Transforming
patient care
through data-driven
innovation

Len A Usvyat

PURPOSE OF REVIEWArtificial intelligence (AI) and machine learning in dialysis have matured beyond the initial phase of algorithm development. Recent work has shifted toward clinical implementation, evaluation, and governance. During the past year, the first randomized trial of AI-guided dosing in maintenance hemodialysis has been published, a meta-analysis of technology-guided dry-weight adjustment has appeared, and the first professional society framework for the responsible use of AI in kidney care has been released.SUMMARYThe major challenge for AI in dialysis is no longer developing accurate prediction models. Instead, success will depend on effective workflow integration, validation across diverse patient populations, ongoing postdeployment monitoring, and demonstrating meaningful improvements in patient outcomes.RECENT FINDINGSThe Anemia Control Model is now supported by hospitalization outcomes and cost-effectiveness analyses. A randomized trial demonstrated that AI-guided erythropoiesis-stimulating agent dosing achieved outcomes comparable to physician-directed dosing. Technology-guided dry-weight adjustment is supported by a meta-analysis of 21 randomized trials. Machine learning for vascular access is evolving toward preoperative clinical decision support, while peritoneal dialysis is beginning to receive greater attention from the machine learning community. Large language models are entering nephrology along two complementary paths: open generative systems for education and clinical support, and deterministic, protocol-driven frameworks designed for safety-critical bedside applications. At the same time, the U.S. Food and Drug Administration has released comprehensive lifecycle guidance for AI-enabled medical devices.

Since its inception, the Renal Research Institute (RRI) has been built on collaboration, bringing together cross-disciplinary teams with a shared purpose: translating scientific research into clinical practice...