AI Clinical Support in Disaster Medicine: A Delphi Study

Background

Since the 1950s, artificial intelligence (AI) technologies were mostly unavailable to disaster medicine (DM) practitioners. The introduction of ChatGPT in 2022 led to a surge of proposed AI applications in disaster medicine. However, AI development is primarily driven by vendors in high-income countries, and practitioner needs are not well understood. This study offers a global perspective on the clinical challenges that DM practitioners want AI to address.

Methods

An online Delphi study was conducted with 131 international DM experts over three rounds. In the first round, experts were asked: "What specific clinical questions or problems in Disaster Medicine would you like to see addressed by artificial intelligence guided clinical decision support?" The statements from the initial round were analyzed and compiled for subsequent rounds. Participants then rated these statements for importance using a 7-point linear scale.

Results

In round one, 77 participants proposed 539 statements, which were then collated into 47 statements for later rounds. Round two saw 89 participants provide 3008 ratings, but consensus was not reached on any statement. In round three, 63 participants submitted 2942 ratings, and five statements achieved consensus: hospital distribution of disaster patients, estimating the size of the affected population, hazard vulnerability analysis, resource acquisition and distribution, and transportation routing. Experts generally disagreed with using AI for ethics, mental health, cultural sensitivity, or complex treatment decisions.

Conclusions

This online Delphi study revealed that DM practitioners favor AI tools that support the logistical aspects of their clinical duties. Participants showed less support for AI involvement in difficult or critical decisions. The development of AI for clinical decision support should prioritize user needs and be informed by a global viewpoint.