Applications of Artificial Intelligence Guided Clinical Decision Support in Disaster Medicine: A Delphi Study

Abstract

Introduction: The rapid evolution of technology has spurred diverse applications in medicine, including clinical decision support. Despite numerous studies validating the potential of AI-assisted clinical decision support (AI-CDS), the desired applications by Disaster Medicine clinicians remain unclear. This study aims to address this gap through consensus-building research with international disaster medicine experts. Methods: Using the traditional Delphi method, an international panel of disaster medicine practitioners was assembled. An open-ended questionnaire in Round 1 elicited prospective consensus statements on AI-CDS applications. These responses were organized into consensus statements for subsequent rounds. A 7-point linear numeric scale was employed in Round 2 to rank the statements. Statements with a standard deviation of 1.0 or less were considered consensus. Results from Round 2 will be shared with each expert, and they will be asked to reconsider their ranking as part of Round 3. An interim analysis will be conducted to determine if a 4th round is necessary. Statements that pass the consensus cutoff in either rounds 3 or 4 will be included in the final analysis. Results: In Round 1, 539 statements were obtained from 77 participants (38% female; 62% male) across 47 countries representing all 7 World Bank Global Regions. Key concerns included triage, training, communication, mental health, organization, and disaster planning. These proposals were condensed into 47 statements for Round 2, with 56 participants completing their assessment so far. Statements scoring above 5.7 out of 7 highlighted AI support would be beneficial in estimating population at risk, disaster medicine training, resource coordination, HVA assistance, surge capacity, patient distribution, and enhancing culturally sensitive and multilingual communication. Conclusion: This Delphi study highlights the varied and critical applications of AI-CDS desired by disaster medicine experts. The findings will provide a foundation for future research and innovation, aligning AI development with the priorities of disaster medicine practitioners.

Ähnliche Arbeiten