JMIR Publications

ABSTRACT

Background:

Clinical gait assessment is essential for monitoring functional progress in children with cerebral palsy (CP), yet traditional visual observation remains inherently subjective and labour-intensive. Digital gait assessment tools powered by artificial intelligence hold significant potential for providing objective and automated assessment. However, a primary bottleneck in digital health is ensuring that such tools possess high usability and seamless alignment with clinical environments. This "implementation gap" persists because technical developments are frequently decoupled from the practical workflows and specific needs of healthcare professionals.

Objective:

This study aimed to iteratively develop and evaluate a clinician-centred, automated gait analysis system for children with CP by employing a structured co-design process, ensuring the tool effectively supports clinical decision-making and integrates into routine practice.

Methods:

This study adopted the FRamework for co-dESign of Clinical practice tOols (FRESCO) to guide a five-step iterative development process. A multidisciplinary advisory group, comprising rehabilitation physicians, therapists, biomechanics experts, and human factors engineers, collaborated throughout the study. The process involved: (1) identifying baseline clinical needs; (2) developing an initial system prototype; (3) conducting think-aloud evaluations and System Usability Scale (SUS) assessments; (4) testing workflow integration and safety in clinical simulations; and (5) finalizing specifications via a consensus workshop.

Results:

By targeting workflow alignment, interpretability, and the delivery of useful system outputs, the iterative co-design process effectively enhanced system usability. The prototype demonstrated good usability during the evaluation phase (mean SUS = 75.8) and was confirmed to be stable and safe in subsequent clinical simulations. This work culminated in two key outputs: a refined prototype ready for large-scale clinical testing and a comprehensive implementation guideline. Conceptually, the study articulated transferable design principles: prioritising interpretability over raw data, ensuring compatibility with time-constrained workflows, and minimising operational barriers.

Conclusions:

This study demonstrates that FRESCO framework successfully bridges the gap between technical feasibility and clinical utility. By iteratively involving a multidisciplinary team, we transformed a video-based analysis tool into a stable, interpretable, and workflow-compatible decision-support system. Broadly, the identified principles offer a generalisable perspective for developing clinician-facing digital health tools intended to support high-stakes decision-making in routine practice.