Assessment of Artificial Intelligence in Preclinical Management of ENT Emergencies
Authors List
You, W., Chen, J., Nightingale, J., Anderson, D., Morrissey, D.,
ENT Department, Toowoomba Base Hospital, Queensland, Australia
Introduction
Artificial intelligence (AI) based platforms are gaining popularity across all areas of life. This includes AI’s ability to act as a medical resource for clinicians, patients, and their families. There is potential application in answering medical questions and vignettes, thus providing patients with targeted information that is readily available, especially in acute settings.
Aim
This study aims to assess performance of AI platforms, including ChatGPT and Google’s Bard AI, as well as online resources in providing patient advice regarding otorhinolaryngological (ENT) emergency case vignettes.
Methods
Sixteen fictional case vignettes were designed to resemble queries of patients in the community that describe potential ENT emergencies. These were entered into ChatGPT and Google’s Bard in triplicate, followed by an online search for patient resources. Results were assessed for appropriateness by two ENT consultants, and presented in terms of diagnostic accuracy, recommendation to seek medical review, appropriateness of triage categorisation, and appropriate of preclinical measures.
Results
Both AI systems diagnosed 100% of conditions and all instances suggested seeking face-to-face medical review. However, there were limitations in differential diagnosis, and appropriate triage occurred in 76.7% of cases. Appropriate preclinical measures were outlined in 84.4% of instances, with scoring detailing no statistical difference between groups (P=0.5634) for grouped data with the Kruskal-Wallis test.
Conclusions
This study suggest that AI systems cannot replace medical review, but may help augment patient understanding. Although advices provided were appropriate in most cases, there is potential narrow focus and differential diagnoses provided by AI, and creates an issue for patients who may need to be considered for more serious, but rare, differential diagnosis. With further development, there is no doubt AI will play a greater role in pre-clinical management of ENT emergencies.
You, W., Chen, J., Nightingale, J., Anderson, D., Morrissey, D.,
ENT Department, Toowoomba Base Hospital, Queensland, Australia
Introduction
Artificial intelligence (AI) based platforms are gaining popularity across all areas of life. This includes AI’s ability to act as a medical resource for clinicians, patients, and their families. There is potential application in answering medical questions and vignettes, thus providing patients with targeted information that is readily available, especially in acute settings.
Aim
This study aims to assess performance of AI platforms, including ChatGPT and Google’s Bard AI, as well as online resources in providing patient advice regarding otorhinolaryngological (ENT) emergency case vignettes.
Methods
Sixteen fictional case vignettes were designed to resemble queries of patients in the community that describe potential ENT emergencies. These were entered into ChatGPT and Google’s Bard in triplicate, followed by an online search for patient resources. Results were assessed for appropriateness by two ENT consultants, and presented in terms of diagnostic accuracy, recommendation to seek medical review, appropriateness of triage categorisation, and appropriate of preclinical measures.
Results
Both AI systems diagnosed 100% of conditions and all instances suggested seeking face-to-face medical review. However, there were limitations in differential diagnosis, and appropriate triage occurred in 76.7% of cases. Appropriate preclinical measures were outlined in 84.4% of instances, with scoring detailing no statistical difference between groups (P=0.5634) for grouped data with the Kruskal-Wallis test.
Conclusions
This study suggest that AI systems cannot replace medical review, but may help augment patient understanding. Although advices provided were appropriate in most cases, there is potential narrow focus and differential diagnoses provided by AI, and creates an issue for patients who may need to be considered for more serious, but rare, differential diagnosis. With further development, there is no doubt AI will play a greater role in pre-clinical management of ENT emergencies.
