictive modeling, operational algorithms, and ethical considerations. We conducted a comprehensive literature search with no restrictions on publication year or country across biomedical databases (PubMed/MEDLINE, Embase), AI repositories (arXiv, IEEE Xplore), and clinical trial registries. Sixteen studies met inclusion criteria, encompassing retrospective cohort analyses, simulation studies, and algorithmic frameworks. Data were extracted on model performance, clinical outcomes, and fairness metrics, with quality assessed via modified QUADAS-2 and PROBAST tools. Findings revealed that AI/ML models-particularly deep learning and ensemble methods-outperform traditional risk scores (e.g., KDRI, EPTS) in predicting graft survival (C-index: 0.65-0.72) and waitlist outcomes. However, only a minority of studies integrated these predictions into actionable allocation policies, with most limited to simulation environments. Ethical frameworks were inconsistently applied; while fairness and transparency were frequently cited, few studies embedded them algorithmically. Key gaps included real-world validation, prospective bias audits, and standardized reporting of subgroup impacts. AI holds immense potential to enhance kidney allocation but requires rigorous clinical translation and ethical governance. Future research must prioritize multidisciplinary collaboration to bridge the divide between predictive accuracy and equitable implementation.