Overview of Systematic Reviews Evaluating Artificial Intelligence Models in Nephrology

Background: Artificial intelligence (AI) is a rapidly evolving field aiming to replicate human-like reasoning, learning, and problem-solving, increasingly applied in healthcare. In nephrology, AI aids real-time disease prediction for conditions like acute kidney injury (AKI), chronic kidney disease (CKD), and renal graft rejection. Techniques like Machine Learning and Deep Learning show promise for early prognosis and risk assessment, but their superiority over traditional methods is uncertain. This overview highlights research gaps, evaluates methods, and explores AI’s future in nephrology. Methods: This review followed PRISMA guidelines and a PROSPERO-registered protocol (CRD42023482793). Comprehensive searches of MEDLINE, EMBASE, Cochrane, Scopus, and Web of Science were conducted through January 1, 2024. Two independent reviewers performed title, abstract, and full-text screening, with disagreements resolved through consensus or third-party arbitration. Findings were summarized using a PRISMA flowchart. ROBIS, GRADE and AMSTAR-2 tools were used for evaluating the risk of bias, quality of evidence and methodological quality of the included systematic reviews respectively. Results: Out of 911 screened systematic reviews, 92 studies underwent full-text review to include 30 studies. These systematic reviews included 941 original studies, out of which 39 were prospective cohort studies and the rest were retrospective studies. This included 15,396,654 patients. All of the SR were performed between 2019-2024. The systematic reviews covered diverse Nephrology condition like AKI (n=6), CKD (n=5), Renal Transplant (n=5), Renal Tumors(n=9) and Urological pathologies. Total of 32 models were used. Most common ML models were Support Vector machine(SVM) and Random Forest(RF). The most common Deep Learning Model was Convolutional Neural Network (CNN). QUADAS 2 was the most frequently used tool to assess Risk of Bias. Conclusion: AI-driven management strategies hold great promise for the future and provide an essential step forward in providing more personalized patient care and improving shared decision making. But this field remains relatively immature, and all the reviewed studies in the field rely on retrospective analyses, with no large-scale prospective trials, a critical requirement for the implementation of this technology in clinical practice.