Explainable Artificial Intelligence-PREDICT: Development and Validation of an Explainable Artificial Intelligence Tool for Early Prediction of Diabetic Complications in Saudi Arabia
Year: 2025
Type: article
Study design: Observational Study
Abstract: One of the major diseases in Saudi Arabia is Diabetes mellitus, with an estimated prevalence rate of 18.3% for adults. Early detection of diabetic complications is crucial for initiating fast medical reactions and improving patient outcomes. Although artificial intelligence (AI) has great promise in predicting disease courses, its clinical usage has been limited by the absence of interpretability. This study aims to build Explainable AI (XAI)-PREDICT, an AI-based predictive tool for the early detection of diabetic complications in Saudi Arabia. A retrospective cohort study was conducted using electronic health records from 12 hospitals in Saudi Arabia, tracking 87,542 patients with diabetes mellitus from January 2021 to January 2025. Five major diabetic complications were predicted using many machine learning models built and assessed over 24 months. Interpretability components were included in model development, considering situational variables particular to Saudi medical institutions. The final XAI-PREDICT system achieved high predictive accuracy with area under the receiver operating characteristic curve values of 0.89 (95% confidence interval [CI]: 0.87–0.91) for nephropathy, 0.86 (95% CI: 0.84–0.88) for retinopathy, 0.84 (95% CI: 0.82–0.86) for neuropathy, 0.88 (95% CI: 0.86–0.90) for cardiovascular occasions and 91 (95% CI: 0.89–0.93) for diabetic foot. Using AI to control diabetes in Saudi Arabia has advanced significantly with XAI-PREDICT. It provides both a useful interpretation and great predictive power. This study shows the importance of using XAI-PREDICT for early detection, management of complications and lowering the burden of complications.
Institution: King Faisal University
Language: English
Subfield: Health Information Management
Field: Health Professions
Domain: Health Sciences
Open Access status: hybrid
OpenAlex ID: W4413280482