8273673 Innovative AI Ergonomic risk assessment in emergency operators: a multimodal approach

<h3>Objective</h3> This study proposes a multidisciplinary approach to evaluate biomechanical risk for Emergency Medical Technicians (EMT’s), combining standardised ergonomic assessment methods, AI-based posture analysis and field data for a more comprehensive strategy. <h3>Material and Methods</h3> A three-person EMT crew performed emergency operations in a simulated environment. Postures adopted during these activities were assessed by an ergonomist using standardised methods (REBA, OWAS, SUVA). Push and pull forces were measured with dynamometer and Snook-Ciriello guidelines. An artificial intelligence system integrating YOLOv8 (person detection), Deep SORT (tracking) and Media Pipe Pose (joint landmark estimation) was implemented to automatically extract joint angles and identify high-risk postures from video recordings. Operational data from 28,936 emergency calls in 2024 at 15 emergency stations were analysed to estimate exposure levels based on the frequency of calls per shift and, with MAPO method adapted for emergency medical services, tried to assess the risks associated with patient handling. <h3>Results</h3> In the simulated activities, approximately 40% of the postures observed were classified as high risk according to REBA, SUVA and OWAS criteria. Measurements of push and pull forces were generally within Snook-Ciriello recommended limits, although several critical issues were identified. The AI-based system assigned a numerical value to each joint angle for each body segment frame-by-frame, calculating a risk score to each frame, showing reasonable agreement with the risk classifications assigned by the professional ergonomist, allowing a more precise quantitative analysis of workers single postures. Analysis of operational data allowed an appraisal based on average interventions per shift (≤1, between 1-2, ≥2). Stations with more than two interventions per shift had MAPO scores ≥2.5, suggesting a potential increase in biomechanical overload risk due to patient handling. <h3>Conclusions</h3> The methodology proposed in this study integrates standardised risk evaluation methods, field measurements, AI-driven video analysis and frequency data to obtain more and detailed specific risk profiles. This innovative approach may provide a useful contribution to improve tailored preventive strategies to reduce overload risk.