Summary
Managing diabetes often involves using wearable devices that track health data over time, but understanding this long-term information can be tricky. The researchers designed HealthLoopQA, a test to see how well AI models can interpret complex patterns and unusual events in diabetes monitoring data across a month. They also created simulated scenarios with faulty devices or cyber attacks to check if these models notice safety problems. Their results show current AI tools struggle with analyzing long, complicated health records accurately. They also found that when asked to consider a lot of information all at once, the models tend to simplify too much, pointing to big challenges ahead for AI in healthcare.
Large language modelsQuestion answeringMedical wearablesDiabetes monitoringLongitudinal dataAnomaly detectionProcess miningCyber-physical attacksTemporal pattern miningIn-context learning
Authors
Yuchen Niu, Yanan Ma, Srinivasan Nandakumar, Maolin Chen, Viktor Schlegel, Kexin Wei, Ling Cheng, Anna Bird, Anil Anthony Bharath, Siew-Kei Lam
Abstract
As medical wearables become integrated into daily chronic disease care, effectively interpreting longitudinal monitoring data is essential for patients and clinicians to understand health trends, detect safety-critical events, and make informed decisions. While large language models (LLMs) show promise for transforming this streaming physiological data into personalized health insights, evaluating their reasoning capability and analytical rigor in diverse monitoring tasks remains a fundamental challenge. Existing medical wearable question answering (QA) benchmarks primarily assess short-horizon classification or statistical summaries, largely ignoring the long-term patterns, therapeutic and behavioural contexts, and potential system failures inherent in real-world deployments. To address this, we introduce HealthLoopQA, a comprehensive diagnostic benchmark for evaluating LLM reasoning over continuous diabetes monitoring data. Grounded in a novel taxonomy of eleven atomic reasoning abilities, HealthLoopQA comprises 127 tasks and over 1,500 QA instances spanning process mining, anomaly detection, and prediction over 30-day horizons. To systematically evaluate safety awareness, we complement real-world datasets with a fault-injected simulation testbed modeling diverse device malfunctions and cyber-physical attacks to generate physiologically plausible hazard scenarios. Evaluating state-of-the-art LLMs across prompting and agentic frameworks reveals severe limitations in complex temporal pattern mining. Furthermore, we identify a broader phenomenon of In-context Laziness under long-context prompting, highlighting critical open challenges in deploying LLMs for rigorous long-horizon medical reasoning.