A research team at The Hong Kong Polytechnic University (PolyU) has successfully developed Hong Kong’s first “AI Agent for Precision Diabetes Management – PIPE-AI” (AI Agent), designed specifically for Asian populations, together with a related disease risk prediction model. Leveraging artificial intelligence (AI) and large-scale local electronic health data, the system can more accurately predict the risk of deterioration of complications such as chronic kidney disease in patients with type 2 diabetes over the next 10 years, enabling healthcare professionals and patients to intervene early and improve disease management. The research findings have been published in the leading international journal npj Digital Medicine.
To promote the technology’s wider application in the community, the PolyU research team has partnered with the Department of Family Medicine and Primary Healthcare of the Hospital Authority’s New Territories West Cluster, as well as the Yuen Long District Health Centre, to recruit patients with prediabetes and type 2 diabetes in New Territories West from early July to participate in a clinical study of the AI Agent system. Participants will experience personalised risk assessment and health management recommendations firsthand. The study is funded by the Health and Medical Research Fund.
Chronic kidney disease is one of the most common and serious complications of type 2 diabetes. As its early symptoms are often not obvious, it is commonly referred to as a “silent killer”. Without early detection and treatment, the condition may progress to uraemia, requiring long-term dialysis or even kidney transplantation. According to local data, the prevalence of diabetes among people aged 65 to 84 is as high as 19%. Diabetes and its complications not only affect patients’ quality of life but also place a heavy burden on families and the public healthcare system.
At present, many chronic kidney disease risk prediction models have been developed primarily based on Western population data and may not be fully applicable to Asian populations. A multidisciplinary team led by Prof. YANG Lin, Professor of the School of Nursing at PolyU, used 17 years of electronic health records from the Hospital Authority Data Collaboration Laboratory, covering more than 560,000 diabetes patients, to develop an AI prediction model that is more suitable for Asian populations and which achieves an accuracy rate of 87.1%. The model can analyse patients’ health data and estimate their future risk of developing diabetic complications such as kidney disease, helping healthcare professionals make earlier clinical judgements and arrange follow-up care.
In addition to the risk prediction model, the research team has also developed an AI Agent as a “clinical interface” to address the challenge of translating data into concrete action. Focusing on diabetes and complication management, the AI Agent can convert complex medical information into language that is easier for patients to understand, helping them better grasp their health conditions while improving communication between patients and healthcare professionals. The system can be applied in four major scenarios: supporting family medicine and primary healthcare in preliminary screening and risk stratification; assisting specialist outpatient clinics in making more precise referrals for high-risk or complex cases; supporting district health centres in providing 24-hour health consultation services; and helping patients manage their own health, including through diet control, exercise, timely medication adherence and recording of health indicators.
To ensure the safety of the system in clinical use, the research team has incorporated a nurse oversight mechanism. When the AI detects abnormal risk levels or important health alerts, the system will automatically notify a registered nurse for further review and follow-up, thereby enhancing the reliability and safety of the system’s application.
Prof. David SHUM, Dean of the Faculty of Health and Social Sciences, Yeung Tsang Wing Yee and Tsang Wing Hing Professor in Neuropsychology, Chair Professor of Neuropsychology at PolyU, said, “Diabetes management is not only about treating a single disease, but is also closely related to the long-term allocation of healthcare resources and public health strategy. By translating advanced AI technology into a tool for clinical application, the PolyU research team has extended risk prediction capabilities, which were previously largely confined to hospitals, to primary healthcare and community services. This will help allocate healthcare resources more precisely and drive a shift in the healthcare model from ‘passive treatment’ to ‘proactive prevention’, which in the long run is expected to alleviate the healthcare burden posed by chronic diseases in Hong Kong.”
Prof. Yang Lin said, “The patient recruitment programme in New Territories West, launched in early July, marks an important step in bringing smart healthcare into the community in a tangible way. Looking ahead, the team will further integrate imaging and wearable device data to enhance predictive accuracy and will promote the integration of the model into electronic health record systems and district health centre platforms. The initiative will also be expanded to cover other related chronic disease areas, such as the cardiovascular-kidney-metabolic syndrome, so that more members of the public can benefit.”
