Enhancing health assessments with large language models: A methodological approach

Area of Science:

• Health Psychology
• Artificial Intelligence
• Natural Language Processing

Background:

• Health assessments are crucial for evaluating physical and mental health status using subject scales.
• Traditional statistical methods lack accuracy due to linear scoring.
• Machine learning (ML) offers potential but suffers from poor interpretability and data dependency.

Purpose of the Study:

• To investigate the application of large language models (LLMs) for improving health psychology assessments.
• To introduce ScaleLLM, a novel framework utilizing LLMs for enhanced health scale evaluation.

Main Methods:

• Developed ScaleLLM, a system that leverages language and knowledge alignment.
• Trained LLMs to function as expert evaluators for health psychology scales.

Main Results:

• ScaleLLM demonstrated improved accuracy in health assessments.
• The approach enhanced the interpretability of health assessment results.
• LLMs offer a promising solution to limitations of traditional and ML methods.

Conclusions:

• LLMs, particularly through frameworks like ScaleLLM, can significantly advance health psychology assessments.
• ScaleLLM provides a more accurate and interpretable method for evaluating physical and mental health status.