JMIR Publications

ABSTRACT

Background:

Stroke affects over 12 million people annually, with up to 80% experiencing gait impairments three months post-event. While rehabilitation can improve mobility, outcomes often vary due to limited personalization of conventional therapy. Real-time gait biofeedback, defined as the delivery of immediate sensory information about motor performance to guide movement, shows promise for enhancing motor learning. However, current biofeedback designs are typically developed without end-user input, potentially reducing end-users’ understanding, enjoyment, and engagement.

Objective:

The objective of this study was to investigate the user experience of people with stroke involved in co-designing personalized real-time gait biofeedback and its associated virtual environment for rehabilitation. A mixed-methods approach was used to identify preferred biofeedback features and assess the impact of personalization on task load and motivation.

Methods:

Ten post-stroke participants completed multiple sessions categorized as kick-off, discovery, or consolidation, each lasting approximately 1 hour. After an initial assessment of gait deficits, an iterative co-design process was used to progressively refine the biofeedback and virtual environment based on individual input. During these sessions, participants walked in the CAREN while receiving real-time biofeedback based on a therapist-selected gait parameter and collaborated with the experimenter to modify feedback modalities and design. Qualitative data were collected via unstructured interviews to determine the preferred biofeedback features, while quantitative data were collected through the raw NASA-Task Load Index and Intrinsic Motivation Inventory.

Results:

Each participant developed a unique biofeedback and virtual environment setup. Despite this, common preferences emerged, including positive feedback, avatars that mirrored participant’s movements, and multimodal visual-auditory hit/error indicators. Age-related differences in motivation were observed across sessions, with participants over 60 showing greater motivation as personalization evolved. When examining the percent change in scores between the first discovery and last consolidation sessions, no significant changes were found in task load (t₈=–1.13, adjusted P=.366), value/usefulness (z=21.00, adjusted P=.858), or effort/importance (t₉=1.29, adjusted P=.366). However, a marginally significant 13% increase in perceived competence (z=0.00, adjusted P=.058) suggested participants felt more capable of achieving the target parameter value with personalized feedback. Success/error counters led to minor individual-level changes in enjoyment, effort, and competence, though no single feature significantly influenced overall outcomes.

Conclusions:

This study highlights the value of co-design in developing personalized biofeedback for gait rehabilitation, revealing consistent preferences for features such as positive feedback, avatars, and multimodal cues. Age-related differences in motivation also emerged, suggesting that personalization may be particularly beneficial for older adults. Clinical Trial: Ethical approval was obtained from the ETH Zurich Ethics Commission (EK-2022-N-173).