Chat GPT excels in diagnosing patients’ conditions in emergencies

Dutch researchers have concluded that the AI-powered chatbot "Chat GP T" excels in diagnosing emergency room cases with accuracy comparable to, and in some cases surpassing, doctors. They suggest that artificial intelligence could "revolutionize the medical field."

However, the study's authors, whose results were published on Wednesday, emphasize that the jobs of emergency room doctors are not threatened. The AI chatbot may accelerate diagnosis, but it doesn't replace the human ability for discernment and experience.

A review of thirty cases treated in Dutch emergency departments in 2022 was conducted by feeding "Chat GP T" with patient data, laboratory tests, and doctor's notes, asking the AI chatbot to provide five possible diagnoses. Doctors correctly provided the accurate diagnosis in 87% of cases, compared to 97% for "Chat GP T" version 3.5.

Heidi ten Berge from the emergency department at Geroen Bosch Hospital in southern Netherlands noted that the chatbot "was able to make medical diagnoses just like a human doctor."

Researcher Steve Korstens, who participated in the study, emphasized that it does not suggest computers will one day run emergency rooms. 

Still, artificial intelligence can play a vital role in assisting doctors under pressure. He told AFP that the chatbot "can help in diagnosing and might propose ideas that haven't occurred to the doctor."

However, he pointed out that such tools are not designed as medical devices, sharing concerns about the confidentiality of sensitive medical data within the chatbot.

As is the case in other areas, "Chat GP T" showed some limitations. The study indicates that its logic "was occasionally unreasonable or medically inconsistent, which could lead to incorrect information or diagnoses, with significant consequences."

The scientists also acknowledge certain shortcomings in their research, such as the small sample size. Furthermore, the study was limited to relatively straightforward cases, involving patients with only one primary issue. Therefore, the effectiveness of AI in more complex cases remains unclear.

At times, "Chat GP T" failed to provide the correct diagnosis among the five possibilities, especially in the case of abdominal aortic aneurysm, a life-threatening complication with a swollen abdominal artery. 

However, in this instance, the human doctor also failed to diagnose it correctly.

The report also highlights "medical errors" committed by the AI chatbot, such as diagnosing anemia (low hemoglobin levels in the blood) in a patient with normal hemoglobin levels.

The study's results will be presented at the 2023 European Emergency Medicine Conference in Barcelona and were published in the specialized journal "Annals of Emergency Medicine."



AFP