Javascript must be enabled for the correct page display
Faculty of Medical Sciences

Role of ChatGPT-4 as a Tool to Assist Human Researchers in Qualitative Coding

Khan, Zainab (2024) Role of ChatGPT-4 as a Tool to Assist Human Researchers in Qualitative Coding. thesis, Medicine.

Full text available on request.

Abstract

Background: Qualitative research is essential for understanding patient experiences and healthcare processes. This study explores the use of ChatGPT-4, an AI language model, to assist human researchers in coding interview transcripts and conducting thematic analysis, aiming to improve the efficiency and accuracy of qualitative research. Methods: The study used existing datasets from two qualitative studies on Overactive Bladder (OAB) and Recurrent Urinary Tract Infections (RUTI). ChatGPT and human researchers independently coded the same interview transcripts. ChatGPT coded each study twice. The identified codes and themes from both rounds were compared with the codes assigned by human researchers through Venn diagrams and qualitative analysis to assess similarity, correctness and comprehensiveness. Results: ChatGPT identified additional themes and codes not recognized by human researchers, suggesting its potential to uncover deeper insights. The refining of the prompts in the second round with integration of more of the prompt engineering principles along with manual checking of the generated codes, enhances the performance of ChatGPT compared to round one. This showcases the degrees in ChatGPT’s robustness and that higher levels of performance can be achieved by the user by acquiring a deeper understanding of how ChatGPT works. However, ChatGPT's stochastic nature led to variability in outputs and as ChatGPT in incapable of critical thinking human oversight is necessary. Conclusion: ChatGPT shows promise as a supplementary tool in qualitative research, capable of identifying significant themes and codes while reducing manual labour. However, human oversight remains crucial for interpreting AI-generated data and ensuring the applicability of findings to clinical practice.

Item Type: Thesis (UNSPECIFIED)
Supervisor name: Witte, Dr. L. P. W. (Bart)
Faculty: Medical Sciences
Date Deposited: 10 Dec 2025 15:00
Last Modified: 10 Dec 2025 15:00
URI: https://umcg.studenttheses.ub.rug.nl/id/eprint/3870

Actions (login required)

View Item View Item