Data-driven framework for adverse manual wheelchair event risk classification and prevention

Manual wheelchairs are frequently associated with adverse incidents such as component failures, falls, and injuries. Despite existing research, there is a lack of systematic frameworks to identify high-risk scenarios and root causes across different wheelchair models. This study integrates Failure Mode and Effects Analysis (FMEA) with natural language processing (NLP) to generate a risk classification framework and propose guidelines to reduce risks. We analysed 752 adverse incidents reported by NHS WestMARC from 2020 to 2025. Events were categorised by five manual wheelchair models: standard, lightweight, ultralight, tilt-in-space, and paediatric. Relative risk (RR) metrics were computed for falls and injuries following component failures. FMEA was used to calculate severity, occurrence, and criticality (severity multiplied by occurrence) scores for each component. Relative risk analysis revealed that tilt-in-space, ultralight, and paediatric models had elevated but not significant risks of falls and injuries following failures (RR = 1.08-1.29). NLP identified contextual factors such as transfer-related incidents, outdoor navigation challenges, and belt misuse. High-risk components included casters (criticality = 2.35-6.8), belts (criticality = 2.27-5.62), brakes (criticality = 2.35-6.8), and frames (criticality = 1.88-5.62), along with incidents suffered by the wheelchair as a system (criticality = 4.09-17.82). Regression analyses confirmed associations between fall status, injury severity, and component failures. High-risk scenarios based on FMEA scores emphasise risk identification, user training, appropriate provision, design improvement, and technology development. The findings support a tiered prevention strategy involving design improvements, user training, and routine wheelchair servicing. Future opportunities include integrating NLP and FMEA into incident-reporting systems to inform clinical practice, policy, and product design, thereby enhancing wheelchair safety and user outcomes.