Healie: A Personalized Medical Content Generation System Integrating Knowledge Graphs and Large Language Models

Healie is an innovative healthcare information system that integrates knowledge graphs, cognitive factors, social determinants of health, and large language models (LLMs) to generate personalized medical content for patients from diverse backgrounds. It enhances health literacy and self-management capabilities, addresses the pain point of traditional one-size-fits-all health education that ignores individual differences, and enables patient-centered information delivery.

Effective communication of medical information is a global challenge. Traditional methods ignore differences in patients' educational backgrounds, cultural environments, and cognitive abilities, leading to poor information absorption, misunderstandings, and anxiety. Healie's vision is to build an intelligent system that dynamically adjusts the form and depth of content expression based on patients' cognitive characteristics and social determinants of health (such as economic status and living environment).

Healie's architecture is based on three core components:

1. Medical Knowledge Graph: Structurally presents medical concepts and their relationships (e.g., diseases and symptoms, drugs and side effects), supports logical reasoning, and ensures content accuracy;
2. Patient Profile Modeling: Integrates clinical data (diagnoses, medical history), cognitive factors (health literacy, learning preferences), and social determinants (education, economic accessibility) to build a comprehensive patient profile;
3. LLM: Translates knowledge and patient profiles into natural, easy-to-understand content, and can adjust language styles to adapt to different users (using professional terminology or common metaphors).

The generation process includes:

1. Data Collection and Profile Construction: Extract clinical information from electronic health records, collect cognitive and social data through questionnaires, and generate dynamic profiles after data desensitization;
2. Knowledge Query: Acquire medical knowledge through semantic reasoning, and make personalized adjustments based on patients' complications, cultural habits, etc. (e.g., dietary advice for diabetes);
3. Content Generation: LLM generates customized content based on knowledge and patient profiles. Strategies include visual aids (for patients with low health literacy), reassuring tones (for anxious patients), and daily checklists (for patients with chronic diseases).

Healie improves health literacy (defined by WHO as the ability to access, understand, and use information), bridges the information gap between doctors and patients, serves as a digital assistant for doctors to provide post-consultation support, and enhances patients' self-management abilities (understanding their condition, following medical advice, and identifying warning signs). It also considers social determinants to promote equity: for example, recommending low-cost solutions for economically disadvantaged individuals and connecting elderly living alone with community resources.

Challenges: Accuracy and timeliness of medical knowledge (needs continuous updates), privacy and security (strict data protection), and LLM hallucinations (ensuring content reliability). Future directions: Multimodal content (text/images/videos), real-time dialogue, integration with wearable devices, and cross-language support. Healie aims to enhance doctor-patient communication and enable patients to receive tailored health guidance.