Diabetes is one of the world’s most prevalent chronic diseases and can lead to serious complications, including chronic kidney disease, cardiovascular problems, and long-term organ damage. Now, researchers at The Hong Kong Polytechnic University have developed an artificial intelligence-based solution that could shift diabetes care from reactive treatment to proactive, personalized prevention.
The research team has created PIPE-AI (AI Agent for Precision Diabetes Management), Hong Kong’s first AI agent specifically designed for precision diabetes management. The system, developed for Asian populations, combines an AI risk prediction model with an intelligent clinical interface that helps patients and healthcare professionals better understand health risks and make informed decisions.
To develop the technology, researchers analyzed 17 years of electronic health record data from more than 560,000 diabetes patients collected through the Hospital Authority Data Collaboration Laboratory. The AI model can assess patients’ health information and predict their future risk of developing diabetes-related complications, including chronic kidney disease, with an accuracy rate of 87.1%.
A major advantage of the system is its focus on Asian populations. Many existing chronic disease prediction models have been developed using data from Western populations and may not fully reflect the health characteristics of Asian patients. By using large-scale local health data, PolyU researchers developed a model better suited for regional healthcare needs.
Beyond risk prediction, the PIPE-AI system acts as an AI-powered healthcare assistant that translates complex medical information into easier-to-understand language for patients. The technology supports several healthcare scenarios, including early screening and risk assessment in primary care, more accurate referrals for high-risk patients in specialist clinics, 24-hour health consultation through community health centers, and personalized self-management support covering diet, exercise, medication adherence, and health monitoring.
To improve safety in clinical applications, the research team has integrated a nurse supervision mechanism into the system. When the AI identifies abnormal risk levels or important health alerts, it automatically notifies a registered nurse for further review and follow-up, ensuring greater reliability and patient safety.
The technology has now entered a clinical study phase, with patient recruitment beginning in Hong Kong’s New Territories West. The project, conducted in collaboration with the Hospital Authority’s New Territories West Cluster and Yuen Long District Health Centre, aims to evaluate the real-world effectiveness of AI-assisted diabetes management among patients with prediabetes and type 2 diabetes.
Researchers plan to further improve the system by integrating medical imaging data and information from wearable health devices to enhance prediction accuracy. They also aim to connect the AI model with electronic health record systems and expand its application to related chronic conditions, including cardiovascular-kidney-metabolic syndrome.
Experts believe that AI-driven healthcare technologies such as PIPE-AI could transform chronic disease management by enabling earlier intervention, more efficient allocation of healthcare resources, and a transition from a healthcare model focused on treating disease after it occurs to one centered on prevention and personalized care.
Re: Na