Public Perceptions and Barriers to Tuberculosis Treatment in Korea: A Large Language Model-Based Analysis of Naver Knowledge-iN Data from 2002 to 2024.

English summary

This study used large language models (LLMs) to analyze 44,174 tuberculosis-related questions posted on Naver Knowledge-iN in Korea from 2002 to 2024. It examined public perceptions and concerns about tuberculosis treatment and evaluated LLMs as tools for analyzing unstructured healthcare data.

Key points

• Among 919 questions that mentioned specific antitubercular medications, rifampin and isoniazid were the most frequently referenced, accounting for 31.8% and 31.6%, respectively.
• Among 10,044 questions about antitubercular medications, treatment and management challenges formed the largest category at 44.8%.
• Analysis of 583 questions concerning infectivity and social implications identified previously unrecognized concerns about blood donation and immigration eligibility. Employment-related concerns were the largest distinct subgroup at 20.6%.

Audience

Clinicians, public-health researchers, health data scientists, and advanced students interested in public concerns about tuberculosis treatment, online healthcare data, and LLM-based text analysis.

Scope and limitations

The dataset consisted of online questions posted on Naver Knowledge-iN and therefore cannot be assumed to represent the perceptions or experiences of the Korean population as a whole. Based on the supplied abstract, the sampling composition, detailed procedures for comparison with human researchers and traditional methods, and specific criteria used to evaluate the LLMs cannot be determined.

Research context

The study applies LLMs, text embeddings, dimensionality reduction, and clustering to unstructured online healthcare data, connecting it thematically with medical AI and data-science approaches to public-health research.