Introduction: Addressing the Growing Challenge of Chronic Diseases
Chronic non-communicable diseases, such as diabetes, hypertension, and kidney disease, have become significant global public health challenges. These conditions not only require long-term medical care but also impose heavy physiological, psychological, and economic burdens on patients and their families. Traditional chronic disease management models often face difficulties such as insufficient healthcare manpower, poor patient adherence, and a lack of continuous follow-up.
At BIT Research Alliance, we firmly believe that Artificial Intelligence (AI), particularly advanced frameworks combining Multi-Agent Systems and Large Language Models (LLMs), can bring revolutionary breakthroughs to chronic disease management, enabling more personalized, proactive, and efficient care models.
Overview of BIT Research Alliance’s Chronic Disease Management Framework
Our AI-driven chronic disease management framework is a comprehensive solution designed to empower clinical nursing staff, enhance patients’ self-management capabilities, and optimize overall care quality. Its core components and concepts include:

- Patient-Centric Personalized Information Input and Health Management Plans:
- Starting Point: The system begins by collecting personalized patient information, including medical history, lifestyle habits, physiological data (e.g., collected via wearable devices), etc.
- AI-Formulated Plans: Based on this data, AI can assist in generating personalized health management plans covering dietary advice, exercise guidance, medication reminders, and more.
- Line Chatbot: Your 24/7 Health Companion :
- Real-time Interaction and Education: We have developed a GenAI customer service chatbot based on the Line platform that can answer patients’ questions about disease management, dietary control (e.g., managing hypotension diet for dialysis patients) at any time, providing instant health education.
- Reminders and Feedback: The chatbot can proactively send medication reminders, follow-up appointment notifications, and collect patient feedback on their adherence to health plans.
- Reducing Nursing Burden: Nursing staff often cannot provide continuous reminders and answer health education questions 24/7. The AI chatbot effectively fills this gap, allowing nurses to focus on more complex clinical decisions.
- Multimodal Data Fusion and Analysis:
- Diverse Data Sources: Our framework can integrate multimodal data from various sources, including Electronic Health Records (EHR), Electrocardiograms (ECG), medical IMAGES, and even daily dietary records input by patients via voice or text (identified through Multi-Modal ANYGPT for food recognition).
- Edge Computing Capabilities: Combined with edge computing platforms like NVIDIA Jetson Orin, some data analysis and model inference can be performed locally or proximally, improving response speed and ensuring data privacy.
- Multi-Agent Large Language Model (LLM) System:
- LLM Agent Core Processing: Receives information from user dialogues and queries, performing semantic understanding and initial processing.
- Chinese Dialog Agent (Fine-tuned LLM): An LLM (e.g., Breeze-7B-Instruct-v1) fine-tuned for Chinese context and specific diseases (like diabetes, kidney disease comorbidities) to understand and respond to patient-specific needs more accurately. We have used extensive chronic disease health education documents and QA data on diabetes and kidney comorbidities for fine-tuning.
- Multi-Modal ANYGPT: Responsible for handling multimodal inputs, such as food image recognition and voice companionship.
- Evaluation Agent: Includes LLMs simulating “Doctor” and “Nurse” roles to assess the professionalism and readability of system-generated responses, ensuring output quality.
- Two-Step Rephrase and Respond (RaR) Agent: Optimizes the clarity and professionalism of system responses through a two-step “rephrase and respond” mechanism .
- Advanced Knowledge Graph Augmented Retrieval (Advanced GRAPH RAG):
- Structured Knowledge Support: Integrates with a knowledge graph containing numerous medical nodes (diseases, treatments, drugs, diets, side effects) to provide LLMs with structured, validated medical knowledge support.
- Improving Response Accuracy: Through retrieval techniques and the knowledge base, it ensures that when LLMs answer professional questions, they can provide more accurate and contextually relevant answers, avoiding “hallucinations” that might arise from relying solely on training text.
- Personalized Feedback and Digital Health Reports :
- Real-time Feedback: The system can provide personalized advice and encouragement based on patient adherence and data changes.
- Automated Reports: Generates execution statistics graphs, sentiment scores, historical conversation summaries, etc., helping patients and healthcare providers better understand health trends and management effectiveness. AI models can generate personalized health education reports, reducing the report-writing burden on healthcare professionals.


Benefits and Impact of BIT Research Alliance’s Framework
Our AI-driven chronic disease management framework aims to deliver multifaceted benefits:
- Enhancing Patient Self-Management and Health Literacy : Empowering patients to more actively participate in their health management through real-time Q&A, personalized advice, and easy-to-understand health information.
- Reducing Healthcare Professional Workload : Automating repetitive reminders, education, and reporting tasks, allowing nursing staff more time to focus on core care tasks requiring professional judgment. Clinical trial data shows the platform significantly reduces staff workload .
- Improving Doctor-Patient Communication and Continuity of Care: Providing continuous interaction channels and recording complete health histories helps healthcare providers gain a more comprehensive understanding of patient conditions to make wiser decisions.
- Lowering Complication Risks and Improving Quality of Life : Through early warnings and precise interventions, it aims to reduce the risk of chronic disease complications, thereby improving patients’ quality of life. Clinical trial data shows a 30% reduction in complication risk for chronic disease patients.
- Increasing Satisfaction for Healthcare Professionals and Patients : More efficient workflows and a better care experience have increased satisfaction by 20% for healthcare professionals and 25% for patients, respectively.

Challenges and Future Outlook
Although AI shows immense potential in chronic disease management, challenges such as data privacy protection, model interpretability, integration between different healthcare systems, and regulatory compliance still exist.
BIT Research Alliance is committed to overcoming these challenges by continuously optimizing model algorithms, strengthening data security measures, and actively collaborating with medical institutions to promote the implementation of AI technology in real clinical environments. We believe that with technological advancements and ecosystem improvements, AI-driven chronic disease management will become a standard configuration in future healthcare, bringing benefits to hundreds of millions of chronic disease patients worldwide.
Conclusion: AI — Illuminating New Hope for Chronic Disease Management
Facing the severe challenges of chronic diseases, the AI-driven multi-agent LLM framework proposed by BIT Research Alliance paints a hopeful future. Through technological innovation, we are endowing chronic disease management with higher efficiency, stronger personalization, and warmer humanistic care. We will continue to explore this path, striving tirelessly for a healthier and more equitable society.