A multinational cross-sectional survey on the use of AI-based voice assistance systems in emergency medical services.

Abstract

Background: Emergency medical services (EMS) worldwide face increasing demands that require improvements in efficiency and quality of care. Voice assistance systems (VAS) based on Artificial Intelligence (AI) can support staff in prehospital settings through contactless, intuitive operation. However, evidence regarding their application in EMS is limited but desired. To date, no large-scale study has systematically explored potential functions, user acceptance, or implementation challenges of voice assistance systems in prehospital emergency medicine. This study aimed to explore the possibilities and challenges of designing and implementing a VAS tailored for EMS from a user perspective. Emergency medical personnel from three European countries were invited to share their requirements, expectations, and anticipated obstacles across different systems and countries.

Methods: An exploratory, cross-sectional online survey was conducted in Germany, Norway, and Switzerland during the summer of 2024 using a 34-item self-administered questionnaire. Eligible participants were EMS professionals, including physicians, paramedics, emergency medical technicians, and students. Collected data were analyzed using descriptive statistics and a large language model. The reporting of this study adhered to the CROSS checklist.

Results: A total of 587 responses were received, of which 186 (32%) were excluded, leaving 401 responses for final analysis. Participants reported occasional use of AI applications and voice assistants in personal or work settings and demonstrated a high level of technical proficiency. In EMS, several digital tools are in use today, mainly concerning documentation, knowledge access, hospital pre-notification and occupancy checks. Most respondents reported a rather positive attitude towards the use of voice assistants during missions, expecting reduced workload and improved quality of care; however, none were aware of an EMS-specific voice assistant to date. Desired features included language translation, patient history summarization, hospital pre-notification, medical information retrieval, and case documentation. Key requirements for adoption were reliability, usability, and fast system response times, while major implementation challenges involved staff acceptance, funding, and data protection.

Conclusion: Developing AI-based voice assistants with emergency medical service-specific functionalities is desirable and likely to be well accepted if user requirements are addressed. Future steps include prototype-based field testing and evaluating measurable improvements in care delivery.

Clinical trial number: Not applicable.