USE OF ARTIFICIAL INTELLIGENCE TOOLS IN EMERGENCY MEDICINE: A CROSS-SECTIONAL SURVEY AND NARRATIVE REVIEW

Abstract

The use of artificial intelligence (AI) tools in emergency medicine is expanding rapidly, with clinicians increasingly integrating platforms such as ChatGPT, Gemini, and DeepSeek into their daily workflows. Despite the growing prevalence of these tools, there remains a lack of formal guidance around their use, particularly in relation to data governance, medico-legal accountability, and clinical safety. Understanding current utilisation patterns, clinician perceptions, and safety concerns is essential to guide safe and effective AI adoption in emergency departments (EDs).

Objective: To evaluate awareness, utilisation, and clinician attitudes toward AI tools among emergency medicine professionals in the UK, and to identify practical concerns and recommendations for integrating AI safely into NHS emergency care.

Methods: A cross-sectional survey was conducted among emergency medicine clinicians across various NHS hospital types. A total of 31 participants completed the structured questionnaire, covering awareness, frequency of use, perceived safety, cognitive workload impact, and preferred safeguards. Responses were analysed descriptively, with data represented as frequencies and percentages.

Results: AI awareness was high, with 93.5% of respondents reporting familiarity with tools such as ChatGPT and Gemini. Over 83% reported occasional or regular use, predominantly for documentation (63%), clinical decision support (38.7%), prescribing guidance (38.7%), triage stratification (38.7%), and patient explanation (38.7%). While 77.4% believed AI could reduce cognitive workload, 87.1% cited data protection and GDPR compliance as the primary concern. Other concerns included medico-legal accountability (67.7%) and algorithmic bias (61.3%). All respondents supported NHS-hosted, GDPR-compliant AI tools, and 93.5% recommended formal training by Trusts. Open-ended responses emphasised the need for rigorous testing, clinician oversight, local hosting, and integration into existing EPR systems.

Conclusion: Emergency clinicians demonstrate strong awareness and emerging reliance on AI tools, primarily for documentation and clinical decision support. However, widespread adoption is hindered by significant data governance concerns and the absence of NHS-sanctioned infrastructure. There is urgent need for policy frameworks, training pathways, and secure AI integrations that align with legal, ethical, and clinical standards in the UK emergency care context.