Artificial Intelligence–Driven Approaches for Prediction, Management, and Complication Risk in Type 2 Diabetes: A Systematic Review

Abstract

Type 2 Diabetes (T2D) is a global health challenge with significant morbidity and economic costs, driven by aging populations and lifestyle changes. Despite advancements in management, T2D complications remain prevalent, necessitating early and personalized interventions. Artificial intelligence (AI) has emerged as a transformative tool for T2D prediction, management, and complication risk assessment. This systematic review aimed to synthesize contemporary evidence on AI-driven approaches for T2D, focusing on methodologies, data modalities, and clinical applications. Following PRISMA 2020 guidelines, a comprehensive search of databases (PubMed, Scopus, Web of Science, etc.) from 2018 to 2024 identified 28 eligible studies. Results highlighted the superiority of deep learning models (e.g., LSTM, Transformers) in glycemic forecasting and risk prediction, achieving high accuracy (e.g., 89–94.7%) and improved clinical outcomes in randomized trials. Multimodal data integration (EHR, CGM, genomics) enhanced predictive performance, while interpretability techniques like attention mechanisms increased clinician trust. However, gaps persist in prospective validation and scalability. The review underscores AI's potential to revolutionize T2D care but calls for rigorous clinical trials and standardized reporting to ensure real-world applicability.