Artificial Intelligence and Machine Learning in Kidney Disease, Dialysis, and Transplantation: Opportunities, Validation Gaps, and Equity Challenges for Arab and African Populations— Narrative Review

Elmukhtar Habas¹, Ala Habas², Amnna Rayani³, Aml Habas⁴

Keywords: artificial intelligence, machine learning, deep learning, large language models, chronic kidney disease, dialysis, kidney transplantation, health equity

DOI: 10.63475/yjm.v5i2.0455

DOI URL: https://doi.org/10.63475/yjm.v5i2.0455

Publish Date: 25-08-2026

Download PDF

Pages: 348 - 354

Citation: 0

Author Affiliation:

1 Professor/Senior Consultant, Department of Medicine, Hamad General Hospital, Qatar University, Doha, Qatar

2 Resident, Tripoli Central Hospital, University of Tripoli, Tripoli, Libya

3 Professor/Consultant, Department of Hematology, University of Tripoli, Tripoli, Libya

4 Specialist, Tripoli Pediatric Hospital, The Open Libyan University, Tripoli, Libya

Abstract

Artificial intelligence (AI)—including machine learning, deep learning, and large language models—is being rapidly adopted across nephrology, spanning chronic kidney disease and acute kidney injury prediction, dialysis management, organ allocation, graft monitoring, and patient education; reported performance is often strong, yet most models are developed and validated on data from a small number of high-income countries and can degrade substantially in populations that differ from the training distribution. The removal of race from estimated glomerular filtration rate equations offers an instructive precedent for how embedded assumptions in clinical algorithms can entrench disparities. This narrative review synthesizes AI applications across kidney disease, dialysis, and transplantation through the lens of external validity and health equity, with particular attention to Arab and African populations, who remain largely absent from both the relevant datasets and the surrounding discussion. To our knowledge, it is among the first reviews to apply this combined framework to these populations and to translate that analysis into a concrete regional agenda. We argue that the principal barrier to equitable benefit is not algorithmic capability but data representativeness, validation, and governance.

---