Pediatr Nephrol. 2025 Nov 25. doi: 10.1007/s00467-025-07019-2. Online ahead of print.
ABSTRACT
Artificial intelligence (AI) has the potential to significantly improve the practice of medicine. However, its application in pediatric critical care nephrology remains underdeveloped. This scoping review investigates the current state of AI/machine learning (ML) algorithms developed and implemented in the field of pediatric critical care nephrology over more than two decades (2000-2024). We identified 24 articles related to commonly encountered pathologies including acute kidney injury (AKI), electrolyte and acid-base imbalances, fluid management, and dialysis. Twenty (80%) of these articles focused on AKI (prediction and/or detection) and only two articles (4%) investigated the impact of implemented AI/ML models on clinical care outcomes. However, significant limitations exist primarily because of the single-center study design and the need for external validation. Other challenges impeding the broader application of AI/ML models in pediatric critical care nephrology include: (1) the heterogeneity of patient diseases, organ maturation, and age-based norms within pediatrics, necessitating larger datasets that have only recently been developed; (2) lack of portability of algorithms across different settings and electronic health records; and (3) limited scalability due to varying computational and bioinformatics infrastructure and the absence of standardized regulations, which hinder further external validation and implementation of data pipelines. In pediatric critical care nephrology, there remains a significant gap between the development of AI/ML models and their implementation/application at the bedside. Closing this gap will require a concerted and collaborative effort across multiple stakeholders.
PMID:41291255 | DOI:10.1007/s00467-025-07019-2
