HybReDe

The global need for rehabilitation is increasing. WHO estimates that about 2.4 billion people live with health conditions that may benefit from rehabilitation, and this need is expected to grow as populations age and chronic diseases become more common. Rehabilitation systems must therefore provide continuous, accessible and person-centred support with limited professional resources.

Repeated one-to-one rehabilitation can no longer meet current and future needs on its own. Distances, workforce shortages, busy everyday lives and changing service models challenge traditional delivery. Rehabilitation is also moving from primarily hands-on treatment towards more consultative and everyday-life-oriented support. Clients need guidance not only during therapy sessions, but also between sessions.

Hybrid rehabilitation, which combines face-to-face and remote support, is one response to this change. However, it does not remove the need for practical information and guidance between sessions, especially when a professional is not immediately available.

In HybReDe project, we develop AI-based virtual assistants to support hybrid rehabilitation. From the beginning, we involve service organisers, service providers, rehabilitation professionals, clients, and researchers and developers from rehabilitation and technology fields. This participatory approach is central because AI is still unfamiliar and sometimes intimidating to many actors in rehabilitation. AI raises questions about information sources, reliability, accountability, data protection and the appropriate role of technology in rehabilitation. We address these questions throughout the development process, not only at the final stage of implementation.

We conducted focus groups and development activities in three participating countries, and the results show that users value a virtual assistant when it is practical, understandable and easy to use. It should gather fragmented information into one place, support preparation for appointments, lower the threshold for asking rehabilitation-related questions and offer guidance between sessions. At the same time, users emphasise that it must complement human rehabilitation, not replace clinical decision-making.

Trust has emerged as a decisive condition. Users need to see what the assistant’s responses are based on. Visible sources, traceable information, technical stability, data security, clear responsibilities and staff training are therefore essential. The AI solution must fit real rehabilitation workflows and must not remain an isolated technology experiment.

We have developed and tested the virtual assistants step by step in three countries. We have tested them with rehabilitation students, clients and professionals, and the third and fourth versions are now entering or already in the pilot phase. We continue to refine the requirement specifications based on user experiences, feedback and pilot data.

In the pilots, we evaluate acceptability, usability and feasibility. We examine whether clients and professionals perceive the virtual assistant as useful, relevant, trustworthy and aligned with their needs. The pilots address MS rehabilitation in Finland, COPD rehabilitation in Poland and persistent pain rehabilitation in Estonia.

We document regulated AI-related questions, including data protection, ethics, accountability, technical safety and integration into clinical work. We also learn from different rehabilitation systems in the Baltic Sea region and document lessons learned for wider European use. HybReDe does not seek to show that AI can replace rehabilitation professionals. Instead, we examine how AI-based virtual assistants can complement human expertise in a way that is useful for clients, credible for professionals and realistic for implementation.