User Preferences and Information Gaps for AI-Assisted Selection of Engineered Devices for Home Rehabilitation After...

There has been a proliferation of engineered devices for movement rehabilitation, but their uptake remains sporadic. Using a mixed-methods approach with qualitative interviews and quantitative surveys, we asked nine post-stroke individuals to identify their information needs and to inform the design of an AI-powered decision support system to help select devices for home movement training after stroke. Key themes that emerged were: 1) difficulties finding information about device efficacy and motivational features and 2) openness to being helped by an AI chatbot. In a follow-up ranking exercise, participants indicated that the decision support system should provide information on the frequency of use required to get a benefit, ease of use, and the amount of expected benefit. They also highly rated potential features such as being able to provide personal information for customized recommendations, ask questions, and receive non-sales-driven guidance. In parallel, we analyzed 67 FDA-listed devices for movement rehabilitation. Of those with cost information, 43% were under $1,000, and 45 % had insurance reimbursement. We emailed companies to get more data about motivational features and expected efficacy, but only 30 % responded. These results confirm that post-stroke users have numerous devices available for home therapy but face substantial barriers to obtaining the information they want. Further, they are enthusiastic about using an AI chatbot to help find devices.