Practical Applications of Large Language Models for Health Care Professionals and Scientists

Abstract

With the popularization of large language models (LLMs), strategies for their effective and safe usage in health care and research have become increasingly pertinent. Despite the growing interest and eagerness among health care professionals and scientists to exploit the potential of LLMs, initial attempts may yield suboptimal results due to a lack of user experience, thus complicating the integration of artificial intelligence (AI) tools into workplace routine. Focusing on scientists and health care professionals with limited LLM experience, this viewpoint article highlights and discusses 6 easy-to-implement use cases of practical relevance. These encompass customizing translations, refining text and extracting information, generating comprehensive overviews and specialized insights, compiling ideas into cohesive narratives, crafting personalized educational materials, and facilitating intellectual sparring. Additionally, we discuss general prompting strategies and precautions for the implementation of AI tools in biomedicine. Despite various hurdles and challenges, the integration of LLMs into daily routines of physicians and researchers promises heightened workplace productivity and efficiency.

Keywords: AI; LLM; applications; artificial intelligence; chatGPT; health care; healthcare; large language model; physicians; prompting; scientists.

Figure 1.. Tailored summaries of an uploaded document depending on the intended target audience. Two different prompts asking for a summary of the same uploaded document (blue). A full-length PDF of one of the authors’ open access scientific articles was uploaded as the input [8]. The prompt on the left specifies word count, output format, and audience for a summary to be used in a professional context. The prompt on the right exemplifies how large language models can be used to extract key information and make the content widely accessible for any audience. Answers given by OpenAI’s ChatGPT-4 are shown in gray.

Figure 2.. Designing customized teaching material to facilitate medical education. An example prompt asking for a course schedule on a given topic for a specific audience (blue). The response by ChatGPT-4 is shown in gray. The output is presented in tabular format.

Figure 3.. Using ChatGPT as a personal sparring partner to simulate interactions and discussions. Exemplary prompts asking for a study plan on rheumatology (blue) and then to be tested on the contents of the proposed plan. The task is broken down into 2 prompts, resulting in a conversational interaction. First, ChatGPT-4 replies with a detailed schedule and specific advice for 21 study sessions over the course of 1 week (gray). The first answer is truncated and shown in smaller font due to limited space. Next, ChatGPT-4 is asked to test the user on the proposed topics. Upon receiving an answer, the large language model first corrects the user’s answer and then proceeds to pose the next question, thus correctly referring to the previously defined task. EULAR: European Alliance of Associations for Rheumatology.

Figure 4.. Setting the output temperature via the application programming interface (API) to generate deterministic results. An identical prompt was entered into the OpenAI API 3 times using the GPT-4 model (chat was cleared between prompts). When the temperature (Temp) was set to 0, the output showed only minimal variation as the model behaved deterministically. At higher values of Temp=1 or Temp=2, the output variety also increased.