Circle packing charts generated by ChatGPT to identify the characteristics of articles by anesthesiology authors in 2022: Bibliometric analysis

• Published in: Medicine (Baltimore) Vol. 102; no. 50; p. e34511
• Main Authors: Ho, Sam Yu-Chieh, Chien, Tsair-Wei, Chou, Willy
• Format: Journal Article
• Language: English
• Published: Hagerstown, MD Lippincott Williams & Wilkins 15.12.2023
• Subjects: Anesthesiology; Bibliometrics; China; Humans; Medical Subject Headings; PubMed; Systematic Review and Meta-Analysis; United States; United States; China; author-weighted scheme; absolute advantage coefficient; ChatGPT; impact beam plot; Sankey diagram; circle packing charts; network chart
• ISSN: 0025-7974; 1536-5964; 1536-5964
• DOI: 10.1097/MD.0000000000034511

Background: The ChatGPT (Open AI, San Francisco, CA), denoted by the Chat Generative Pretrained Transformer, has been a hot topic for discussion over the past few months. A verification of whether the code for drawing circle packing charts (CPCs) with R can be generated by ChatGPT and used to identify characteristics of articles by anesthesiology authors is needed. This study aimed to provide insights into article characteristics in the field of anesthesiology and to highlight the potential of ChatGPT for data visualization techniques (e.g., CPCs) in bibliometric analysis. Methods: A total of 23,012 articles were indexed in PubMed in 2022 by authors in the field of anesthesiology. The code for drawing CPCs with R was generated by ChatGPT and then modified by the authors to identify the characteristics of articles in 2 forms: 23,012 and 100 top-impact factors in journals (T100IF). Using CPCs and 3 other visualizations-network charts, impact beam plots, and Sankey diagrams-we were able to display article features commonly used in bibliometric analysis. The author-weighted scheme and absolute advantage coefficient were used to assess dominant entities, such as countries, institutes, authors, and themes (defined by PubMed and MeSH terms). Results: Our findings indicate that: further modifications should be made to the code generated by ChatGPT for drawing CPCs in R; publications in the field of anesthesiology are dominated by China, followed by the United States and Japan; Capital Medical University (China) and Showa University Hospital (Japan) dominate research institutes in terms of publications and IF, respectively; and COVID-19 is the most frequently reported theme in T100IF, accounting for 29%. Conclusions: No such articles with CPCs regarding bibliometrics have ever been found in PubMed. The code for drawing CPCs with R can be generated by ChatGPT, but further modification is required for implementation in bibliometrics. CPCs should be used in future studies to identify the characteristics of articles in other areas of research rather than limiting them to anesthesiology, as we did in this study.