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: Jun 25, 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

Decreasing Administrative Effort Related to Non-Approval of Image-guidED Procedures Using Large Language Models - The

Colin J McCarthy1, Vijay Ramalingam1, Yiftach Barash1

  • 1Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts (C.J.M., V.R., Y.B., A.S.).

Academic Radiology
|May 7, 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

Vascular Access-Related Heart Failure: Pathophysiological Insights, Diagnostic Challenges, and Management Strategies.

Kidney360·2026
Same author

Socioeconomic Factors Impacting Survival in Patients with Hepatocellular Carcinoma Treated with Locoregional Therapies: A Single-Center Study.

Journal of vascular and interventional radiology : JVIR·2026
Same author

Homoharringtonine Impedes Migration and Invasion by Inhibiting EphB4/SRI/EMT Signaling and Enhances the Antimetastatic Abilities of Erlotinib in Pancreatic Cancer.

Current cancer drug targets·2026
Same author

How to Prepare for Multidisciplinary Tumor Board Conferences: A Radiology Resident Guide to Success.

Radiographics : a review publication of the Radiological Society of North America, Inc·2026
Same author

Technical advances and clinical impact of photon-counting computed tomography angiography for peripheral artery disease.

Journal of vascular surgery cases and innovative techniques·2026
Same author

Multidisciplinary clinics in cancer-models, metrics, and meaning: a review.

The oncologist·2026

Large language models (LLMs) can create usable insurance appeal letters for radiology denials, but accuracy issues like fabricated references and hallucinations require human oversight before clinical application.

Area of Science:

  • Artificial Intelligence in Healthcare
  • Medical Informatics
  • Radiology Workflow Optimization

Background:

  • Insurance denials for radiology procedures present a significant administrative burden.
  • Developing efficient tools to streamline the appeals process is crucial for healthcare providers.

Purpose of the Study:

  • To assess the accuracy, clinical validity, and usability of large language models (LLMs) in generating insurance appeal letters for radiology procedures.
  • To compare the performance of different LLMs and generation techniques (zero-shot, few-shot, retrieval-augmented generation).

Main Methods:

  • Four LLMs (Claude 3.5, Nova Pro, Llama-3.1-70B, ChatGPT-4o) were employed to generate appeal letters for a simulated clinical scenario.
  • Interventional radiologists evaluated letters based on content, grammar, and usability, with reference verification and hallucination checks.
Keywords:
Administrative burdenArtificial intelligenceInsurance denialInterventional radiologyLarge language models

More Related Videos

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
05:33

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Published on: July 11, 2025

Related Experiment Videos

Last Updated: Jun 25, 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

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System
05:33

Introduction of an Integrated Pathology Image Management, Artificial Intelligence, and Reporting System

Published on: July 11, 2025

  • Statistical analyses, including ANOVA and Fleiss' Kappa, were used to assess model performance and interrater reliability.

    • Main Results:

    • LLM-generated letters received adequate mean scores for content (3.9/5) and grammar (4.3/5), with no significant differences between models or techniques.
    • Reviewer agreement was poor, indicating subjective interpretation of letter quality.
    • Hallucinations and fabricated references were prevalent, particularly in offline models, necessitating critical human review.

    Conclusions:

    • Large language models show potential for generating usable insurance appeal letters for radiology denials.
    • Despite high usability, the significant prevalence of fabricated references and hallucinations underscores the need for rigorous human validation prior to clinical deployment.
    • Further research is needed to improve the factual accuracy and reliability of LLM-generated medical documentation.