Clinical application of large language models : Does ChatGPT replace medical report formulation? An experience report

Abstract

Artificial intelligence (AI)-based language models, such as ChatGPT offer an enormous potential for research and medical care but also for clinical workflow optimization by making medical documentation easier and more efficient in taking over standardized routine tasks. With their ability to guess a text's content using word statistics and thus outputting contextually relevant results in chat dialogues, large language models (LLM) can provide appropriate summaries of medical documentation for different target groups. For instance, text generation in easy to understand language could potentially contribute to an increase in patients' health literacy and, consequently, to increased adherence to treatment. Subsequent, the function of AI-based chatbot models to improve user experiences and enhance competence in the use of AI-based language models will be adressed. Current limitations and chances in creating epicrises are presented as an experience report. In the future, the implementation of local LLMs in medical management systems (hospital information systems, HIS and practice administration systems, PAS) and in conjunction with the electronic patient records (ePA) can fundamentally change clinical and outpatient care.