Chatbot dược sĩ ảo tích hợp cơ sở dữ liệu thuốc để hỗ trợ quản lý bệnh nhân đái tháo đường

The increasing prevalence of chronic diseases like diabetes mellitus necessitates innovative and accessible healthcare solutions. Digital health, particularly artificial intelligence (AI) applications, holds significant promise in supporting disease management. However, existing tools often lack personalized, continuous engagement aligned with behavioral change models and practical needs in community settings. This research addresses the critical gap in providing timely, individualized advice on medication use, lifestyle adjustments, and blood glucose monitoring, especially in resource-constrained community pharmacies. This study develops and conducts an early-stage feasibility assessment of a virtual pharmacist chatbot integrated with a drug database and lifestyle guidance, aiming to enhance support for people with diabetes and improve their self-management capabilities.

Healthcare researchers, pharmacists, medical professionals, AI developers in healthcare, and public health policymakers interested in digital health solutions for chronic disease management.

This study details the development and early-stage feasibility assessment of an AI-powered virtual pharmacist chatbot designed to support diabetes mellitus management in community pharmacies. The methodology involved comprehensive user requirement analysis, followed by the construction of a robust drug and health-education knowledge base adhering to Ministry of Health guidelines and international recommendations from ADA/EASD. A total of 200 personalized dialogue scenarios were designed to align with behavioral change stages. The chatbot was implemented using an NLP pipeline (Rasa/Dialogflow) combined with the Gemini API, enabling real-time data capture for blood glucose and pressure. Rigorous internal alpha testing and a 7-day feasibility trial conducted in one community pharmacy demonstrated its practical utility. Key results show the knowledge base encompassed 40 topics with 200 scenarios, yielding 94% accuracy for in-scope queries and 91% useful responses for out-of-scope questions. The system achieved a mean response time of 1.8 seconds, with 88% of users rating it "useful/very useful." These findings affirm the chatbot's content readiness, high accuracy, and strong usability within the community pharmacy context. The study concludes that this virtual pharmacist chatbot presents a significant potential for scaled deployment and further evaluation of its impact on chronic disease self-management and pharmacy counseling practices, offering a promising solution for personalized, accessible healthcare.