Automated video-based AVPU assessment within a FHIR-enabled clinical decision support framework

Area of Science:

• Medical Informatics
• Computer Vision in Healthcare
• Clinical Decision Support Systems

Background:

• Inconsistent documentation of the AVPU scale (Alert, Verbal, Pain, Unresponsive) due to high workload and fragmented workflows impacts patient safety.
• Accurate consciousness assessment is crucial for emergency triage, care escalation, and overall patient safety.

Purpose of the Study:

• To develop and evaluate Consc.ia, a video-based clinical decision-support platform for automated AVPU scale inference.
• To enable seamless, interoperable documentation through HL7 FHIR integration.
• To assess technology adoption and deployment scenarios for emergency and hospital settings.

Main Methods:

• Developed a video-based platform (Consc.ia) using edge-computing computer vision for real-time AVPU assessment.
• Integrated a clinician-in-the-loop validation layer and HL7 FHIR for EHR interoperability.
• Simulated AVPU dataset with 136 videos from 58 healthcare professionals; modelled technology adoption using Rogers' and Bass Diffusion models.

Main Results:

• The Consc.ia architecture achieves low-latency inference with privacy-by-design.
• Stakeholder validation confirmed workflow fit but highlighted documentation gaps during EMS-to-hospital transitions.
• Bass modelling projects gradual adoption, reaching ~50% of Intermediate Care wards by 2037, with early clinical evidence and FHIR integration as key adoption accelerators.

Conclusions:

• Consc.ia presents a feasible, interoperable proof-of-concept for standardizing AVPU documentation and enhancing early warning systems.
• The platform addresses digitalization gaps in emergency care by combining video analytics, edge computing, clinician validation, and FHIR integration.
• Further validation through empirical model evaluation against expert-annotated clinical recordings is required for clinical translation.