Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Experiment Video

Updated: Jul 5, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

Multimodule Human-Artificial Intelligence Collaboration Pipeline for Large Language Model-Assisted Thematic Analysis

Yunbing Bai1, Joseph Finkelstein1

  • 1Arizona Center for Telemedicine and Digital Health, College of Medicine, University of Arizona, Tucson, AZ, United States.

JMIR Medical Informatics
|July 3, 2026
PubMed
Summary

Related Concept Videos

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Clinico-genomic nomograms for estimation of survival risk in metastatic castrate-resistant prostate cancer.

JNCI cancer spectrum·2026
Same author

Extracting Cardiorespiratory Symptoms From Clinical Notes Using Open-Weight Large Language Models: Method Development and Validation Study.

JMIR cardio·2026
Same author

Perception of AI Symptom Models in Oncology Nursing: Mixed Methods Evaluation Study.

JMIR nursing·2026
Same author

Telemedicine-Supported Exercise Interventions in Hematopoietic Stem Cell Transplantation: A Scoping Review.

International journal of general medicine·2025
Same author

Deep Learning Approaches to Forecast Physical and Mental Deterioration During Chemotherapy in Patients with Cancer.

Diagnostics (Basel, Switzerland)·2025
Same author

Exploring Artificial Intelligence Biases in Predictive Models for Cancer Diagnosis.

Cancers·2025
Same journal

Graph Network Feature Space Fusion for Predicting Irregularly Sampled Medical Time-Series Data: Deep Learning Model Development and Validation Study.

JMIR medical informatics·2026
Same journal

Intrasystem Repeatability of S-Detect for Breast Ultrasound Classification With Identical Static Images: Single-Center Retrospective Repeatability Study.

JMIR medical informatics·2026
Same journal

Clinician Perspectives on Ambient AI Scribes in the Intensive Care Unit: Qualitative Interview Study.

JMIR medical informatics·2026
Same journal

IdeaDistiller-AI Support for Idea Synthesis in Concept Mapping: Algorithm Development and Validation Study.

JMIR medical informatics·2026
Same journal

Pregnancy-Related Clinical Codes in Unlikely Populations in Primary Care.

JMIR medical informatics·2026
Same journal

Selecting, Scaling, and Measuring the Value of Ambient AI in a Nonacademic Health System: Multiphase Pilot Study.

JMIR medical informatics·2026
See all related articles
This summary is machine-generated.

Large language models (LLMs) can assist qualitative health research, but performance varies by model and workflow. Opus demonstrated the strongest alignment with human themes, while Gemini and ChatGPT offered different speed and usability trade-offs.

Area of Science:

  • Health Informatics
  • Qualitative Research Methods
  • Artificial Intelligence in Healthcare

Background:

  • Qualitative thematic analysis is crucial for patient experience research but is time-consuming.
  • Large language models (LLMs) offer potential to accelerate thematic analysis in health research.
  • Limited evidence exists on LLM performance across different workflows and health datasets.

Purpose of the Study:

  • Evaluate a modular human-AI pipeline for LLM-assisted thematic analysis.
  • Compare LLM choice and workflow strategies for theme alignment in health studies.
  • Assess AI-generated themes against human-generated themes across diverse qualitative health datasets.

Main Methods:

  • Applied a human-AI pipeline to analyze interview transcripts from three health studies (interstitial lung disease, POTS, COPD).
Keywords:
LLMgenerative AIhealth informaticshuman-AI collaborationlarge language modelqualitative studythematic analysis

Related Experiment Videos

Last Updated: Jul 5, 2026

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness
03:14

Augmenting Large Language Models via Vector Embeddings to Improve Domain-Specific Responsiveness

Published on: December 6, 2024

  • Compared three LLMs: Gemini, ChatGPT, and Opus, using a workflow of code extraction, combination, and theme generation.
  • Evaluated theme alignment using cosine similarity and assessed runtime and output consistency.

    • Main Results:

    • Opus showed the strongest and most consistent alignment with human-generated themes across studies.
    • ChatGPT was competitive in specific settings, while Gemini had the shortest runtime but slightly lower similarity.
    • Opus and ChatGPT demonstrated better formatting consistency and usability than Gemini.

    Conclusions:

    • A modular human-AI pipeline effectively supports thematic analysis in digital health research.
    • LLM performance is highly dependent on model selection and workflow design.
    • LLMs serve as valuable human-supervised analytic assistants, not replacements for qualitative researchers.