Viability of Open Large Language Models for Clinical Documentation in German Health Care: Real-World Model Evaluation Study

• Published in: JMIR medical informatics Vol. 12; p. e59617
• Main Authors: Heilmeyer, Felix, Böhringer, Daniel, Reinhard, Thomas, Arens, Sebastian, Lyssenko, Lisa, Haverkamp, Christian
• Format: Journal Article
• Language: English
• Published: Canada JMIR Publications 28.08.2024
• Subjects: Advanced Data Analytics in eHealth; AI Language Models in Health Care; Artificial intelligence; Big Data; Case Study; Clinical Informatics; Documentation; Electronic Health Records; General Data Protection Regulation; Health care industry; Large language models; Machine Learning; Medical records; Natural Language Processing; New Technologies; Open Source Software; Original Paper; Outpatient care facilities; Software; Working hours; Writing; algorithms; models; writing assistance; natural language processing; LLMs; AI; practical model; LLM; large language model; machine learning; artificial intelligence; writing assistance for physicians; analytics; medical documentation; deep learning; NLP; medical administration; large language models; model; practical models; ML; algorithm
• ISSN: 2291-9694; 2291-9694
• DOI: 10.2196/59617

The use of large language models (LLMs) as writing assistance for medical professionals is a promising approach to reduce the time required for documentation, but there may be practical, ethical, and legal challenges in many jurisdictions complicating the use of the most powerful commercial LLM solutions. In this study, we assessed the feasibility of using nonproprietary LLMs of the GPT variety as writing assistance for medical professionals in an on-premise setting with restricted compute resources, generating German medical text. We trained four 7-billion-parameter models with 3 different architectures for our task and evaluated their performance using a powerful commercial LLM, namely Anthropic's Claude-v2, as a rater. Based on this, we selected the best-performing model and evaluated its practical usability with 2 independent human raters on real-world data. In the automated evaluation with Claude-v2, BLOOM-CLP-German, a model trained from scratch on the German text, achieved the best results. In the manual evaluation by human experts, 95 (93.1%) of the 102 reports generated by that model were evaluated as usable as is or with only minor changes by both human raters. The results show that even with restricted compute resources, it is possible to generate medical texts that are suitable for documentation in routine clinical practice. However, the target language should be considered in the model selection when processing non-English text.