Scholarlink

Introduction <p>With the rapid advancement of artificial intelligence (AI), particularly large language models (LLMs), there is an emerging opportunity to enhance qualitative health research by streamlining the analytical process.</p> Methods <p>This study explored how GPT-4 and Claude analyzed text from four online brain cancer support forums, focusing on coding, theme identification, quotation selection, and AI-generated visualizations using DALL·E, with researcher feedback helping refine their outputs.</p> Results <p>While both LLMs generated fewer and less detailed codes than traditional thematic analysis (both with or without using qualitative analysis software), their main themes and codes were largely comparable, differing slightly in titles and classifications. GPT-4 often chose deeper, more reflective quotations similar to human-led analysis, whereas Claude focused more on structured medical and clinical details. Notable differences also appeared in how GPT-4, Claude, and DALL·E structured visual representations, reflecting variations in analytical depth and methods.</p> Discussion <p>These findings suggest that AI-powered tools, like LLMs and generative models such as DALL·E, can improve coding, theme identification, quotation selection, and data visualization in qualitative research, while ongoing advances in AI offer further opportunities for efficiency and transparency, requiring continued methodological refinement and human oversight.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Using large language models to support qualitative health research through coding, theming, quotation selection, and data visualization

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年3月16日

查看全文
  • Christy Muasher-Kerwin,
  • M. Courtney Hughes,
  • Samantha M. Econie

摘要

Introduction

With the rapid advancement of artificial intelligence (AI), particularly large language models (LLMs), there is an emerging opportunity to enhance qualitative health research by streamlining the analytical process.

Methods

This study explored how GPT-4 and Claude analyzed text from four online brain cancer support forums, focusing on coding, theme identification, quotation selection, and AI-generated visualizations using DALL·E, with researcher feedback helping refine their outputs.

Results

While both LLMs generated fewer and less detailed codes than traditional thematic analysis (both with or without using qualitative analysis software), their main themes and codes were largely comparable, differing slightly in titles and classifications. GPT-4 often chose deeper, more reflective quotations similar to human-led analysis, whereas Claude focused more on structured medical and clinical details. Notable differences also appeared in how GPT-4, Claude, and DALL·E structured visual representations, reflecting variations in analytical depth and methods.

Discussion

These findings suggest that AI-powered tools, like LLMs and generative models such as DALL·E, can improve coding, theme identification, quotation selection, and data visualization in qualitative research, while ongoing advances in AI offer further opportunities for efficiency and transparency, requiring continued methodological refinement and human oversight.