Clinician's Dilemma: Navigating AI and Data Challenges in Early Detection of Metabolic Health Disorders

Early detection of metabolic disorders, such as insulin resistance, is critical to prevent irreversible physiological damage. However, clinicians face significant challenges in identifying these conditions at a preclinical stage due to the limitations of current tools, which often rely on late-stage biomarkers and fragmented data. In this compelling session, the speaker explores how Artificial Intelligence (AI) can enhance early diagnosis while also addressing the inherent complexities of deploying these technologies in real-world clinical settings. The talk introduces SATS (Stress Assessment Tracking System), an innovative AI-driven model that leverages subjective patient-reported data alongside emerging objective inputs to create probabilistic insights ahead of traditional standards of care. Attendees will examine the limitations of current AI systems—such as the unreliability of Large Language Models (LLMs) and the dependency of Retrieval-Augmented Generation (RAG) models on curated data—and consider a path forward that integrates human context with machine-driven precision. Key learning objectives: Understand the diagnostic limitations in early detection of metabolic disorders. Explore the role and pitfalls of AI models (LLMs, RAG) in clinical settings. Learn about the SATS framework and its application in early, probabilistic detection.