The potential of LLMs in generating questions and answers with EHRs - Takeaways - MDSpire

The potential of LLMs in generating questions and answers with EHRs

  • By

  • Yunqi Zhu

  • Wen Tang

  • Huayu Yang

  • Jinghao Niu

  • Liyang Dou

  • Yifan Gu

  • Yuanyuan Wu

  • Wensheng Zhang

  • Ying Sun

  • Xuebing Yang

  • July 15, 2026

Share

  • 1

    The study evaluated the ability of eight large language models (LLMs) to generate medical exam questions and answers from electronic health records.

  • 2

    ERNIE 4 achieved the highest score for question generation, while human experts excelled in providing sufficient key information.

  • 3

    LLMs outperformed human experts in information correctness, with ERNIE showing significantly higher correctness scores.

  • 4

    For answer generation, human experts scored higher overall, but Doubao outperformed other LLMs in coherence and professionalism.

  • 5

    The study highlights the potential of LLMs to assist in medical education, despite current limitations in some performance aspects.

Original Source(s)

The potential of LLMs in generating questions and answers with EHRs

  • by Yunqi Zhu, Wen Tang, Huayu Yang, Jinghao Niu, Liyang Dou, Yifan Gu, Yuanyuan Wu, Wensheng Zhang, Ying Sun, Xuebing Yang

  • July 15, 2026

Related Content

Clinical cognition in the age of cardiovascular AI

Low IMA Ligation Did Not Significantly Reduce Leaks

A randomized trial found similar short-term morbidity and 12-month functional outcomes with both approaches, while highlighting the need to balance perfusion with a tension-free anastomosis.

Contextual recommendation modeling in eCoaching with machine learning, X-AI, and semantic ontology