Year of HybReDe

Project background
The HybReDe project explores how virtual assistants can support hybrid rehabilitation in a safe, practical, and user-friendly way. The project brings together partners from four countries to develop and evaluate digital solutions that can support both rehabilitation professionals and people participating in rehabilitation.

What we did during the first year
It has been about a year since the kickoff meeting at Turku University of Applied Sciences, and a great deal has been accomplished in the HybReDe project. At the kickoff meeting, partners from Finland, Sweden, Estonia, and Poland came together. In line with the project’s theme, the kickoff meeting itself was organized in a hybrid format. Partners who were unable to attend onsite joined via video call.

After the meeting, we began building a shared understanding of hybrid rehabilitation by conducting a literature review and interviews with management-level representatives of relevant professional groups. In addition, a pre-pilot study of VA was conducted with people undergoing rehabilitation, and professionals, and VA was developed based on the results of that study . Hybrid rehabilitation, which combines in-person visits with remote services, offers a promising model for improving the accessibility and continuity of rehabilitation services. New technologies, such as artificial intelligence and virtual assistants, can support both rehabilitation professionals and people participating in rehabilitation, but their implementation requires clear operational models, appropriate safeguards, and user acceptance.

All four countries recognize the potential of hybrid rehabilitation to improve accessibility, continuity, and patient engagement. Similarly, virtual assistants are seen as valuable tools for supporting daily routines, providing reminders, and strengthening communication and motivation. These findings underline a shared understanding that digital solutions can complement traditional rehabilitation models rather than replace them.
During the first year, we also developed the first two versions of our virtual assistant demos for national focus groups. There is separate version for rehabilitation clients and separate version for rehabilitation professionals. Also the focus groups were conducted for clients and professionals, separately. The VA prototypes were developed for different purposes: in Estonia, for people with chronic pain; in Poland, for people with COPD; and in Finland, for people with multiple sclerosis (MS). Although the content differed between countries and target groups, the core principles of the virtual assistants remained the same. Pilot activities are being carried out in Estonia, Poland, and Finland, while Sweden is responsible for developing the evaluation framework for the pilots.

In Finland, the virtual assistant is designed to support people with MS in managing fatigue-related symptoms in everyday life. Users are encouraged to ask practical questions related to daily routines, coping strategies, and symptom management.
The national focus group activities provided valuable insight into both the opportunities and the challenges associated with virtual assistants from the perspectives of different patient and professional groups. Some rehabilitation clients were very impressed by the VA and praised it highly. One participant described the VA as “being somewhere between Google and a doctor”. At the same time, important concerns were raised, which is valuable from a development perspective. Participants. especially rehabilitation professionals, questioned the reliability of the VA and whether the information it provides is evidence-based.

Developing and testing the virtual assistants
The development of the VA began with a Retrieval-Augmented Generation (RAG) approach. In practice, this means that the VA, built on the OpenAI ChatGPT-4 model, was provided with evidence-based background material related to the relevant patient group, such as people with multiple sclerosis, and guided to base its answers on that material. The aim was to ensure that the VA uses verified information as the basis for its responses.
We then continued the development process by moving on to a fine-tuning method to further improve the VA and increase its reliability. In the fine-tuning process, multiple question–answer pairs were created, and the answers were verified by professionals with extensive experience in treating the relevant patient groups. This method is expected to improve the consistency of the VA’s responses and, consequently, strengthen its reliability.
One key element in preparing for the next phase of the project has been the development of evaluation procedures, an analytical framework, and a regulatory basis for pilot testing. In the pilot phase, we aim to assess the acceptability, usability, and feasibility of the VA designed to support rehabilitation-related information and behaviour. These three dimensions are complementary and together determine whether a digital health intervention is likely to be adopted and sustained in practice.

What happens next
In the next phase, we will begin the official pilot phase, during which rehabilitation clients and professionals will use the virtual assistant as part of the rehabilitation process. We will evaluate whether users feel comfortable using the VA, whether it is easy to use, and whether it can realistically be integrated into rehabilitation practice.

Programme Manual Online

Choose