MEDEC: a benchmark for detecting and correcting medical errors in clinical notes using LLM

LLMs have demonstrated impressive capabilities in answering medical questions accurately, even surpassing average human scores on some medical exams. However, its adoption in medical documentation tasks, such as clinical note generation, faces challenges due to the risk of generating incorrect or inconsistent information. Studies reveal that 20% of patients who read clinical notes identified errors, and 40% considered them serious, often related to misdiagnosis. This raises important concerns, especially as LLMs increasingly support medical documentation tasks. While these models have demonstrated great performance in answering medical exam questions and mimicking clinical reasoning, they are prone to generating hallucinations and potentially harmful content, which could negatively impact clinical decision-making. This highlights the critical need for robust validation frameworks to ensure the accuracy and security of medical content generated by LLM.

Recent efforts have explored benchmarks for consistency evaluation in general domains, such as semantic, logical, and factual consistency, but these approaches often fail to ensure reliability in all test cases. While models like ChatGPT and GPT-4 show improved reasoning and language understanding, studies show that they struggle with logical coherence. In the medical field, LLM assessments such as ChatGPT and GPT-4 have demonstrated accurate performance on structured medical exams such as the USMLE. However, limitations arise when handling complex medical consultations, and drafts generated by LLM in patient communication have shown potential risks, including serious harm if errors are not corrected. Despite advances, the lack of publicly available benchmarks to validate the accuracy and consistency of medical texts generated by LLMs underscores the need for reliable and automated validation systems to effectively address these challenges.

Researchers at Microsoft and the University of Washington have developed MEDEC, the first publicly available benchmark for detecting and correcting medical errors in clinical notes. MEDEC includes 3,848 clinical texts covering five types of errors: diagnosis, management, treatment, pharmacotherapy, and causative organism. Assessments using advanced LLMs, such as GPT-4 and Claude 3.5 Sonnet, revealed their ability to address these tasks, but are outperformed by human medical experts. This benchmark highlights the challenges of validating and correcting clinical texts, emphasizing the need for models with sound medical reasoning. Insights from these experiments provide guidance for improving future error detection systems.

The MEDEC dataset contains 3,848 clinical texts, annotated with five error types: diagnosis, treatment, treatment, pharmacotherapy, and causative organism. Errors were introduced by leveraging medical board (MS) exams and modifying actual clinical notes from the University of Washington (UW) hospitals. Annotators manually created errors by injecting incorrect medical entities into the text while ensuring consistency with other parts of the note. MEDEC is designed to evaluate error detection and correction models, divided into error prediction, error phrase identification, and correction generation.

The experiments used several small LLMs, including Phi-3-7B, Claude 3.5 Sonnet, Gemini 2.0 Flash, and OpenAI's GPT-4 series, to evaluate their performance in medical error detection and correction tasks. These models were tested on subtasks such as identifying errors, pointing out erroneous sentences, and generating corrections. Metrics such as precision, recall, ROUGE-1, BLEURT, and BERTScore were used to evaluate its capabilities, along with an aggregate score that combines these metrics to determine the quality of the correction. Claude 3.5 Sonnet achieved the highest accuracy in detecting error indicators (70.16%) and sentences (65.62%), while o1-preview excelled in error correction with a total score of 0.698. Comparisons with expert medical annotations highlighted that although LLMs performed well, they were still outperformed by physicians in detection and correction tasks.

The performance gap is likely due to the limited availability of error-specific medical data in prior LLM training and the challenge of analyzing pre-existing clinical texts rather than generating responses. Among models, preview o1 demonstrated superior recall across all error types, but struggled with precision, often overestimating error occurrences compared to medical experts. This accuracy deficit, coupled with the models' reliance on public datasets, resulted in a performance disparity between subsets, with models performing better on public datasets (e.g. MEDEC-MS) than on private collections. as MEDEC-UW.

Verify he Paper and GitHub page. All credit for this research goes to the researchers of this project. Also, don't forget to follow us on <a target="_blank" href="https://twitter.com/Marktechpost”>twitter and join our Telegram channel and LinkedIn Grabove. Don't forget to join our SubReddit over 60,000 ml.

UPCOMING FREE ai WEBINAR (JANUARY 15, 2025): <a target="_blank" href="https://info.gretel.ai/boost-llm-accuracy-with-sd-and-evaluation-intelligence?utm_source=marktechpost&utm_medium=newsletter&utm_campaign=202501_gretel_galileo_webinar”>Increase LLM Accuracy with Synthetic Data and Assessment Intelligence–<a target="_blank" href="https://info.gretel.ai/boost-llm-accuracy-with-sd-and-evaluation-intelligence?utm_source=marktechpost&utm_medium=newsletter&utm_campaign=202501_gretel_galileo_webinar”>Join this webinar to learn practical information to improve LLM model performance and accuracy while protecting data privacy..

Sana Hassan, a consulting intern at Marktechpost and a dual degree student at IIT Madras, is passionate about applying technology and artificial intelligence to address real-world challenges. With a strong interest in solving practical problems, he brings a new perspective to the intersection of ai and real-life solutions.