The anatomy of a personal health agent (www.nature.com)

A new comprehensive personal health agent has been developed, leveraging large language models (LLMs) to reason over multimodal data from consumer wellness devices and medical records. This innovative agent includes three specialized sub-agents: a data science agent that analyzes wearable and medical data, a health domain expert agent that integrates personalized and contextual information for tailored insights, and a health coach agent that guides users in setting health goals and making behavioral changes. This multi-agent framework addresses diverse individual health needs identified through user-centered research, including general health knowledge, personal data interpretation, wellness advice, and symptom assessments. Significantly, this personal health agent represents a step toward more accessible and personalized health support, combining modular functionality with an extensive evaluation process involving over 7,000 annotations and 1,350 hours of expert and user input. The integration of these specialized agents aims to provide coherent and actionable health insights while maintaining a positive user experience. This advancement not only reflects the potential of AI in consumer health but also paves the way for future developments in intelligent health management systems that empower individuals to better understand and manage their well-being.