Epidemiological surveys indicate that approximately one-quarter of Taiwanese adults aged 20 to 64 meet the diagnostic criteria for metabolic syndrome. Evidence shows that this population faces a two- to six-fold higher risk of developing chronic conditions such as hypertension, diabetes, and dyslipidemia. However, individuals often seek care across multiple healthcare settings, resulting in fragmented data and delayed interventions.
To address this gap, Cathay Health Management has developed NutriX, a software-based integrated health promotion platform that combines one-on-one consultations with registered dietitians and continuous support from Buddy, an AI-driven health management assistant. The platform was co-developed by physicians, nurses, dietitians, and digital innovation specialists, ensuring both clinical applicability and professional rigor. Unlike conventional services that are constrained by healthcare professionals’ appointment schedules, NutriX enables on-demand access to suitable health guidance, offering 24/7 support to users.
Prototype validation demonstrated improvements in over 80% of key clinical indicators associated with metabolic syndrome, including blood pressure, fasting glucose, lipid levels, and waist circumference. These findings have been presented at international scientific conferences across Europe and Asia, highlighting the potential of human–AI collaboration in advancing health promotion.
This product does not require a product license number.
Cathay Healthcare Management Co.,Ltd.
The NutriX prototype has been validated through research studies. Two research posters — Beyond Intervention Duration: How Adherence and Personalization Drive Metabolic Health Improvements and Precision Health Model Using Modular Clustering: An 8-Week Intervention Settings — were presented at the 2025 International Forum on Quality and Safety in Healthcare, Singapore, and the 2025 IHF World Hospital Congress, Switzerland, respectively. These presentations received recognition from international organizations in Asia and Europe, validating that the human-AI collaborative model and stratified intervention strategies can improve user adherence and metabolic syndrome outcomes.
The project demonstrated improvements in more than 80% of key metabolic syndrome indicators, including reductions in blood pressure, triglycerides, and fasting blood glucose. Lipid-related outcomes peaked at Week 8, with LDL-C decreasing by 9.3 mg/dL and triglycerides by 22.6 mg/dL. High-adherence participants achieved greater improvements than low-adherence participants, including larger reductions in waist circumference (3.3 cm vs. 1.5 cm), triglycerides (26.6 mg/dL vs. 14.8 mg/dL), and LDL-C (9.9 mg/dL vs. 8.3 mg/dL). These findings highlight adherence as a key factor in improving metabolic syndrome outcomes.
During the development and operation of NutriX, a comprehensive quality assurance and monitoring framework was established to ensure service accuracy and reliability:
Data sources are derived from people who have provided informed consent. Based on intervention protocols designed by registered dietitians, user health condition labels are generated to match appropriate health plans, ensuring that the large language model prioritizes professional dietitian perspectives when responding to users’ daily health inquiries.
During the semi-finished product stage, users are randomly selected for validation to verify the accuracy of the labeling logic. Registered dietitians also directly evaluate model-generated responses to assess both recommendation accuracy and the level of human-centered interaction.
During the final product acceptance stage, a set of professional questions is used to evaluate the model’s human-like interaction, memory capability, and professional accuracy. Deployment is approved only after registered dietitians confirm that the model achieves an accuracy rate exceeding 70%.
In terms of information security, the platform has obtained ISO 27001 Information Security Management System certification, ensuring that the overall operational process is protected by established security standards.
To mitigate potential risks during product operation, NutriX has implemented multiple control mechanisms to ensure system stability and the appropriate use of resources:
To prevent excessive AI token consumption caused by malicious or abnormal usage, each user is assigned a daily question limit to ensure fair allocation of system resources.
User behavior is continuously monitored in the backend, particularly for high-frequency activity. Abnormal operations may be restricted or blocked when necessary to prevent system misuse.
To maintain a health-focused user experience, the system provides guidance when users submit inquiries unrelated to health topics. Users are reminded to focus on health-related issues, while the platform clearly communicates its intended purpose and proper usage to reduce the risk of misuse.
To ensure continuous optimization and reliability, each system version update is required to pass the same set of professional evaluation questions, establishing a rigorous monitoring mechanism. This process enables NutriX to maintain high standards, traceability, and effective quality control throughout its entire operational workflow.
NA
NA
