The AI dispatcher copilot: beyond cardiac arrest recognition to dynamic large language model-assisted Tele-CPR

Area of Science:

• Emergency medicine
• Artificial intelligence
• Cardiopulmonary resuscitation

Background:

• Out-of-hospital cardiac arrest (OHCA) is a major cause of mortality.
• Current AI in emergency medical dispatch has a "translation gap," showing diagnostic skill but limited clinical impact.
• Existing systems lack dynamic, multimodal capabilities for real-time decision support.

Purpose of the Study:

• To propose a paradigm shift in Tele-CPR systems using a multimodal large language model (LLM)-enabled "AI dispatcher copilot."
• To move beyond passive AI-assisted OHCA recognition towards a dynamic, integrated AI decision-support architecture.
• To align with European Resuscitation Council (ERC) Guidelines 2025 for emergency medical dispatch.

Main Methods:

• Narrative synthesis of recent advancements in LLMs, computer vision, and cognitive load theory.
• Evaluation of the conceptual feasibility of an AI-driven decision-support architecture for emergency medical dispatch.
• Integration of acoustic signal analysis and caller narrative interpretation for OHCA recognition.

Main Results:

• Multimodal LLMs can integrate acoustic data (e.g., agonal breathing) and semantic analysis to reduce OHCA recognition ambiguity.
• The proposed system offers adaptive instruction generation to lower caller cognitive load.
• Features include video-assisted CPR coaching with feedback and automated resource orchestration (first responders, AED routing).

Conclusions:

• The AI dispatcher copilot offers a transformative potential for Tele-CPR systems.
• Clinical translation necessitates ethical governance for bias and privacy concerns.
• Prospective validation is crucial to confirm improvements in neurologically intact survival.