JMIR Publications

ABSTRACT

Background:

Artificial intelligence (AI) is increasingly being considered for rehabilitation applications including personalised exercise, progress monitoring, risk identification, documentation, remote rehabilitation, and technology-assisted interventions. However, successful implementation depends not only on technical performance but also on patient and carer acceptability, trust, and willingness to engage with AI-supported care.

Objective:

This study aimed to evaluate patient and carer perspectives on AI-supported rehabilitation within a National Health Service (NHS) rehabilitation service, with particular focus on perceived benefit, willingness to use AI-supported rehabilitation, and the influence of clinician oversight on trust.

Methods:

A cross-sectional paper-based survey was conducted at a single NHS rehabilitation service between May and July 2026. Adult patients and relatives or carers completed items assessing AI familiarity, digital confidence, perceived benefits and acceptability of AI applications, willingness to choose and recommend AI-supported rehabilitation, and trust with and without clinician review. Questionnaires with missing responses were excluded using complete-case analysis. Paired trust responses were compared using the Wilcoxon signed-rank test, with an exact McNemar test used as a sensitivity analysis. A composite perceived-benefit score was evaluated using internal consistency and exploratory factor analysis. Associations with willingness were examined using Spearman correlations and exploratory proportional-odds ordinal logistic regression.

Results:

Of 112 questionnaires collected, 106 complete questionnaires were included, comprising 46 patients (43.4%) and 60 relatives or carers (56.6%). Attitudes toward AI-supported rehabilitation were cautiously favourable, with agreement highest for personalisation of rehabilitation goals and exercises (59.4%) and exploration of AI-supported rehabilitation technologies (56.6%). Trust was substantially higher when clinician oversight was present: median trust increased from 3.0 (IQR 2.0–4.0) without clinician review to 4.0 (IQR 3.0–4.0) with clinician review. Overall, 60/106 participants (56.6%) reported greater trust with clinician oversight, and agreement or strong agreement increased from 29.2% to 61.3% (Wilcoxon W=120.0, P<.001; rank-biserial correlation=0.885). The perceived-benefit items demonstrated excellent internal consistency (Cronbach α=0.952; McDonald ω=0.953) and supported a single-factor composite. Perceived benefit was strongly associated with willingness to choose AI-supported rehabilitation (Spearman ρ=0.784, P<.001) and remained the only independently associated predictor in the adjusted ordinal regression model (adjusted OR 12.44, 95% CI 6.36–24.33; P<.001).

Conclusions:

Patients and carers showed cautious but meaningful openness to AI-supported rehabilitation. Trust was strongly dependent on continued clinician involvement, supporting a human-in-the-loop model rather than autonomous AI implementation. Perceived practical benefit appeared central to willingness to adopt AI-supported rehabilitation. Implementation strategies should therefore combine visible clinician oversight with clear demonstration of patient benefit and support tailored to individual digital confidence. Clinical Trial: N/A