Balsa, JoãoFélix, IsaCláudio, Ana PaulaCarmo, Maria BeatrizSilva, Isabel Costa eGuerreiro, AnaGuedes, MariaHenriques, AdrianaGuerreiro, Mara Pereira2020-06-162020-06-162020Journal of Medical Systems (2020) 44: 130http://hdl.handle.net/10451/43839© Springer Science+Business Media, LLC, part of Springer Nature 2020In the context of the VASelfCare project, we developed an application prototype of an intelligent anthropomorphic virtual assistant. Designed as a relational agent, the virtual assistant has the role of supporting older people with Type 2 Diabetes Mellitus (T2D) in medication adherence and lifestyle changes. Our paper has two goals: describing the essentials of this prototype, and reporting on usability evaluation. We describe the general architecture of the prototype, including the graphical component, and focus on its main feature: the incorporation, in the way the dialogue flows, of Behavior Change Techniques, identified through a theoretical framework, the Behaviour Change Wheel. Usability was experimentally evaluated in field tests in a purposive sample of 20 participants (11 older adults with T2D and 9 experts). The Portuguese version of the System Usability Scale was employed, supplemented with qualitative data from open questions, diaries, digital notes and telephone follow-ups. The aggregated mean SUS score was 73,75 (SD 13,31), which corresponds to a borderline rating of excellent. Textual data were content analyzed and will be prioritized to further improve usability.engIntelligent virtual assistantsUsabilityRelational agentsBehavior changemHealth applicationsDiabetesjournal article10.1007/s10916-020-01583-w1573-689X