Artificial intelligence for early detection, risk prediction, and prevention-oriented decision support across diabetes-related disability pathways: A structured narrative review

Diabetes mellitus can lead to long-term disability through complications affecting vision, mobility, kidney function, neurologic recovery, physical capacity, and cognition.This review examined how artificial intelligence (AI) supports early detection, risk prediction, and prevention-oriented decision support across diabetesrelated disability pathways.PubMed and Google Scholar were searched through 14 February 2026 using a two-stage strategy combining contextual disability evidence with AI-focused searches.Eligible studies were English-language human studies with accessible full text and relevance to predefined disability pathways; AI-focused studies additionally required explicit use of AI, machine learning, deep learning, or computational decision support.The merged corpus included 101 studies: 38 contextual non-AI studies and 63 AI-focused studies.Diagnostic and detection evidence was appraised using Quality Assessment of Diagnostic Accuracy Studies 2-informed criteria, prediction and prognostic evidence using the Prediction model Risk of Bias Assessment Tool-informed criteria, and heterogeneous findings were synthesized narratively.The strongest AI evidence concerned diabetic retinopathy, diabetic foot complications, and diabetic kidney disease, whereas neuropathy-related falls, frailty, sarcopenia, and cognitive decline were less mature and less frequently externally validated.Overall, evidence is stronger for detection and risk prediction than for demonstrated prevention of disability.Clinical translation therefore requires external validation, calibration, transportability, workflow integration, and prospective evaluation of patient-centered disability outcomes.