Gastroenterologists at the University of Miami Miller School of Medicine and the University of Pennsylvania Perelman School of Medicine have developed a large language model for surveillance colonoscopy intervals.
Manual processes for applying guidelines can have wide variability and inconsistencies, which are reduced when using the LLM, according to a Dec. 3 news release from the University of Miami.
Physicians asked ChatGPT to determine the appropriate surveillance interval for 1,000 examples. The average accuracy of the AI across 10 experiments was 94.6%.
The LLM can be used to improve workflow efficiencies, and the next step is to integrate it into an electronic health record.
Daniel Sussman, MD, and Amar Deshpande, MD, co-authored the study.
