Artificial Intelligence Chatbots in Pediatric Emergencies: A Reliable Lifeline or a Risk?

Area of Science:

• Digital Health
• Artificial Intelligence in Medicine
• Emergency Medicine

Background:

• AI chatbots, including ChatGPT, are increasingly used for health information dissemination.
• While effective for adult guidelines, AI chatbot performance in pediatric basic life support (BLS) is not well-established.
• Parents frequently seek online health guidance for pediatric emergencies, highlighting the need for reliable AI chatbot information.

Purpose of the Study:

• To compare the accuracy and completeness of ChatGPT-4o and ChatGPT-4o mini responses against American Heart Association (AHA) pediatric BLS guidelines.
• To assess the suitability of AI chatbots for providing critical guidance during pediatric emergencies.
• To identify areas for AI chatbot improvement in emergency response frameworks.

Main Methods:

• A prospective comparative content analysis was performed.
• ChatGPT (versions 4o and 4o mini) responses were evaluated against the 2020 AHA Guidelines for pediatric BLS.
• 13 questions covering key pediatric BLS components were used, with responses assessed for completeness and conformity to AHA standards.

Main Results:

• Only 9.61% of ChatGPT responses were fully addressed and 5.77% fully conformed to AHA pediatric BLS guidelines.
• A significant portion of responses were partially addressed (61.54%) or superficially conformed (59.61%).
• ChatGPT-4o provided more comprehensive responses than ChatGPT-4o mini, though neither fully met AHA standards.

Conclusions:

• AI chatbots lack the necessary accuracy, depth, and practical instruction for life-saving pediatric BLS procedures.
• Reliance on incomplete or inaccurate AI chatbot information in emergencies can lead to critical errors.
• Widespread formal basic life support training is essential for effective emergency response.