Clinical Report: Evaluation of Tailored GPT Models for Systematic Reviews in Rheumatology
Overview
This study evaluates the performance of customized GPT-based models in data extraction and risk-of-bias appraisal for systematic reviews in rheumatology.
Background
Systematic reviews are essential for evidence-based decision-making in rheumatology, yet they are resource-intensive, particularly in data extraction and risk-of-bias assessment. The complexity of diseases like systemic lupus erythematosus (SLE) adds to the challenges faced by reviewers. Recent advancements in large language models (LLMs) present an opportunity to streamline these processes, but their effectiveness in real-world applications requires thorough evaluation.
Data Highlights
No numerical data was provided in the source material.
Key Findings
Customized GPT-based models were evaluated against human reviewers in rheumatology systematic reviews.
The study focused on two components: data extraction from metabolomics studies and QUIPS-based risk-of-bias appraisal.
Fifteen full-text articles were selected for the data extraction analysis, reflecting diverse study designs.
Nineteen full-text studies were included for the QUIPS component, requiring independent assessments by trained reviewers.
Clinical Implications
It is crucial to maintain human oversight to ensure the accuracy and reliability of data extraction and risk assessments.
Conclusion
The study emphasizes the importance of human expertise in the review process.
by Pamela Munguía-Realpozo, Edith Ramírez-Lara, Claudia Mendoza-Pinto, Ivet Etchegaray-Morales, Juan Carlos Solis-Poblano, Marco Alejandro Trinidad-González, Jorge Ayón-Aguilar, Máximo Alejandro García-Flores, Álvaro José Montiel-Jarquín
