1. Bern University of Applied Sciences
  2. Research + Consulting services
  3. Projects
  4. RehaBot – AI as an interface between patients and reha professionals
See table of contents See table of contents

RehaBot – AI as an interface between patients and reha professionals

This project explores how patients can be supported in rehabilitative aftercare by a chatbot based on artificial intelligence. The project is part of the Care@Home initiative of the Canton of Bern.

Factsheet

Situation

After discharge from inpatient rehabilitation, everyday life is challenging for many patients. Failure to adjust to the new situation and, if necessary, adapt their lifestyle can lead to a worsening of the underlying disease or rehospitalisation. A chatbot based on artificial intelligence ("RehaBot") linked to health data would give patients the opportunity to receive targeted answers to their individual questions relating to their personal situation. The RehaBot is intended to help patients reintegrate at home, actively promote a healthy lifestyle and, if necessary, network with the relevant players in aftercare to ensure a seamless transition from secondary to tertiary prevention. The improved and more sustainable care should reduce re-admissions. It should also ensure the resource-efficient deployment of specialists.

Course of action

This pilot project explores the use of artificial intelligence and large language models such as GPT-4 in the context of a highly individualised eHealth application. A new model of care for rehabilitative aftercare is being addressed. The needs of those affected (patients and relatives) and professionals in rehabilitative aftercare are recorded, a RehaBot prototype for an exemplary use case "Management of malnutrition and sarcopenia" is created and piloted, and the quality is evaluated. A new RehaBot is intended to provide patients with needs-based, customised information according to their individual health history, regardless of the appointment, without direct consultation with a specialist. However, patients' specific questions and interactions with the RehaBot could recommend or trigger the involvement of appropriate specialists and thus support interdisciplinary aftercare. The opportunities and limitations of the application will also be explored.

Result

The project explored the possibilities and challenges of introducing an AI-based digital aftercare programme. A functioning prototype was developed (figure) that can conduct personalized conversations based on relevant health data. The chatbot summarizes the conversations. Through targeted prompt engineering, the RehaBOT has been specially designed for patient needs and medical conversations. The conversation history is saved in order to be able to provide contextually precise answers over longer periods of time. In addition, the chatbot was connected to clinical data processing software (REDCap), allowing patient-specific data to be integrated. The prototype showed solid technical potential, but was received differently by patients. Patients with cardiac diseases showed the highest willingness (32%) to use a programme such as RehaBOT after discharge. Functions such as the automatic recognition and forwarding of critical health values to specialists were perceived with particular interest. Furthermore, approximately half of the patients indicated a desire for occasional contact with specialists at the rehabilitation clinic. Here, too, the RehaBOT could serve as an intermediary. With various adaptations, the RehaBOT could become an innovative tool to personalize and sustainably improve rehabilitation aftercare.

Contact

Prof. Dr. Kai-Uwe Schmitt

  • Leiter APP Insel Gruppe
  • +41 31 848 37 10
  • Show e-mail

This project contributes to the following SDGs

  • 3: Good health and well-being