Creation of AI-powered virtual assistants for monitoring and therapeutic adherence in chronic diseases

Abstract

This study evaluated the development and implementation of "GlucoGuide," a virtual assistant with artificial intelligence (VAI), to improve therapeutic adherence and monitoring in patients with type 2 diabetes mellitus (T2DM). Using a convergent mixed-methods design combining a randomized controlled trial (RCT) with qualitative research, the VAI was proven to be an effective tool. The intervention group showed a significant 12.3% improvement in objective pharmacological adherence (measured by MEMS®) and a 0.6% reduction in HbA1c, compared to the control group. However, qualitative analysis revealed that the primary mechanism of action was not the emulation of a human relationship, but the VAI's function as a tool for "cognitive offloading" that structured routines, managed reminders, and provided feedback, facilitating habit formation. Users developed operational trust based on technical reliability, not an emotional connection. The main barriers identified for scaling its implementation were the lack of integration with clinical information systems and the "digital affective gap," where purely instrumental interaction could exacerbate feelings of isolation in some patients. It is concluded that the VAI is an effective augmentation technology for self-care, but its sustainable success requires seamless integration into clinical workflows and design that acknowledges its limits in addressing complex psychosocial needs