US researchers concluded in a study that ChatGPT, a linguistic model developed using artificial intelligence (AI), has 72% accuracy in making general clinical decisions. The report was published in a prestigious journal Journal of Medical Internet Research.
The study showed that the famous chat program obtained “good results in both primary care and emergency situations and in various specialties.” Apparently ChatGPT “comprehensively evaluates decision support from the beginning of working with the patient and throughout the entire care scenario“From differential diagnosis to testing, diagnosis and treatment,” said Mark Suchi, one of the authors of the paper.
The software’s accuracy was measured using a system that records the chatbot’s responses. And the researchers found that ChatGPT provides “60.3% accuracy in determining differential diagnosis based solely on HPI (History of Present Illness), PE (Physical Examination) and ROS (Review of Systems) On the other hand, 76.9% accuracy in final diagnosis with additional information managementsuch as related diagnostic tests.
from here, They concluded that ChatGPT had an average performance of 71.8% across all question typesIt thus positions itself as a potential ally of health professionals in the context of clinical diagnosis.
“There are no real standards, but we estimate ChatGPT performance to be at the level of someone who just graduated from medical school, such as an intern or residentsaid the study authors.
