JMIR Publications

ABSTRACT

Background:

A large language model (LLM) is a machine learning model inferred from text data that captures subtle patterns of language use in context. Modern LLMs are based on neural network architectures that incorporate transformer methods that allow the model to relate words together through attention to multiple words in a text sequence [1]. LLMs have been shown to be highly effective for a range of tasks in Natural Language Processing (NLP), including classification and information extraction tasks as well as generative applications such as machine translation and summarization. In late 2022, the public gained widespread awareness of such models with the release of the Artificial Intelligence system ChatGPT that is based on a generative pre-trained transformer (GPT) model.

Objective:

The aim of this adapted Delphi study was to gain insights into opinions of how researchers think LLMs might influence healthcare and what are the strengths, weaknesses, opportunities and threats (SWOT) of the use of LLMs in healthcare.

Methods:

We invited researchers in the field of health informatics, nursing informatics, and medical NLP to share their opinions on the use of LLMs in healthcare. We started the first round with open questions based on our SWOT framework. In the second and third round, the participants scored these items.

Results:

The first, second, and third rounds had 28, 23, and 21 participants, respectively. Almost all participants were affiliated with academic institutions. Agreement was reached on 103 items related to use cases, benefits, risks, reliability, adoption aspects, and the future of LLMs in healthcare. Participants offered several use cases including supporting clinical tasks, documentation tasks, and medical research and education; and agreed that LLM-based systems will act as virtual health assistants for patient education. The agreed benefits included increased efficiency in data handling and extraction, improved automation of processes, improved quality of healthcare services and overall health outcomes, provision of personalized care, accelerated diagnosis and treatment processes, and improved interaction between patients and healthcare professionals. Five risks to healthcare in general were identified: cybersecurity breaches, the potential for patient misinformation, ethical concerns, the likelihood of biased decision making, and the risk associated with inaccurate communication. Overconfidence in LLM-based systems was recognized as a risk for the medical profession. The six agreed privacy risks included the use of unregulated cloud services that compromise data security, exposure of sensitive patient data, breaches of confidentiality, fraudulent use of information, vulnerabilities in data storage and communication, and inappropriate access or use of patient data.

Conclusions:

Future research related to LLMs should not only focus on testing their possibilities for NLP-related tasks, but should also consider the workflows the methods could contribute to and the requirements regarding quality, integration, and regulations needed for successful implementation in practice.