Medical Implications of LLM Based Clinical Decision Support Systems in Healthcare
Cagatay Umut Ogdu, Selen Gurbuz, Mehmet Karaköse, Eray Hanoğlu
Abstract
the development of large language models (LLMs) has also opened new horizons in the healthcare sector. In this study, the potential contributions of large language models to clinical decision support systems (CDSS) are presented in detail. The capabilities of these models, which can be used in critical tasks such as diagnosis, disease prediction and optimization of patient management processes, have been comprehensively analyzed. In the study, general purpose large language models (ChatGPT, Gemma, Meta Llama, Mixtral) and models developed specifically for the healthcare field (BioBART, GatorTron) have been comparatively examined. The performance of the models has been evaluated on the basis of criteria such as semantic similarity, disease prediction accuracy and generalization capacity; MedMCQA and PubMedQA datasets have been utilized in this process. The results clearly reveal the strengths and weaknesses of large language models in the healthcare sector and provide concrete suggestions for their more effective use in clinical applications.