Large Language Models for Health Care Text Classification: Systematic Review.

Area of Science:

• Artificial Intelligence in Medicine
• Natural Language Processing
• Health Informatics

Background:

• Large language models (LLMs) offer advanced capabilities for natural language processing tasks.
• Accurate and efficient text classification is vital in healthcare for applications like clinical note analysis and diagnosis coding.
• Existing systematic reviews lack a specific focus on LLMs for healthcare text classification.

Purpose of the Study:

• To systematically review and critically evaluate the literature on LLM applications for text classification in healthcare.
• To synthesize current evidence regarding the performance and methodologies of LLMs in healthcare text classification.

Main Methods:

• A systematic literature search was conducted across major databases (PubMed, Scopus, etc.) for papers published between 2018 and 2024.
• Studies were categorized by classification type, application, methodology, text type, and evaluation metrics, adhering to PRISMA guidelines.
• 65 eligible research articles were included in the review.

Main Results:

• A significant increase in publications from 2020 to Q3 2024, with 28 papers in the first three quarters of 2024.
• Fine-tuning (35 papers) and prompt engineering (17 papers) were the most common LLM approaches.
• LLMs consistently outperformed traditional machine learning methods in healthcare text classification, particularly in clinical decision support applications.

Conclusions:

• LLMs demonstrate superior performance in healthcare text classification compared to traditional machine learning.
• Further research is needed to address identified gaps in the literature and explore future directions.
• The review provides a comprehensive overview of LLM utilization in healthcare text classification.