A Scoping Review of Large Language Model Applications in Healthcare

The release of ChatGPT has driven widespread adoption of generative large language models (LLMs) in healthcare. This scoping review analyzed 415 studies from PubMed, published between January 2023 and July 2024, focusing on LLM performance in healthcare applications. Applications included clinical decision-making (26.7%), patient information (23.9%), education (18.1%), research (16.1%), and workflow support (12.5%). Key NLP tasks were question answering (36.1%), text classification (27.5%), and text generation (26.3%). GPT-4 (51.3%) was the most common model. Accuracy was the primary metric, while safety and efficiency were less evaluated. Standardized evaluation criteria are essential for integrating LLMs into healthcare workflows.