Research News

Meta-analysis of medical research with LLMs reveals diagnostic accuracy

Generative AI and diagnostic accuracy

Large language models show potential on par with non-specialist doctors in the medical field.

Credit: Osaka Metropolitan University

The use of generative AI for diagnostics has attracted attention in the medical field and many research papers have been published on this topic. However, because the evaluation criteria were different for each study, a comprehensive analysis was needed to determine the extent AI could be used in actual medical settings and what advantages it featured in comparison to doctors.

A research group led by Dr. Hirotaka Takita and Associate Professor Daiju Ueda at Osaka Metropolitan University’s Graduate School of Medicine conducted a meta-analysis of generative AI’s diagnostic capabilities using 83 research papers published between June 2018 and June 2024 that covered a wide range of medical specialties. Of the large language models (LLMs) that were analyzed, ChatGPT was the most commonly studied.

The comparative evaluation revealed that medical specialists had a 15.8% higher diagnostic accuracy than generative AI. The average diagnostic accuracy of generative AI was 52.1%, with the latest models of generative AI sometimes showing accuracy on par with non-specialist doctors.

“This research shows that generative AI’s diagnostic capabilities are comparable to non-specialist doctors. It could be used in medical education to support non-specialist doctors and assist in diagnostics in areas with limited medical resources.” stated Dr. Takita. “Further research, such as evaluations in more complex clinical scenarios, performance evaluations using actual medical records, improving the transparency of AI decision-making, and verification in diverse patient groups, is needed to verify AI’s capabilities.”

The findings were published in npj Digital Medicine.

Paper information

Journal: npj Digital Medicine
Title: A systematic review and meta-analysis of diagnostic performance comparison between generative AI and physicians
DOI: 10.1038/s41746-025-01543-z
Authors:  Hirotaka Takita, Daijiro Kabata, Shannon L Walston, Hiroyuki Tatekawa, Kenichi Saito, Yasushi Tsujimoto Yukio Miki, Daiju Ueda
Published: 22 March 2025
URL: https://doi.org/10.1038/s41746-025-01543-z

Contact

Hirotaka Takita
Graduate School of Medicine
z21227o[at]omu.ac.jp

*Please change [at] to @.