CARE-LLM – Clear and Accessible Resources for Empowering patients using multi-agent LLMs

The CARE-LLM project focuses on building an LLM-based framework to make medical information about kidney diseases more accessible to patients and their caregivers by simplifying complex medical texts, such as diagnostic reports, treatment plans, and relevant medical research. The project addresses the critical need for understandable health information to enable patients with rare kidney diseases to make informed decisions, manage their conditions effectively, and navigate the healthcare system more efficiently. By translating complex medical information into easy-to-understand language, CARE-LLM contributes also to saving valuable time for healthcare professionals. This contributes to better patient engagement and participation in their own care and treatment processes, which is especially vital given the complexity and rarity of their conditions.

Even patients with high educational backgrounds struggle to understand medical texts due to the pervasive use of technical terminology and complex information inherent in such documents.  Another challenge is the time constraint faced by health professionals, who often find it difficult to devote sufficient time to thoroughly explain medical details to each patient. This can hinder effective communication and patient understanding, emphasizing the need for a solution that can bridge the information gap efficiently and make these texts more accessible without diluting their medical accuracy.

Large Language Models (LLMs) present a promising solution to automate the simplification of medical texts. However, these technologies come with their own set of challenges, primarily related to accuracy and reliability. LLMs are prone to errors and can generate "hallucinations" (i.e. false or misleading information not supported by the source data). Such inaccuracies can have serious negative consequences, potentially leading to misinformation and inappropriate health decisions by patients.

The CARE-LLM project aims to develop a robust framework that integrates multi-agent large language models (LLMs), each specialized in different tasks to ensure the simplification of medical texts is both accurate and tailored to individual patient profiles. These LLMs work in a harmonized manner, using external medical evidence to verify the medical validity of the information presented.

An essential component of this solution is a user-friendly interface (UI) that facilitates the interaction between patients and the simplified medical texts. This UI is designed to be intuitive, allowing patients to easily access and understand their medical information, which is crucial for enhancing patient engagement and ensuring the information is accessible to those with varying levels of health literacy.

FIT is responsible for developing the backend framework and integrating external medical evidence to ensure the accuracy and reliability of simplified medical texts.

• Fraunhofer Institute for Applied Information Technology FIT (Lead)
• Fraunhofer Institute for Industrial Engineering IAO
• University Hospital of Cologne (Associate)
• expandAI (Associate)