While increasing patients’ access to medical documents improves medical care, this benefit is limited by varying health literacy levels and complex medical terminology. Large language models (LLMs) offer solutions by simplifying medical information. However, evaluating LLMs for safe and patient-friendly text generation is difficult due to the lack of standardized evaluation resources. To fill this gap, we developed MeDiSumQA. MeDiSumQA is a dataset created from MIMIC-IV discharge summaries through an automated pipeline combining LLM-based question-answer generation with manual quality checks. We use this dataset to evaluate various LLMs on patient-oriented question-answering. Our findings reveal that general-purpose LLMs frequently surpass biomedical-adapted models, while automated metrics correlate with human judgment. By releasing MeDiSumQA on PhysioNet, we aim to advance the development of LLMs to enhance patient understanding and ultimately improve care outcomes.

Clinical documents are essential to patient care, but their complexity often makes them inaccessible to patients. Large Language Models (LLMs) are a promising solution to support the creation of lay translations of these documents, addressing the infeasibility of manually creating these translations in busy clinical settings. However, the integration of LLMs into medical practice in Germany is challenging due to data scarcity and privacy regulations. This work evaluates an open-source LLM for lay translation in this data-scarce environment using datasets of German synthetic clinical documents and real tumor board protocols. The evaluation framework used combines readability, semantic, and lexical measures with the G-Eval framework. Preliminary results show that zero-shot prompts significantly improve readability (e.g., FREde: 21.4 → 39.3) and few-shot prompts improve semantic and lexical fidelity. However, the results also reveal G-Eval’s limitations in distinguishing between intentional omissions and factual inaccuracies. These findings underscore the need for manual review in clinical applications to ensure both accessibility and accuracy in lay translations. Furthermore, the effectiveness of prompting highlights the need for future work to develop applications that use predefined prompts in the background to reduce clinician workload.

This paper presents our solution to the MEDIQA-M3G Challenge at NAACL-ClinicalNLP 2024. We participated in all three languages, ranking first in Chinese and Spanish and third in English. Our approach utilizes LLaVA-med, an open-source, medical vision-language model (VLM) for visual question-answering in Chinese, and Mixtral-8x7B-instruct, a Large Language Model (LLM) for a subsequent translation into English and Spanish. In addition to our final method, we experiment with alternative approaches: Training three different models for each language instead of translating the results from one model, using different combinations and numbers of input images, and additional training on publicly available data that was not part of the original challenge training set.

Recent advances in natural language processing (NLP) can be largely attributed to the advent of pre-trained language models such as BERT and RoBERTa. While these models demonstrate remarkable performance on general datasets, they can struggle in specialized domains such as medicine, where unique domain-specific terminologies, domain-specific abbreviations, and varying document structures are common. This paper explores strategies for adapting these models to domain-specific requirements, primarily through continuous pre-training on domain-specific data. We pre-trained several German medical language models on 2.4B tokens derived from translated public English medical data and 3B tokens of German clinical data. The resulting models were evaluated on various German downstream tasks, including named entity recognition (NER), multi-label classification, and extractive question answering. Our results suggest that models augmented by clinical and translation-based pre-training typically outperform general domain models in medical contexts. We conclude that continuous pre-training has demonstrated the ability to match or even exceed the performance of clinical models trained from scratch. Furthermore, pre-training on clinical data or leveraging translated texts have proven to be reliable methods for domain adaptation in medical NLP tasks.

Traditionally, large language models have been either trained on general web crawls or domain-specific data. However, recent successes of generative large language models, have shed light on the benefits of cross-domain datasets. To examine the significance of prioritizing data diversity over quality, we present a German dataset comprising texts from five domains, along with another dataset aimed at containing high-quality data. Through training a series of models ranging between 122M and 750M parameters on both datasets, we conduct a comprehensive benchmark on multiple downstream tasks. Our findings demonstrate that the models trained on the cross-domain dataset outperform those trained on quality data alone, leading to improvements up to 4.45% over the previous state-of-the-art.

Writing the conclusion section of radiology reports is essential for communicating the radiology findings and its assessment to physician in a condensed form. In this work, we employ a transformer-based Seq2Seq model for generating the conclusion section of German radiology reports. The model is initialized with the pretrained parameters of a German BERT model and fine-tuned in our downstream task on our domain data. We proposed two strategies to improve the factual correctness of the model. In the first method, next to the abstractive learning objective, we introduce an extraction learning objective to train the decoder in the model to both generate one summary sequence and extract the key findings from the source input. The second approach is to integrate the pointer mechanism into the transformer-based Seq2Seq model. The pointer network helps the Seq2Seq model to choose between generating tokens from the vocabulary or copying parts from the source input during generation. The results of the automatic and human evaluations show that the enhanced Seq2Seq model is capable of generating human-like radiology conclusions and that the improved models effectively reduce the factual errors in the generations despite the small amount of training data.