earticle

영어

The goal of this study was to confirm the consistency between the answers recommended in the KTAS instructor manual and the KTAS results classified by ChatGPT. The consistency between ChatGPT's classification results and the instructor manual's recommended results was evaluated for nine adult cases, five pediatric cases, and ten special cases provided in the KTAS instructor manual. As a result of the analysis, the agreement between the KTAS instructor manual and ChatGPT was Cohen’s kappa of .518, Cohen’s kappa of .429 for adults, Cohen’s kappa of .286 for pediatric, and Cohen’s kappa of .625 for special cases. In addition, the kappa of the disease was .511 and the kappa of the exclusion disease was .474. The ICC for agreement between the KTAS instructor‘s manual and ChatGPT was .798, confirming a high overall agreement. This study confirmed that ChatGPT has a certain level of reliability as a KTAS classification tool in the field of emergency medicine, but it was found that additional data is needed, especially for cases pediatric or excluding diseases.

한국어

본 연구는 KTAS 강사 매뉴얼의 추천답안과 ChatGPT가 분류한 KTAS 결과 일치도를 확인하는 것을 목표로 하였다. KTAS 강사 매뉴얼에 제공된 성인 9사례, 소아 5사례, 특수상황 10사례, 총 24사례를 대상으로 ChatGPT가 분류한 결과와 KTAS 강사매뉴얼의 추천답안 간의 일치도를 평가하였다. 분석 결과, KTAS 강사매뉴얼과 ChatGPT 간의 일치에 대한 Cohen’s kappa .518로 중증도의 일치도를 확인하였 다. 성인 Cohen’s kappa .429, 소아 Cohen’s kappa .286, 특수 상황 Cohen’s kappa .625이었다. 또한, 질병 Cohen’s kappa는 .511, 질병 외 Cohen’s kappa는 .474이었다. KTAS 강사매뉴얼과 ChatGPT 간의 일치에 대한 ICC는 .798로 전반적인 일치도가 높음을 보여주었다. 본 연구는 ChatGPT가 응급의료 분야에서 KTAS 분류 도구로서 일정 수준의 신뢰성을 보일수 있음을 확인하였지만, 특히 소아나 질병 외의 사례에 대한 추가적인 자료가 필요함을 알 수 있었다.