Skip to main content
QUICK REVIEW

[论文解读] MEDEC: A Benchmark for Medical Error Detection and Correction in Clinical Notes

Asma Ben Abacha, Wen-wai Yim|arXiv (Cornell University)|Dec 26, 2024
Electronic Health Records Systems被引用 8
一句话总结

MEDEC 是首个公开可用的基准,用于检测和纠正临床笔记中的医疗错误,在五种错误类型下评估 LLMs 和医生在错误检测、定位和纠正方面的表现。

ABSTRACT

Several studies showed that Large Language Models (LLMs) can answer medical questions correctly, even outperforming the average human score in some medical exams. However, to our knowledge, no study has been conducted to assess the ability of language models to validate existing or generated medical text for correctness and consistency. In this paper, we introduce MEDEC (https://github.com/abachaa/MEDEC), the first publicly available benchmark for medical error detection and correction in clinical notes, covering five types of errors (Diagnosis, Management, Treatment, Pharmacotherapy, and Causal Organism). MEDEC consists of 3,848 clinical texts, including 488 clinical notes from three US hospital systems that were not previously seen by any LLM. The dataset has been used for the MEDIQA-CORR shared task to evaluate seventeen participating systems [Ben Abacha et al., 2024]. In this paper, we describe the data creation methods and we evaluate recent LLMs (e.g., o1-preview, GPT-4, Claude 3.5 Sonnet, and Gemini 2.0 Flash) for the tasks of detecting and correcting medical errors requiring both medical knowledge and reasoning capabilities. We also conducted a comparative study where two medical doctors performed the same task on the MEDEC test set. The results showed that MEDEC is a sufficiently challenging benchmark to assess the ability of models to validate existing or generated notes and to correct medical errors. We also found that although recent LLMs have a good performance in error detection and correction, they are still outperformed by medical doctors in these tasks. We discuss the potential factors behind this gap, the insights from our experiments, the limitations of current evaluation metrics, and share potential pointers for future research.

研究动机与目标

  • 推动对由 LLMs 生成或协助生成的医学文本进行严格验证,以防止在临床笔记中出现幻觉和错误。
  • 创建一个包含注入五种错误类型的海量、多样化临床笔记数据集。
  • 提供检测、定位和纠正在笔记中医疗错误的任务和评估协议。
  • 对当代 LLMs 与医疗专家进行基准测试,以量化差距并指导未来改进。

提出的方法

  • 用来自两个来源和三个美国医院系统的 3,848 份临床文本构建 MEDEC,涵盖五种错误类型(Diagnosis, Management, Treatment, Pharmacotherapy, Causal Organism)。
  • 两条错误注入管线生成正确笔记和错误笔记;MS (M#1) 使用医师资格考试叙述,而 UW (M#2) 将错误注入真实的 UW 笔记并进行去识别化处理。
  • 定义三项子任务:A) 错误标记预测,B) 错误句子提取,C) 修正句子生成。
  • 对三项子任务评估 LLMs 和两个提示(P#1,P#2),并计算错误纠正指标(ROUGE-1、BLEURT、BERTScore、AggregateScore)。
  • 在同一 MEDEC 测试集上比较 LLMs 与两位医生的表现,以量化模型与专业临床医生之间的差距。

实验结果

研究问题

  • RQ1LLMs 能否可靠地检测出临床笔记是否包含错误?
  • RQ2LLMs 能否识别笔记中包含错误的确切句子?
  • RQ3LLMs 能否生成与医学专家修正一致的修正句子?
  • RQ4LLMs 在五种错误类型上与医生在错误检测、定位和纠正方面的表现有何差异?

主要发现

  • Claude 3.5 Sonnet 在 LLMs 中实现最高的错误标记检测准确率(70.16%)和错误句子检测准确率(65.62%)。
  • o1-mini 在错误标记检测方面排名第二(69.08%)。
  • 在错误纠正方面,最佳 Aggregate Score 由 o1-preview(0.698)获得,领先于 GPT-4o 变体(≈0.639)。
  • 医学医生在错误标记和句子检测准确性方面整体优于 LLMs,医生 #2 在 MS/UW 划分中达到 0.7161(错误标记)和 0.6677(错误句子),并且修正分数更高。
  • 在模型之间存在检测准确性与修正质量的排序不一致,凸显评估挑战以及在临床文本纠正中需要改进的指标。
  • MEDEC 的表现差距表明模型在预训练数据中对罕见医疗错误类型的暴露有限,以及编辑现有笔记的复杂性高于从头起草。

更好的研究,从现在开始

从阅读论文到最终审阅,大幅缩短您的研究时间。

无需绑定信用卡

本解读由 AI 生成,并经人工编辑审核。