Next Article in Journal
Relationship between Cardiovascular Risk Factors and Composite Cardiovascular Outcomes in Patients Hospitalized with Takotsubo Syndrome: A Nationwide Analysis
Previous Article in Journal
Evaluation of a Multi-Gene Methylation Blood-Test for the Detection of Colorectal Cancer
 
 
Search for Articles:
Title / Keyword
Author / Affiliation / Email
Journal
Article Type
 
 
Advanced Search
 
Section
Special Issue
Volume
Issue
Number
Page
 
Journals
Medical Sciences
Volume 11
Issue 3
10.3390/medsci11030061
medsci-logo
Submit to this Journal Review for this Journal Propose a Special Issue
► Article Menu

Article Menu

share Share announcement Help format_quote Cite question_answer Discuss in SciProfiles thumb_up
...
Endorse textsms
...
Comment
first_page
settings
Order Article Reprints
Font Type:
Arial Georgia Verdana
Font Size:
Aa Aa Aa
Line Spacing:
Column Width:
Background:
Review

A Bibliometric Analysis of the Rise of ChatGPT in Medical Research

by
Nikki M. Barrington
1,*,
Nithin Gupta
2,
Basel Musmar
3,
David Doyle
4,
Nicholas Panico
5,
Nikhil Godbole
6,
Taylor Reardon
7 and
Randy S. D’Amico
8
1
Chicago Medical School, Rosalind Franklin University, North Chicago, IL 60064, USA
2
School of Osteopathic Medicine, Campbell University, Lillington, NC 27546, USA
3
Faculty of Medicine and Health Sciences, An-Najah National University, Nablus P.O. Box 7, West Bank, Palestine
4
Central Michigan College of Medicine, Mount Pleasant, MI 48858, USA
5
Lake Erie College of Osteopathic Medicine, Erie, PA 16509, USA
6
School of Medicine, Tulane University, New Orleans, LA 70112, USA
7
Department of Neurology, Henry Ford Hospital, Detroit, MI 48202, USA
8
Department of Neurosurgery, Lenox Hill Hospital, New York, NY 10075, USA
*
Author to whom correspondence should be addressed.
Med. Sci. 2023, 11(3), 61; https://doi.org/10.3390/medsci11030061
Submission received: 14 August 2023 / Revised: 4 September 2023 / Accepted: 11 September 2023 / Published: 17 September 2023
(This article belongs to the Section Translational Medicine)
Browse Figures
Versions Notes

Abstract

:
The rapid emergence of publicly accessible artificial intelligence platforms such as large language models (LLMs) has led to an equally rapid increase in articles exploring their potential benefits and risks. We performed a bibliometric analysis of ChatGPT literature in medicine and science to better understand publication trends and knowledge gaps. Following title, abstract, and keyword searches of PubMed, Embase, Scopus, and Web of Science databases for ChatGPT articles published in the medical field, articles were screened for inclusion and exclusion criteria. Data were extracted from included articles, with citation counts obtained from PubMed and journal metrics obtained from Clarivate Journal Citation Reports. After screening, 267 articles were included in the study, most of which were editorials or correspondence with an average of 7.5 +/− 18.4 citations per publication. Published articles on ChatGPT were authored largely in the United States, India, and China. The topics discussed included use and accuracy of ChatGPT in research, medical education, and patient counseling. Among non-surgical specialties, radiology published the most ChatGPT-related articles, while plastic surgery published the most articles among surgical specialties. The average citation number among the top 20 most-cited articles was 60.1 +/− 35.3. Among journals with the most ChatGPT-related publications, there were on average 10 +/− 3.7 publications. Our results suggest that managing the inevitable ethical and safety issues that arise with the implementation of LLMs will require further research exploring the capabilities and accuracy of ChatGPT, to generate policies guiding the adoption of artificial intelligence in medicine and science.

1. Introduction

Large Language Models (LLMs) are machine learning models designed to generate text that resembles human language. LLMs have attracted a great deal of interest in the medical field [1] as they have the potential to revolutionize medical research, patient care, and education by processing vast quantities of data faster and more precisely than humans can [1,2,3]. One of the most common LLMs is OpenAI’s ChatGPT (Chat Generative Pre-trained Transformer) [4] which has been increasingly investigated for its use in medical research [5], practice [3], and education [2]. As its name implies, ChatGPT employs transformer architecture, which was introduced in 2017 [6] as a means to overcome the limitations of previous Artificial Neural Network (ANN) models, such as Recurrent Neural Networks (RNNs) and Long Short-Term Memory Networks (LSTMs) [7], specifically their limitations with regard to understanding the context of a given language input (for example, understanding that the phrase “The bark was loud” refers to an animal noise, rather than the bark of a tree). Transformer architecture employs self-attention to contextualize words, matching one vector known as a query for a given word to a vector known as a key for another word, thereby indicating that the second word is relevant to the first in terms of context [6,8]. The weight of this relevance is indicated by a third vector known as the value, and helps to determine which keys are most relevant to the query (in our example above, these vectors would allow for understanding that “bark” in the context of “loud” refers to a noise, rather than an object) [6,8].
Since its introduction in November 2022, ChatGPT (version 3.5), armed with the capabilities of transformer architecture, has garnered a great deal of interest as a means to increase efficiency in medical practice and accelerate medical research. As an LLM, ChatGPT employs vast amounts of training data to “learn” and improve its predictive capabilities based on the context of the inputs provided, producing accurate and seemingly thoughtful responses to text prompts based on complex algorithms [9]. As such, it is intended to simulate human intelligence processes. However, despite its increasing ability to understand context, one of the biggest criticisms of the platform is that ChatGPT’s goal is not necessarily to produce correct answers to user queries, but rather to produce text that reads as though a human wrote it [9].
ChatGPT reached over 100 million users within two months of its launch [10] and was quickly trialed in both academic and medical settings [11,12], prompting a variety of ethical concerns around plagiarism [13], authorship [14,15], and dissemination of misinformation [16]. With its remarkable abilities, ChatGPT has already demonstrated success in passing specialized exams [17,18,19], including medical board exams [2], and has been credited with authorship on research articles [20], forcing educators and publishers to rapidly evolve teaching methods and publication policies to keep up with AI [21].
Despite these limitations, ChatGPT shows promise for use in healthcare settings, provided it is integrated mindfully. ChatGPT can be used as a search engine for both patients and healthcare professionals, for medical education, and for patient monitoring [14,22]. For healthcare professionals, ChatGPT could function as a powerful tool to remain up to date on the vast amounts of new scientific literature published daily, and to structure their critical analysis of such literature [23,24]. In medical education, ChatGPT could be similarly employed to structure preparation for standardized exams and summarize information for high-yield review [25,26]. For patients, ChatGPT may provide a tool to access health information on simple topics or conditions, or to clarify and summarize complex medical literature that would otherwise be challenging to comprehend [27,28]. Advancing ChatGPT’s transformer architecture by training it on additional parameters with each subsequent release (117 million for GPT-1, 1.5 billion for GPT-2, 175 billion for GPT-3) expands its ability to interpret the context of a string or block of text input and respond accordingly [29], allowing for increased precision in healthcare settings where accuracy is vital to patient safety and provider knowledge.
The release of ChatGPT-4 in February 2023 further expanded ChatGPT’s capabilities, particularly with regard to its accuracy, increasing ChatGPT test scores on standardized exams and decreasing the frequency of fabricated information [29]. ChatGPT-4 also allows for inputs in the form of images or data rather than text only [29], expanding its ability to generate text based on a body of data, which certainly has implications for scientific research, particularly manuscript writing, and may also improve ChatGPT’s diagnostic capabilities in response to patient data input (demographics, exam findings, lab values, imaging, etc.).
Given the exponential increase in articles published on the use of ChatGPT in academia and medicine, this bibliometric analysis seeks to report on the current state of ChatGPT literature in these settings, evaluating reports of ChatGPT’s use in health care and research endeavors in various countries and specialties. Further, it will examine the advantages and pitfalls of ChatGPT in patient care, medical education, and scientific research.

2. Materials and Methods

A title, abstract, and keyword search of PubMed, Embase, Scopus, and Web of Science databases was conducted using the search term “(“ChatGPT”) AND (med* OR surg* OR physician OR doctor OR patient)”. Criteria for inclusion were: (1) primary literature on ChatGPT in the medical field (including patient care, medical research, and medical education), (2) peer-reviewed, and (3) published in English. Articles were excluded based on the following exclusion criteria: (1) not primary literature on ChatGPT in the medical field, (2) primary literature on ChatGPT use in basic science and non-medically focused academic applications, (3) articles written by ChatGPT (and not about ChatGPT itself), (4) not peer reviewed (abstract, poster, published in preprint server, etc.), (5) not published in English.
Articles were uploaded to Rayyan [30] and duplicates were removed. Title and abstract screening were conducted by two independent and blinded reviewers (NG and BM). Afterwards, conflicts were resolved by a third independent reviewer (NB).
Included articles were subjected to a data extraction process. To ensure homogenous data, citation counts for all articles were obtained from PubMed and journal metrics were obtained from Clarivate Journal Citation Reports. Further variables collected included date of publication, journal, country of origin for first and last author, total citations, medical specialty/subspecialty, topic of article, study type, and journal metrics. For publication data, the month of official e-publication was taken (rather than online-ahead-of-print or early online preview dates). Journal metrics were obtained using the 2023 Journal Citation Report [31]. Data analysis and figure generation were completed using Microsoft Excel, Google Sheets, and Python.

3. Results

3.1. Search Results

The initial search results yielded 889 total results and, after duplicates were removed, 393 articles remained for screening. Upon the initial title and abstract screen, 196 articles were excluded, followed by 33 exclusions upon full text review. Finally, 267 articles were included in the study (Figure 1).

3.2. General Characteristics, Temporal and Geographic Trends

The types of articles included were mainly letters to the editor/correspondence (n = 91, 35.2%) and commentary/editorials (n = 81, 31.3%). Articles, including observational and survey studies, comprised 78 articles (30.2%). Finally, case reports comprised six articles (2.3%). The included articles had a total of 1975 citations, ranging from 0 to 147 citations. The mean number of citations was 7.5 +/− 18.4 per publication, with a median of one citation. Finally, there were 177 unique journals represented.
As the use of ChatGPT is relatively novel, all publications included in this study were from 2023. Therefore, temporal trends were evaluated by month, rather than year (as traditionally seen in bibliometric analyses) (Figure 2). The number of publications on ChatGPT in medicine steadily increased from January 2023 (4 publications) to March 2023 (42 publications), reaching a peak in April 2023 (90 articles). Our search indicated a lower number of articles published in May and June 2023 (78 and 27, respectively).
For geospatial trends in publications on ChatGPT in medicine, the country of the senior author was recorded and is presented in Figure 3. The United States had the highest number of publications (89), followed by India (24), China (21), and the United Kingdom (15).

3.3. Topics and Medical Specialties

Articles were probed to determine the focus of the content, with most articles mainly concerned about the use of ChatGPT in research (62). The second most common topic was the evaluation of ChatGPT capabilities and accuracy (59). Examples of these included studies such as “Artificial Hallucinations in ChatGPT: Implications in Scientific Writing” and “Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum”. Other common topics included seeking to understand the use of ChatGPT in patient counseling (20) and medical education (19). Examples of studies which looked at these topics included “Artificial intelligence chatbots will revolutionize how cancer patients access information: ChatGPT represents a paradigm-shift” and “How Does ChatGPT Perform on the United States Medical Licensing Examination? The Implications of Large Language Models for Medical Education and Knowledge Assessment”, respectively. There were 55 articles which fit multiple topics, with common combinations including the use in research and evaluation of the accuracy of ChatGPT, and patient counseling and diagnostic/treatment plans. These data are summarized in Figure 4.
For the application of ChatGPT, specifically to aid in certain medical fields, there were 20 non-surgical and 12 surgical specialties represented (Table 1). For non-surgical specialties, the application of ChatGPT was most published in radiology (including both diagnostic and interventional) (21, 25.3%). The second most common was internal medicine/primary care (10, 12.0%), followed by oncology (6, 7.2%). For surgical applications/specialties, plastic surgery had the highest number of articles (18, 26.9%). General surgery had the second most (15, 22.4%), followed by orthopedic surgery and ophthalmology, (7, 10.4% and 6, 9.0%, respectively). For neurosurgery specifically, there were four articles (6.0%).

3.4. Citation Metrics and Characteristics of Top 20 Cited Studies

The top 20 most-cited articles for the use of ChatGPT in medicine ranged in citation count from 147 (“How Does ChatGPT Perform on the United States Medical Licensing Examination? The Implications of Large Language Models for Medical Education and Knowledge Assessment”) to 25 citations (“Assessing the performance of ChatGPT in answering questions regarding cirrhosis and hepatocellular carcinoma”) (Table 2). The average number of citations was 60.1 +/− 35.3 and the majority (14) of articles were published during February and March of 2023. Geographic analysis of the top articles revealed substantial contributions from countries such as the United States and United Kingdom, with six and four articles, respectively. Countries with a more minor contribution included European countries (such as Italy, Belgium, and France), Australia, Pakistan, and Taiwan. Finally, the most common type of article in the top 20 most cited were editorials and letters to the editor.
The most common topic within the top cited articles was the use of ChatGPT in research (9 of 20). Other topics which were covered included the evaluation of ChatGPT’s capabilities, diagnosis and treatment plans, and the use of ChatGPT for reducing the burden of administrative tasks for physicians. As the most common topic was the application of ChatGPT to research in general, 12 of the 20 studies were not specific to a singular medical specialty. Among the papers that investigated the application of ChatGPT to a specific specialty, the most common was surgery (two general surgery, one orthopedic surgery, and one neurosurgery). Interestingly, both the neurosurgical and orthopedic surgery articles focused mainly on research, a large aspect of both specialties. The three other medical specialties represented in the top 20 included internal medicine, hepatology, and infectious disease.

3.5. Top Journals and Keywords

The top five journals with the highest number of related publications contained on average 10 +/− 3.7 publications, with an average impact factor and h-index of 150.8 +/− 117.6 and 8.6 +/− 8.3, respectively (Table 3). The journal with the highest number of articles was The Annals of Biomedical Engineering, containing 14 articles (IF = 3.8). Following this was the Aesthetic Surgery Journal (13 articles, IF = 2.9), and Cureus (10 articles, IF = 1.15).
Within the top 20 most-cited articles, the most common journal was The Lancet, with three articles published in The Lancet Digital Health and one article published in The Lancet Infectious Disease. The top five journals with the highest number of related publications contained on average 10 +/− 3.7 publications, with an average h-index and impact factor of 150.8 +/− 117.6 and 8.6 +/− 8.3, respectively (Table 3). The journal with the highest number of articles was The Annals of Biomedical Engineering, containing 14 articles (h-index = 141, IF = 3.8). Following this was the Aesthetic Surgery Journal (13 articles, h-index = 71, IF = 2.9), and Cureus (10 articles, h-index = NA, IF = 1.15). Other journals with significant contributions included Radiology (8 articles, h-index = 320, IF = 19.7) and the International Journal of Surgery (5 articles, h-index = 71, IF = 15.3).
Finally, to visually represent the top keywords which were used for articles included in this study, a keyword cloud map was generated (Figure 5).

4. Discussion

The abrupt emergence of LLMs into public awareness and their rapid adoption in academics and medicine has highlighted their applicability and compelled careful evaluation of the ethical concerns raised by this powerful technology. Although there are a multitude of LLMs which are emerging, OpenAI’s ChatGPT is among the most popular, surpassing 100 million users within two months of release [32]. As such, its use among the medical research community and patients has led to the production of large amounts of literature. Therefore, we performed a comprehensive bibliometric analysis of ChatGPT to better understand the response of the scientific and medical community to the use of LLMs for both academic and clinical pursuits.
The increasing use of LLMs in academic and medical settings is highlighted by the types of publications appearing in our analysis, with editorials and correspondence publications identified as the most common article types. This is likely the result of an urgent need for dialogue on the implications of these unique rapidly materializing technologies, that in the case of ChatGPT became the fastest-growing consumer application in mere months [10]. This dialogue was necessary to initiate discussions on the potential benefits and risks of incorporating LLMs, and the need to build a foundation within medicine for its use. These commentaries initially outpaced the ability to perform observational and controlled studies evaluating the technology’s capabilities and accuracy. The editorials and correspondences published have often emphasized concerns regarding the use of ChatGPT in academics and medicine, particularly surrounding authorship, plagiarism, and patient safety issues such as the spread of misinformation. Given these concerns, it is unsurprising that the most-cited articles were those that put ChatGPT to the test using medical board exams or questions about specific medical conditions. As pointed out in many of the editorial publications included in this study, there is a paucity of information about ChatGPT’s accuracy regarding medical topics, and the few studies that tested ChatGPT’s capabilities in a controlled way to better understand its limitations were cited by many other articles. As such, the immense potential of ChatGPT must be carefully tempered to safeguard against misinformation and potential biases generated by the datasets used to train ChatGPT.
In terms of geography and medical specialties, the United States, India, and China had the highest number of ChatGPT-related publications, while European countries had lower numbers of articles. This is likely representative of the significant research output of these countries in general and may also be reflective of the prevalence of artificial intelligence research in the U.S. and China in particular [33,34,35]. Interestingly, radiology was the most represented specialty among ChatGPT-related articles, which may suggest an interest in exploring whether artificial intelligence could be used to accelerate the creation of radiology reports, enabling clinicians to redistribute their time toward assessing images [36,37]. Second to radiology was internal medicine, a field in which artificial intelligence could be used similarly to expedite clinical charting [38], generate differential diagnoses, or provide patients with basic information about their diagnoses, provided the information is accurate [39,40]. Given the hands-on and highly technical nature of surgical specialties, it is interesting, although perhaps not surprising, that most articles emerging from these fields focused on using ChatGPT in research settings. In plastic surgery for example, ChatGPT may not be particularly useful in the operating room outside of generating operative notes, but it can accelerate academic productivity in fields that are highly competitive and value publication output. According to the 2022 NRMP Match Data [41], the highest number of publications were produced by applicants matching in plastic surgery and neurological surgery. For osteopathic medical students, neurological surgery had the highest number of publications per matched applicant. Our data reflect the large number of plastic surgery publications; however, it is interesting to note the lack of neurosurgical publications—even with the “publication arms race” for medical students applying for neurosurgery [42].
Though the potential of ChatGPT is vast, our analysis makes clear the need for observational controlled studies to fully evaluate the capabilities and accuracy of ChatGPT in academic, medical, and research settings. Many of the editorials included in our analysis discussed the importance of purposeful and ethical integration of artificial intelligence in these settings to avoid compromising intellectual integrity. In competitive fields such as plastic surgery and neurosurgery, ChatGPT’s capabilities may add fuel to the “publication race” by encouraging the use of artificial intelligence to increase publication output, as it may be impossible for authors who choose not to use artificial intelligence to keep up with those who do [43]. While this is likely to occur more prominently among students and trainees seeking to match into residency or build their academic careers, as ChatGPT’s capabilities expand this will certainly impact the upper echelons of academic medicine and science, where eligibility for academic promotions or funding for scientific endeavors is often directly tied to publication output and grant writing productivity, both of which can be accelerated via artificial intelligence.
Ethical integration of ChatGPT into medicine and scientific research was among the most discussed topics in our study. These ethical considerations will not only require significant further study of ChatGPT itself, but will also require institutions, journals, and medical associations to formulate clear guidelines for the use of ChatGPT in each setting [20,44,45,46]. ChatGPT has incredible potential to reduce administrative burdens in medicine by assisting with charting and record-keeping [38], potentially enabling providers to focus more on the clinician-patient relationship that is often challenging to prioritize in modern medicine. However, it is essential to ensure that recorded information is accurate, and that diagnoses are not missed due to an over-reliance on artificial intelligence. Similarly, ChatGPT has the potential to curb the impact of our present healthcare worker shortage [47], specifically by providing patients with basic information that they might otherwise seek in primary care settings. In the age of misinformation, using ChatGPT in patient-facing formats must be evaluated and integrated with caution to ensure patients are given accurate information about their health [32], and appropriately directed to human providers when needed.
In research settings, transparency is vital to ensuring intellectual integrity, which may take the form of statements specifying where and how artificial intelligence was used within a publication, or stated limitations on the appropriate use of artificial intelligence when submitting a manuscript for publication. The scientific community will need to grapple with difficult questions when determining what is acceptable: where do we draw the line between ChatGPT writing assistance (outlining, syntax suggestions) and articles written entirely by ChatGPT? Is ChatGPT use allowable in training settings (from primary school through to graduate school and beyond) where the writing process itself is intended to be a learning experience? None of these questions have an easy answer, but as ChatGPT and other LLM platforms become ubiquitous, it may become necessary to redefine our conceptions of plagiarism, originality, and innovation to make room for both the enormous potential and dangerous pitfalls of artificial intelligence.

Limitations

A recognized constraint in the methodology of bibliometric analysis is the necessity to rely on citations as a metric for analysis, as various factors influence the likelihood of being cited, such as the journal’s impact factor, the reputation of the authors involved, and the recency of publication [48]. Specifically, within the context of a novel technology such as ChatGPT, it is believed that the recentness of its adaptation has impacted the ability of relevant publications to be recognized, resulting in some potent articles likely being overlooked. Collectively, this could potentially lead to a lack of citations within these meaningful works, falsely deflating their relevance.
Another limitation of our study is that we focused primarily on ChatGPT, which is currently the most prevalent LLM in the medical field. Other LLMs, such as Meta’s large language modeled meta-AI (LLaMA), have not been extensively reported on due to relatively recent deployment or less adoption by the medical community. For non-specialized LLMs such as LLaMA there are a small but growing number of reports in the literature which have opted to use it over ChatGPT because it is open source [49]. There are other, more specialized LLMs, such as the recently reported HealthSearchQA [50], Stanford’s BioMedLM [51], and RadBERT which is specifically for radiology [52], that could have significant impacts on the medical research community as well. Similarly, neurosurgical journals are utilizing novel and innovative approaches to artificial intelligence, by employing specialized conversational LLMs to aid readers in understanding manuscripts [53]. Given the novelty of these emerging platforms, these specialized LLMs were not evaluated in our study, as there are not enough studies to warrant an accurate bibliometric analysis. The extensive range of applications for large language models such as ChatGPT warrant extensive validation of these technologies across diverse settings in medicine and science, while concurrently establishing universal guidelines and ethical models for their efficient and responsible utilization.

5. Conclusions

As ChatGPT has had an explosive impact on the field of medicine, this study sought to evaluate the exponential increase in publications since the release of ChatGPT and its impact on the medical field. Our results suggest that there is a large focus on the application of ChatGPT in medical research, the accuracy and capabilities of ChatGPT, and its use in patient education/counseling. It is critical that future studies seek to thoroughly evaluate the impact and potential uses and misuses of ChatGPT, including ethical and legal implications. While current research has sought to exhibit the wide range of capabilities ChatGPT possesses, our review establishes that gaps exist in the literature with regards to demonstrating its specific limitations. Potential applications should also be further established in a specialty-specific context, as wide variations in research activity among specialties reveal significant potential for an exploration of ChatGPT as a patient care tool within many nonsurgical specialties. Together, these developments in research will likely further establish and optimize the niche of LLM in medicine. Our analysis suggests LLM accuracy and precision in healthcare and research settings will be the primary topic of discussion in this body of literature moving forward, and, given the rapid rollout of updated versions of ChatGPT and its peer LLM platforms, it will be essential for medical professionals and scientists to remain vigilant in ensuring the safe and ethical implementation of these technologies in an ever-changing healthcare landscape.

Author Contributions

All authors contributed to the study conception and design. Material preparation, data collection and analysis were performed by N.M.B., N.G. (Nithin Gupta), B.M., N.P., D.D. and N.G. (Nikhil Godboyle). T.R. and R.S.D. guided the data collection and analysis. The first draft of the manuscript was written by N.M.B. and N.G. (Nithin Gupta), who had equal contribution to the work. All authors commented on and edited previous versions of the manuscript. All authors have read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable.

Informed Consent Statement

Not applicable.

Data Availability Statement

All data generated during the study are contained in this manuscript.

Conflicts of Interest

The authors declare no conflict of interest.

References

  1. Sejnowski, T.J. Large Language Models and the Reverse Turing Test. Neural Comput. 2023, 35, 309–342. [Google Scholar] [CrossRef]
  2. Kung, T.H.; Cheatham, M.; Medenilla, A.; Sillos, C.; De Leon, L.; Elepaño, C.; Madriaga, M.; Aggabao, R.; Diaz-Candido, G.; Maningo, J.; et al. Performance of ChatGPT on USMLE: Potential for AI-assisted medical education using large language models. PLoS Digit. Health 2023, 2, e0000198. [Google Scholar] [CrossRef] [PubMed]
  3. Jeblick, K.; Schachtner, B.; Dexl, J.; Mittermeier, A.; Stüber, A.T.; Topalis, J.; Weber, T.; Wesp, P.; Sabel, B.; Ricke, J.; et al. ChatGPT Makes Medicine Easy to Swallow: An Exploratory Case Study on Simplified Radiology Reports. arXiv 2022, arXiv:2212.14882. Available online: http://arxiv.org/abs/2212.14882 (accessed on 1 August 2023).
  4. Introducing ChatGPT. Available online: https://openai.com/blog/chatgpt (accessed on 4 September 2023).
  5. Li, H.; Moon, J.T.; Purkayastha, S.; Celi, L.A.; Trivedi, H.; Gichoya, J.W. Ethics of large language models in medicine and medical research. Lancet Digit. Health 2023, 5, e333–e335. [Google Scholar] [CrossRef] [PubMed]
  6. Vaswani, A.; Shazeer, N.; Parmar, N.; Uszkoreit, J.; Jones, L.; Gomez, A.N.; Kaiser, L.; Polosukhin, I. Attention is all you need. arXiv 2017, arXiv:1706.03762. Available online: https://arxiv.org/abs/1706.03762 (accessed on 1 August 2023).
  7. Bahdanau, D.; Cho, K.; Bengio, Y. Neural Machine Translation by Jointly Learning to Align and Translate. arXiv 2016, arXiv:1409.04732014. Available online: https://arxiv.org/abs/1409.0473 (accessed on 1 August 2023).
  8. Transformer Architecture: The Engine Behind ChatGPT. Available online: https://www.thoughtspot.com/data-trends/ai/what-is-transformer-architecture-chatgpt (accessed on 28 August 2023).
  9. Carman, J.; Cui, J. Inside ChatGPT: How AI chatbots work. NBC News, 17 May 2023. Available online: https://www.nbcnews.com/data-graphics/chat-gpt-artificial-intelligence-how-chatbot-work-rcna83266 (accessed on 1 August 2023).
  10. Hu, K. ChatGPT sets record for fastest-growing user base—Analyst note. Reuters, 2 February 2023. Available online: https://www.reuters.com/technology/chatgpt-sets-record-fastest-growing-user-base-analyst-note-2023-02-01/ (accessed on 1 August 2023).
  11. Tamayo-Sarver, J. I’m an ER doctor, here’s how I’m using ChatGP to help treat patients. Fast Company, 15 May 2023. Available online: https://www.fastcompany.com/90895618/how-a-doctor-uses-chat-gpt-to-treat-patients (accessed on 1 August 2023).
  12. Subbaraman, N. ChatGPT Will See You Now: Doctors Using AI to Answer Patient Questions. WSJ, 28 April 2023. Available online: https://www.wsj.com/articles/dr-chatgpt-physicians-are-sending-patients-advice-using-ai-945cf60b (accessed on 1 August 2023).
  13. Barnett, S. ChatGPT Is Making Universities Rethink Plagiarism. WIRED, 30 January 2023. Available online: https://www.wired.com/story/chatgpt-college-university-plagiarism/ (accessed on 1 August 2023).
  14. Dave, T.; Athaluri, S.A.; Singh, S. ChatGPT in medicine: An overview of its applications, advantages, limitations, future prospects, and ethical considerations. Front. Artif. Intell. 2023, 6, 1169595. [Google Scholar] [CrossRef]
  15. Wen, J.; Wang, W. The future of ChatGPT in academic research and publishing: A commentary for clinical and translational medicine. Clin. Transl. Med. 2023, 13, e1207. [Google Scholar] [CrossRef]
  16. Sebastian, G. Exploring Ethical Implications of ChatGPT and Other AI Chatbots and Regulation of Disinformation Propagation. SSRN Electron. J. 2023, 1–16. [Google Scholar] [CrossRef]
  17. Varanasi, L. AI models like ChatGPT and GPT-4 are acing everything from the bar exam to AP Biology. Here’s a list of difficult exams both AI versions have passed. Business Insider, 25 June 2023. Available online: https://www.businessinsider.com/list-here-are-the-exams-chatgpt-has-passed-so-far-2023-1 (accessed on 1 August 2023).
  18. Oztermeli, A.D.; Oztermeli, A. ChatGPT performance in the medical specialty exam: An observational study. Medicine 2023, 102, e34673. [Google Scholar] [CrossRef] [PubMed]
  19. Gilson, A.; Safranek, C.W.; Huang, T.; Socrates, V.; Chi, L.; Taylor, R.A.; Chartash, D. How Does ChatGPT Perform on the United States Medical Licensing Examination? The Implications of Large Language Models for Medical Education and Knowledge Assessment. JMIR Med. Educ. 2023, 9, e45312. [Google Scholar] [CrossRef]
  20. Stokel-Walker, C. ChatGPT listed as author on research papers. Nature 2023, 613, 620–621. [Google Scholar] [CrossRef]
  21. Gupta, N.K.; Doyle, D.M.; D’amico, R.S. Response to “Large language model artificial intelligence: The current state and future of ChatGPT in neuro-oncology publishing”. J. Neuro Oncol. 2023, 163, 731–733. [Google Scholar] [CrossRef] [PubMed]
  22. Will ChatGPT Transform Healthcare? Nat. Med. 2023, 29, 505–506. [CrossRef] [PubMed]
  23. Qureshi, R.; Shaughnessy, D.; Gill, K.A.R.; Robinson, K.A.; Li, T.; Agai, E. Are ChatGPT and large language models “the answer” to bringing us closer to systematic review automation? Syst. Rev. 2023, 12, 72. [Google Scholar] [CrossRef] [PubMed]
  24. Mahuli, S.A.; Rai, A.; Mahuli, A.V.; Kumar, A. Application ChatGPT in conducting systematic reviews and meta-analyses. Br. Dent. J. 2023, 235, 90–92. [Google Scholar] [CrossRef] [PubMed]
  25. Feng, S.; Shen, Y. ChatGPT and the Future of Medical Education. Acad. Med. 2023, 98, 867. [Google Scholar] [CrossRef]
  26. Mohammad, B.; Supti, T.; Alzubaidi, M.; Shah, H.; Alam, T.; Shah, Z.; Househ, M. The Pros and Cons of Using ChatGPT in Medical Education: A Scoping Review; Mantas, J., Gallos, P., Zoulias, E., Hasman, A., Househ, M.S., Charalampidou, M., Magdalinou, A., Eds.; Studies in Health Technology and Informatics; IOS Press: Amsterdam, The Netherlands, 2023. [Google Scholar] [CrossRef]
  27. Mondal, H.; Mondal, S.; Podder, I. Using ChatGPT for writing articles for patients’ education for dermatological diseases: A pilot study. Indian Dermatol. Online J. 2023, 14, 482. [Google Scholar] [CrossRef]
  28. Zhao, J.; Shao, C.; Li, H.; Liu, X.-L.; Li, C.; Yang, L.-Q.; Zhang, Y.-J.; Luo, J. Appropriateness and Comprehensiveness of Using ChatGPT for Perioperative Patient Education in Thoracic Surgery in Different Language Contexts: Survey Study. Interact. J. Med Res. 2023, 12, e46900. [Google Scholar] [CrossRef]
  29. GPT-4 vs. ChatGPT-3.5: What’s the Difference? PCMag, 16 March 2023. Available online: https://www.pcmag.com/news/the-new-chatgpt-what-you-get-with-gpt-4-vs-gpt-35 (accessed on 28 August 2023).
  30. Ouzzani, M.; Hammady, H.; Fedorowicz, Z.; Elmagarmid, A. Rayyan—A web and mobile app for systematic reviews. Syst. Rev. 2016, 5, 210. [Google Scholar] [CrossRef]
  31. First Time Journal Citation Reports Inclusion List 2023. Clarivate. Available online: https://clarivate.com/first-time-journal-citation-reports-inclusion-list-2023/ (accessed on 1 August 2023).
  32. Shah, N.H.; Entwistle, D.; Pfeffer, M.A. Creation and Adoption of Large Language Models in Medicine. JAMA 2023, 330, 866–869. [Google Scholar] [CrossRef]
  33. Savage, N. The race to the top among the world’s leaders in artificial intelligence. Nature 2020, 588, S102–S104. [Google Scholar] [CrossRef]
  34. Country/Territory Tables | Nature Index. Available online: https://www.nature.com/nature-index/country-outputs/generate/all/global (accessed on 4 September 2023).
  35. Fontelo, P.; Liu, F. A review of recent publication trends from top publishing countries. Syst. Rev. 2018, 7, 147. [Google Scholar] [CrossRef]
  36. Grewal, H.; Dhillon, G.; Monga, V.; Sharma, P.; Buddhavarapu, V.S.; Sidhu, G.; Kashyap, R. Radiology Gets Chatty: The ChatGPT Saga Unfolds. Cureus 2023, 15, e40135. [Google Scholar] [CrossRef]
  37. Bosbach, W.A.; Senge, J.F.; Nemeth, B.; Omar, S.H.; Mitrakovic, M.; Beisbart, C.; Horváth, A.; Heverhagen, J.; Daneshvar, K. Ability of ChatGPT to generate competent radiology reports for distal radius fracture by use of RSNA template items and integrated AO classifier. Curr. Probl. Diagn. Radiol. 2023, in press. [Google Scholar] [CrossRef]
  38. Harris, J.E. An AI-Enhanced Electronic Health Record Could Boost Primary Care Productivity. JAMA 2023, 330, 801. [Google Scholar] [CrossRef]
  39. Harris, E. Large Language Models Answer Medical Questions Accurately, but Can’t Match Clinicians’ Knowledge. JAMA 2023, 330, 792. [Google Scholar] [CrossRef]
  40. Walker, H.L.; Ghani, S.; Kuemmerli, C.; Nebiker, C.A.; Müller, B.P.; Raptis, D.A.; Staubli, S.M. Reliability of Medical Information Provided by ChatGPT: Assessment Against Clinical Guidelines and Patient Information Quality Instrument. J. Med. Internet Res. 2023, 25, e47479. [Google Scholar] [CrossRef]
  41. NRMP. Residency Data & Reports. Available online: https://www.nrmp.org/match-data-analytics/residency-data-reports/ (accessed on 1 August 2023).
  42. Wadhwa, H.; Shah, S.S.; Shan, J.; Cheng, J.; Beniwal, A.S.; Chen, J.-S.; Gill, S.A.; Mummaneni, N.; McDermott, M.W.; Berger, M.S.; et al. The neurosurgery applicant’s “arms race”: Analysis of medical student publication in the Neurosurgery Residency Match. J. Neurosurg. 2020, 133, 1913–1921. [Google Scholar] [CrossRef] [PubMed]
  43. Noy, S.; Zhang, W. Experimental evidence on the productivity effects of generative artificial intelligence. Science 2023, 381, 187–192. [Google Scholar] [CrossRef] [PubMed]
  44. Cascella, M.; Montomoli, J.; Bellini, V.; Bignami, E. Evaluating the Feasibility of ChatGPT in Healthcare: An Analysis of Multiple Clinical and Research Scenarios. J. Med. Syst. 2023, 47, 33. [Google Scholar] [CrossRef]
  45. Cifarelli, C.P.; Sheehan, J.P. Large language model artificial intelligence: The current state and future of ChatGPT in neuro-oncology publishing. J. Neuro Oncol. 2023, 163, 473–474. [Google Scholar] [CrossRef] [PubMed]
  46. Goto, A.; Katanoda, K. Should We Acknowledge ChatGPT as an Author? J. Epidemiol. 2023, 33, 333–334. [Google Scholar] [CrossRef]
  47. Howard, J. Concern Grows around US Health-Care Workforce Shortage: ‘We don’t Have Enough Doctors’. CNN, 16 May 2023. Available online: https://www.cnn.com/2023/05/16/health/health-care-worker-shortage/index.html (accessed on 1 August 2023).
  48. Reddy, V.P.; Singh, R.; McLelland, M.D.; Barpujari, A.; Catapano, J.S.; Srinivasan, V.M.; Lawton, M.T. Bibliometric Analysis of the Extracranial-Intracranial Bypass Literature. World Neurosurg. 2022, 161, 198–205.e5. [Google Scholar] [CrossRef] [PubMed]
  49. Li, Y.; Li, Z.; Zhang, K.; Dan, R.; Jiang, S.; Zhang, Y. ChatDoctor: A Medical Chat Model Fine-Tuned on a Large Language Model Meta-AI (LLaMA) Using Medical Domain Knowledge. Cureus 2023, 15, e40895. [Google Scholar] [CrossRef] [PubMed]
  50. Singhal, K.; Azizi, S.; Tu, T.; Mahdavi, S.S.; Wei, J.; Chung, H.W.; Scales, N.; Tanwani, A.; Cole-Lewis, H.; Pfohl, S.; et al. Large language models encode clinical knowledge. Nature 2023, 620, 172–180. [Google Scholar] [CrossRef] [PubMed]
  51. BioMedLM: A Domain-Specific Large Language Model for Biomedical Text. Available online: https://www.mosaicml.com/blog/introducing-pubmed-gpt (accessed on 28 August 2023).
  52. Yan, A.; McAuley, J.; Lu, X.; Du, J.; Chang, E.Y.; Gentili, A.; Hsu, C.-N. RadBERT: Adapting Transformer-based Language Models to Radiology. Radiol. Artif. Intell. 2022, 4, e210258. [Google Scholar] [CrossRef]
  53. Engaging a Scientific ‘Conversational Manuscript’ With Artificial Intelligence: A Test and Reader Invitation. Available online: https://www.newswise.com/articles/engaging-a-scientific-conversational-manuscript-with-artificial-intelligence-a-test-and-reader-invitation (accessed on 1 August 2023).
Medsci 11 00061 g001
Figure 1. PRISMA diagram demonstrating screening strategy for included articles.
Figure 1. PRISMA diagram demonstrating screening strategy for included articles.
Medsci 11 00061 g001
Medsci 11 00061 g002
Figure 2. Number of publications on ChatGPT in the medical field per month since the beginning of 2023.
Figure 2. Number of publications on ChatGPT in the medical field per month since the beginning of 2023.
Medsci 11 00061 g002
Medsci 11 00061 g003
Figure 3. Geospatial map showing the location of senior authors publishing on ChatGPT in medicine.
Figure 3. Geospatial map showing the location of senior authors publishing on ChatGPT in medicine.
Medsci 11 00061 g003
Medsci 11 00061 g004
Figure 4. Topics of ChatGPT publications.
Figure 4. Topics of ChatGPT publications.
Medsci 11 00061 g004
Medsci 11 00061 g005
Figure 5. Visual map of the top keywords used in included studies. Larger font indicates higher frequency of keyword usage.
Figure 5. Visual map of the top keywords used in included studies. Larger font indicates higher frequency of keyword usage.
Medsci 11 00061 g005
Table 1. Number of publications produced by surgical and non-surgical specialties.
Table 1. Number of publications produced by surgical and non-surgical specialties.
Medical Specialty Number of Publications
Non-SurgicalN (%)
Radiology21 (25.3%)
Internal Medicine/Primary Care10 (12.0%)
Oncology6 (7.2%)
Gastroenterology5 (6.02%)
Rheumatology5 (6.02%)
Dermatology4 (4.8%)
Emergency Medicine4 (4.8%)
Endocrine4 (4.8%)
Pediatrics4 (4.8%)
Psychiatry4 (4.8%)
Neurology3 (3.6%)
Anesthesia2 (2.4%)
Infectious Disease2 (2.4%)
Pathology2 (2.4%)
Pulmonology/Critical Care2 (2.4%)
Cardio1 (1.2%)
Fam Med1 (1.2%)
Hepatology1 (1.2%)
Sports Med1 (1.2%)
Toxicology1 (1.2%)
Surgical SpecialtiesN (%)
Plastic Surgery18 (26.9%)
General Surgery15 (22.4%)
Orthopedic Surgery7 (10.4%)
Ophthalmology6 (9.0%)
Obstetrics and Gynecology 5 (7.5%)
Neurosurgery4 (6.0%)
Otolaryngology4 (6.0%)
Oral and Maxillofacial Surgery2 (3.0%)
Surgical Oncology2 (3.0%)
Urology2 (3.0%)
Bariatric Surgery1 (1.5%)
Colorectal Surgery 1
Table 2. Top 20 most-cited publications regarding the use of ChatGPT in medicine.
Table 2. Top 20 most-cited publications regarding the use of ChatGPT in medicine.
RankArticle Name Number of CitationsJournal of Article
1How Does ChatGPT Perform on the United States Medical Licensing Examination? The Implications of Large Language Models for Medical Education and Knowledge Assessment.147JMIR Medical Education
2A Conversation on Artificial Intelligence, Chatbots, and Plagiarism in Higher Education119Cellular and Molecular Bioengineering
3Artificial Hallucinations in ChatGPT: Implications in Scientific Writing118Cureus Journal of Medical Science
4ChatGPT: the future of discharge summaries?94The Lancet Digital Health
5Can artificial intelligence help for scientific writing?87Critical Care
6Evaluating the Feasibility of ChatGPT in Healthcare: An Analysis of Multiple Clinical and Research Scenarios.68Journal of Medical Systems
7Role of Chat GPT in Public Health61Annals of Biomedical Engineering
8ChatGPT: evolution or revolution?57Medicine, Health Care, and Philosophy
9Generating scholarly content with ChatGPT: ethical challenges for medical publishing56The Lancet Digital Health
10ChatGPT—Reshaping medical education and clinical management.51Pakistan Journal of Medical Sciences
11The future of medical education and research: Is ChatGPT a blessing or blight in disguise?48Medical Education Online
12Can ChatGPT draft a research article? An example of population-level vaccine effectiveness analysis.45Journal of Global Health
13Comparing Physician and Artificial Intelligence Chatbot Responses to Patient Questions Posted to a Public Social Media Forum.41JAMA Internal Medicine
14ChatGPT and other artificial intelligence applications speed up scientific writing.37Journal of the Chinese Medical Association
15Revolutionizing radiology with GPT-based models: Current applications, future possibilities and limitations of ChatGPT.32Diagnostic and Interventional Imaging
16Using ChatGPT to write patient clinic letters.32The Lancet Digital Health
17Artificial intelligence bot ChatGPT in medical research: the potential game changer as a double-edged sword.29Knee Surgery, Sports Traumatology, Arthroscopy
18ChatGPT and antimicrobial advice: the end of the consulting infection doctor?28The Lancet Infectious Diseases
19To ChatGPT or not to ChatGPT? The Impact of Artificial Intelligence on Academic Publishing26The Pediatric Infectious Disease Journal
20Assessing the performance of ChatGPT in answering questions regarding cirrhosis and hepatocellular carcinoma.25Clinical and Molecular Hepatology
Table 3. Characteristics of journals with the top 5 highest numbers of related publications.
Table 3. Characteristics of journals with the top 5 highest numbers of related publications.
Journal NameNumber of Related Publicationsh-Index Impact Factor
Annals of Biomedical Engineering141413.8
Aesthetic Surgery Journal13712.9
Cureus10NA1.15
Radiology832019.7
International Journal of Surgery57115.3
Average +/− SD10 +/− 3.7150.8 +/− 117.68.6 +/− 8.3
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.

Share and Cite

MDPI and ACS Style

Barrington, N.M.; Gupta, N.; Musmar, B.; Doyle, D.; Panico, N.; Godbole, N.; Reardon, T.; D’Amico, R.S. A Bibliometric Analysis of the Rise of ChatGPT in Medical Research. Med. Sci. 2023, 11, 61. https://doi.org/10.3390/medsci11030061

AMA Style

Barrington NM, Gupta N, Musmar B, Doyle D, Panico N, Godbole N, Reardon T, D’Amico RS. A Bibliometric Analysis of the Rise of ChatGPT in Medical Research. Medical Sciences. 2023; 11(3):61. https://doi.org/10.3390/medsci11030061

Chicago/Turabian Style

Barrington, Nikki M., Nithin Gupta, Basel Musmar, David Doyle, Nicholas Panico, Nikhil Godbole, Taylor Reardon, and Randy S. D’Amico. 2023. "A Bibliometric Analysis of the Rise of ChatGPT in Medical Research" Medical Sciences 11, no. 3: 61. https://doi.org/10.3390/medsci11030061

APA Style

Barrington, N. M., Gupta, N., Musmar, B., Doyle, D., Panico, N., Godbole, N., Reardon, T., & D’Amico, R. S. (2023). A Bibliometric Analysis of the Rise of ChatGPT in Medical Research. Medical Sciences, 11(3), 61. https://doi.org/10.3390/medsci11030061

Note that from the first issue of 2016, this journal uses article numbers instead of page numbers. See further details here.

Article Metrics

Back to TopTop