Potential of Large Language Models in Health Care: Delphi Study.

Area of Science:

• Health Informatics
• Medical Natural Language Processing (NLP)

Background:

• Large language models (LLMs) are advanced AI systems trained on vast text data.
• LLMs utilize neural networks with transformer architectures and attention mechanisms.
• They excel in various NLP tasks, including classification, information extraction, and generation.

Purpose of the Study:

• To gather researchers' perspectives on the influence of LLMs in healthcare.
• To identify the strengths, weaknesses, opportunities, and threats associated with LLM implementation in healthcare settings.

Main Methods:

• An adapted Delphi study involving researchers from health informatics, nursing informatics, and medical NLP.
• Three rounds of data collection, starting with open-ended questions and progressing to item scoring.
• Inclusion of 28, 23, and 21 participants across the three rounds, respectively.

Main Results:

• Agreement was reached on 103 items concerning LLM use cases, benefits, risks, and future integration in healthcare.
• Identified use cases include clinical task support, documentation, medical research, and patient education.
• Key benefits encompass increased efficiency, improved automation, enhanced care quality, personalized medicine, and accelerated diagnosis.
• Identified risks include cybersecurity breaches, patient misinformation, ethical concerns, biased decision-making, and privacy vulnerabilities.

Conclusions:

• Future LLM research must extend beyond NLP tasks to address practical workflow integration.
• Essential considerations include quality standards, system integration, and regulatory frameworks for successful healthcare implementation.