Large language models in health care: Development, applications, and challenges

Area of Science:

• Artificial Intelligence
• Natural Language Processing
• Health Informatics

Background:

• The emergence of advanced AI chatbots like ChatGPT highlights the potential of large language models (LLMs).
• LLMs are increasingly recognized as a significant area of development across various fields, including healthcare.
• LLMs in healthcare can be categorized into biomedical and clinical domains based on their training data.

Purpose of the Study:

• To explore the capabilities and potential applications of LLMs in healthcare.
• To differentiate between general and domain-specific LLMs in the context of healthcare.
• To discuss the implications, opportunities, and challenges of integrating LLMs into healthcare systems.

Main Methods:

• Review of recent advancements in LLM development and performance.
• Analysis of domain-specific LLMs (biomedical and clinical) and their performance on natural language processing tasks.
• Discussion of potential applications in clinical practice, medical education, and writing.

Main Results:

• Domain-specific LLMs have demonstrated superior performance compared to general LLMs in recent years.
• LLMs show exceptional capabilities in language comprehension and content generation.
• The performance of specialized LLMs emphasizes the need for domain-specific development.

Conclusions:

• LLMs hold significant promise for transforming healthcare, with potential applications in preconsultation, diagnosis, management, medical education, and writing.
• Domain-specific LLMs are crucial for achieving high performance in healthcare-related natural language processing tasks.
• Healthcare systems must proactively address the challenges and ethical considerations associated with LLM implementation.