1

1 Exploring the Role of Chat GPT in patient care (diagnosis and Treatment) and medical research: A Systematic Review
2
3
4
5
6
7 Ravindra Kumar Garg
8 Vijeth L Urs
9 *Akshya Anand Agrawal
10 Sarvesh Kumar Chaudhary
11 **Vimal Paliwal
12 ***Sujita Kumar Kar
13
14
15 Department of Neurology
16 King George’s Medical University
17 Lucknow, India: 226003
18
19 *Department of Surgery
20 King George’s Medical University
21 Lucknow, India: 226003
22
23 **Department of Neurology
24 Sanjay Gandhi Institute of Medical Sciences
25 Lucknow, India: 226001
26
27 ***Department of Psychiatry
28 King George’s Medical University
29 Lucknow, India: 226003
30
31
32
 2

33 Correspondence
34 Ravindra Kumar Garg
35 Department of Neurology
36 King George’s Medical University
37 Lucknow, India: 226003
38 Mobile: +91 9335901790
39 Email: [email protected]
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55

56

57

58

59
 3

60 Abstract

61 Background

62 ChatGPT(Chat Generative Pre-trained Transformer) is an artificial intelligence (AI) based on a natural language processing tool

63 developed by OpenAI (California, USA). This systematic review examines the potential of Chat GPT in diagnosing and treating

64 patients and its contributions to medical research.

65

66 Methods

67 In order to locate articles on ChatGPT's use in clinical practise and medical research, this systematic review used PRISMA

68 standards and conducted database searches across several sources. Selected records were analysed using ChatGPT, which also

69 produced a summary for each article. The resultant word document was transformed to a PDF and handled using ChatPDF. The

70 review looked at topics pertaining to scholarly publishing, clinical practise, and medical research.

71

72 Results

73 We reviewed 118 publications. There are difficulties and moral conundrums associated with using ChatGPT in therapeutic settings

74 and medical research. Patient inquiries, note writing, decision-making, trial enrolment, data management, decision support,

75 research support, and patient education are all things that ChatGPT can help with. However, the solutions it provides are frequently
 4

76 inadequate and inconsistent, presenting issues with its originality, privacy, accuracy, bias, and legality. When utilising ChatGPT for

77 academic writings, there are issues with prejudice and plagiarism, and because it lacks human-like characteristics, its authority as

78 an author is called into question.

79

80 Conclusions

81 ChatGPT has limitations when used in research and healthcare. Even while it aids in patient treatment, concerns regarding

82 accuracy, authorship, and bias arise. Currently, ChatGPT can serve as a "clinical assistant" and be a huge assistance with

83 research and scholarly writing.

84

85

86 Key words: Artificial intelligence; Plagiarism; Authorship; Scholarly publishing

87

88

89

90

91
 5

92 Introduction

93 ChatGPT(Chat Generative Pre-trained Transformer) is an artificial intelligence (AI) based on a natural language processing tool

94 developed by OpenAI (California, USA). ChatGPT is chat boat based technology. A chatbot is in fact a type of software creates text

95 akin to human-like conversation. ChatGPT has the capacity to respond to follow-up questions, recognise errors, debunk unfounded

96 theories, and turn down inappropriate requests. Large Language Models (LLMs), which are frequently abbreviated as LLMs, are

97 extremely complex deep-learning programmes that are capable of comprehending and producing text in a manner that is strikingly

98 comparable to that of humans. LLMs can recognise, summarise, translate, predict, and create text as well as other sorts of

99 information by using the large knowledge base they have amassed from massive datasets.1,2

100

101 The possible uses of ChatGPT in medicine is currently under intense investigation. ChatGPT is considered to have enormous

102 capability in helping experts with clinical and laboratory diagnosis to planning and execution of medical research. Another

103 significant use of ChatGPT in medical researchers is the creation of virtual assistants to physicians helping them in writing

104 manuscripts in more efficient way. Usage of ChatGPT in medical writing is considered to have associated with several ethical and

105 legal issues. Possible copyright violations, medical-legal issues, and the demand for openness in AI-generated content are a few of

106 these.3-5

107
 6

108 In this systematic review we aimed to review published article and explore the potential of ChatGPT in facilitating patient care,

109 medical research and medical writing. We will also focus on ethical issues associated with usage of ChatGPT.

110 Methods

111 We performed a systematic review of published articles on ChatGPT. The protocol of the systematic review was registered with

112 PROSPERO (PROSPERO 2023 CRD42023415845).6 Our systematic review was conducted following the Preferred Reporting

113 Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines.

114

115 Search strategy

116 We searched four databases, PubMed, Scopus, Embase, and Google Scholar. Our search was aimed at identifying all kinds of

117 articles on ChatGPT and its application in medical research, scholarly and clinical practice, published till 24 May 2023. Articles

118 related to medical education was not considered. The search item that we used was “ChatGPT”. We reviewed all kinds of

119 publications including original articles, reviews, editorial/ commentaries and even letter to the editor describing ChatGPT.

120

121 Data extraction

122 The selection of the papers that were published was done in two steps. Two reviewers (RK and VKP) reviewed the titles and

123 abstracts in the initial phase. Two reviewers (VU and SKC) then examined the entire texts of the chosen papers to determine their
 7

124 eligibility. A third author (SK) settled any differences that arose between the two authors. Two reviewers (RK and VKP) assessed

125 the information available in the included publication for the suitability of the article to be included in the review. Any disagreement

126 between them was resolved by mutual agreement. If a dispute persisted, it was resolved via consultation with a third reviewer (SK).

127

128 EndNote 20 web tool (Clarivate Analytics) was used to handle duplicate records. This process was carried out by two reviewers

129 independently (RK and VKP). Any issue that arose was resolved with a discussion with another reviewer. The number of retrieved

130 and assessed records at each stage was provided in the form of a PRISMA flow chart. EndNote 20 (Clarivate Analytics) was used

131 to make a PRISMA flow chart.

132

133 Quality assessment

134 Quality assessment was not done.

135

136 Data analysis

137 ChatGPT was extensively used for analysing the selected records and writing this manuscript. A table was made with six columns

138 (First author/sole author, country of origin, status of peer review (peer-reviewed or preprint), title of the paper and short point wise

139 summary of full text. Short point wise summary of full text of each and every article was created with the help of ChatGPT. The
 8

140 voluminous word file was then converted to a pdf file and was processed with the sister software “ ChatPDF” (OpenAI, California,

141 USA available at https://www.chatpdf.com/). Following questions were asked from ChatPDF.

142

143 1. What are potential role of ChatGPT in medical writing and research?

144 2. What could be the role of ChatGPT in clinical practice?

145 3. What are ethical issues associated with paper writing?

146 4. Can ChatGPT be an author?

147 5. Can ChatGPT write text in good English and free of plagiarism?

148 6. Role of ChatGPT so far in neurological disorders related clinical practice and research.

149 7. Effectiveness and efficiency of ChatGPT in medical research and clinical settings

150 8. Potential benefits and limitations of ChatGPT in medical research and clinical applications

151 9. The ethical implications of using ChatGPT in medical research and clinical practice

152 10. Identify the gaps in the current research on ChatGPT and suggest areas for further investigation.

153 11. Provide insights into the potential future applications of ChatGPT in medical research and clinical practice

154 12. Recommendations for researchers, clinicians, and policymakers on the use of ChatGPT in medical research and clinical

155 practice
 9

156

157 All the responses were compiled in a word file.

158

159 Results

160 Our data collection followed PRISMA guidelines. (Additional file 1: PRISMA checklist) The PRISMA flowchart for our systematic

161 review is shown in Figure 1. We reviewed 118 publications. ChatGPT related publications are available from across the globe.

162 There were 33 original articles and rest were commentary/ editorial, review articles, research letters or letter to the editors. Out of

163 118 articles, 18 articles were available as preprint only. Summaries of 118 articles and answers to 12 questions have been

164 provided in form of tables. (Table-1 and Table-2)

165

166 Discussion

167 We looked into two main uses of ChatGPT: in healthcare settings and for medical writing and research. We studied 118 articles -

168 most were opinion pieces, commentaries, and reviews. Another group, Ruksakulpiwat et al, also did a similar study. They analyzed

169 six articles out of 114 that met their criteria. These articles covered a variety of ways to use ChatGPT, such as finding new drugs,

170 writing literature reviews, improving medical reports, providing medical info, bettering research methods, analyzing data, and

171 personalizing medicine.7

 10

172 Levin et al, on the other hand, conducted an analysis of the first batch of publications about ChatGPT. They found 42 articles

173 published in 26 journals in the 69 days after ChatGPT was launched. Only one was a research article. The rest were mostly

174 editorials and news pieces. Five publications focused on studies on ChatGPT. There were no articles on its use in Obstetrics and

175 Gynecology.In terms of where these articles were published, Nature was the top journal. Radiology and Lancet Digital Health came

176 next. The articles mostly discussed the quality of ChatGPT's scientific writing, its features, and its performance. Some also talked

177 about who should get credit for the work and ethical concerns. Interestingly, when comparing the articles that described a study to

178 the others, the average impact factor (a measure of the influence of a journal) was significantly lower for the study articles.8

179

180 In our review, we identified several potential advantages of using ChatGPT in the medical field. It appears to enhance productivity

181 and expedite research workflows by aiding in data organization, assisting in the selection of trial candidates, and supporting overall

182 research activities. Furthermore, ChatGPT's capacity to review manuscripts and contribute to editing may potentiate the efficiency

183 of academic publishing. Beyond the scope of research, it could also prove beneficial for patient education, fostering scientific

184 exploration, and shaping clinical decision-making. However, we also need to consider certain limitations and ethical concerns

185 associated with the use of ChatGPT. The model, as sophisticated as it is, lacks the capability to offer comprehensive diagnoses

186 and cannot replace the human qualities inherent to medical practice. Ethical issues also arise, specifically in relation to potential

187 biases in the machine learning model and potential breaches of privacy. Moreover, while ChatGPT can process and generate
 11

188 information, it might not exhibit the level of originality, creativity, and critical thinking that are often required in the medical field.

189 However, the use of ChatGPT in producing scholarly articles is raising questions in the academic publishing. While these tools can

190 greatly enhance the clarity and fluency of written material, it is crucial that human oversight is maintained throughout the process.

191 This is because AI can potentially produce content that is authoritative-sounding, yet it might be inaccurate, incomplete, or biased.

192 Incorrect GPT-4 responses, known as "hallucinations," can be harmful, particularly in the field of medicine. Therefore, it is essential

193 to check or validate GPT-4's output. ChatGPT can generate references to made-up research publications. Therefore, authors must

194 thoroughly check and modify the output of these tools. Furthermore, it is not appropriate to recognize AI or AI-assisted tools as

195 authors or co-authors in the by-line of publications. Instead, their use should be transparently acknowledged within the manuscript.

196 For example, according to Elsevier's policy on AI for authors, the responsibility and accountability for the work ultimately still lie with

197 the human authors, despite any technological assistance they may have received.9-11

198

199 In conclusion, ChatGPT has a great potential. Its full potentials are still evolving. ChatGPT as a source of information can not be

200 trusted, many ethical issues are associated with it. Certainly, ChatGPT can be credited with authorship. However, ChatGPT is

201 certainly a good clinical assistant. ChatGPT is nowhere near to replace human brain.

202

203
 12

204 Acknowledgment

205 The concept, data collection analysis, writing, and reporting of this article were solely done by authors. ChatGPT was extensively

206 utilized as mentioned in the methods section.

207

208 Conflict of Interest

209 All authors have no conflict of interest to report.

210

211 Human/Animal Studies informed consent statement

212 No human or animal subjects were involved.

213

214 Financial support

215 None

216

217

218

219
 13

220 References

221 1. OpenAI. (2023). Conversational AI Model for Medical Inquiries. ChatGPT. https://www.openai.com/chatgpt.

222

223 2. Lee P, Bubeck S, Petro J. Benefits, Limits, and Risks of GPT-4 as an AI Chatbot for Medicine. N Engl J Med. 2023 Mar

224 30;388(13):1233-1239. doi: 10.1056/NEJMsr2214184.

225

226 3. Bommasani R, Liang P, Lee T. Holistic Evaluation of Language Models. Ann N Y Acad Sci. 2023 May 25. doi:

227 10.1111/nyas.15007. Epub ahead of print.

228

229 4. Waisberg E, Ong J, Masalkhi M, Kamran SA, Zaman N, Sarker P, Lee AG, Tavakkoli A. GPT-4: a new era of artificial

230 intelligence in medicine. Ir J Med Sci. 2023 Apr 19. doi: 10.1007/s11845-023-03377-8. Epub ahead of print.

231

232 5. Ong CWM, Blackbourn HD, Migliori GB. GPT-4, artificial intelligence and implications for publishing. Int J Tuberc Lung Dis. 2023

233 Jun 1;27(6):425-426. doi: 10.5588/ijtld.23.0143.

234
 14

235 6. Garg RK, Paliwal V, Kar SK, Urs VL, Chaudhary SK. Exploring the Role of Chat GPT in patient care (diagnosis and Treatment)

236 and medical research: A Systematic Review. PROSPERO 2023 CRD42023415845 Available

237 from: https://www.crd.york.ac.uk/prospero/display_record.php?ID=CRD42023415845.

238

239 7. Ruksakulpiwat S, Kumar A, Ajibade A. Using ChatGPT in Medical Research: Current Status and Future Directions. J Multidiscip

240 Healthc. 2023 May 30;16:1513-1520. doi: 10.2147/JMDH.S413470.

241

242 8. Levin G, Brezinov Y, Meyer R. Exploring the use of ChatGPT in OBGYN: a bibliometric analysis of the first ChatGPT-related

243 publications. Arch Gynecol Obstet. 2023 May 24. doi: 10.1007/s00404-023-07081-x. Epub ahead of print.

244

245 9. Anonymous. Tools such as ChatGPT threaten transparent science; here are our ground rules for their use. Nature. 2023

246 Jan;613(7945):612. doi: 10.1038/d41586-023-00191-1.

247

248 10. Hill-Yardin EL, Hutchinson MR, Laycock R, Spencer SJ. A Chat(GPT) about the future of scientific publishing. Brain Behav

249 Immun. 2023 May;110:152-154. doi: 10.1016/j.bbi.2023.02.022. Epub 2023 Mar 1.

250
 15

251 11. Thorp HH. ChatGPT is fun, but not an author. Science. 2023 Jan 27;379(6630):313. doi: 10.1126/science.adg7879. Epub 2023

252 Jan 26.

253

254

255

256

257

258

259

260

261

262

263

264

265

266
 16

267 Legend

268 Figure 1: The study's PRISMA flow diagram shows how articles are selected for this systematic review.

269

270

271

272

273

274

275

276

277

278

279

280

281

282
 17

283 Table-1: Summaries of full text of 118 selected article. Summaries were created with the help of ChatGPT. PROMPT, that
284 given was “Provide point wise summary” .
285
Reference Country Type of Status of Title of the paper Main findings of the paper
publication peer
review
Ali and Djalilian India Editorial Peer Readership awareness The question of whether a chatbot like ChatGPT can be regarded as an author
2023 reviewed series–paper 4: gives rise to ethical dilemmas concerning accountability for the content produced
Chatbots and ChatGPT- and the accuracy of source attributions.
ethical considerations in
scientific publications
Ali et al 2023 United Commentary Peer Using ChatGPT to write CHATBOTS, such as ChatGPT, were employed by researchers to generate a
Kingdom reviewed patient clinic letters variety of clinical communication scenarios that were designed to be
comprehensible for individuals at or below a sixth-grade level (approximately 11-12
years of age).
Alser and Egypt Commentary Peer Concerns with the ChatGPT, a language model, was credited authorship in four medical papers. It
Waisberg 2023 reviewed Usage of ChatGPT in met one of four ICMJE authorship criteria by contributing to writing parts of the
Academia and papers. Plagiarism was found in its contributions. Unclear sources of information
Medicine: A and bias due to tuning also limit its use in scientific contexts.
Viewpoint
Anderson et al Germany and Editorial Peer AI did not write this The ability of AI to produce academic papers has been tested and found wanting.
2003 other reviewed manuscript, or did The papers produced by AI lacked fresh ideas, in-depth knowledge of the subject,
European it? Can we trick the AI and precise referencing. There were problems with equality, accuracy, poor AI
countries text detector into detection, unethical content creation, and plagiarism. To maintain the quality of
generated texts? The scientific literature, the application of AI tools needs to be carefully studied. AI
potential future of misuse can be avoided with the aid of current plagiarism-detection tools and
ChatGPT and AI in possible cutting-edge software for AI-generated text detection.
Sports & Exercise
Medicine manuscript
generation
Arun Babu and India Letter to the Peer Using artificial The use of AI-generated content in medical journals raises ethical concerns
Sharmila 2023 editor reviewed intelligence chatbots regarding accuracy, bias, authorship, and disclosure, necessitating new guidelines
like 'ChatGPT' to draft and standards for publication integrity.
articles for medical
journals - Advantages,
limitations, ethical
concerns and way
forward
Asch USA Commentary Peer An Interview with ChatGPT has potential in healthcare, assisting with tasks like virtual assistance,
2023 Interview reviewed ChatGPT About Health documentation, research, education, and patient engagement. Concerns such as
and chat Care privacy, bias, regulation, and ethics must be addressed. It can both improve and
with reduce equity. Advancements should focus on data quality, NLP, integration,
 18

ChatGPT security, explainability, and ethics. It enhances, not replaces, healthcare

professionals.
Athaluri et al 2023 India Original Peer Exploring the This study investigated AI hallucination in research proposals drafted by ChatGPT.
reviewed Boundaries of Reality: Analysis of 178 references revealed limitations in generating reliable references,
Investigating the including the absence of DOIs and inaccessible articles. The findings emphasize
Phenomenon of the need for improvements in training and caution when relying solely on
Artificial ChatGPT's generated references.
Intelligence
Hallucination in
Scientific Writing
Through ChatGPT
References
Balas and Ing Canada Original Peer Conversational AI When it came to accurately recognising the diagnosis in 9 out of 10 cases and
2023 reviewed Models for ophthalmic including it in all differential diagnosis lists, ChatGPT performed better than Isabel
diagnosis: Comparison Pro at diagnosing ocular illnesses. ChatGPT has potential for primary care
of ChatGPT physicians.
and the Isabel Pro
Differential Diagnosis
Generator
Barker and Rutka Editor-in- Editorial Peer Generative artificial n a comparative study, ChatGPT answered 60.2% of neurosurgical questions
2023 Chief, reviewed intelligence, chatbots, correctly, while average human users scored 69%. However, caution is necessary
Journal of and the Journal of as chatbot outputs may contain biases, inaccuracies, and even fabrications.
Neurosurgery Neurosurgery Trusting chatbot-generated text without verification can lead to misleading
Publishing Group information.
Bauchner 2023 former editor- Original Peer ChatGPT: Not An The ICMJE's requirements for authorship are not currently met by ChatGPT.
in-chief reviewed Author, But A Tool However, it might be utilised by researchers as a tool to design early publications,
of JAMA possibly complying to reporting standards and offering a more objective viewpoint.
USA Additionally encouraging diversity, it might help non-native English speakers
prepare papers for publication. ChatGPT can be recognised in articles along with
other pertinent information as a method to improve scientific communication.
Baumgartner Austria Commentary Peer The potential impact of The utilization of ChatGPT in clinical and translational medicine has the potential to
2023 reviewed ChatGPT in clinical and bring about a transformative impact, enhancing patient engagement, reducing
translational medicine healthcare providers' workload, and delivering current information. Nonetheless, it
is imperative to confront challenges such as data privacy risks, limitations in
accuracy, and potential bias originating from training data. Ongoing research and
development efforts are crucial to ensure the safe and effective implementation of
ChatGPT within the healthcare domain.
Benoit Canada Original Preprint ChatGPT for Clinical This study examined the capabilities of ChatGPT in generating, rewriting, and
2023 Vignette Generation, evaluating clinical vignettes. ChatGPT demonstrated the ability to follow
Revision, and instructions, generate varied contexts, and match patient demographics. However,
Evaluation there were limitations and caveats, requiring monitoring and review. The findings
highlight the potential of ChatGPT with appropriate oversight.
Bhattacharya et al India Review Peer ChatGPT in Surgical ChatGPT in anaesthesia and surgical care has potential for clinical decision-
 19

2023 reviewed Practice—a New Kid on making, preoperative education, medical record transcription, surgical planning,
the Block and medical education. However, limitations include potential misinformation, bias,
reliance on outdated data, copyright concerns, technical issues, and lack of
personalization. Proper evaluation, oversight, and addressing legal and ethical
considerations are necessary for optimal use.
Biswas 2023 USA Commentary Peer ChatGPT and the future Article introduces AI and ChatGPT.
reviewed of medical writing
Boßelmann et al USA Commentary Peer Are AI language models Artificial Intelligence (AI) can significantly improve epilepsy diagnosis and
2023 reviewed such as ChatGPT ready treatment. Examples of epilepsy-related AI research include EEG data analysis,
to improve the care of MRI-based lesion detection, and Clinical Decision Support Systems. While AI
individuals with language models like ChatGPT show promise in improving patient care, they
epilepsy? require rigorous testing, validation, and ethical consideration, along with the
expertise of medical professionals.
Brainard 2023 USA Commentary Peer Journals take up arms Publishing directors and journal managers are concerned about the impact of AI-
reviewed against AI-written text powered chatbots like ChatGPT on scholarly literature. These tools can produce
accurate-sounding, AI-generated reports and scientific manuscripts, raising issues
of accuracy, authorship, and potential flooding of AI-produced manuscripts.
Publishers are formulating policies requiring disclosure of AI tool use and are
exploring technology to detect synthetic text.
Cahan and Switzerland Editorial Peer A conversation with Computational biology and systems biology have made significant contributions to
Treutlein 2023 reviewed ChatGPT on the role of stem cell research by enabling the analysis of large datasets, identifying patterns
computational systems and trends, optimizing stem cell culture conditions, and creating detailed models of
biology in stem cell stem cell behavior. Challenges include the need for extensive data collection and
research processing, but advancements in data standardization and accessibility have
facilitated data reuse and meta-analyses. Integrating computational techniques with
stem cell research holds great promise for advancing our understanding and
applications of stem cells.
Nasrallah 2023 Editor-in- Editorial Peer A ‘guest editorial’ … The article discusses the introduction of ChatGPT, an AI program with potential
Chief Guest reviewed generated by societal impact. It highlights concerns about accuracy and references, tests
Current editorial ChatGPT? ChatGPT's abilities, and reflects on the era of AI-generated articles.
Psychiatry generated
by ChatGPT
Cascella et al Italy Review Peer Evaluating the feasibility This study examines ChatGPT's possible uses in clinical and research settings,
2023 reviewed of ChatGPT in including supporting clinical practise, doing scientific work, preventing potential
healthcare: an analysis abuse, and debating public health issues. The study highlights ChatGPT's
of multiple clinical and shortcomings and emphasises the need for careful use and further development to
research scenarios avoid dangers, such as disinformation and misuse, even if it shows promise in a
number of areas.
Chen et al 2023 USA Original Preprint The utility of ChatGPT This study found that while ChatGPT correctly aligned with National
for cancer treatment Comprehensive Cancer Network (NCCN) guidelines in 98% of cases, it also
information provided partially non-concordant advice in 34.3% of them. Therefore, it concluded
that ChatGPT is not entirely reliable for cancer treatment advice, urging patients
and clinicians to recognize its limitations.
 20

Cheng et al 2023 China Letter to the Peer- ChatGPT/GPT-4: The potential application of ChatGPT/GPT-4 in surgical oncology, from aiding in
editor reviewed Enabling a new era of clinical trial design, case management, data analysis, to preoperative preparations
surgical oncology and intraoperative processes, is discussed in this study. Despite potential
challenges and ethical concerns, the authors advocate for responsibly harnessing
this AI technology to improve surgical oncology outcomes.
Chervenak et al USA Original Peer- The promise and peril ChatGPT, tested against established fertility sources, provided responses
2023 reviewed of using a large comparable in quality to the CDC's FAQs and scored highly on validated fertility
language model to knowledge surveys. Although it effectively reproduced missing facts from the
obtain clinical American Society for Reproductive Medicine's committee opinion, its clinical utility
information: might be limited by occasional incorrect information and lack of source citation.
ChatGPT performs
strongly as a fertility
counseling tool with
limitations
Cifarelli and USA Commentary Peer- Large language model However, ethical considerations arise when utilizing ChatGPT in scholarly
Sheehan 2023 reviewed artificial intelligence: the publishing. These encompass potential bias in training data and the perpetuation of
current state and future prejudice, concerns about authorship, copyright, and plagiarism, implications for
of ChatGPT in neuro- citation practices and the "Matthew Effect," and impacts on academic job
oncology publishing expectations, tenure, and promotion.

Addressing ownership of generated content and adherence to copyright laws is

crucial, necessitating careful source attribution. Plagiarism is a concern, but proper
citation practices can help mitigate it.

ChatGPT can streamline citation processes and assist researchers in identifying

and formatting citations accurately. However, relying solely on automated tools
may overlook important literature and raise questions about the value of human
expertise.

Academic institutions and publishers can implement measures such as anti-

ChatGPT software, encourage creative and innovative research, and reconsider
tenure evaluation criteria to tackle the challenges posed by ChatGPT.

Thoughtful consideration of ethical concerns is essential to ensure responsible and

ethical use of ChatGPT in academia and scholarly publishing. Further research is
warranted to explore its implications and limitations.
Corsello and Italy Review Peer- May Artificial This interview with ChatGPT explores the transformative potential of AI in pediatric
Santangelo 2023 reviewed Intelligence Influence research, highlighting advantages such as improved clinical decision-making,
Future Pediatric enhanced education, faster drug development, and better research outcomes. It
Research?—The Case also probes into potential challenges like bias, safety issues, overreliance on
of ChatGPT technology, and ethical concerns. It underscores the importance of careful
consideration of these technologies' implications and usage, ensuring they are
responsibly leveraged for beneficial outcomes.
 21

D'Amico et al USA Editorial Peer- I asked a ChatGPT to This editorial discusses the potential and challenges of incorporating AI chatbots in
2023 reviewed write an editorial about neurosurgery for data collection, patient care, and communication. However, issues
how we can incorporate like potential misinformation, privacy, ethical concerns, bias, legal liabilities, content
chatbots into validity, and effectiveness need addressing. It emphasizes human responsibility in
neurosurgical research verifying machine-generated content for moral and ethical standards. Neurosurgery
and patient care… should lead in responsibly integrating these AI technologies.
Darkhabani et al Turkey Review Peer- ChatGPT and This paper explores the role of AI language model ChatGPT in clinical medicine,
2023 reviewed autoimmunity - A new specifically autoimmunity, discussing its capabilities, limitations, and potential cyber
weapon in the risks. It highlights the importance of continuous evaluation, as ChatGPT and similar
battlefield of knowledge technologies evolve rapidly, emphasizing the need for healthcare professionals to
stay abreast of these developments.

Dave 2023 UK Commentary Peer- Plagiarism software Turnitin, the academic integrity firm used by 98% of UK universities, can now
reviewed now able to detect detect plagiarism in content generated by artificial intelligence systems like
students using ChatGPT. This development aims to maintain academic integrity in the face of
ChatGPT advancements in AI technology, preventing dishonest use of AI-generated work.
Dave et al 2023 UK Review Peer- ChatGPT in medicine: ChatGPT can be a valuable tool for medical research and patient care, providing
reviewed an overview of its support in data analysis and diagnosis generation. However, it should supplement,
applications, not replace, human expertise. Ethical considerations, like data privacy and
advantages, limitations, accuracy of recommendations, need to be addressed. Future prospects include
future prospects, and integration with other AI technologies, prioritizing a balanced, cautious, and
ethical considerations ethically mindful approach.
Day et al 2023 Canada Original Peer- A Preliminary ChatGPT provide citations as well although recent scrutiny revealed that these
article reviewed Investigation of Fake references are often incorrect or non-existent, according to a study. Despite the
Peer-Reviewed impressive capabilities of the chatbot, the presence of fake references and citations
Citations and is a concern for both academic research and student academic integrity. While
References Generated these inaccuracies present a problem for those who might use the bot for sourcing,
by ChatGPT they could serve as a useful tool for identifying academic misconduct or as prompts
for further research. Despite these issues, the chatbot has been utilized effectively
for generating course materials in certain subject areas, suggesting potential for its
continued use in education. However, the tool requires subject expertise and its
use should be approached with caution.
de Oliveira and Brazil Editorial Peer- The future of Pediatric The use of AI like ChatGPT in medical writing is revolutionary, but has ethical
Ballestero 2023 reviewed Neurosurgery and implications and potential for errors. AI's usage should be transparent, and authors
ChatGPT: an editor's should ensure their publications' scientific integrity.
perspective
De Vito 2023 Argentina Editorial Peer- Artificial intelligence and This editorial discusses the implications of using AI, particularly OpenAI's GPT-3, in
reviewed chatGPT. Would you scientific and academic writing. While acknowledging the benefits of AI tools in
read an artificial author? producing coherent text, it raises concerns about authenticity, potential misuse, and
the need for transparency. It also stresses the importance of human responsibility
in ensuring the accuracy of scientific content and the ethical use of AI technology in
academia.
Dergaa et al 2023 Multiple Original Peer- From human writing to This study explores the benefits and potential threats of NLP technologies like
 22

countries article reviewed artificial intelligence ChatGPT in academic writing and research. While these tools can enhance
generated text: efficiency, concerns arise about the authenticity and credibility of work. The study
examining the underscores the need for ethical considerations, transparency, and human
prospects and potential intelligence in their use to maintain academic integrity.
threats of ChatGPT
in academic writing
Donato et al 2023 Portugal Editorial Peer The Transparency of This article discusses ethical implications of large language models (LLMs) like
reviewed Science with ChatGPT ChatGPT in scientific publishing. It highlights concerns about authorship,
and the Emerging transparency, and integrity in utilizing AI tools. The need to declare AI's role and
Artificial Intelligence limit its misuse is emphasized. It also discusses future prospects of detecting AI-
Language Models: generated text and potential applications of LLMs in non-English language
Where Should Medical contexts.
Journals Stand?

Dunn et al 2023 USA Original Peer Artificial intelligence- This letter discusses a study comparing the quality and readability of dermatology-
article reviewed derived dermatology based case reports produced by human authors and AI models like ChatGPT.
case reports are Findings suggest that AI-generated reports were often indistinguishable from
indistinguishable from human-created ones, though some limitations were noted. AI detection tools varied
those written by in effectiveness, raising ethical implications about the reliability, accuracy, and
humans: A single- transparency of AI-generated content. The authors argue for the need to maintain
blinded observer study integrity of AI-produced content, and for editorial teams to consider policies around
the use of AI and Large Language Models.
Fatani 2023 Saudi Arabia Review Peer ChatGPT for Future This narrative review discusses the application and implications of AI language
reviewed Medical and Dental model, ChatGPT, in medical and dental research. ChatGPT can assist in writing
Research scientific papers, summarizing data, and translating languages. While AI can
enhance clinical workflow and help generate quick responses, it's cautioned that
over-reliance could lead to papers lacking critical thinking. Ethical concerns such as
plagiarism and data integrity are also highlighted. The review concludes that while
ChatGPT can support research and potentially transform clinical medicine, there
are limitations and potential risks that need to be carefully managed.
Galland 2023 France Editorial Peer Chatbots and internal In November 2022, OpenAI® made headlines by releasing version 3.5 of its AI
reviewed medicine: Future ChatGPT, a revolutionary conversational AI capable of generating text on any
opportunities and topic, including in the medical field. It was joined in 2023 by other advanced
challenges generative AIs such as Google's BardAI and image generators like Dall-E 2 and
Mid Journey. The use cases in healthcare are increasing, and chatbots could be
valuable tools for prevention, appointment scheduling, symptom information,
administrative tasks, and more. However, concerns about data privacy, user safety,
and accuracy of information arise with the use of chatbots in healthcare. The
adoption of these technologies by patients may outpace the establishment of
scientific consensus on their use. The article emphasizes the need for caution,
evaluation, and defining the indications and limitations of their utilization in the
medical field.
Gandhi India Letter to the Peer ChatGPT: roles and The article discusses the advancements in AI, particularly ChatGPT, an advanced
 23

Periaysamy et al editor reviewed boundaries of the new language model developed by OpenAI. It highlights its potential in medical
2023 artificial intelligence tool education and health research. ChatGPT can provide answers and information for
in medical education medical students, assist researchers in writing and reviewing content, and even
and health research - generate scientific abstracts. However, it also raises concerns about potential
correspondence misuse, integrity of generated content, and ethical considerations regarding
authorship in research papers involving AI. Proper guidelines and reporting
standards are recommended to ensure responsible and transparent use of AI tools
in the medical field.
Gao et al 2023 USA Original Peer- Comparing scientific The article assesses ChatGPT's capacity to produce scientific abstracts, a sizable
article reviewed abstracts generated by language model. Although neither was flawless, both human reviewers and AI
ChatGPT to original output detectors were able to recognise some created abstractions. The created
abstracts using an abstracts were in the same range as actual abstracts but contained made-up
artificial intelligence numbers. The study is aware of its limitations, including the sample size and the
output detector, type of cues employed. The technology has the potential to be used in both ethical
plagiarism detector, and and immoral ways, such as to fabricate research or aid in research writing. The
blinded human biases in training data are underlined, as is the requirement for more investigation
reviewers into potential biases.
Goedde et al 2023 Germany Original Preprint ChatGPT in medical This concise review focuses on ChatGPT (Chat Generative Pre-trained
article literature-a concise Transformer) and its role in medical literature. The study analyzed the literature
review and SWOT published on ChatGPT from December 2022 to March 2023 and conducted a
analysis SWOT analysis. The findings highlight the strengths, weaknesses, opportunities,
and threats associated with ChatGPT. The review emphasizes the need for further
research, regulations, and policies to address the potential of ChatGPT in medical
literature.
Gordijn and Have Ireland Editorial Peer- ChatGPT: evolution or This editorial explores the implications of large language models (LLMs),
2023 reviewed revolution? specifically ChatGPT by OpenAI, on academic publishing and research. It
questions whether the development of LLMs will necessitate new editorial policies
to address the challenges they pose. Viewing ChatGPT as either evolutionary or
revolutionary, the article suggests the utility of the AI model as a research tool may
be limited due to deficiencies. It acknowledges the potential for LLMs to improve
research but also highlights potential risks such as plagiarism and factual
inaccuracies. Current editorial policies may already handle these issues, but further
review and adjustments could be necessary.
Gottlieb et al 2023 USA Commentary Peer- ChatGPT and AI and machine learning are significantly impacting healthcare and research,
reviewed conversational artificial enhancing precision medicine through big data analysis and personalized
intelligence: Friend, foe, treatment plans.
or future of research? Large Language Models (LLMs) like ChatGPT serve as conversational interfaces,
aiding in patient communication and assisting researchers in content generation.
Advantages include accelerating content creation, language translation, simplifying
complex topics, synthesizing literature, creating abstracts, identifying limitations,
and disseminating knowledge.
Potential harms encompass limitations in the extant knowledge of LLMs, potential
misinterpretation of study methods/results, overconfidence in AI-generated content,
 24

risk of plagiarism, inaccurate references, and biases.

Current trends indicate the use of AI for language editing and translation of
academic papers, increasing global access.
Controversy persists regarding AI's authorship status in academic research; some
argue that AI fails to meet authorship criteria as it can't be accountable for the work.
A balance must be struck, leveraging AI benefits while ensuring ethical standards,
accuracy, and acknowledging inherent limitations.
Graf and Bernardi Germany Editorial Peer- ChatGPT in Research:
2023 reviewed Balancing Ethics, ChatGPT can generate realistic text, propose research questions, design studies,
Transparency and analyze data, write and edit documents, among other tasks, offering valuable
Advancement starting points or improvements for researchers' work.

The ethics and appropriateness of listing ChatGPT as a co-author in research

papers has stirred debate in the scientific community.

Some argue that if ChatGPT contributes to the planning, execution or manuscript

writing/editing of a study, it should be credited as a co-author.

However, editors-in-chief from major scientific journals disagree, arguing that

authorship requires the capacity for accountability and consent, which ChatGPT, as
an AI, lacks.

ChatGPT agrees with this sentiment, stating it doesn't possess the ability to
consent to co-authorship, but can assist with writing and editing.

Despite these discussions, the AI's contributions are currently undetected by

plagiarism checker software, raising concerns about originality and authorship in
academic work.

A recent study showed that ChatGPT-generated abstracts were often

indistinguishable from "original work," both by plagiarism checkers and human
academic reviewers.

As AI continues to evolve rapidly, it poses significant questions for publishers,

researchers and educators about how to best utilize, regulate and attribute its
capabilities
Graham 2023 UK Commentary Peer- ChatGPT and other AI As advanced AI technology becomes more accessible, the MDU (Medical Defence
reviewed tools put students at Union) is emphasizing the importance of students ensuring that their academic
risk of plagiarism work remains "above suspicion."
allegations, MDU warns
Ellie Mein, a medicolegal adviser at MDU, cautions against taking shortcuts to meet
deadlines and advises struggling students to seek support or extensions instead.
 25

Plagiarism allegations can lead to fitness to practice investigations according to the

guidelines set by the General Medical Council (GMC), which assesses whether
students meet the required competency standards.

Universities have the authority to impose conditions, suspension, or expulsion on

students found to be involved in plagiarism, even if it was unintentional.

When applying for full registration, students are obligated to disclose any fitness to
practice investigations to the GMC.

Mein highlights that some students have been taken aback when contacted by the
GMC about undisclosed investigations.

To help students avoid plagiarism in the "age of AI," the MDU has provided tips that
address common errors such as "para-plagiarizing" and insufficient citation.
Gravel et al 2023 Canada Original Preprint Learning to fake it: The objective of this study was to assess the quality and suitability of responses
article limited responses and and references provided by ChatGPT when addressing medical questions. The
fabricated references study selected 20 diverse medical questions derived from recent research articles
provided by ChatGPT published in high-impact factor medical journals. These questions were related to
for medical questions the main objectives of the articles or framed within a broader context to ensure a
comprehensive range of references.

The study followed a specific methodology where the questions were posed to
ChatGPT without imposing any word limit or constraint. After receiving the
response, a follow-up query was made to request references, with only the first
three references being considered for analysis. To ensure domain expertise, the
corresponding authors of the selected articles were invited to rate the responses.
The primary outcomes measured in the study were the appropriateness of the
references and the quality of the responses.

The findings of the study revealed that the majority of references provided by
ChatGPT were fabricated. These references appeared credible, featuring authors
with previous publications or affiliations with reputable organizations. However,
69% of the references were found to be nonexistent. Among the 59 references
included in the analysis, only 18 were authentic, although they contained minor or
major citation errors. The remaining references were completely fictional.
Additionally, the study reported that 95% of the references listed authors with prior
publications or had affiliations with recognized organizations.

Among the participating corresponding authors, seventeen agreed to evaluate the

responses, assigning a median score of 60% for the quality of the answers.
Notably, five responses were identified by the raters to contain major factual errors.
 26

To summarize, the study revealed that the references provided by ChatGPT were
predominantly fabricated, and the quality of the responses varied, including
instances of major factual errors. These findings raise concerns regarding the
reliability and accuracy of the references and responses generated by ChatGPT
when addressing medical questions.

Guo et al 2023 Canada Original Preprint neuroGPT-X: Towards This study enhanced OpenAI's GPT-3 model for answering questions about
article an Accountable Expert vestibular schwannoma. The context-enriched GPT model provided faster and non-
Opinion Tool for inferior responses compared to expert neurosurgeons. The model's responses
Vestibular were often rated higher and included in-text citations and references. However,
Schwannoma expert surgeons expressed concerns about the model's reliability in addressing
nuanced and controversial aspects of management. The study also introduced
neuroGPT-X, a chat-based platform for clinical support with accurate and reliable
information.
Gurha et al 2003 USA Commentary Peer- ChatGPT and other An AI chatbot or AI writer should not be listed as an author in scientific publications
reviewed artificial intelligence because authors provide nuanced interpretation and insight beyond synthesizing
chatbots and sentences.
biomedical writing Authors cannot claim ownership of text generated by chatbots and should avoid
plagiarism by acknowledging their use.
Chatbots are limited in generating in-depth and critical scientific writings and lack
the essential insight required for review articles, commentaries, or perspectives.
Large language models (LLMs) have potential in scientific writing but should be
carefully evaluated for validity and edited for incorrect statements.
Chatbots have various biomedical and clinical applications but require expert
review and verification.
LLM-based programs face challenges in extracting reliable data from biased and
low-quality sources.
As programs advance, commercialization may occur, but scientists should maintain
access to platforms for synthesizing and extracting advanced information.
JCA editors prioritize articles with nuanced insight and new understanding.
Hurley 2023 USA Commentary Peer- Your AI Program Will ChatGPT, the language model developed by OpenAI, is creating ripples in the
reviewed Write Your Paper Now: academic world, with several journals considering guidelines for its usage. Experts
Neurology Editors on note the AI's remarkable writing ability and its potential to generate data and
Managing Artificial references, which raises concerns about authenticity. While some editors suggest
Intelligence there could be a place for ChatGPT to assist with language barriers and manuscript
Submissions restructuring, there's consensus that the AI should not be credited as an author,
and its usage should be disclosed. Additionally, experts emphasize the need for
human oversight and responsibility for content generated with AI's help.
Haemmerli et al Switzerland Original Preprint ChatGPT in glioma In this study, the performance of ChatGPT, an AI-based chatbot, was assessed in
2023 article patient adjuvant therapy providing treatment recommendations for glioma patients within a multidisciplinary
decision making: ready decision-making context. ChatGPT demonstrated accuracy in identifying cases as
to assume the role of a gliomas; however, it faced challenges when specifying tumor subtypes. The
doctor in the tumour treatment recommendations and regimens generated by the tool were generally
 27

board? rated as good, while its ability to consider functional status received a moderate
rating. Experts found ChatGPT's involvement in the CNS Tumor Board to be
valuable, recognizing its potential for improvement and learning.
Harskamp and Netherlands Original Preprint Performance of The AMSTELHEART-2 study evaluated the performance of the ChatGPT language
Clercq 2023 article ChatGPT as an AI- model in answering cardiovascular trivia questions and interpreting case vignettes.
assisted decision The model achieved an overall accuracy of 74% in answering multiple-choice
support tool in questions and provided appropriate advice for patients reaching out to primary
medicine: a proof-of- care. However, it had limitations in addressing medical questions from physicians
concept study for seeking expert consultation. The study's strengths include using various
interpreting symptoms approaches to evaluate ChatGPT, while its limitations include a small sample size.
and management of Further refinements of the model are needed before considering its use in real-life
common cardiac scenarios.
conditions
(AMSTELHEART-2)
Hill-Yardin et al Australia Viewpoint- Peer- A Chat(GPT) about the The use of AI-generated software, such as ChatGPT, in academic publishing has
2023 Commentary reviewed future of scientific sparked discussions and debates.
publishing The generated text can be accurate, logical, and grammatically correct, but it lacks
a distinct "voice" and is generic.
The writing style of the generated text may resemble the bland and formulaic style
often found in scientific articles.
Introducing more diversity in writing styles and discussing mistakes and
unsupported hypotheses could help differentiate human writing from AI-generated
writing.
AI may eventually learn individual writing styles and inject more nuanced diction.
The use of AI and predictive text software raises questions about language
patterns, vocabulary, and linguistic diversity.
ChatGPT's generated text is too shallow for the detailed knowledge and
interpretation required in neuroscience.
The lack of accountability and transparency in AI-generated text is a fundamental
concern for its use in academic publishing.

Hirani et al 2023 USA Commentary Peer- Experimenting with The functionality of ChatGPT in academic medical publishing was examined,
reviewed ChatGPT: Concerns for revealing the potential for fabrication of inaccurate research.
Academic Medicine The lack of transparency and accountability in ChatGPT's output raised concerns
and prompted guidelines from publishing companies and journal editors.
Bias is a significant problem as well because word embeddings, which are used in
ChatGPT training, introduce biases originating from humans into the original data.
Algorithmic bias in medical AI systems, which also leads to care disparities, may
have a negative impact on clinical decision-making.An analysis uncovered biases
in ChatGPT's responses, such as professions and jobs that are linked with
particular genders. Physicians should advocate for transparency, regulation, and
accountability in LLM training datasets and algorithms to address the issues of
fabrication and bias.
 28

Homolak Croatia Commentary Peer- Opportunities and risks The release of ChatGPT has generated excitement about the potential of AI in
2023 reviewed of ChatGPT in various fields, including academic publishing.
medicine, science, and AI has the potential to revolutionize healthcare but is unlikely to replace physicians
academic publishing: a due to limitations in context and nuance.
modern Promethean Ethical issues surround conversational AI in medical practice, including biased data
dilemma sets and accountability for mistakes.
ChatGPT's ability to generate scientific manuscripts raises concerns about
reliability and the potential for misinformation.
Acknowledging limitations and addressing ethical challenges are necessary before
implementing AI in publishing.
While AI has the potential to improve efficiency, a mindful approach and open
debate about risks and benefits are essential.
Hosseini et al USA Original Preprint ChatGPT versus the A hybrid panel discussion about the use of ChatGPT in education, research, and
2023 article neurosurgical written healthcare was conducted.
boards: a comparative 420 responses were received from 844 participants, with a response rate of 49.7%.
analysis of artificial 40% of the audience had used ChatGPT, with more trainees than faculty.
intelligence/machine Interest in wider application was higher among ChatGPT users.
learning performance Greatest uncertainty was around its use in education.
on neurosurgical Varied perspectives were present among different roles (trainee, faculty, staff).
board–style questions The need for further discussion about LLM usage was highlighted.
A thoughtful, measured approach in adopting ChatGPT was suggested to reduce
potential risks.
Howard et al 2023 UK Commentary Peer ChatGPT and ChatGPT has been noted for its potential in medical scenarios, stimulating urgent
reviewed antimicrobial advice: the discussions in the medical community.
end of the consulting Despite its limitations, it's considered as capable as an average human physician in
infection doctor? answering some open-ended medical queries.
The study tested ChatGPT in eight hypothetical infection scenarios.
It recognized natural language effectively but struggled with complex situational
aspects.
Its responses were coherent, clear, and reflective of its information sources.
Some inconsistencies and erroneous advice emerged on repeated questioning.
Deficits in situational awareness, inference, and consistency are major barriers to
ChatGPT's clinical implementation.
Despite no access to specific medical databases, it provides compelling responses.
The authors propose a modifiable qualitative framework for future improvements.
Hsu et al 2023 Taiwan Original Preprint Plagiarism, Quality, and This study aimed to evaluate the quality and accuracy of medical abstracts
article Correctness of generated by ChatGPT, specifically the ChatPDF version, using a selection of 20
ChatGPT-Generated vs randomised controlled trials in the field of psychiatry. Two types of abstracts,
Human-Written Abstract structured and unstructured, were considered.
for Research Paper
The generated abstracts were analyzed in terms of similarity, plagiarism, AI-
content, and subheading proportion. They were also evaluated by five experts in
psychiatry for quality, using a Likert scale. Further, the validity of the conclusions
 29

generated by ChatPDF was checked.

The findings showed a similarity (duplicate content) between the original and the
generated abstracts of 16.35%. The plagiarism percentage was 18.75%, with
11.20% being self-plagiarism. After adjusting for self-plagiarism, the actual
plagiarism percentage was 7.55%. The AI-content percentage was notably higher
in unstructured abstracts (75.58%) compared to structured ones (31.48%).

The quality of generated abstracts was found to be lower than the original
abstracts, especially for the unstructured format. Interestingly, a structured format
and a higher H-index were associated with higher quality scores in generated and
original abstracts, respectively.

Lastly, while the ability of experts to accurately identify original authors was fairly
low (40%) for structured abstracts, it was significantly higher (73%) for unstructured
ones. However, a noteworthy concern was that 30% of the abstracts generated by
ChatGPT presented incorrect conclusions.

These findings suggest that while the performance of ChatGPT in generating

medical abstracts has potential, improvements are necessary, particularly in terms
of the quality of unstructured abstracts and the accuracy of conclusions.
Huang et al 2023 China Review Peer- The role of ChatGPT in Topic Selection: ChatGPT generates keywords and suggests research areas,
Article reviewed scientific helping scientists choose relevant topics.
communication: writing Literature Search: The AI provides relevant search queries and database
better scientific review suggestions for efficient literature searches.
articles Article Selection: ChatGPT generates summaries of articles, providing context for
relevance in a review.
Citation and Referencing: It aids in accurate citation and referencing, generating
appropriate formats.
Efficiency: By automating tedious tasks, ChatGPT facilitates comprehensive,
efficient reviews, enhancing manuscript quality.
Time-saving: The AI tool streamlines the writing process, saving time and effort.
Caution Required: Despite its benefits, scientists should exercise care, reviewing
and editing generated text to avoid plagiarism and fabrication. Outline
Development: Scientists input topic, ChatGPT generates and organizes related
subtopics into a logical outline.
Adding Details: ChatGPT suggests key points and relevant literature to add depth
to the outline.
Text Improvement: Scientists input text, ChatGPT analyzes and provides
suggestions for improvements in grammar, sentence structure, and style.
Suggested Improvements: ChatGPT offers advice to enhance clarity, precision and
avoid jargon.
Example Provision: The AI tool provides examples of well-written scientific articles
 30

or sentences to illustrate the improvements.

Feedback Incorporation: Scientists can apply ChatGPT's suggestions and adjust as
necessary for tone and style.
Enhancing Quality: By adopting ChatGPT's suggestions, scientists can improve
their writing's clarity, precision, and effectiveness, leading to higher quality
manuscripts. Assisting Non-Native English Speakers: ChatGPT aids with correct
grammar, sentence structure, and appropriate vocabulary, even offering translation
help.
Drawbacks: AI may lack context, introduce bias, encourage over-reliance, struggle
with technical terms, and require substantial investment.
Plagiarism Risk: AI-generated text can resemble other sources; to minimize
plagiarism risk, use AI as supplement, properly attribute sources, and use
plagiarism-detection tools.
Avoiding Plagiarism: Understand text sources, use multiple references, ensure
accurate citation, employ plagiarism-detection software, and carefully review and
edit AI-generated text.
Summary: While beneficial, ChatGPT use necessitates vigilance against plagiarism
and adherence to proper citation and editing practices. Contextual Understanding:
ChatGPT lacks contextual understanding; humans ensure generated content is
appropriate for its purpose.
Checking Accuracy: ChatGPT can produce inaccuracies; human oversight verifies
content accuracy.
Editing and Formatting: AI-generated text needs editing and formatting to meet
article requirements; human oversight ensures compliance with these standards.
Conclusion: ChatGPT significantly enhances scientific writing but requires human
oversight to prevent plagiarism and verify accuracy.
Future of AI: Advanced AI tools will further streamline scientific research
processes, enabling impactful research output.
Janssen et al The Commentary Peer- The use of ChatGPT Surgical scientists conducting clinical trials and research on artificial intelligence in
2023 Netherlands reviewed and other large surgery are currently investigating the potential impact of Large Language Models
language models in (LLMs), such as ChatGPT, which is set to be published on November 30, 2022.
surgical science
From the perspective of surgeon-scientists and editors, LLMs have the potential to
enhance productivity and efficiency in writing tasks. They can be employed for data
extraction and clinical decision-making purposes as well.

LLMs can assist in generating drafts for study ideas, research protocols,
manuscripts, grant proposals, instructional materials, and patient education
materials. They can also help improve text and rectify errors for non-native
speakers.

Integration of LLMs into electronic health record systems could enable information
extraction from patient records and other data repositories, automating data
 31

collection for research purposes. This could potentially alleviate challenges

associated with manual data extraction.

Advanced LLMs could find application in clinical workflows for evaluating patient
information and generating patient management plans, leading to more efficient
patient care and improved outcomes.

However, there are limitations and risks associated with using LLMs. These include
the possibility of "neural hallucinations," where the model generates incorrect or
nonsensical text, and the potential introduction of bias into the model's output.

Establishing precise criteria and guidelines for LLM usage is crucial to ensure their
trustworthiness and security in surgical science, particularly when considering the
ethical and legal implications of their use in patient management.

Despite these limitations, LLMs have the potential to be valuable tools in surgical
science and clinical practice by augmenting human expertise rather than replacing
it.

The article emphasizes the importance of ongoing monitoring of developments in

this field and evaluating the impact of language models on surgical science.
Johnson et al USA Original Peer- Using ChatGPT to In this study, the accuracy of cancer information provided by ChatGPT, an AI
2023 article reviewed evaluate cancer myths chatbot, was compared to the responses from the National Cancer Institute (NCI)
and misconceptions: regarding common cancer myths and misconceptions.
artificial intelligence and
cancer information Thirteen questions related to cancer were posed to ChatGPT, and the precision of
its answers was assessed. Five scientists, who had experience in cancer treatment
and false information, reviewed both the NCI's responses and ChatGPT's
responses independently and blinded. The results showed that the NCI's answers
received a 100% accuracy rating from all five reviewers for all 13 questions,
indicating full interrater agreement. ChatGPT's answers received a 96.9% accuracy
rating from all five reviewers for 11 out of the 13 questions.

There were minimal differences observed in word count and readability grade level
between the responses from NCI and ChatGPT. The study suggests that ChatGPT
provides accurate information on common cancer myths and misconceptions
without disseminating misinformation or harmful information.

The researchers also conducted additional rounds of questioning to assess

whether repetitive questioning could lead to semantic variations and potential
misinformation. The answers remained consistent, indicating that the accuracy of
the responses was not affected by repeated questioning.
 32

The study underscores the positive potential of ChatGPT and AI systems in

delivering accurate cancer-related information. However, it highlights the
importance of monitoring and evaluating the use of such tools in online
communication settings to address potential biases and health disparities.

Limitations of the study include the evaluation of only common cancer

misinformation in English and the possibility of outdated scientific information in
ChatGPT's training data.

Further research is needed to assess the accuracy of other chatbots and AI-driven
systems, their performance in addressing diverse claims about cancer, and the
development of an ideal infrastructure to monitor and ensure the accuracy of
cancer information online.

Juhi et al 2023 India Original Peer The Capability of The study aimed to investigate the effectiveness of ChatGPT in predicting and
article reviewed ChatGPT in Predicting explaining common drug-drug interactions (DDIs).
and Explaining A total of 40 DDI pairs were used to converse with ChatGPT using a two-stage
Common Drug-Drug question approach.
Interactions The first question asked whether two specific drugs can be taken together, and the
second question asked why those drugs should not be taken together.
The responses from ChatGPT were checked by two pharmacologists and
categorized as "correct" or "incorrect." The "correct" responses were further
classified as "conclusive" or "inconclusive."
Among the 40 DDI pairs, one answer was incorrect in the first question, and one
answer was wrong in the second question.
For the first question, 19 answers were classified as conclusive and 20 as
inconclusive. For the second question, 17 answers were conclusive, and 22 were
inconclusive.
Using Flesch reading ease and Flesch-Kincaid grade level, the readability ratings
of the answers were assessed. The answers were found to have substantially
higher grade levels than anticipated for a fictitious sixth grade, making them rather
challenging to read.
According to the study, ChatGPT can be used to predict and explain DDIs to a
limited extent. Patients without rapid access to medical facilities may receive some
guidance from it, but it may also provide insufficient information. It need further
work before patients might possibly use it to comprehend DDIs.
Kaneda Japan Commentary Peer- In the Era of Prominent A large language model (LLM) called ChatGPT, created by OpenAI, has become
2023 reviewed AI, What Role Will well-known for producing conversational responses.
Physicians Be Expected An AI chat application called Chat Doctor has proven to be quite accurate at
to Play? prescribing prescriptions based on talks between patients and doctors.
The ability of LLM models to aid patients and medical personnel in the sphere of
healthcare is underlined.
As AI models have limits in offering the most recent research, ethical judgement,
 33

and clinical intuition, doctors will continue to play an important role in healthcare
settings. Collaboration between medical professionals, administrators, and
stakeholders is crucial to advance healthcare reform and effectively utilize AI.
Kim Republic of Original Preprint Search for Medical The study evaluated the ability of ChatGPT, an artificial intelligence chatbot, to
2022 Korea Information and provide medical information and treatment options for subacromial impingement
Treatment Options for syndrome (SIS).
Musculoskeletal ChatGPT was able to generate responses that included correct definitions,
Disorders through an prevalence, and risk factors of SIS based on input messages.
Artificial Intelligence It also provided information on symptoms, diseases with similar symptoms, and
Chatbot: Focusing on orthopedic tests related to SIS.
Shoulder Impingement ChatGPT suggested treatment options such as rest, medication, physical therapy,
Syndrome and potential surgical intervention.
However, some of the information provided by ChatGPT was biased or
inappropriate depending on the individual's condition, indicating the limitations of
relying solely on AI chatbots for accurate medical information.
The study concluded that while ChatGPT can offer overall helpful information about
SIS, caution is necessary, and it still falls short in providing accurate and
individualized medical information and treatment options.
With further advancements in natural language processing technology, it is
expected that AI models like ChatGPT can deliver more detailed and precise
medical information in the future.
Kim 2023 China Original Peer Using ChatGPT for The use of AI chatbots as authors in scientific articles is an ethical issue and has
reviewed language editing in raised concerns in scientific journals.
scientific articles ChatGPT, developed by OpenAI, is a popular AI-powered chatbot available for free
to anyone.
The study submitted questions to ChatGPT regarding the effect of streptozotocin-
induced diabetes on facial bone growth in rats.
ChatGPT provided a detailed response with references to studies, but upon
verification, all the references were found to be fake.
ChatGPT's performance in editing English grammar was found to be excellent.
Language editing services like ChatGPT can be advantageous for authors, but they
should not be listed as co-authors in scientific articles.
Koo China Letter to the Peer- The importance of ChatGPT has several helpful applications in the writing process, including
2023 editor reviewed proper use of ChatGPT paraphrasing difficult sentences, translation, spell-checking, grammar correction,
in medical writing improving clarity and flow of text, generating outlines and abstracts, and formatting
references.
The author of the article in Radiology found that ChatGPT could write an entire
article based on inputted headings and subheadings.
The potential risks of using AI in the writing process were discussed, emphasizing
the importance of proper and ethical use of ChatGPT.
ChatGPT is a versatile tool that can streamline the writing process for researchers,
but it should be used responsibly to maintain ethical standards.
Kumar 2023 Ireland Original Peer- Analysis of ChatGPT he study aimed to evaluate the effectiveness of ChatGPT, an AI chatbot, in the
 34

article reviewed tool to assess the context of academic writing within the field of biomedical sciences. Five random
potential of its utility for topics were inputted into ChatGPT, and the study assessed factors such as
academic writing in response time, content quality, and reliability. The responses generated by
biomedical domain ChatGPT were exported to a Word file and analyzed for originality using Urkund
software.

The findings revealed that ChatGPT had a fast response rate, producing 300-500
words of text in under 2 minutes. However, the content fell short of the expected
quality and depth typically found in academic writing. Limitations observed included
deficiencies in word count, referencing errors, and a lack of academic merit.

In conclusion, while ChatGPT has the potential to serve as a useful tool for training
and improving academic writing skills, it cannot replace human intelligence. When
used appropriately and under academic mentoring, it can assist in refining human
capabilities. Nevertheless, it is crucial to acknowledge the limitations of ChatGPT
and complement its usage with human expertise to ensure the production of high-
quality academic work.
Lee et al 2023 USA Review Peer- Benefits, Limits, and The use of artificial intelligence (AI) in medicine is expanding, including applications
article reviewed Risks of GPT-4 as an AI in medical image analysis, drug interaction detection, patient risk identification, and
Chatbot for Medicine medical note coding.
This article focuses on a specific type of AI called a medical AI chatbot, utilizing the
GPT-4 system developed by OpenAI.
A chatbot consists of a general-purpose AI system and a chat interface, allowing
users to enter queries in natural language and receive relevant responses.
Prompt engineering is essential for optimizing chatbot performance, as current
systems are sensitive to the wording and form of prompts.
While GPT-4 excels in providing correct answers to questions with firm answers, its
most interesting interactions occur when prompts have no single correct response.
Chatbot errors, referred to as hallucinations, can be dangerous in medical
scenarios, so it is crucial to verify the output.
GPT-4 can help catch mistakes, not only in its own work but also in human-
generated content, providing a means of error detection and correction. GPT-4 is a
general-purpose AI system developed by OpenAI, designed to have broad
cognitive skills rather than being programmed for specific tasks in medicine.
Prompts for GPT-4 can be in the form of questions or directives to perform specific
tasks, and they can be written in various human languages and include data inputs
like spreadsheets, research papers, and equations.
Microsoft Research and OpenAI have been studying the applications of GPT-4 in
healthcare and medicine, focusing on areas such as documentation, data
interoperability, diagnosis, research, and education.
Other notable AI chatbots like LaMDA (Google) and GPT-3.5 have also been
explored for medical applications, despite not being explicitly trained for healthcare
purposes.
 35

Three scenario-based examples are provided to demonstrate potential medical use

of GPT-4, including medical note-taking, performance on a USMLE problem, and
providing advice in a "curbside consult" situation.
GPT-4's responses have shown improvements over time, and it has the ability to
generate medical notes, answer questions, generate orders, and provide feedback
to clinicians and patients.
However, GPT-4 is not infallible and can make errors or hallucinations, so
mechanisms for error detection and correction should be implemented in future
deployments. GPT-4, when tested with questions from the USMLE, answers
correctly over 90% of the time, showcasing its innate medical knowledge.
GPT-4 can be used for medical consultation, providing useful responses and aiding
in tasks like diagnosis, research, and education.
It can generate medical notes, provide technical analysis, summarize research,
identify prior work, and ask follow-up research questions.
GPT-4 has the potential to write computer programs, translate languages, decipher
medical documents, and provide emotional support.
While GPT-4 has limitations and can make mistakes, it also has the ability to catch
mistakes made by both AI and humans.
The evaluation of GPT-4's general intelligence and the level of trust users can
place in it remain important questions for further discussion.
GPT-4 is a powerful tool with evolving capabilities, and its responsible use can
bring new possibilities while acknowledging associated risks.
Levin et al 2023 Canada Original Peer- ChatGPT-written The goal of the study was to determine how well doctors with different degrees of
articles reviewed OBGYN abstracts fool publication experience could distinguish between abstracts created by ChatGPT
practitioners and human-written abstracts.
In the study, 20 abstracts—10 authored by ChatGPT and 10 by humans—were
submitted for examination by eight reviewers.
30% of ChatGPT-written abstracts and 43.7% of human-written abstracts were
correctly identified by the reviewers. The accuracy in identifying ChatGPT-written
abstracts was associated with publication experience rather than years of
experience.
A modified Grammarly-based tool showed higher accuracy in identifying abstracts
compared to the reviewers.
Web-based AI output detectors were able to detect ChatGPT text, suggesting their
potential use in screening submissions to journals.
The study highlights the need for improved means of detecting ChatGPT-written
scientific abstracts.
The limitations of the study include a small sample size of reviewers and
referencing a preprint report, while the novelty and total number of abstracts tested
are considered strengths.
Efforts should be focused on optimizing methods for the detection of ChatGPT-
written abstracts.
Li et al 2023 USA Commentary Peer- Ethics of large language The medical industry has become interested in large language models (LLMs),
 36

reviewed models in medicine and such as ChatGPT, because they may be utilised for text production,
medical research summarization, and correction activities.
LLMs can produce content for patient-friendly communication, standardised
reporting, presentation outlines, sample cover letters, and clinical documentation.
As LLMs mirror the training data, which may perpetuate prejudices and
underrepresent some perspectives, ethical problems include the possibility of
hidden biases in LLMs.
Trust and credibility of LLM-generated content are challenges, as LLM outputs are
difficult to trace and may be inaccurate. Current models do not evaluate quality or
provide measures of uncertainty.
Authorship becomes a complex issue when LLMs are used, as LLMs lack
intentionality and cannot approve or be held accountable for the final work. New
guidelines may be required.
Varying payment models for LLMs may widen digital divides, and affordability and
accessibility should be ensured.
Privacy concerns arise regarding the collection, use, and potential dissemination of
data inputted into LLMs. Strict controls and informed consent are necessary.
It is important to establish guidelines that promote responsible and effective use of
LLMs, rather than imposing an outright ban on their use.
In summary, while LLMs have the potential to revolutionize medicine and medical
research, careful consideration of their biases, trustworthiness, authorship,
equitability, and privacy implications is crucial. Responsible guidelines should be
established to ensure their ethical use.
Loh 2023 Australia Review Peer- ChatGPT and ChatGPT, an AI chatbot tool released by OpenAI, has gained significant popularity
article reviewed generative AI chatbots: and raised concerns about fraud and plagiarism due to its ability to generate text
challenges and similar to human-authored content.
opportunities for The risks of generative AI tools in health include challenges of fraud and plagiarism
science, medicine and in educational settings, the potential for biased outputs, and privacy and
medical leaders cybersecurity concerns.
AI tools like ChatGPT have been used to generate scientific articles, leading to
changes in guidelines by scientific meetings and publishers to address authorship
and credibility.
Despite the risks, generative AI tools offer opportunities for scientific research,
including assistance in experimental design, peer review, and administrative tasks
such as grant applications and research article editing.
In clinical practice, AI chatbots like ChatGPT can support frontline clinicians by
automating administrative tasks, such as generating discharge summaries and
medical letters.
AI chatbots can also assist in patient communication by providing accurate
information and answering low-risk health questions, although ethical
considerations need to be addressed.
Some health systems have already started using generative AI technology to
improve healthcare provision and alleviate administrative burdens and clinician
 37

burnout.
Li 2023 China Commentary Peer ChatGPT has made the ChatGPT is a language generation model developed by OpenAI and is capable of
reviewed field of surgery full of generating high-quality text content.
opportunities and It has been trained on a large amount of preprocessed and filtered text data.
challenges IBM's Watson robot is an AI technology specifically designed for the medical field,
providing diagnosis and treatment suggestions based on clinical guidelines.
In the field of surgery, the Da Vinci robotic surgical system is widely used, but
surgeons still operate the surgeries and AI technology assists in surgical planning,
navigation, and support.
While ChatGPT can provide valuable diagnostic and treatment suggestions,
doctors must evaluate and judge these suggestions and comply with legal
regulations.
ChatGPT is suitable for document polishing and has the potential to be a powerful
medical assistant in the future.
The current application of ChatGPT in the medical field may not have a disruptive
effect yet, but its potential is promising and it can become a valuable tool with
further development and training.
Liebrenz et al UK Commentary Peer Generating scholarly The use of generative AI, such as ChatGPT, in medical publishing practices has
2023 reviewed content with ChatGPT: unknown implications, but it raises substantial ethical concerns.
ethical challenges for ChatGPT is an AI chatbot developed by OpenAI that generates responses based
medical publishing on internet sources and has been used to create university essays and scholarly
articles.
The imperceptibility of AI-generated content to human readers and anti-plagiarism
software raises concerns about copyright, attribution, plagiarism, and authorship.
There is a debate about whether AI can fulfill the criteria for authorship set by
organizations like the International Committee of Medical Journal Editors.
Organisations like the Committee on Publication Ethics and the International
Association of Scientific, Technical, and Medical Publishers have given ethical
principles and guidelines for AI in publishing.
As AI technologies proliferate, comprehensive talks about authorship laws and AI-
generated work are seen as urgent and crucial.
The usage of ChatGPT and comparable AI platforms might result in a paywall-
based business model, escalating already-existing inequities in scientific
publication. ChatGPT's accessibility and ease of use have the potential to increase
scholarly output and democratize knowledge dissemination in multiple languages.
However, there are concerns about the potential harm caused by ChatGPT in
producing misleading or inaccurate content, contributing to scholarly
misinformation.
Competitors to ChatGPT may arise, amplifying the challenges associated with AI-
generated content in scholarly publishing.
The Lancet Digital Health and the Lancet family are called upon to carefully
consider the ethical implications of publishing articles produced by AI and initiate
discussions on comprehensive guidance for AI-generated content in scholarly
 38

publishing.
Lin 2023 Hong Kong Commentary Preprint Modernizing authorship The conventional requirement for authorship, based on "substantial contributions,"
criteria: Challenges is out of sync with current practices and emerging trends in scientific research.
from exponential The exponential increase in the number of authors per paper poses challenges to
authorship inflation and the notion of "substantial contributions," as each author's contribution becomes
generative artificial increasingly small.
intelligence The rise of generative AI tools, like ChatGPT, in scientific research raises questions
about whether they should be listed as co-authors when they substantially
contribute to a paper.
The International Committee of Medical Journal Editors (ICMJE) and other
authoritative guidelines emphasize substantial contributions to the research and
writing process as criteria for authorship.
However, the current criteria do not align with the increasing number of authors and
the potential substantial contributions of AI tools.
A new approach is proposed, where authorship is seen as proportional, role-
specific credit, reflecting the specific contributions made by individuals in various
roles.
The CRediT system, which defines specific roles in research, can be used to
specify and quantify individual contributions in a standardized manner.
Organizations, publishers, and authors should consider reevaluating and updating
authorship criteria to better reflect current research practices and the involvement
of generative AI tools.
Recommendations include revising the requirement of "substantial contributions,"
requiring authors to specify contributions using a standardized system like CRediT,
and actively documenting and discussing contributions throughout the research
process.
Maeker and France Commentary Peer- ChatGPT: A solution for Artificial intelligence (AI) can produce grammatically correct and coherent text
Maeker-Poquet reviewed producing medical based on patterns it has learned, but it lacks a deep understanding of the content.
2023 literature reviews?
AI like ChatGPT operates by processing natural language and following patterns in
data, not by truly understanding the content.

AI's capability to synthesize complex thoughts or scientific content is limited due to

its lack of true understanding.

An experiment was conducted to see if ChatGPT could write a literature review,

drawing upon its learning from diverse sources including PubMed.

The initial trial showed that while ChatGPT could provide clear, well-written content,
it lacked sufficient references, often invented content, and did not meet established
literature review guidelines (e.g., PRISMA and ICMJE).

A second trial, with more structured input and guidance, led to more coherent
 39

content but still had shortcomings in terms of depth, reference handling, and
staying on topic.

Despite the limitations, ChatGPT could be a useful tool in drafting early versions of
literature reviews or helping novice writers get started, owing to its speed and
ability to generate content.

AI could also help suggest initial manuscript structure, rephrase and summarize
content, and suggest references, although it can only draw from information up to
2021 and can't compare and interpret results.

There are ethical considerations, such as plagiarism detection and authorship

issues, that arise when using AI tools for writing scientific literature.

Guidelines set by publishers like Elsevier allow for the use of AI for improving
readability and language but discourage it from replacing key research steps like
producing scientific information, analyzing data, or formulating conclusions.

The use of AI such as ChatGPT in writing scientific literature has potential, but
should be considered a supplement to, not a replacement for, human expertise and
judgement. Ethical issues around this usage remain under debate.
Marchandot et al France Commentary Peer- ChatGPT: the next ChatGPT, developed by OpenAI, is a powerful language model capable of
2023 reviewed frontier in academic generating human-like text. It can understand context and generate coherent, fluent
writing text that is often indistinguishable from text written by humans.
for cardiologists or a
pandora’s box of ethical The model has potential applications in academic research, such as assisting in
dilemmas data analysis, literature reviews, and drafting research papers. It can analyze vast
amounts of data quickly and can be trained on specific topics to help find and
summarize relevant literature.

ChatGPT can assist in revising scientific manuscripts, identifying and correcting

grammar and spelling errors, suggesting alternative phrasing, and even proposing
additional experiments or data analysis.

There are potential downsides to using ChatGPT in academic research. For

example, if trained on biased data, the model might produce biased results. It may
also not understand the nuances of a specific field and produce inaccurate results.

Dependence on models like ChatGPT could lead to a decrease in critical thinking

and creativity among researchers. There are also ethical concerns related to
plagiarism, as the model may generate text similar to existing content.

There is ongoing debate on the use of AI tools like ChatGPT in academic research.
 40

It's controversial whether it's appropriate to cite ChatGPT as an author in published

literature.

Despite concerns and potential downsides, the authors of the commentary

acknowledge the invaluable assistance of ChatGPT in writing and editing the text,
advocating for its recognition as a co-author. They call for academic institutions and
publishers to establish guidelines for using AI-generated text in academic research.
Martínez-Sellés Spain Review Peer- Current and Future Use AI technology is rapidly being integrated into electrocardiography (ECG) for the
and Marina- article reviewed of Artificial Intelligence detection, categorization, and treatment of heart disease.
Breysse 2023 in Electrocardiography
AI algorithms have the ability to assist clinical professionals in analyzing and
identifying ECG abnormalities such as arrhythmias, ST-segment changes, and QT
prolongation.

AI can be utilized alone or in combination with clinical factors to predict the

occurrence of arrhythmias, sudden cardiac death, stroke, and other cardiovascular
events.

Real-time monitoring of ECG signals from wearable electronic devices and cardiac
implants can be facilitated by AI, alerting medical professionals or patients when
significant changes occur based on timing, duration, and context.

By reducing noise, artifacts, and interference, as well as extracting data that may
not be visible to the human eye, such as heart rate variability, beat-to-beat
intervals, and wavelet transforms, AI can enhance signal processing, leading to
improved ECG quality and accuracy.

AI can support therapy decision-making by assisting with patient selection,

treatment optimization, timing of symptom-to-treatment, and cost-effectiveness.
Furthermore, by predicting the outcome of antiarrhythmic drug or cardiac implanted
device therapy, AI can reduce the risk of cardiac toxicity and facilitate the early
activation of emergency protocols in patients with ST-segment elevation.

The fusion of ECG data with information from other modalities, including imaging,
genomics, proteomics, and biomarkers, can be facilitated by AI.

As more data becomes available and AI algorithms continue to advance, it is

expected that AI will play an increasingly significant role in the diagnosis and
management of ECG-related conditions in the future.

Mehnen et al Austria Original Preprint ChatGPT as a medical The study tested the diagnostic accuracy of ChatGPT using 50 clinical case
2023 article doctor? A diagnostic vignettes, including 40 common complaints and 10 rare diseases.
accuracy study on
 41

common and rare The case vignettes were collected from previous research and from the first
diseases matching articles on PubMed for randomly selected rare disorders with an
approved orphan drug.

ChatGPT was prompted with the phrase “What are the 10 most likely diagnoses for
this patient?” followed by the full text of the clinical case vignette.

Each vignette was prompted three times in independent chats using both ChatGPT
3.5 and ChatGPT 4, totaling 300 prompts and 3000 suggested diagnoses.

The 50 case vignettes were also presented to three human medical doctors for
diagnosis without the use of search engines.

The correctness of suggested diagnoses was assessed by comparing them with

the correct diagnoses provided in the case vignettes.

For common cases, ChatGPT 3.5 solved more than 90% of all cases within the top
two suggested diagnoses, while ChatGPT 4 solved 100% of all cases within the top
three suggested diagnoses.

For rare cases, ChatGPT 3.5 achieved an accuracy of 60% within the top 10
suggestions, while ChatGPT 4 achieved an accuracy of 90% within eight or more
suggestions.

Fleiss' kappa was used for diagnosis agreement hypothesis-testing between

different versions of ChatGPT and the correct diagnoses, indicating a significant
similarity to the correct diagnoses.

Despite the high diagnostic accuracy on common cases, caution should be

exercised when using ChatGPT for medical advice, especially for non-
professionals, due to the lower performance on rare cases.
Mello and Guha USA Commentary Peer- ChatGPT and Large language models (LLMs) like ChatGPT have the potential to augment or
2023 reviewed Physicians' Malpractice replace human decision-making in various fields, including medical decision-
Risk making. However, they also bring with them a risk of liability for practitioners who
use them.

The legal position of physicians using these models is unclear. However, the
precedent of clinical practice guidelines may provide some guidance. In cases
where physicians have followed these guidelines, it has often constituted
exculpatory evidence in malpractice suits.

LLMs have some distinctive features that make them different from clinical practice
guidelines. They have a tendency to generate incorrect outputs (hallucinations),
 42

often without clear references for physicians to assess the reliability of the
information.

Most LLMs are trained on a broad array of web text, which does not discern
between reputable medical articles and less reliable sources such as online
forums. Efforts are being made to train LLMs on authoritative medical texts, but
these are still in their infancy.

LLMs' outputs are not fixed, they can yield different responses for the same query,
and also vary depending on the date and the phrasing of the question. This can
make it difficult to determine the reasonableness of a physician's actions based on
the LLM's output.

While clinical decision support tools often undergo rigorous validation processes,
the same can't always be said for LLMs. They are mainly validated as text
generators and their design and evaluation is often performed by computer
scientists, not clinicians.

Despite these drawbacks, LLMs have their benefits. They can provide more
patient-specific recommendations than other decision-support tools and may also
suggest diagnostic and treatment possibilities that may be overlooked by
physicians.

There are studies evaluating LLMs' performance in clinical scenarios, and although
the results show some accuracy, there are also instances where they provide
completely incorrect answers.

At the present time, physicians should only use LLMs to supplement more
traditional forms of information seeking, such as searching for reputable sources on
Google or using clinical decision support systems.

Future reliable LLMs may likely come from specialized systems rather than
generalist systems like ChatGPT. These specialized systems can be validated
more easily and are less prone to the problems of indiscriminate information
sourcing.

While the potential of LLMs is promising, it is important to carefully understand and

manage the risks associated with their use in medical practice. Continued research
and development is needed to improve these systems for practical clinical use.
Mese 2023 Turkey Letter to the Peer The imperative of a To promote safety, efficient use, and ongoing development, the study in Clinical
editor reviewed radiology AI deployment Radiology suggests setting up a UK registry for radiology artificial intelligence (AI)
registry and the applications.
potential of ChatGPT
 43

The writers of the article propose a precise description of the registry's objective,
target market, and domain, as well as frequent evaluations for preserving its worth
and applicability.

Due to its remarkable natural language interpretation and generation capabilities,

ChatGPT, developed by OpenAI, can be a useful tool in this situation.

By condensing voluminous material and highlighting salient aspects, ChatGPT can

promote more effective talks and decision-making processes between
stakeholders.

In order to raise awareness of the register among members of the Royal College of
Radiologists (RCR) and other interested parties, ChatGPT can also provide
precise, coherent, and correct text.

By evaluating textual information like as user comments, performance reports, and

research articles, ChatGPT can be helpful in the continuing monitoring and
assessment of AI applications in the registry. This information can help pinpoint
possible areas for improvement or concern.

The AI model's comprehension of English may also help with the registry's long-
term upkeep by detecting pertinent updates, summarising fresh research, and
creating rules to make sure the registry is kept current.

The incorporation of ChatGPT in the creation and upkeep of the registry may
simplify monitoring, assessment, and communication procedures, potentially
enhancing patient safety, resource effectiveness, and learning retention.

The introduction of ChatGPT into the radiology AI deployment registry has the
potential to make a substantial impact on the development of AI technology.
Nastasi et al 2023 USA Commentary Preprint Does ChatGPT Provide A huge language model from OpenAI called ChatGPT has demonstrated amazing
Appropriate and ability in conversational activities like successfully choosing the right radiological
Equitable Medical studies, writing research papers, and passing medical licencing examinations.
Advice?: A Vignette-
Based, Clinical However, due to a lack of publicly available training data and unsolved concerns
Evaluation Across Care regarding the safety, fairness, and regulation of AI systems, there are doubts about
Contexts its capacity to offer competent and equitable medical advice in a variety of clinical
scenarios.

The study evaluated the clinical appropriateness of ChatGPT's answers to

questions about seeking guidance across a range of settings and demographics.

It was discovered that ChatGPT alters its responses depending on the patient's
 44

insurance status, race, and gender, frequently resulting in varied and clinically
unsuitable recommendations that, if followed, could worsen health inequities.

The responses lacked the personalised nuance or follow-up questions that a

clinician would generally offer, even though the AI frequently gave useful
recommendations. Concerns concerning reliability were raised when it was
discovered that responses varied greatly and were inconsistent.

The study's limitations include the fact that just three distinct clinical settings were
tested, and that within-vignette variance was not measured.

Despite these drawbacks, the study offers crucial data that contextualises
ChatGPT's capacity to deliver accurate and impartial guidance along the treatment
continuum.

Conclusion: While ChatGPT can offer generic information on clinical subjects, it is

currently unable to provide reliable personalised or suitable medical advice. Future
improvements and training might help future LLMs give better medical
recommendations.
Nguyen and France Commentary Peer- Artificial intelligence and Artificial intelligence (AI), specifically deep learning, is revolutionizing various fields,
Costedoat- reviewed internal medicine: The including medicine.
Chalumeau 2023 example of ChatGPT is a conversational AI system developed by OpenAI, capable of providing
hydroxychloroquine natural and fluent communication with users.
according to ChatGPT ChatGPT has been pretrained on a vast amount of textual data, enabling it to
generate precise and relevant responses.
ChatGPT can be used as an additional source of information for patients seeking
medical information on topics like hydroxychloroquine (HCQ).
While ChatGPT's responses can be informative, they should be validated and
supplemented with information from reliable sources.
Patients increasingly rely on the internet for medical information, and
conversational AI systems like ChatGPT can provide insights into the content
patients may encounter online.
The example of HCQ information provided by ChatGPT demonstrates both
accurate and inaccurate information, highlighting the need for caution and
verification.
ChatGPT can also be used by prescribers to obtain information on topics like
dosing and monitoring requirements.
However, prescribers should consult healthcare professionals for precise dosing
and specific guidelines.
Conversational AI systems have the potential to improve patient education and
enhance healthcare delivery, but their limitations and potential risks should be
considered.
Further research and development are needed to refine and enhance the
 45

capabilities of conversational AI in the field of medicine.

Nógrádi et al 2023 Hungary Original Preprint ChatGPT M.D.: Is there This preprint research paper discusses the use of ChatGPT, an AI language model,
article any room for generative for diagnosing neurological disorders. The study involved 12 medical doctors from
AI in neurology and the University of Szeged, Hungary, who were divided into two groups: neurologist
other specialists and general medical doctors. Synthetic cases representing various
medical areas? neurological diseases were generated by neurological experts and fed into
ChatGPT to obtain diagnostic predictions.

The results showed that ChatGPT's diagnostic accuracy was higher than that of the
general medical doctors but lower than that of the neurologist specialists. However,
when considering the three most likely diagnoses or the five most likely diagnoses
provided by ChatGPT, the accuracy reached a level comparable to the experts.
The study also found that ChatGPT performed equally well in diagnosing acute
neurological disorders compared to human doctors, but its accuracy decreased for
non-acute cases.

The researchers observed a significant overlap in incorrect diagnoses between

ChatGPT and human doctors, indicating similarities in diagnostic errors. In a subset
of unsolved cases where even the experts failed to provide an accurate diagnosis,
ChatGPT's most probable diagnosis matched the original diagnosis in 40% of
cases, and the five most probable diagnoses included the correct diagnosis in 60%
of cases.

The paper emphasizes that ChatGPT should be used as an augmentation tool in

healthcare, with all suggestions made by the AI further evaluated by medical
experts. It also highlights the potential of AI, such as ChatGPT, in supporting the
triaging process for acute neurological scenarios and diagnosing rare and atypical
cases.

The study acknowledges the ethical, integrity, and data safety concerns associated
with using AI in medicine. While AI has the potential to improve healthcare, careful
consideration of these concerns is necessary.
North UK Commentary Peer Plagiarism re-imagined The author of this personal account reflects on a time when scientific writing was
2023 reviewed done manually, with quills, pens, and typewriters. Manuscripts were often typed by
someone else, and research was conducted using limited resources like tape
recorders and index cards. The author received a request to review a manuscript
from the journal Life Sciences, which turned out to have significant portions copied
verbatim from a paper the author had previously published. The author contacted
the editor, who took the matter seriously and barred the authors from submitting
further work to the journal.

Later, the author came across another paper published in Neuroscience Letters,
which also contained substantial portions copied from the author's earlier work. The
 46

author contacted the editor, who published a disclaimer acknowledging the overlap
and the prior publication of the author's work. The author reflects on the irony of
receiving a plagiarized manuscript for review and the increased ease of plagiarism
in the modern age with AI-based text generators like ChatGPT. The author also
highlights the detrimental impact of simplistic numerical methods, such as the h-
index, on scientific integrity and the motivation behind academic dishonesty.
Oh et al 2023 Korea Original Peer ChatGPT goes to the The goal of this study was to assess how well the ChatGPT language model
article reviewed operating room: performed using a dataset of questions from the Korean general surgery board
evaluating GPT-4 exam. 280 questions from the exam's first stage between 2020 and 2022 made up
performance and its the dataset. The dataset did not include any questions that requested visual data.
potential in surgical For model testing, the study used ChatGPT versions GPT-3.5 and GPT-4.
education and
training in the era of The study findings revealed a notable disparity in performance between the GPT-
large language models 3.5 and GPT-4 models, with GPT-4 exhibiting significant improvement. GPT-4
achieved an overall accuracy rate of 76.4%, surpassing the accuracy rate of GPT-
3.5, which was 46.8%. The accuracy rates varied across different subspecialties,
with transplantation, breast, and hepatobiliary and pancreas demonstrating higher
accuracy rates.

The study highlights the potential of language models like ChatGPT in

comprehending complex surgical clinical information. It emphasizes the
significance of the findings, as the models achieved high accuracy without fine-
tuning and using prompts exclusively in the Korean language.

These results suggest that language models can have implications for surgical
education and training, providing a tool for evaluating surgical knowledge and skills.
However, further research and fine-tuning of the models are needed to enhance
their performance and applicability in the surgical field.
Okan 2023 Turkey Commentary Peer AI and Psychiatry: The Modern language model ChatGPT, created by OpenAI, simulates human
reviewed ChatGPT Perspective conversation and offers precise information and responses on a range of subjects.

A clinical decision support tool, medical content generator, drug information

retrieval tool, medical translator, and patient educator are just a few of the potential
uses for ChatGPT in the healthcare industry.

ChatGPT can support the selection of treatments and long-term patient monitoring
in the field of psychiatry. It can also help with the diagnostic process, support
psychiatric education, simulate case examples, assist researchers in data
collecting and analysis, and more.

The diagnosis, treatment, and monitoring of psychiatric diseases could be

revolutionised by integrating artificial intelligence, natural language processing, and
GPT technologies.
 47

These technologies can help psychiatrists make more accurate diagnoses, design
effective personalized treatments, improve clinical trial accuracy and speed, and
increase accessibility to mental health care through chatbots and virtual assistants.
The use of AI in psychiatry has the potential to advance the field and improve
patient outcomes.

AI technologies like ChatGPT have exceeded expectations and are expected to

provide significant support under human control.
The future of psychiatry and mental health care looks promising with the continued
evolution and improvement of AI technologies in the field.
Parsa and USA Editorial Peer ChatGPT in Medicine; a ChatGPT, developed by OpenAI, has gained immense popularity as a fluent
Ebrahimzadeh reviewed Disruptive Innovation or chatbot with over 100 million users.
2023 Just One Step Incorporating AI technologies like ChatGPT in medical education has great
Forward? potential, improving patient education, health literacy, and facilitating clinical
workflows.
ChatGPT can assist with patient inquiries, help with writing medical notes and
discharge summaries, aid in decision-making and treatment planning, and support
medical educators in various tasks.
It can generate creative ideas, write essays and homework assignments, and even
contribute as an author on research papers.
Challenges exist with the use of ChatGPT, such as the need for access to current
information, potential biases in the training data, privacy concerns, and issues
related to authorship in academic papers.
Regulations and control mechanisms should be established to address ethical
utilization of ChatGPT and similar tools in medicine.
Despite challenges, AI tools like ChatGPT are expected to be widely adopted in the
medical field in the future.
Patel et al 2023 USA Editorial Peer ChatGPT: friend or foe ChatGPT, developed by OpenAI, has garnered significant attention and excitement
reviewed due to its ability to generate text in response to prompts.
It has demonstrated potential applications in healthcare, such as generating patient
discharge summaries and simplifying radiology reports, but errors and limitations
have been observed.
OpenAI acknowledges that ChatGPT's output can be incorrect, biased, and
respond to harmful instructions, despite implementing guardrails to mitigate risks.
Concerns have been raised regarding algorithmic bias, the spread of
misinformation, and the potential for introducing errors or plagiarized content into
scientific publications.
The World Association of Medical Editors has issued recommendations on the use
of ChatGPT and other chatbots in scholarly publications, emphasizing the need for
new tools to detect AI-generated content.
Editorial policies are evolving to address the use of AI in scientific writing, including
the requirement for manual checks of AI-generated output and the exclusion of AI
 48

tools as authors or co-authors.

Careful and responsible use of ChatGPT, along with investment in robust AI output
detectors, is necessary to prevent potential negative impacts on society and
scholarly publishing.
While ChatGPT is considered a game changer, there is a need for more
forethought, oversight, and technological advancements before its widespread
adoption.
Pourhoseingholi USA Editorial Peer Does chatGPT (or any ChatGPT, developed by OpenAI, is a chatbot powered by an AI neural network that
et al 2023 reviewed other artificial can provide responses to user prompts.
intelligence language It has been used in various fields, including healthcare, where it can assist with
tool) tasks such as accessing medical records and providing patient services.
deserve to be included ChatGPT has also been shown to be a useful tool for education and research,
in authorship list? aiding in learning and generating high-quality manuscripts.
The question of whether ChatGPT should be considered an author in scientific
publications has sparked debate and ethical concerns.
ChatGPT is seen as a tool rather than a collaborator, and it lacks the ability to
participate in the research process and take responsibility for the work.
International guidelines for authorship criteria, such as those from ICMJE and
COPE, are not met by ChatGPT.
The placement of ChatGPT as an author, the issue of consent, and conflicts of
interest need to be addressed.
Policies should be implemented to prevent the inclusion of ChatGPT as an author,
and transparent guidelines should be established by publishers and journals.
International organizations like ICMJE and COPE should consider adapting their
criteria to address AI authorship.
Prohibition of the use of ChatGPT in research seems impractical, and efforts
should focus on developing clear policies and criteria for AI authorship.
Rao 2023 India Letter to Peer- The Urgent Need for Effective implementation of AI in healthcare necessitates substantial training and
editor reviewed Healthcare Workforce upskilling of healthcare providers to ensure their proficiency in utilizing these
Upskilling and Ethical technologies.
Considerations in the
Era of AI-Assisted Addressing ethical and privacy concerns associated with AI in healthcare requires
Medicine the establishment of stringent guidelines and robust safeguards to protect patient
data.

Language generation models such as ChatGPT have the capability to assist

healthcare professionals in composing discharge summaries, leading to improved
efficiency and accuracy in documentation.

The responsible and conscientious utilization of AI technologies can enhance

healthcare outcomes and alleviate the workload burden on healthcare providers
who often face significant demands.
 49

The integration of AI in India's healthcare system holds the potential to bring about
transformative changes in healthcare delivery. However, it is crucial to exercise
caution and consider all relevant factors to ensure a well-planned and regulated
implementation process.
Ray and India Letter to Peer- AI Tackles Pandemics: The research examines ChatGPT's possible uses as an AI language model for
Majumder 2023 editor reviewed ChatGPT's Game- infectious disease management and control.
Changing Impact on ChatGPT should be regarded as an additional tool for clinical practise rather than a
Infectious Disease replacement for healthcare providers' knowledge.
Control The problems of ChatGPT's accessibility, lack of human interaction, and inaccurate
information are all acknowledged.

ChatGPT can assist in disseminating accurate and up-to-date information about

infectious diseases, monitoring outbreaks, and aiding in diagnosis, treatment, and
research.
Limitations of ChatGPT include potential inaccuracies from biased or incorrect input
data, lack of human interaction, ethical considerations, and language barriers.
Optimization of ChatGPT can be achieved through continuous data collection,
refining training data, and validation by human experts.
Making ChatGPT more accessible can be achieved by expanding language
capabilities and incorporating translation features.
Ethical guidelines and regulations are needed to ensure responsible use of
ChatGPT, fostering accountability, providing transparency, and protecting privacy.
The effectiveness of ChatGPT in managing infectious diseases can be increased
through seamless interaction with current healthcare infrastructure and operations.
The potential for ChatGPT and other models to play a substantial role in healthcare
is made possible by the ongoing advancements in AI technology, but strict
oversight and regulation are required.
Ros-Arlanzón and Spain Letter to Peer- [ChatGPT: a novel tool ChatGPT, developed by OpenAI, is a language model that has sparked significant
Pérez-Sempere editor reviewed for writing scientific debate due to its potential applications in various fields, including scientific writing
2023 articles, but not an and medicine.
author (for the time ChatGPT can generate coherent and relevant text, making it a valuable tool for
being)] researchers in neurology and other disciplines.
However, it is important to note that ChatGPT cannot be considered an author of a
scientific article. Authorship is reserved for individuals who have made substantial
contributions to the research, and as an AI language model, ChatGPT cannot
contribute to the conception, design, execution, analysis, or interpretation of the
research.
ChatGPT can be acknowledged in the acknowledgments section of a scientific
article or recognized as a tool used in the research process, but it should not
replace human contribution in scientific research and writing.
The continuous development of AI technology may lead to future considerations
regarding the role of ChatGPT and other AI models as coauthors or even as the
main author of scientific articles.
 50

The potential of AI, including ChatGPT, is significant, but it is important to

recognize the current limitations and the need for researchers to conduct the
research and take responsibility for the content of scientific publications.
Sabry Abdel- Portugal Letter to Peer- ChatGPT in Clinical Since its public launch on November 30, 2022, ChatGPT, developed by OpenAI in
Messih and Kamel editor reviewed Toxicology San Francisco, has made significant strides in the medical field, despite not being
Boulos 2023 explicitly trained in medicine.

Powered by deep learning artificial intelligence (AI) techniques and based on the
GPT-3 architecture, ChatGPT provides human-like responses to natural language
queries.

Integration of ChatGPT into the Microsoft Bing search engine is currently

underway, which will make it accessible to a wide range of online users worldwide,
including clinicians, medical and nursing students, as well as patients.

The authors of a study conducted a test on ChatGPT using a clinical toxicology

vignette focused on acute organophosphate poisoning, a relatively straightforward
case that professionals in the field are unlikely to overlook.

ChatGPT successfully answered all questions related to the case, offering well-
explained reasoning. Both the initial response and a regenerated response
provided satisfactory results.

While ChatGPT demonstrated competence with straightforward cases, the authors

highlight that the true challenge in clinical medicine often lies in accurate history
taking, identification of signs, and eliciting pertinent information—areas where
junior clinicians may encounter difficulties.

There is potential for future iterations of ChatGPT, specifically adapted for medical
use, to be valuable in handling less common clinical cases that experts may
sometimes overlook.

The authors envision a future where AI serves as a tool utilized by clinicians, rather
than a replacement for them. They predict that in the coming years, clinicians who
incorporate AI into their practice will replace those who do not.

Sallam Jordan Review Preprint The Utility of ChatGPT A thorough analysis was done to determine the possible benefits and drawbacks of
2023 as an Example of Large the "ChatGPT" AI-based large language model (LLM) in healthcare instruction,
Language Models in study, and practise.
Healthcare Education,
Research and Practice: The review employed PRIMSA standards and utilised the term "ChatGPT" to
Systematic Review on search English records in PubMed/MEDLINE and Google Scholar. Any preprints or
the Future Perspectives published studies considering ChatGPT in the context of healthcare were eligible
 51

and Potential for inclusion.

Limitations
60 records out of a total of 280 were deemed eligible for inclusion after a careful
screening process.

ChatGPT was mentioned in 85% of the records that were collected, with its use in
scientific writing and medical research being the most common. It was praised for
its ability to reduce costs, facilitate paperwork, enable personalised therapy, and
promote more health literacy in healthcare practise.

Nearly 97% of the records mentioned worries or potential hazards related to

ChatGPT. The most frequent problems were ethical ones, including bias risk,
plagiarism, copyright issues, transparency problems, legal concerns, a lack of
originality, erroneous responses, a lack of understanding, and improper citations.

Despite these reservations, ChatGPT demonstrates promising applications that

have the potential to revolutionise healthcare research, teaching, and practise.
These apps should be used cautiously, though.

ChatGPT could be a useful resource for individualised learning in the field of health
education and promote a move towards critical thinking and problem-based
learning.

ChatGPT could improve personalised medication and streamline workflows in the

healthcare industry.

ChatGPT could speed up the experimental design process in scientific research

and improve the equity and adaptability of the research.

ChatGPT does not meet the criteria to be an author in scientific journals as of right
now. However, modifications to the ICMJE/COPE standards may cause this to
change.

The study urges the creation of a code of ethics and conduct for the responsible
use of ChatGPT and other LLMs, engaging all parties involved in healthcare
education, research, and practise.
Sallam et al Jordan Review Peer ChatGPT Output ChatGPT could be a useful resource for individualised learning in the field of health
2023 reviewed Regarding Compulsory education and promote a move towards critical thinking and problem-based
Vaccination and learning.
COVID-19 Vaccine
Conspiracy: A In light of its impending integration into web search engines, the study sought to
Descriptive Study at the assess the output of the AI-based language model ChatGPT in relation to COVID-
Outset of a Paradigm 19 vaccine conspiracy theories and mandatory vaccination.
 52

Shift in Online Search

for Information On January 14, 2023, the investigation was carried out utilising ChatGPT, an
OpenAI creation. Two writers assessed the correctness, clarity, conciseness, and
bias of the ChatGPT-generated responses.

Findings revealed that ChatGPT replies rejected conspiracy theories regarding the
origins of the SARS-CoV-2 virus as untrue and devoid of scientific support.
Additionally, ChatGPT answers substantially refuted the claims of COVID-19
vaccine conspiracies.

ChatGPT exhibited objectivity when discussing mandatory vaccination, presenting

both the benefits and drawbacks of such a policy. The benefits included preserving
herd immunity, preventing the spread of disease, being cost-effective, and having a
legal obligation. The drawbacks included moral and legal dilemmas, opposition and
mistrust, logistical difficulties, and a lack of information and resources.

According to the study's findings, ChatGPT may provide important information to

dispel COVID-19 vaccine conspiracies. It presents both arguments for and against
mandatory vaccination in a straightforward, succinct, and impartial manner.

The study cautions that, despite ChatGPT's usefulness, content produced by it

shouldn't take the place of trustworthy sources of vaccine information like the World
Health Organisation (WHO) and the Centres for Disease Control and Prevention.
(CDC).

Salvagno et al Italy Review Peer Can artificial ChatGPT has demonstrated its usefulness in supporting researchers across
2023 article reviewed intelligence help for various aspects of scientific writing, including drafting articles, abstracts, literature
scientific writing? research, data summarization, structure suggestions, and language review.

However, there have been no published articles so far on the use of this technology
in critical care medicine, leaving it uncertain whether its implementation would
simplify or complicate the writing process within this discipline.

While ChatGPT cannot generate entirely new ideas, it can assist researchers by
developing their existing ideas and creating initial drafts. It has the potential to aid
in the review process by identifying and summarizing academic papers and
highlighting areas of uncertainty.

During the writing process, ChatGPT can generate an initial draft of a scientific
paper, suggest titles, justify sample sizes, and describe data analysis techniques.

Furthermore, ChatGPT can provide assistance in the editing process, including

formatting, language editing, and summarizing the text to create a suitable abstract,
 53

thereby saving researchers time.

However, it is important to carefully evaluate the quality of AI-generated content, as

it cannot substitute human expertise, creativity, and critical thinking.

Future applications of AI may involve the automatic generation of figures, tables,

and other visual elements for manuscripts, contributing to data summarization.

While AI can quickly comprehend information and establish connections between

evidence, it lacks the nuance, style, and originality that human authors provide. It
also carries the risk of perpetuating existing biases and inaccuracies present in the
data.

Ethical concerns arise regarding AI-generated writing, including issues of

plagiarism, bias, accuracy, and authorship.

In addition to scientific writing, ChatGPT has the potential to support physicians in

their hospital work by providing information on recognized ICU protocols,
generating clinical notes, and assisting in patient communication.

Despite its potential benefits, the technology's limitations must be recognized. The
accuracy, current knowledge, and precision of the AI-generated content may be
variable and should always be validated by human experts.

As AI and chatbot tools continue to develop, international academic regulations are

needed to govern their use in scientific writing and establish mechanisms for
identifying and penalizing unethical usage.

Ultimately, AI and chatbots like ChatGPT should not replace human researchers'
expertise, judgment, and personality, but rather supplement them.
Sanmarchi et al Italy Original Preprint A step-by-step Aim of the Study: The purpose of this research was to evaluate how early-stage AI-
2023 article Researcher's Guide to based transformer models, specifically OpenAI's ChatGPT, can assist researchers
the use of an AI-based in the multiple steps of an epidemiological study. The study was simulated using
transformer in the STROBE framework and the responses of the transformer were assessed for
epidemiology: an coherence and relevance.
exploratory analysis of
ChatGPT using the Methods: The study involved posing questions derived from the STROBE checklist
STROBE checklist for to ChatGPT and assessing its answers in the context of conducting an
observational studies epidemiological study. The answers were evaluated by three independent senior
researchers.
 54

STROBE Guidelines: The STROBE guidelines were used prospectively in the

design and conduct of the study. These guidelines provide a framework for
transparent reporting of observational studies, ensuring the quality and
transparency of the research.

ChatGPT: ChatGPT is a third-generation, autoregressive language model by

OpenAI that generates human-like text. It was used to provide support to the
researchers conducting the epidemiological study.

Use of ChatGPT: ChatGPT was utilized by formulating questions from the

STROBE recommendations and contextualizing them with a real epidemiological
study. The STROBE recommendations were transformed into specific questions for
ChatGPT, which then provided answers.

Expert Assessment: The outputs of the transformer were assessed through "human
evaluation," where annotators were asked to rate the responses produced by the
model on a 1 to 5 points Likert scale, taking into account the coherence and
relevance of the responses.

Results: From the 35 STROBE recommendations, 35 specific questions were

derived for ChatGPT. Three domain experts evaluated the answers. The overall
mean coherence score was 3.6 out of 5.0, and for relevance, it was 3.3 out of 5.0.
Some responses received high scores for coherence and relevance, while others
scored poorly.
Sarink et al 2023 The Letter to the Peer A Study on the An AI chatbot called ChatGPT employs deep learning to produce discourse that
Netherlands editor reviewed Performance of sounds human. By contrasting its performance with the recommendations of
ChatGPT in Infectious clinical microbiologists (CM) or ID specialists in 40 clinical consultations at a tertiary
Diseases Clinical hospital in the Netherlands, its use in the field of infectious disease (ID) was
Consultation investigated.

In the study, the output of ChatGPT was compared to the diagnosis or treatment
recommendations made by the ID or CM specialists in the identical clinical settings.
On a scale of 1 (bad, wrong advise) to 5 (outstanding, entirely correlating with the
specialist's recommendations), the AI's output was evaluated.

The ChatGPT had a mean score of 2.8 and a standard deviation of 1.5. It
performed best when interpreting positive blood culture results (mean 3.3, SD 1.3)
and poorly when providing guidance on osteomyelitis or infections of prosthetic
joints (mean 1.3, SD 0.5). These scores' differences were statistically significant.

ChatGPT occasionally supplied unclear treatment options and didn't always take
into account given data. It occasionally made incorrect claims about the origin of its
guidance, which raises questions about its veracity and accuracy.
 55

ChatGPT's recommendations fully matched those of the experts in a few difficult

situations, suggesting that it may be helpful in difficult circumstances.

The study's limitations include the exclusion of chronic cases and the possibility of
differences between the data provided to ChatGPT and the reference standard
clinical consultation.

The study comes to the conclusion that while ChatGPT can generate moderately
accurate diagnostic or therapeutic suggestions, it cannot take the role of clinical
expertise. Clinicians are required to direct and interpret ChatGPT's
recommendations.
Schulte 2023 USA Original Peer Capacity of ChatGPT to This study aimed to evaluate the capability of ChatGPT in suggesting suitable
article reviewed Identify Guideline- systemic treatments for new cases of advanced solid malignancies, in comparison
Based Treatments for with the recommendations found in the National Comprehensive Cancer Network
Advanced Solid Tumors (NCCN) guidelines.

Researchers used standardized prompts to assess ChatGPT's ability and

introduced a measurement called the valid therapy quotient (VTQ), which is a ratio
of medications listed by ChatGPT to those recommended by the NCCN guidelines.

The study used 51 different diagnoses and ChatGPT was able to identify 91
different medications for the treatment of advanced solid tumors.

The overall VTQ was found to be 0.77, indicating a high level of concordance with
the NCCN guidelines.

For each case, ChatGPT was able to provide at least one systemic therapy
example that matched the NCCN's suggestions.

There was a weak correlation identified between the incidence of each malignancy
and the VTQ.

The study concluded that ChatGPT showed significant ability in recognizing and
suggesting medications used in the treatment of advanced solid tumors according
to the NCCN guidelines.

However, the practical usefulness of ChatGPT in assisting oncologists and patients

in deciding the treatment remains uncertain.

The study anticipates that future versions of ChatGPT will have improved accuracy
and consistency in this area and emphasizes the need for further studies to
quantify its capabilities more effectively.
 56

Singh 2023 India Commentary Peer Artificial intelligence in OpenAI launched ChatGPT, an AI-based chatbot, on November 30, 2022, and it
reviewed the era of ChatGPT - has already amassed over 100 million users. Numerous other AI applications are
Opportunities and on the horizon, including Google Bard, Microsoft Bing AI, Chinese Ernie bot,
challenges in mental Korean SearchGPT, Russian YaLM 2.0, Chatsonic, Jasper Chat, Character AI,
health care Perplexity AI, and YouChat.

AI platforms like ChatGPT are recognized for their vast potential across various
domains, including mental health. They are being utilized for chatting, gaming,
computer programming, and even generating scientific papers and abstracts.

There exists a substantial treatment gap in mental health care, particularly in

developing and lower-income countries, where 76%-85% of mental disorders
remain untreated. AI interfaces such as ChatGPT are viewed as potential tools to
bridge this gap, providing companionship, support, and therapy for individuals with
limited accessibility and affordability in mental health care.

However, while AI chatbots like ChatGPT have immense promise, there are
concerns regarding their use in mental health. These concerns include the potential
for providing incorrect information or advice, issues related to confidentiality,
privacy, and data security, as well as challenges in accurately diagnosing mental
health conditions.

Other concerns involve the lack of standardization, the possibility of misdiagnosis,

the provision of inappropriate advice, and limitations in handling crises. These
ethical concerns surrounding the use of AI-based applications in mental health are
significant.

In response to these concerns, organizations like the American Psychiatric

Association (APA) are establishing task forces to monitor and regulate AI-based
mental health apps. The APA has developed an App Evaluation Model known as
App Advisor, which is being adopted by other healthcare organizations for the
assessment of health apps.

Given the significant variations in awareness, education, language, and

understanding within the Indian population, the Indian Psychiatric Society and other
stakeholders should begin evaluating and regulating AI-based global and local
apps for their safety, efficacy, and tolerability. They should also provide guidance to
the general public regarding proper and safe usage of these apps.
Singh et al 2023 India Original Peer ChatGPT and he study examined the ability of the OpenAI model, ChatGPT, to generate
article reviewed Ophthalmology: operative notes and discharge summaries across various ophthalmic surgical
Exploring Its Potential specialties, maintaining the tenets of the Declaration of Helsinki. The surgical
with Discharge specialties included cataract surgery, penetrating keratoplasty, LASIK,
Summaries and trabeculectomy, pars plana vitrectomy, scleral buckle, pneumatic retinopexy,
 57

Operative Notes intraocular injections, open-globe injury repair, extraocular muscle surgery, Levator
resection for ptosis, and dacryocystorhinostomy. The AI model's outputs were
evaluated for their evidence-based content, specificity, factual accuracy, and ability
to recognize mistakes.

Key findings from the study include:

The AI model could quickly generate detailed discharge summaries and operative
notes in response to provided prompts, with responses usually returned in less
than 20 seconds.

The generated summaries contained no disclaimers, indicating confidence in the

provided information.

ChatGPT showed a capability to incorporate specific information such as

medications, follow-up instructions, and consultation times into the discharge
summaries when prompted.

While the AI model's summaries were generally accurate, they contained some
generic text, particularly in the 'discharge instructions' section.

The operative notes were divided into several sections, including pre- and post-
operative diagnosis, procedure, anesthesia, indication, procedure details, post-
operative care, estimated blood loss, complications, and summary.

ChatGPT demonstrated a capacity to learn from its mistakes. When confronted

with a potential error regarding the administration of sedation prior to intraocular
injection of Ranibizumab, the AI model immediately apologized for the error and
correctly generated the procedure without sedation when re-prompted.

Despite some shortcomings, the authors noted that the model's responses could be
customized to the user's needs with further training. This study demonstrates the
potential of AI in assisting with documentation tasks in the field of ophthalmology.
Tang et al 2023 USA Review Preprint Evaluating Large This extensive study assesses the performance of large language models (LLMs)
Language Models on in medical evidence summarization using Cochrane Reviews across six clinical
Medical Evidence domains, including heart failure, kidney illness, esophageal cancer, Alzheimer's
Summarization disease, and neurological problems.

The study compares the zero-shot performance of two models: GPT-3.5 and
ChatGPT. These models were given different parts of the review abstracts to
summarize. The models' output was assessed on several metrics including
ROUGE-L, METEOR, BLEU, and manual human evaluation for factors like
coherence, factual consistency, comprehensiveness, and harmfulness.
 58

Findings indicate that all models perform similarly on automatic metrics with higher
ROUGE scores showing that key information from the source document was
effectively captured. However, the models were found to be more extractive
compared to human-written summaries, indicating a lower level of abstraction.

Human tests revealed that ChatGPT's summaries were more cohesive than
GPT3.5's. It is noteworthy that less than 10% of the summaries created by
ChatGPT-MainResult included factual errors. In terms of thoroughness, ChatGPT-
MainResult and ChatGPT-Abstract both delivered more than 75% of the time, with
the former also producing the fewest summaries that were medically dangerous.

According to the survey, ChatGPT-MainResult performed much better than its

competitors, delivering the most preferred summaries about half of the time among
the three LLM setups. It was chosen because it produced the most thorough
summary and contained more important information.
Temsah et al 2023 Saudi Arabia Original Peer Overview of Early Introduction: OpenAI's ChatGPT has emerged as a powerful AI model capable of
article reviewed ChatGPT's Presence in generating human-like text in response to user queries, with applications spanning
Medical Literature: various domains, including medical education.
Insights From a Hybrid
Literature Review by Methodology: This review utilizes a hybrid narrative methodology, combining
ChatGPT and Human traditional narrative review methods with the assistance of ChatGPT to analyze and
Experts synthesize abstracts.

Inclusion and Exclusion Criteria: Articles discussing ChatGPT within the context of
medical education, literature, or practice were included. Non-medical or unrelated
contexts, non-English sources, and articles published outside the specified date
range were excluded.

Results: A total of 65 papers were identified in PubMed and 110 papers in Europe
PMC. These papers highlighted the utilization of ChatGPT in diverse medical fields
such as medical education, scientific writing, research, and diagnostic decision-
making.

Key Findings: The review identified eight main themes: (1) medical writing, (2)
medical education, (3) diagnostic decision-making, (4) public health, (5) scientific
research, (6) ethical considerations of ChatGPT use, (7) ChatGPT's potential to
automate medical tasks, and (8) criticism of ChatGPT's usage.

ChatGPT in Medical Writing: ChatGPT demonstrates the ability to generate

coherent and readable content, but concerns arise regarding its accuracy and the
ethical implications it introduces.
 59

ChatGPT in Medical Education: ChatGPT has the potential to enhance learning,

interpretation, and recall of medical information, but concerns persist regarding the
undermining of clinical reasoning and the lack of context or tailoring to individual
learning needs.

ChatGPT in Diagnostic Decision-Making: ChatGPT can improve efficiency and

reduce errors in diagnostic decision-making, but concerns exist regarding
accuracy, potential bias, and the need for human oversight.

ChatGPT in Public Health: ChatGPT can provide reliable estimates in public health
research, provided the input data is valid and accurate.

ChatGPT in Science and Research: It can streamline scientific writing processes

and assist with literature reviews, data interpretation, and hypothesis generation.
However, concerns are raised regarding its accuracy and potential to undermine
human expertise.
Ethical Considerations of ChatGPT: It raises concerns about authorship,
accountability, and transparency, along with the potential for generating misleading
or inaccurate information.

ChatGPT's Potential to Automate Medical Tasks: It can potentially improve

efficiency and speed of programming but undermines the value of human
expertise.

Criticism of ChatGPT Usage: The model's limitations, such as the production of

erroneous content, susceptibility to bias, and need for stringent regulatory
safeguards, are discussed.

Overall, this review highlights the transformative potential of ChatGPT.

Thorp 2023 USA Commentary Peer ChatGPT is fun, but not ChatGPT, developed by OpenAI, has become a cultural sensation within two
reviewed an author months of its release, providing endless entertainment by creating text based on
written prompts.

Despite its popularity, the AI has been noted to sometimes write plausible but
incorrect or nonsensical answers. This includes referencing non-existent scientific
studies.

Concerns have been raised about the impact of ChatGPT on education, as it can
write essays on a range of topics. While the program can generate factual answers,
the quality of scholarly writing is currently lacking.

The emergence of AI tools like ChatGPT is prompting academics to rethink their

course designs and assignments to ensure they aren't easily solvable by AI.
 60

Concerns also exist regarding the influence of AI on scientific paper writing. In a

study, only 63% of abstracts created by ChatGPT were identified as AI-generated
by academic reviewers.

Science journals are now updating their licenses and editorial policies to explicitly
state that text generated by AI tools, such as ChatGPT, cannot be used in their
work. This includes figures, images, or graphics produced by AI tools.

Violating these policies will be considered scientific misconduct, akin to plagiarism

or altering images.

The journals recognize that most instances of scientific misconduct occur due to
lack of adequate human attention, hence the need for scientists to be vigilant in
their work.

The role of AI should ultimately be as a tool to aid people in posing hypotheses,

designing experiments, and making sense of results in the scientific process.
Haq et al 2023 Pakistan Original Preprint Comparing human and The study aimed to investigate the suitability of ChatGPT for scientific writing by
article artificial intelligence in comparing short articles written by human authors and ChatGPT on three different
writing for health topics.
journals: an exploratory
study The research included criteria for testing ChatGPT's use in scientific writing and
developed evaluation criteria to assess the quality of articles written by human
authors and ChatGPT.

The evaluation criteria were adapted from the EASE guidelines, focusing on three
major areas: structure, scientific content, and credibility.

Grammarly, another AI-based tool, was used to score two items: originality and
readability.

The team selected three diverse topics for human authors and ChatGPT to write
about, shared the output as blinded versions, and evaluators scored the articles
independently.

Results showed that while human-authored articles scored perfectly, ChatGPT

articles lacked in areas like structure and organization, did not follow the IMRaD
(Introduction, Methods, Results, and Discussion) structure and lacked a logical
flow.

In terms of credibility, human-authored articles again scored a perfect five, but

ChatGPT articles had modest scores. Citations in ChatGPT articles were minimal
 61

and often pointed to nonexistent resources.

On scientific content, both types of articles scored well on originality, but ChatGPT
articles were weaker in specificity of response, lack of numerical data,
cohesiveness, and inclusion of study limitations.

Efficiency was the only area where ChatGPT scored perfectly, taking significantly
less time to produce the articles compared to human authors.

The study concluded that while ChatGPT can efficiently produce original and
seemingly coherent articles, they lack in-depth scientific basis, lack reproducibility,
and might quote non-existent scholarly work. This raises questions about the
accuracy and reliability of ChatGPT in the context of scientific writing.
Uprety et al 2023 USA Review Peer ChatGPT-A promising ChatGPT is a transformative AI model developed by OpenAI, offering
article reviewed generative AI tool and sophisticated, human-like text-based conversations. With the release of GPT-4,
its implications for ChatGPT Plus has seen widespread interest and potential applications, particularly
cancer care within the healthcare sector.

ChatGPT can be utilized in the medical field for extracting crucial information from
patient records, saving physicians time and enhancing efficiency. This is
particularly beneficial in areas such as oncology, where patients often have
extensive and complicated histories.

The AI can be used for administrative tasks, such as creating insurance letters for
evidence-based authorization of therapies.

ChatGPT could aid in the rapidly evolving field of oncology by staying updated on
recent advances and approvals, aiding in the interpretation of complex Next-
Generation Sequencing (NGS) reports, and making appropriate treatment
recommendations based on the correlation between mutations (biomarkers) and
treatment drugs.

The AI can support oncologists in making suitable clinical trial recommendations for
patients by interpreting their detailed medical history and the range of content
available on ClinicalTrials.gov.

Despite its vast potential, ChatGPT currently has some limitations. It's not
compliant with the Health Insurance Portability and Accountability Act (HIPAA),
posing a risk of violating patient privacy. Additionally, the AI could unintentionally
propagate biases and misinformation inherent in the internet text it was trained on,
and it is vulnerable to adversaries and manipulation.

ChatGPT plugins could enhance cancer care by efficiently summarizing medical

 62

documentation, interpreting NGS testing reports, and suggesting relevant clinical

trials for patients. However, it needs to overcome its current limitations related to
HIPAA compliance, internet bias, and vulnerability to adversarial prompting.

The authors anticipate that future iterations of ChatGPT may overcome these
limitations and become a highly valuable tool for healthcare, particularly in the field
of oncology.
Uz and Umay Turkey. Original Peer "Dr ChatGPT": Is it a The objective of this study was to utilize Google Trends to identify the most
2023 article reviewed reliable and useful frequently searched terms related to common rheumatic disorders (OA, RA, FMS,
source for common AS, SLE, gout, and PSA). Based on the search engine results, keywords were
rheumatic diseases? determined and categorized for each condition. Likert-type ratings were employed
to evaluate the accuracy and value of ChatGPT's responses to these keywords.

The most commonly searched keywords related to OA included "knee

osteoarthritis," "osteoarthritis pain," "osteoarthritis hip," and "osteoarthritis
treatment." For RA, popular terms were "rheumatoid arthritis symptoms,"
"rheumatoid arthritis pain," "rheumatoid arthritis causes," and "rheumatoid arthritis
treatment."

AS-related keywords encompassed "ankylosing spondylitis pain," "ankylosing

spondylitis symptoms," "ankylosing spondylitis test," and "ankylosing spondylitis
treatment." SLE was associated with keywords such as "lupus disease," "lupus
symptoms," "lupus causes," and "lupus treatment." PSA keywords included
"psoriatic arthritis symptoms," "psoriatic arthritis pain," "psoriatic arthritis causes,"
and "psoriatic arthritis treatment." FMS was linked to terms like "what is
fibromyalgia," "fibromyalgia pain," "fibromyalgia symptoms," and "fibromyalgia
treatment."

Using Likert-type scales, the study evaluated the accuracy and value of ChatGPT's
responses for each keyword. The reliability and usefulness scores demonstrated
good to very good agreement, as indicated by Cronbach values and inter-rater
reliability scores.

Overall, the dependability and usefulness scores ranged from 4 to 7. AS received

the highest usefulness score, while OA received the highest reliability score. Some
topics received lower marks in terms of usefulness and reliability. However, there
were no significant discrepancies in the ratings for overall dependability and
usefulness among the raters.

This study comprehensively examined the accuracy and value of ChatGPT's

responses to the most popular search terms related to common rheumatic
conditions. The findings provide insights into the precision and applicability of AI-
generated knowledge for individuals seeking information about these disorders.
 63

Van Dis et al 2023 the Commentary Peer- ChatGPT: five priorities Large Language Models (LLMs) like ChatGPT have the potential to revolutionize
Netherlands reviewed for research the scientific research field, offering assistance in various tasks including writing
essays, summarizing literature, and identifying research gaps.

However, there is controversy around this technology due to its capacity to produce
convincingly wrong information, potentially distorting scientific facts and spreading
misinformation.

LLMs can introduce inaccuracies, bias, and plagiarism. ChatGPT, for instance, has
been found to generate false and misleading responses when asked complex,
nuanced scientific questions.

Like humans, current AI models are susceptible to biases, such as availability,

selection, and confirmation biases, which they can inadvertently propagate.

Researchers using AI like ChatGPT risk being misled by false or biased information
and incorporating it into their work. Unintentional plagiarism is also a concern as AI
reproduces text without reliable citations.

The researchers suggest the necessity of fact-checking and verification processes,

even with AI assistance, to maintain the integrity of scientific practice.

Accountability rules for AI usage in research are proposed, including transparency

in acknowledging AI assistance in papers and the development of policies to
handle the use of AI in manuscript preparation.

There are emerging questions about authorship and rights to texts created with AI
assistance, indicating the need for updated definitions and legal considerations.

The proprietary nature of AI technologies, mostly controlled by a few large tech

companies, raises ethical concerns and hampers transparency. Open-source,
independent non-profit projects for AI development are recommended.

Despite the challenges, the potential benefits of AI in reducing workload, speeding

up publication, and potentially accelerating innovation are recognized. Investment
in the validity and reliability of LLMs for effective use in scientific research is
suggested.

AI could change the academic skill set, optimizing training, reducing the need for
certain skills, and introducing new ones. As AI advances, it might handle more
complex tasks, prompting the need for careful evaluation of AI acceleration vs loss
of human potential.
 64

The research community is urged to engage in a wide-ranging debate on the use

and implications of LLMs in research, with focus on topics like essential academic
skills, steps requiring human verification, and ensuring LLMs promote equity in
research.

Every research group should discuss the use and implications of LLMs, and
educators should introduce discussions about LLM usage and ethics to students.
Accountability for research work will apply regardless of whether it's generated with
ChatGPT or not.

Waisberg et al Ireland Letter to the Peer GPT
4: a new era of A sizable language model called ChatGPT was created by OpenAI. To anticipate
2023 editor reviewed artificial intelligence in words or phrases in context, it employs a deep neural network based on the
medicine Transformer architecture.

The model is trained to anticipate words in a sentence or phrase depending on

words that came before them. Recursively repeating the technique until a whole
sentence or paragraph is generated will accomplish this.

ChatGPT recognises links between words and sentences in natural language to

deliver logical and pertinent responses to user inputs.

The AI processes user input, analyses it, and then responds in accordance with the
patterns it identified during training. This response is intended to be situationally
suitable and can cover a wide range of issues.

Banking, gaming, and healthcare are just a few of the industries where artificial
intelligence is being used. It might help with patient diagnosis, treatment
alternatives, and individualised care plans, for instance, in the healthcare industry.

The AI was instructed to write an easy-to-read discharge narrative for a patient who
had undergone a simple cholecystectomy. Important recommendations on exercise
limitations, dietary modifications, and when to seek medical assistance were all
included in the summary.
The AI was also asked to provide information about the latest clinical trials to treat
interstitial lung disease (ILD). It listed four ongoing phase 2 clinical trials and
provided brief details about each one.

Furthermore, the AI was asked to provide the latest guidelines for AI in medicine. It
listed guidelines from the European Commission, the American Medical
Association, the International Medical Informatics Association, and the National
Institute for Health and Care Excellence, all of which were relevant.
 65

In contrast, the AI's newly implemented image analysis feature demonstrated a lack
of accuracy in identifying fundus photographs of anterior ischemic optic neuropathy
and non-arteritic anterior ischemic optic neuropathy. Instead, it erroneously
classified these images as a schematic representation of the CRISPR-Cas system.

Despite the inaccurate image analysis, GPT-4 is considered an improvement over

its predecessor, with enhanced problem-solving abilities, an expanded knowledge
base, and the ability to process eight times more words. It's also more difficult to
deceive and is less likely to respond to inappropriate requests.
Wen and Wang Australia Commentary Peer- The future of ChatGPT ChatGPT in Medical Research: ChatGPT, a large AI model developed by OpenAI,
2023 reviewed in academic research is being increasingly used in various occupations including clinical and translational
and publishing: A medicine. It has entered the scientific literature with published papers and
commentary for clinical preprints, however, its limitations in medical research, particularly in providing
and translational accurate and up-to-date information, are causing concern.
medicine
Factual Inaccuracies of ChatGPT: The biggest disadvantage of ChatGPT is its
potential to compile inaccurate information. It uses data up until 2021, and does not
consider information after that, which can be a hindrance in fields like medicine
where knowledge and advances are constantly evolving.

ChatGPT: Difficult to Detect: Researchers at Northwestern University conducted a

study where ChatGPT was asked to write 50 medical research abstracts. Medical
researchers had difficulty distinguishing between AI-generated and human-written
abstracts, raising concerns about the potential consequences of relying on
potentially inaccurate research.

ChatGPT as an Aide for Scientific Innovation: Despite its limitations, ChatGPT can
be used beneficially in scientific research to improve completed research papers. It
can increase researchers' productivity, save time, and improve the quality of their
content.

Pros and Cons of ChatGPT in Scientific Research and Publishing: While ChatGPT
can be useful for certain tasks like editing, it is not reliable for providing accurate
facts or references. Using it judiciously while being aware of its limitations can help
researchers streamline their work without risking the publication of false
information.

ChatGPT and Clinical and Translational Medicine: Implementing ChatGPT must be

pursued cautiously due to its limitations in providing reliable information. It can be
useful for tasks like proofreading and manuscript checks, but cannot replace the
specialized knowledge, innovative ideas, and opinions that human input provides.
Further exploration of research ethics and the moral use of AI in health research is
required to establish guidelines.
 66

Xue et al 2023 China Commentary Peer- The potential impact of ChatGPT is an AI model developed by OpenAI that simulates human interaction
reviewed ChatGPT in clinical and and uses a deep learning technique called 'transformer architecture'. It's trained on
translational medicine large datasets of text from the internet.

It is still unclear how ChatGPT will affect clinical medicine. However, similar AI
models could help with patient education, clinical trial recruiting, clinical data
management, research support, clinical decision support, and the automation of
activities like data processing and picture interpretation.

ChatGPT could also assist in drug discovery by recognizing, classifying, and

describing chemical formulas or molecular structures. It can also help in disease
prediction, diagnosis, and assessment of therapeutic targets.

ChatGPT has restrictions despite its advantages. It occasionally gives general or

ambiguous responses in conversations about medicine since it is unable to update
its training data in real-time. ChatGPT's diagnoses could not be thorough or
sufficient, showing that it is still unable to manage the challenging tasks of clinical
practise on its own.

It has been claimed that emphasising ChatGPT's use in human-computer

interaction could increase its usefulness in clinical settings. By speeding data
gathering, recording, and analysis, it could assist with patient questionnaires and
interviews in epidemiological research as well as mental health care.
ChatGPT has already been used in medical education, training, and writing, though
ethical issues around its use have arisen, particularly regarding its authorship of
academic articles.

AI models, including ChatGPT, can aid healthcare by providing a more objective,

evidence-based approach to decision-making and by helping to identify patterns
and correlations in large datasets. They can also assist in disease detection and
prognosis prediction.

AI may have unfavourable effects on society, such as privacy problems and bias or
discrimination. To ensure the responsible and effective use of AI in healthcare,
more research and development is required.

Even so, ChatGPT and other forms of AI will continue to advance. The best course
of action is to embrace it, making use of its potential to advance clinical practise
while simultaneously addressing any potential drawbacks.
Yadava 2023 India Editorial Peer- ChatGPT—a foe or an ChatGPT, an AI platform developed by OpenAI, employs natural language
reviewed ally? processing technology, using advanced machine learning paradigms such as
transfer learning, supervised learning, and reinforced learning.
 67

The model has the capability to generate articulate and meaningful content across
various fields of knowledge, with outputs including music, plays, poetry, and song
lyrics.

Despite its advanced abilities, there are concerns within academia about the impact
of non-human authors on the integrity and validity of scientific publications. This
could potentially lead to an increase in academic plagiarism.

Some renowned scientific journals such as 'Nature' and 'JAMA Network Science'
have decided not to accept articles generated by ChatGPT, demanding full
disclosure of its use.

The technology has also sparked worries among non-medical intelligentsia and
politicians, with fears about potential job losses, discrimination, and uncontrollable
military applications.

Historical examples show that disruptive technologies often face resistance upon
introduction. However, despite initial skepticism, many have gone on to have
profound impacts on society and industry.

Some academic journals and researchers have already adopted the technology,
recognizing its potential benefits for medical writing and document creation.

The technology is expected to streamline scientific writing and administrative tasks,

improving efficiency in clinical roles.

The debate on the implications of AI like ChatGPT is ongoing. It could turn out to
be a dual-use technology, where the ethics and intent behind its use determine its
value.

Regardless of the pros and cons, AI-powered language models are becoming a
reality, with human oversight remaining crucial.

The adoption of AI technology is seen as inevitable, and it is essential to accept,

adapt, and leverage it for the future.
Yeung et al 2023 UK Commentary Peer- AI chatbots not yet A Cambrian explosion of natural language processing (NLP) models,
reviewed ready for clinical use predominantly based on the transformer model, occurred in 2022, offering
unprecedented advancements in natural language generation capabilities and
opening up the potential for their application in healthcare as AI chatbots.

Large Language Models (LLMs), like OpenAI’s GPT3 and Google’s PALM, Gopher,
and Chinchilla, have evolved to exhibit emergent properties: performing tasks they
 68

were not explicitly trained on. This could be due to the models' ability to extract
more knowledge from vast amounts of text data used in their training.

LLMs have shown promise in the healthcare field, exhibiting human-level

performances in medical question answering and summarisation tasks. However,
safety and accuracy should be prioritised over human-like interactivity.

Current LLMs face challenges such as mirroring biases and inaccuracies from their
training data, raising concerns about their suitability for critical fields like healthcare.
Moreover, LLMs trained on biomedical corpora, while being domain-specific, don't
necessarily reflect the realities of actual patients and diseases in healthcare.

Some LLMs, such as GatorTron and BEHRT, are trained on de-identified clinical
notes or disease classification codes to circumvent the problem of sensitive patient
data.

Current medical AI benchmarks do not adequately capture the complexity of real-

world clinical practice. In actual practice, healthcare professionals generate a list of
potential differential diagnoses and then proceed with investigations to gradually
narrow down the possibilities, rather than simply identifying a single correct
diagnosis.

In a comparative study between ChatGPT and Foresight GPT (trained on real-

world hospital data), both models performed well in predicting the 5 most likely
diagnoses from synthetic clinical histories. However, ChatGPT, which is not
domain-specific, often omitted crucial diagnoses.

LLMs can exhibit biases and associations based on their training text, leading to
potential issues of racial-ethnic disparities in treatment recommendations.

LLMs can also generate "hallucinations" or false information, especially when faced
with insufficient or misleading information in the prompt.

Although LLMs have made encouraging strides in natural language processing and
creation, it is still unclear whether they are ready to be used as clinical tools that
patients will see. Focus should be placed on domain-specific training data, expert
clinician fine-tuning, and transparent depiction of output relevancy versus safety
implications in order to use AI chatbots in healthcare in a safe manner.
Young et al 2023 USA Research Peer The utility of ChatGPT The study aimed to assess the effectiveness of OpenAI's ChatGPT, a large
Letter reviewed in generating patient- language model (LLM), as an informational platform for patients seeking
facing and clinical information about melanoma care. To evaluate its suitability, twenty-five
responses for hypothetical patient questions related to melanoma were presented to three board-
melanoma certified dermatologists. The dermatologists assessed the appropriateness of the
 69

responses for a patient-facing informational platform, their sufficiency for clinical

practice, accuracy, and comprehensibility.

The accuracy of ChatGPT's responses consistently received high ratings, with an

average score of 4.88 out of 5. Furthermore, 92% of the responses were
considered appropriate for a patient-facing informational platform.

However, the study found that only 64% of the responses were deemed sufficient
for clinical practice. This was mainly due to important details being omitted in
ChatGPT's advice. For instance, while recommending "regular skin exams" is
accurate, it lacks the specificity of a physician's recommendation for exams every
few months.

Another limitation highlighted by the study was the average readability of

ChatGPT's responses, as measured by the Flesch Reading Ease Score (FRES).
The FRES indicated a college-level comprehension requirement, which is too
advanced for the general public. Health-related materials are typically
recommended to be at a 5th-6th grade reading level.

In conclusion, the study acknowledges that ChatGPT can generate accurate

responses suitable for patient-facing tools. However, it emphasizes significant
limitations, particularly in terms of comprehensibility and sufficiency for clinical
practice. The authors stress the importance of dermatology providers
understanding the potential uses and limitations of AI tools like ChatGPT to
effectively counsel patients as these technologies become more prevalent in
clinical practice.
Zheng and Zhan USA Commentary Peer ChatGPT in Scientific ChatGPT is an AI chatbot that has garnered worldwide excitement and concerns,
2023 reviewed Writing: A Cautionary particularly in the scientific community due to its use in scientific writing and
Tale publishing. The concerns are mainly related to copyright, attribution, plagiarism,
authorship, and the accuracy of the content generated by ChatGPT.

The authors conducted an evaluation of ChatGPT's accuracy in generating content

by using an article on Body Surface Area (BSA) formulas that was not included in
its training data. They summarized key facts from the article and used them as
prompts to ask ChatGPT questions five times, assessing the responses for
accuracy.

While ChatGPT generates responses that are well-written and sound plausible, it
has been observed that these responses often contain significant errors and
fabricated information. Consequently, ChatGPT cannot be considered a reliable
and trustworthy source for scientific writing.

Given the potential for falsifications and fabrications in ChatGPT's output, the study
 70

concludes that there is no clear advantage to using ChatGPT for scientific writing.
Authors utilizing ChatGPT must take the responsibility to manually verify all facts,
statements, and references generated by the AI.

Upholding the highest ethical standards in scientific research is paramount, and

proper data management is essential in maintaining these standards. Poor data
management practices, including the presence of fabricated or falsified findings,
can have serious consequences.

It is important to note that the authors listed on an article are ultimately responsible
for ensuring accuracy and integrity, not ChatGPT itself.
Detecting fabrication or falsification during the peer-review process of manuscripts
containing text generated by ChatGPT is challenging. Studies have shown that
human reviewers can overlook up to 32% of fully fabricated abstracts produced by
ChatGPT.

Directly adopting full text written by ChatGPT may constitute plagiarism and violate
the code of conduct in scientific publishing.

Scientific journals that accept articles involving ChatGPT in the writing process may
face a significant increase in retractions and loss of credibility.

While ChatGPT offers great potential for the future, its current state is not mature
enough for scientific writing. The role of a more advanced ChatGPT in scientific
writing necessitates comprehensive discussions and debates.

Considering the potential for fabricated and inaccurate information, the Science
family of journals has implemented a ban on all ChatGPT-generated content. It is
recommended that this policy becomes standard practice for all scientific
publishing.

Despite its limitations, ChatGPT can still be a useful tool for checking grammar and
syntax errors, as well as refining language, particularly for non-native speakers.
Zhong et al 2023 China Commentary Peer The Artificial Artificial intelligence (AI) can significantly enhance psychiatric research and
reviewed intelligence large practice by improving diagnostic accuracy, optimizing treatment outcomes, and
language models and offering personalized care. AI can analyze extensive patient data, recognize
neuropsychiatry complex patterns, and suggest individualized treatment strategies.
practice and research
ethic AI large language models (LLMs) like ChatGPT, despite their potential benefits,
raise concerns such as reliability and accuracy, transparency, accountability, and
ethical implications. These models can sometimes generate plausible yet incorrect
or nonsensical information, leading to doubts about their scientific reliability.
 71

The transparency of AI LLMs is a challenge due to the complex algorithms and

large datasets they employ. The opaque nature of these models could potentially
undermine trust and credibility in scientific research where transparency is vital.

Accountability is another issue with AI LLMs. They could perpetuate biases present
in the training datasets, and this could lead to biased outputs. For instance, if the AI
model is trained on racially or ethnically biased data, its outputs may reflect these
biases.

Ethical concerns with AI in research include the potential for these models to
generate misleading content. Some cases have reported AI-generated abstracts
that fooled academic reviewers. AI LLMs' increasing usage for tasks like data
analysis, literature reviews, grant proposals, etc., is stirring debate about whether
they should be acknowledged as authors in research papers.

There are concerns over intellectual property rights violations as AI models are
trained on vast datasets that may include copyrighted or proprietary content.
There's a risk that these models could generate content that infringes on these
rights.

The article suggests measures to address these concerns and regulate AI LLMs
use in science. These include insisting on human review of AI-generated content,
establishing accountability mechanisms, investing in open-source LLMs, and
promoting a broad discussion among stakeholders on the implications and
challenges of AI.

Despite the potential issues, AI can be transformative in psychiatric research and

practice. However, it's essential to address these concerns and ensure that AI's
use is ethical, protects patients, and maximizes its benefits.
Zhou et al 2023 China Letter to the Peer The Potential of Summary could not be generated.
editor reviewed Applying ChatGPT to
Extract Keywords of
Medical Literature in 5
Plastic Surgery
Zhou 2023 China Original Peer Evaluation of The study explores the potential of ChatGPT, an OpenAI language model, in
article reviewed ChatGPT's Capabilities supporting healthcare professionals by generating medical reports based on real
in Medical Report patient laboratory results.
Generation
The language model was employed in various medical domains, such as
interpreting lab results and analyzing medical literature, to improve and expedite
the process of writing medical reports.

In this particular case, ChatGPT produced a detailed medical report for a 31-year-
 72

old male patient who presented with abdominal pain and sought medical attention.

Based on routine laboratory tests, the model generated personalized

recommendations for the patient, including suggested lifestyle modifications and
potential medical treatment options.

Additionally, the model advised the patient to seek a gastroenterologist's

consultation for further assessment and consideration of advanced treatment plans.

The structure and organization of the case report were entirely generated by
ChatGPT, utilizing the patient's physical information and lab results as input.

To validate the accuracy and reliability of ChatGPT's recommendations, the

generated report will be compared with an online doctor consultation system.

The study aims to demonstrate that ChatGPT can consistently and accurately
produce coherent, comprehensive, and clinically relevant medical reports,
showcasing its potential in aiding healthcare professionals.
Zhu et al 2023 China Commentary Peer Can the ChatGPT and The researchers conducted an evaluation of multiple large language models
reviewed other Large Language (LLMs), including ChatGPT, YouChat, NeevaAI, Perplexity, and Chatsonic, to
Models with internet- assess their usefulness in providing accurate and comprehensive information about
connected database prostate cancer (PCa).
solve the questions and
concerns of patient with To evaluate the LLMs' performance, the researchers designed 22 questions based
prostate cancer? on patient education guidelines and their own clinical experience. They assessed
the LLMs in terms of accuracy, comprehensiveness, patient readability, humanistic
care, and stability.

Most of the LLMs exhibited accuracy rates above 90%, with ChatGPT
demonstrating the highest accuracy. Interestingly, the free version of ChatGPT
performed slightly better than the paid version.

The LLMs generally provided comprehensive responses, addressing different

aspects such as the significance of PSA levels, detailed treatment comparisons,
and advising patients to consult their doctors.

In terms of readability, most LLM responses were satisfactory and displayed

humanistic care, particularly when discussing PCa's relatively long survival time.
However, not all inquiries received the same level of humanistic care.

One issue identified was the inclusion of outdated or incorrect information in some
LLM responses. Examples include inaccurate comparisons between apalutamide
and enzalutamide, and claims that open surgery was more common than robot-
 73

assisted surgery for radical prostatectomy.

NeevaAI, in particular, tended to simply relay literature content without

summarizing or explaining, resulting in poor readability. The study also found that
real-time internet-connected LLMs did not outperform ChatGPT, which had limited
access to data. This suggests that model training may be more crucial than real-
time internet connection.

Despite their imperfections, LLMs show promise in providing accurate basic

information about PCa. They could potentially be applied in patient education and
consultation, facilitating shared decision-making and democratizing medical
knowledge.

However, the researchers caution against replacing doctors with LLMs at this
stage. LLMs may contain errors, omit important points, and struggle to analyze
specific contexts. They also lack the ability to ask follow-up questions for further
information or provide the same level of comfort as a human healthcare provider.
Zielinski et al 2023 UK Commentary Peer Chatbots, ChatGPT, Chatbots, including ChatGPT, are AI tools used in diverse fields, including
reviewed and Scholarly healthcare, customer service, and education. They can create new content by
Manuscripts: WAME processing and reorganizing existing information.
Recommendations on
ChatGPT and Chatbots Despite its potential, ChatGPT has several limitations. It can generate incorrect or
in Relation to Scholarly nonsensical answers and it may fail to ask clarifying questions. Its knowledge is
Publications restricted to what it learned before 2021.

OpenAI is working on improving ChatGPT. Other companies are also developing

similar generative AI tools.

Chatbots can only generate output based on their training data and they cannot
produce truly original thoughts. They may unintentionally plagiarize from their
training materials.

ChatGPT can produce false statements, though it does not possess the
intentionality to lie in the human sense.

As software tools, chatbots cannot be held legally responsible for their output.
Liability falls on the users.

Including ChatGPT as an author is controversial and potentially legally

indefensible.

ChatGPT represents a threat to scholarly journals due to potential introduction of

false or plagiarized content. Peer review may fail to identify AI-generated content.
 74

AI tools like DALL-E 2 and Imagen, which generate images, have similar concerns
to ChatGPT. The provenance of images should be clearly indicated.

Editors need to establish policies on the use of AI tools like ChatGPT and require
tools for detecting AI-generated content.

Recommendations include disqualifying chatbots from authorship, requiring

transparency when chatbots are used, ensuring authors take responsibility for AI-
generated content, and providing editors with AI-detection tools.
Zimmerman 2023 USA Commentary Peer A Ghostwriter for the Natural Language Processing (NLP) is a subfield of computer science that enables
reviewed Masses: ChatGPT and computers to understand, respond to, and mimic human language in text form. This
the Future technology has been instrumental in developing models such as ChatGPT.
of Writing
ChatGPT is a product of two-phase training: unsupervised pre-training on a large
dataset and supervised fine-tuning on smaller datasets. This results in an effective
model able to provide high-quality responses to diverse tasks and prompts.

NLP is already used in various applications like Siri, Alexa, Grammarly, and even in
automated journalism. In the medical field, NLP is used for tasks such as
information extraction, productivity improvement, and documentation.

ChatGPT, a product of OpenAI, has gained significant attention due to its ability to
generate human-like content, ranging from passing scores on the USMLE exam to
generating convincing medical abstracts.

While tools like Wordtune and Paperpal assist with sentence restructuring,
ChatGPT can help restructure entire manuscripts and provide feedback on
limitations.

ChatGPT is leading in its space, but it is not without competitors. Companies like
Google, Facebook, and Anthropic are working to create models with similar
capabilities.

Despite its impressive capabilities, ChatGPT has been met with criticism, including
the potential for generating confabulatory outputs (hallucinations) when it does not
have an appropriate response and potential bias.

Ethical considerations are also paramount when it comes to the originality of the
content produced by these tools, making transparency of use crucial.

Corporate investment in AI systems has been growing over the years, indicating a
growing interest and trust in the technology. However, the responsibility of using
 75

these tools ethically and responsibly lies with the user.

286
287
288
289
290 References
291
292 1. Ali MJ, Djalilian A. Readership Awareness Series - Paper 4: Chatbots and ChatGPT - Ethical Considerations in Scientific Publications.
293 Semin Ophthalmol. 2023 Mar 21:1-2. doi: 10.1080/08820538.2023.2193444. Epub ahead of print.
294
295 2. Ali SR, Dobbs TD, Hutchings HA, Whitaker IS. Using ChatGPT to write patient clinic letters. Lancet Digit Health. 2023 Apr;5(4):e179-e181.
296 doi: 10.1016/S2589-7500(23)00048-1. Epub 2023 Mar 7.
297
298 3. Alser M, Waisberg E. Concerns with the Usage of ChatGPT in Academia and Medicine: A Viewpoint. American Journal of Medicine Open
299 2023;9. https://doi.org/10.1016/j.ajmo.2023.100036.
300
301 4. Anderson N, Belavy DL, Perle SM, Hendricks S, Hespanhol L, Verhagen E, Memon AR. AI did not write this manuscript, or did it? Can we
302 trick the AI text detector into generated texts? The potential future of ChatGPT and AI in Sports & Exercise Medicine manuscript generation.
303 BMJ Open Sport Exerc Med. 2023 Feb 16;9(1):e001568. doi: 10.1136/bmjsem-2023-001568.
304
305 5. Arun Babu T, Sharmila V. Using artificial intelligence chatbots like 'ChatGPT' to draft articles for medical journals - Advantages, limitations,
306 ethical concerns and way forward. Eur J Obstet Gynecol Reprod Biol. 2023 May 16:S0301-2115(23)00187-2. doi:
307 10.1016/j.ejogrb.2023.05.008. Epub ahead of print.
308
309 6. Asch DA. An Interview with ChatGPT About Health Care. NEJM Catalyst Innovations in Care Delivery 2023;2.
310 https://catalyst.nejm.org/doi/abs/10.1056/CAT.23.0043.
311
312 7. Athaluri SA, Manthena SV, Kesapragada VSRKM, Yarlagadda V, Dave T, Duddumpudi RTS. Exploring the Boundaries of Reality:
313 Investigating the Phenomenon of Artificial Intelligence Hallucination in Scientific Writing Through ChatGPT References. Cureus. 2023 Apr
314 11;15(4):e37432. doi: 10.7759/cureus.37432.
315
316 8. Balas M, Ing EB. Conversational AI models for ophthalmic diagnosis: Comparison of ChatGPT and the isabel pro differential diagnosis
317 generator. JFO Open Ophthalmology. 2023 Mar 1;1:100005. https://doi.org/10.1016/j.jfop.2023.100005.
318
319 9. Barker FG, Rutka JT. Editorial. Generative artificial intelligence, chatbots, and the Journal of Neurosurgery Publishing Group. J Neurosurg.
320 2023 Apr 28:1-3. doi: 10.3171/2023.4.JNS23482. Epub ahead of print.
 76

321
322 10. Bauchner H. ChatGPT: Not An Author, But A Tool. Health Affairs Forefront. 2023. DOI: 10.1377/forefront.20230511.917632.
323
324 11. Baumgartner C. The potential impact of ChatGPT in clinical and translational medicine. Clin Transl Med. 2023 Mar;13(3):e1206. doi:
325 10.1002/ctm2.1206.
326
327 12. Benoit JR. ChatGPT for Clinical Vignette Generation, Revision, and Evaluation. medRxiv 2023.02.04.23285478; doi:
328 https://doi.org/10.1101/2023.02.04.23285478.
329
330 13. Bhattacharya K, Bhattacharya AS, Bhattacharya N, Yagnik VD, Garg P, Kumar S. ChatGPT in surgical practice—a New Kid on the Block.
331 Indian Journal of Surgery. 2023 Feb 22:1-4. https://doi.org/10.1007/s12262-023-03727-x.
332
333 14. Biswas S. ChatGPT and the Future of Medical Writing. Radiology. 2023 Apr;307(2):e223312. doi: 10.1148/radiol.223312.
334
335 15. Boßelmann CM, Leu C, Lal D. Are AI language models such as ChatGPT ready to improve the care of individuals with epilepsy? Epilepsia.
336 2023 May;64(5):1195-1199. doi: 10.1111/epi.17570.
337
338 16. Brainard J. Journals take up arms against AI-written text. Science. 2023 Feb 24;379(6634):740-741. doi: 10.1126/science.adh2762. Epub
339 2023 Feb 23.
340
341 17. Cahan P, Treutlein B. A conversation with ChatGPT on the role of computational systems biology in stem cell research. Stem Cell Reports.
342 2023 Jan 10;18(1):1-2. doi: 10.1016/j.stemcr.2022.12.009.
343
344 18. Nasrallah HA. A ‘guest editorial’… generated by ChatGPT? Current Psychiatry 2023:6. https://cdn.mdedge.com/files/s3fs-
345 public/CP02204006.pdf.
346
347 19. Cascella M, Montomoli J, Bellini V, Bignami E. Evaluating the Feasibility of ChatGPT in Healthcare: An Analysis of Multiple Clinical and
348 Research Scenarios. J Med Syst. 2023 Mar 4;47(1):33. doi: 10.1007/s10916-023-01925-4.
349
350 20. Chen S, Kann BH, Foote MB, Aerts HJ, Savova GK, Mak RH, Bitterman DS. The utility of ChatGPT for cancer treatment information
351 medRxiv 2023.03.16.23287316; doi: https://doi.org/10.1101/2023.03.16.23287316.
352
353 21. Cheng K, Wu H, Li C. ChatGPT/GPT-4: enabling a new era of surgical oncology. Int J Surg. 2023 May 16. doi:
354 10.1097/JS9.0000000000000451. Epub ahead of print.
355
 77

356 22. Chervenak J, Lieman H, Blanco-Breindel M, Jindal S. The promise and peril of using a large language model to obtain clinical information:
357 ChatGPT performs strongly as a fertility counseling tool with limitations. Fertil Steril. 2023 May 20:S0015-0282(23)00522-8. doi:
358 10.1016/j.fertnstert.2023.05.151. Epub ahead of print.
359
360 23. Cifarelli CP, Sheehan JP. Large language model artificial intelligence: the current state and future of ChatGPT in neuro-oncology
361 publishing. J Neurooncol. 2023 May 20. doi: 10.1007/s11060-023-04336-0. Epub ahead of print.
362
363 24. Corsello A, Santangelo A. May Artificial Intelligence Influence Future Pediatric Research?-The Case of ChatGPT. Children (Basel). 2023
364 Apr 21;10(4):757. doi: 10.3390/children10040757.
365
366 25. D'Amico RS, White TG, Shah HA, Langer DJ. I Asked a ChatGPT to Write an Editorial About How We Can Incorporate Chatbots Into
367 Neurosurgical Research and Patient Care…. Neurosurgery. 2023 Apr 1;92(4):663-664. doi: 10.1227/neu.0000000000002414. Epub 2023
368 Feb 9.
369
370 26. Darkhabani M, Alrifaai MA, Elsalti A, Dvir YM, Mahroum N. ChatGPT and autoimmunity - A new weapon in the battlefield of knowledge.
371 Autoimmun Rev. 2023 May 19:103360. doi: 10.1016/j.autrev.2023.103360. Epub ahead of print.
372
373 27. Dave, M. Plagiarism software now able to detect students using ChatGPT. Br Dent J 234, 642 (2023). https://doi.org/10.1038/s41415-023-
374 5868-8.
375
376 28. Dave T, Athaluri SA, Singh S. ChatGPT in medicine: an overview of its applications, advantages, limitations, future prospects, and ethical
377 considerations. Front Artif Intell. 2023 May 4;6:1169595. doi: 10.3389/frai.2023.1169595.
378
379 29. Day T. A Preliminary Investigation of Fake Peer-Reviewed Citations and References Generated by ChatGPT. The Professional
380 Geographer. 2023 Mar 23:1-4. DOI: 10.1080/00330124.2023.2190373.
381
382 30. de Oliveira RS, Ballestero M. The future of Pediatric Neurosurgery and ChatGPT: an editor's perspective. Archives of Pediatric
383 Neurosurgery 2023;5. DOI: https://doi.org/10.46900/apn.v5i2.191.
384
385 31. De Vito EL. Artificial intelligence and chatGPT. Would you read an artificial author? Medicina 2023;83:329-332.
386
387 32. Dergaa I, Chamari K, Zmijewski P, Ben Saad H. From human writing to artificial intelligence generated text: examining the prospects and
388 potential threats of ChatGPT in academic writing. Biol Sport 2023;40:615-622.
389
 78

390 33. Donato H, Escada P, Villanueva T. The Transparency of Science with ChatGPT and the Emerging Artificial Intelligence Language Models:
391 Where Should Medical Journals Stand? Acta Med Port. 2023 Mar 1;36(3):147-148. doi: 10.20344/amp.19694. Epub 2023 Feb 9.
392
393 34. Dunn C, Hunter J, Steffes W, Whitney Z, Foss M, Mammino J, Leavitt A, Hawkins SD, Dane A, Yungmann M, Nathoo R. Artificial
394 intelligence-derived dermatology case reports are indistinguishable from those written by humans: A single-blinded observer study. J Am
395 Acad Dermatol. 2023 Apr 11:S0190-9622(23)00587-X. doi: 10.1016/j.jaad.2023.04.005. Epub ahead of print.
396
397 35. Fatani B. ChatGPT for Future Medical and Dental Research. Cureus. 2023 Apr 8;15(4):e37285. doi: 10.7759/cureus.37285.
398
399 36. Galland J. Les chatbots en médecine interne : opportunités et défis à venir [Chatbots and internal medicine: Future opportunities and
400 challenges]. Rev Med Interne. 2023 May;44(5):209-211. French. doi: 10.1016/j.revmed.2023.04.001. Epub 2023 Apr 29.
401
402 37. Gandhi Periaysamy A, Satapathy P, Neyazi A, Padhi BK. ChatGPT: roles and boundaries of the new artificial intelligence tool in medical
403 education and health research - correspondence. Ann Med Surg (Lond) 2023;85:1317-1318.
404
405 38. Gao CA, Howard FM, Markov NS, Dyer EC, Ramesh S, Luo Y, Pearson AT. Comparing scientific abstracts generated by ChatGPT to
406 original abstracts using an artificial intelligence output detector, plagiarism detector, and blinded human reviewers. bioRxiv
407 2022.12.23.521610; doi: https://doi.org/10.1101/2022.12.23.521610.
408
409 39. Goedde D, Noehl S, Wolf C, Rupert Y, Rimkus L, Ehlers J, Breuckmann F, Sellmann T. ChatGPT in medical literature-a concise review and
410 SWOT analysis. medRxiv 2023.05.06.23289608; doi: https://doi.org/10.1101/2023.05.06.23289608.
411
412 40. Gordijn B, Have HT. ChatGPT: evolution or revolution? Med Health Care Philos. 2023 Mar;26(1):1-2. doi: 10.1007/s11019-023-10136-0.
413
414 41. Gottlieb M, Kline JA, Schneider AJ, Coates WC. ChatGPT and conversational artificial intelligence: Friend, foe, or future of research? Am J
415 Emerg Med. 2023 May 18;70:81-83. doi: 10.1016/j.ajem.2023.05.018. Epub ahead of print.
416
417 42. Graf A, Bernardi RE. ChatGPT in Research: Balancing Ethics, Transparency and Advancement. Neuroscience. 2023 Apr 1;515:71-73. doi:
418 10.1016/j.neuroscience.2023.02.008. Epub 2023 Feb 21.
419
420 43. Graham A. ChatGPT and other AI tools put students at risk of plagiarism allegations, MDU warns. BMJ. 2023 May 17;381:1133. doi:
421 10.1136/bmj.p1133.
422
423 44. Gravel J, D'Amours-Gravel M, Osmanlliu E. Learning to fake it: limited responses and fabricated references provided by ChatGPT for
424 medical questions. medRxiv 2023.03.16.23286914; doi: https://doi.org/10.1101/2023.03.16.23286914.
 79

425
426 45. Guo E, Gupta M, Sinha S, Rössler K, Tatagiba M, Akagami R, Al-Mefty O, Sugiyama T, Stieg PE, Pickett GE, de Lotbiniere-Bassett M.
427 neuroGPT-X: Towards an Accountable Expert Opinion Tool for Vestibular Schwannoma. medRxiv 2023.02.25.23286117; doi:
428 https://doi.org/10.1101/2023.02.25.23286117.
429
430 46. Gurha P, Ishaq N, Marian AJ. ChatGPT and other artificial intelligence chatbots and biomedical writing. J Cardiovasc Aging. 2023;3(2):20.
431 doi: 10.20517/jca.2023.13. Epub 2023 Mar 31.
432
433 47. Haemmerli J, Sveikata L, Nouri A, May A, Egervari K, Freyschlag C, Lobrinus JA, Migliorini D, Momjian S, Sanda N, Schaller K. ChatGPT in
434 glioma patient adjuvant therapy decision making: ready to assume the role of a doctor in the tumour board?. medRxiv
435 2023.03.19.23287452; doi: https://doi.org/10.1101/2023.03.19.23287452.
436
437 48. Harskamp RE, De Clercq L. Performance of ChatGPT as an AI-assisted decision support tool in medicine: a proof-of-concept study for
438 interpreting symptoms and management of common cardiac conditions (AMSTELHEART-2). medRxiv 2023.03.25.23285475; doi:
439 https://doi.org/10.1101/2023.03.25.23285475.
440
441 49. Hill-Yardin EL, Hutchinson MR, Laycock R, Spencer SJ. A Chat(GPT) about the future of scientific publishing. Brain Behav Immun. 2023
442 May;110:152-154. doi: 10.1016/j.bbi.2023.02.022. Epub 2023 Mar 1.
443
444 50. Hirani R, Farabi B, Marmon S. Experimenting with ChatGPT: Concerns for academic medicine. J Am Acad Dermatol. 2023 May 11:S0190-
445 9622(23)00747-8. doi: 10.1016/j.jaad.2023.04.045. Epub ahead of print.
446
447 51. Homolak J. Opportunities and risks of ChatGPT in medicine, science, and academic publishing: a modern Promethean dilemma. Croat Med
448 J. 2023 Feb 28;64(1):1-3. doi: 10.3325/cmj.2023.64.1.
449
450 52. Hosseini M, Gao CA, Liebovitz DM, Carvalho AM, Ahmad FS, Luo Y, MacDonald N, Holmes KL, Kho A. An exploratory survey about using
451 ChatGPT in education, healthcare, and research. medRxiv 2023.03.31.23287979; doi: https://doi.org/10.1101/2023.03.31.23287979.
452
453 53. Howard A, Hope W, Gerada A. ChatGPT and antimicrobial advice: the end of the consulting infection doctor? Lancet Infect Dis. 2023
454 Apr;23(4):405-406. doi: 10.1016/S1473-3099(23)00113-5. Epub 2023 Feb 20.
455
456 54. Hsu TW, Tsai SJ, Ko CH, Thompson T, Hsu CW, Yang FC, Tsai CK, Tu YK, Yang SN, Tseng PT, Liang CS. Plagiarism, Quality, and
457 Correctness of ChatGPT-Generated vs Human-Written Abstract for Research Paper. Available at SSRN:
458 https://ssrn.com/abstract=4429014 or http://dx.doi.org/10.2139/ssrn.4429014
459
 80

460 55. Huang J, Tan M. The role of ChatGPT in scientific communication: writing better scientific review articles. Am J Cancer Res. 2023 Apr
461 15;13(4):1148-1154.
462
463 56. Hurley D. Your AI Program Will Write Your Paper Now: Neurology Editors on Managing Artificial Intelligence Submissions. Neurology Today
464 2023;23:10-11.
465
466 57. Janssen BV, Kazemier G, Besselink MG. The use of ChatGPT and other large language models in surgical science. BJS Open. 2023 Mar
467 7;7(2):zrad032. doi: 10.1093/bjsopen/zrad032.
468
469 58. Johnson SB, King AJ, Warner EL, Aneja S, Kann BH, Bylund CL. Using ChatGPT to evaluate cancer myths and misconceptions: artificial
470 intelligence and cancer information. JNCI Cancer Spectr. 2023 Mar 1;7(2):pkad015. doi: 10.1093/jncics/pkad015.
471
472 59. Juhi A, Pipil N, Santra S, Mondal S, Behera JK, Mondal H. The Capability of ChatGPT in Predicting and Explaining Common Drug-Drug
473 Interactions. Cureus. 2023 Mar 17;15(3):e36272. doi: 10.7759/cureus.36272.
474
475 60. Kaneda Y. In the Era of Prominent AI, What Role Will Physicians Be Expected to Play? QJM. 2023 May 22:hcad099. doi:
476 10.1093/qjmed/hcad099. Epub ahead of print.
477
478 61. Kim J. Search for Medical Information and Treatment Options for Musculoskeletal Disorders through an Artificial Intelligence Chatbot:
479 Focusing on Shoulder Impingement Syndrome. medRxiv 2022.12.16.22283512; doi: https://doi.org/10.1101/2022.12.16.22283512.
480
481 62. Kim SG. Using ChatGPT for language editing in scientific articles. Maxillofac Plast Reconstr Surg. 2023 Mar 8;45(1):13. doi:
482 10.1186/s40902-023-00381-x.
483
484 63. Koo M. The Importance of Proper Use of ChatGPT in Medical Writing. Radiology. 2023 May;307(3):e230312. doi: 10.1148/radiol.230312.
485 Epub 2023 Mar 7.
486
487 64. Kumar AH. Analysis of ChatGPT Tool to Assess the Potential of its Utility for Academic Writing in Biomedical Domain. BEMS Reports
488 [Internet]. 2023Feb.2 [cited 2023Jun.10];9(1):24-30. Available from: https://bemsreports.org/index.php/bems/article/view/132.
489
490 65. Lee P, Bubeck S, Petro J. Benefits, Limits, and Risks of GPT-4 as an AI Chatbot for Medicine. N Engl J Med. 2023 Mar 30;388(13):1233-
491 1239. doi: 10.1056/NEJMsr2214184.
492
493 66. Levin G, Meyer R, Yasmeen A, Young B, Guige PA, Bar-Noy T, Tatar A, Perelstein O, Brezinov Y. ChatGPT-written OBGYN abstracts fool
494 practitioners. Am J Obstet Gynecol MFM. 2023 Apr 29:100993. doi: 10.1016/j.ajogmf.2023.100993. Epub ahead of print.
 81

495
496 67. Li H, Moon JT, Purkayastha S, Celi LA, Trivedi H, Gichoya JW. Ethics of large language models in medicine and medical research. Lancet
497 Digit Health. 2023 Jun;5(6):e333-e335. doi: 10.1016/S2589-7500(23)00083-3. Epub 2023 Apr 27.
498
499 68. Li S. ChatGPT has made the field of surgery full of opportunities and challenges. Int J Surg. 2023 May 17. doi:
500 10.1097/JS9.0000000000000454. Epub ahead of print.
501
502 69. Liebrenz M, Schleifer R, Buadze A, Bhugra D, Smith A. Generating scholarly content with ChatGPT: ethical challenges for medical
503 publishing. Lancet Digit Health. 2023 Mar;5(3):e105-e106. doi: 10.1016/S2589-7500(23)00019-5. Epub 2023 Feb 6.
504
505 70. Lin Z. Modernizing authorship criteria: Challenges from exponential authorship inflation and generative artificial intelligence. 2023.
506 (Preprint). https://psyarxiv. com/s6h58.
507
508 71. Loh E. ChatGPT and generative AI chatbots: challenges and opportunities for science, medicine and medical leaders. BMJ Lead. 2023 May
509 2:leader-2023-000797. doi: 10.1136/leader-2023-000797. Epub ahead of print.
510
511 72. Maeker E, Maeker-Poquet B. ChatGPT : une solution pour rédiger des revues de littérature en médecine ?, Volume 8035, Issue 135,
512 06/2023, Pages 137-212, ISSN 1627-4830, http://dx.doi.org/10.1016/j.npg.2023.03.002
513 (http://www.sciencedirect.com/science/article/pii/S1627-4830(23)00037-5).
514
515 73. Marchandot B, Matsushita K, Carmona A, Trimaille A, Morel O. ChatGPT: the next frontier in academic writing for cardiologists or a
516 pandora's box of ethical dilemmas. Eur Heart J Open. 2023 Feb 13;3(2):oead007. doi: 10.1093/ehjopen/oead007.
517
518 74. Martínez-Sellés M, Marina-Breysse M. Current and Future Use of Artificial Intelligence in Electrocardiography. J Cardiovasc Dev Dis. 2023
519 Apr 17;10(4):175. doi: 10.3390/jcdd10040175.
520
521 75. Mehnen L, Gruarin S, Vasileva M, Knapp B. ChatGPT as a medical doctor? A diagnostic accuracy study on common and rare diseases.
522 medRxiv 2023.04.20.23288859; doi: https://doi.org/10.1101/2023.04.20.23288859.
523
524 76. Mello MM, Guha N. ChatGPT and Physicians' Malpractice Risk. JAMA Health Forum. 2023 May 5;4(5):e231938. doi:
525 10.1001/jamahealthforum.2023.1938.
526
527 77. Mese I. The imperative of a radiology AI deployment registry and the potential of ChatGPT. Clin Radiol. 2023 Jul;78(7):554. doi:
528 10.1016/j.crad.2023.04.001. Epub 2023 Apr 18.
529
 82

530 78. Nastasi NJ, Courtright KR, Halpern SD, Weissman GE. Does ChatGPT Provide Appropriate and Equitable Medical Advice? A Vignette-
531 Based, Clinical Evaluation Across Care Contexts. medRxiv 2023.02.25.23286451; doi: https://doi.org/10.1101/2023.02.25.23286451.
532
533 79. Nguyen Y, Costedoat-Chalumeau N. Les intelligences artificielles conversationnelles en médecine interne : l’exemple de
534 l’hydroxychloroquine selon ChatGPT [Artificial intelligence and internal medicine: The example of hydroxychloroquine according to
535 ChatGPT]. Rev Med Interne. 2023 May;44(5):218-226. French. doi: 10.1016/j.revmed.2023.03.017.
536
537 80. Nógrádi B, Polgár TF, Meszlényi V, Kádár Z, Hertelendy P, Csáti A, et al. Is There Any Room for Generative AI in Neurology and Other
538 Medical Areas?. Available at SSRN: https://ssrn.com/abstract=4372965 or http://dx.doi.org/10.2139/ssrn.4372965.
539
540 81. North RA. Plagiarism Reimagined. Function. 2023; 4(3):zqad014, https://doi.org/10.1093/function/zqad014.
541
542 82. Oh N, Choi GS, Lee WY. ChatGPT goes to the operating room: evaluating GPT-4 performance and its potential in surgical education and
543 training in the era of large language models. Ann Surg Treat Res. 2023 May;104(5):269-273. doi: 10.4174/astr.2023.104.5.269. Epub 2023
544 Apr 28.
545
546 83. Okan Ç. AI and Psychiatry: The ChatGPT Perspective. Alpha Psychiatry. 2023 Mar 1;24(2):41-42. doi:
547 10.5152/alphapsychiatry.2023.010223.
548
549 84. Parsa A, Ebrahimzadeh MH. ChatGPT in Medicine; a Disruptive Innovation or Just One Step Forward? Arch Bone Jt Surg. 2023;11(4):225-
550 226. doi: 10.22038/abjs.2023.22042.
551
552 85. Patel SB, Lam K, Liebrenz M. ChatGPT: friend or foe. Lancet Digit Health 2023;5:e102. DOI: https://doi.org/10.1016/S2589-7500(23)00023-
553 7.
554
555 86. Pourhoseingholi MA, Hatamnejad MR, Solhpour A. Does chatGPT (or any other artificial intelligence language tool) deserve to be included
556 in authorship list? Gastroenterol Hepatol Bed Bench. 2023;16(1):435-437. doi: 10.22037/ghfbb.v16i1.2747.
557
558 87. Rao D. The Urgent Need for Healthcare Workforce Upskilling and Ethical Considerations in the Era of AI-Assisted Medicine. Indian J
559 Otolaryngol Head Neck Surg (2023). https://doi.org/10.1007/s12070-023-03755-9.
560
561 88. Ray PP, Majumder P. AI Tackles Pandemics: ChatGPT's Game-Changing Impact on Infectious Disease Control. Ann Biomed Eng. 2023
562 May 18:1–3. doi: 10.1007/s10439-023-03239-5. Epub ahead of print.
563
 83

564 89. Ros-Arlanzón P, Pérez-Sempere A. ChatGPT: una novedosa herramienta de escritura para artículos científicos, pero no un autor (por el
565 momento) [ChatGPT: a novel tool for writing scientific articles, but not an author (for the time being)]. Rev Neurol. 2023 Apr 16;76(8):277.
566 Spanish. doi: 10.33588/rn.7608.2023066.
567
568 90. Sabry Abdel-Messih M, Kamel Boulos MN. ChatGPT in Clinical Toxicology. JMIR Med Educ. 2023 Mar 8;9:e46876. doi: 10.2196/46876.
569
570 91. Sallam M. The Utility of ChatGPT as an Example of Large Language Models in Healthcare Education, Research and Practice: Systematic
571 Review on the Future Perspectives and Potential Limitations. medRxiv 2023.02.19.23286155; doi:
572 https://doi.org/10.1101/2023.02.19.23286155.
573
574 92. Sallam M, Salim NA, Al-Tammemi AB, Barakat M, Fayyad D, Hallit S, Harapan H, Hallit R, Mahafzah A. ChatGPT Output Regarding
575 Compulsory Vaccination and COVID-19 Vaccine Conspiracy: A Descriptive Study at the Outset of a Paradigm Shift in Online Search for
576 Information. Cureus. 2023 Feb 15;15(2):e35029. doi: 10.7759/cureus.35029.
577
578 93. Salvagno M, Taccone FS, Gerli AG. Can artificial intelligence help for scientific writing? Crit Care. 2023 Feb 25;27(1):75. doi:
579 10.1186/s13054-023-04380-2. Erratum in: Crit Care. 2023 Mar 8;27(1):99.
580
581 94. Sanmarchi F, Bucci A, Nuzzolese AG. et al. A step-by-step researcher's guide to the use of an AI-based transformer in epidemiology: an
582 exploratory analysis of ChatGPT using the STROBE checklist for observational studies. J Public Health (Berl.) (2023).
583 https://doi.org/10.1007/s10389-023-01936-y
584
585 95. Sarink MJ, Bakker IL, Anas AA, Yusuf E. A study on the performance of ChatGPT in infectious diseases clinical consultation. Clin Microbiol
586 Infect. 2023 May 18:S1198-743X(23)00241-0. doi: 10.1016/j.cmi.2023.05.017. Epub ahead of print.
587
588 96. Schulte B. Capacity of ChatGPT to Identify Guideline-Based Treatments for Advanced Solid Tumors. Cureus. 2023 Apr 21;15(4):e37938.
589 doi: 10.7759/cureus.37938.
590
591 97. Singh OP. Artificial intelligence in the era of ChatGPT - Opportunities and challenges in mental health care. Indian J Psychiatry. 2023
592 Mar;65(3):297-298. doi: 10.4103/indianjpsychiatry.indianjpsychiatry_112_23.
593
594 98. Singh S, Djalilian A, Ali MJ. ChatGPT and Ophthalmology: Exploring Its Potential with Discharge Summaries and Operative Notes. Semin
595 Ophthalmol. 2023 May 3:1-5. doi: 10.1080/08820538.2023.2209166. Epub ahead of print.
596
597 99. Tang L, Sun Z, Idnay B, Nestor JG, Soroush A, Elias PA, Xu Z, Ding Y, Durrett G, Rousseau J, Weng C, Peng Y. Evaluating Large
598 Language Models on Medical Evidence Summarization. medRxiv 2023.04.22.23288967; doi: https://doi.org/10.1101/2023.04.22.23288967.
 84

599
600 100. Temsah O, Khan SA, Chaiah Y, Senjab A, Alhasan K, Jamal A, Aljamaan F, Malki KH, Halwani R, Al-Tawfiq JA, Temsah MH, Al-
601 Eyadhy A. Overview of Early ChatGPT's Presence in Medical Literature: Insights From a Hybrid Literature Review by ChatGPT and Human
602 Experts. Cureus. 2023 Apr 8;15(4):e37281. doi: 10.7759/cureus.37281.
603
604 101. Thorp HH. ChatGPT is fun, but not an author. Science. 2023 Jan 27;379(6630):313. doi: 10.1126/science.adg7879. Epub 2023 Jan 26.
605
606 102. Haq ZU, Naeem H, Naeem A, Iqbal F, Zaeem D. Comparing human and artificial intelligence in writing for health journals: an
607 exploratory study. medRxiv 2023.02.22.23286322; doi: https://doi.org/10.1101/2023.02.22.23286322.
608
609 103. Uprety D, Zhu D, West HJ. ChatGPT-A promising generative AI tool and its implications for cancer care. Cancer. 2023 May 14. doi:
610 10.1002/cncr.34827. Epub ahead of print.
611
612 104. Uz C, Umay E. "Dr ChatGPT": Is it a reliable and useful source for common rheumatic diseases? Int J Rheum Dis. 2023 May 23. doi:
613 10.1111/1756-185X.14749. Epub ahead of print.
614
615 105. van Dis EAM, Bollen J, Zuidema W, van Rooij R, Bockting CL. ChatGPT: five priorities for research. Nature. 2023 Feb;614(7947):224-
616 226. doi: 10.1038/d41586-023-00288-7.
617
618 106. Waisberg E, Ong J, Masalkhi M, Kamran SA, Zaman N, Sarker P, Lee AG, Tavakkoli A. GPT-4: a new era of artificial intelligence in
619 medicine. Ir J Med Sci. 2023 Apr 19. doi: 10.1007/s11845-023-03377-8. Epub ahead of print.
620
621 107. Wen J, Wang W. The future of ChatGPT in academic research and publishing: A commentary for clinical and translational medicine.
622 Clin Transl Med. 2023 Mar;13(3):e1207. doi: 10.1002/ctm2.1207.
623
624 108. Xue VW, Lei P, Cho WC. The potential impact of ChatGPT in clinical and translational medicine. Clin Transl Med. 2023
625 Mar;13(3):e1216. doi: 10.1002/ctm2.1216.
626
627 109. Yadava OP. ChatGPT—a foe or an ally?. Indian J Thorac Cardiovasc Surg 2023;39:217–21. https://doi.org/10.1007/s12055-023-01507-
628 6.
629
630 110. Au Yeung J, Kraljevic Z, Luintel A, Balston A, Idowu E, Dobson RJ, Teo JT. AI chatbots not yet ready for clinical use. Front Digit Health.
631 2023 Apr 12;5:1161098. doi: 10.3389/fdgth.2023.1161098.
632
 85

633 111. Young JN, O'Hagan R, Poplausky D, et al. The utility of ChatGPT in generating patient-facing and clinical responses for melanoma. J
634 Am Acad Dermatol. 2023 May 18:S0190-9622(23)00908-8. doi: 10.1016/j.jaad.2023.05.024.
635
636 112. Zheng H, Zhan H. ChatGPT in Scientific Writing: A Cautionary Tale. Am J Med. 2023 Mar 10:S0002-9343(23)00159-6. doi:
637 10.1016/j.amjmed.2023.02.011. Epub ahead of print.
638
639 113. Zhong Y, Chen YJ, Zhou Y, Lyu YA, Yin JJ, Gao YJ. The Artificial intelligence large language models and neuropsychiatry practice and
640 research ethic. Asian J Psychiatr 2023;84:103577.
641
642 114. Zhou J, Jia Y, Qiu Y, Lin L. The Potential of Applying Chatgpt to Extract Keywords of Medical Literature in Plastic Surgery. Aesthet Surg
643 J. 2023 May 20:sjad158. doi: 10.1093/asj/sjad158. Epub ahead of print.
644
645 115. Zhou Z. Evaluation of ChatGPT's Capabilities in Medical Report Generation. Cureus 2023;15:e37589.
646
647 116. Zhu L, Mou W, Chen R. Can the ChatGPT and other large language models with internet-connected database solve the questions and
648 concerns of patient with prostate cancer and help democratize medical knowledge? J Transl Med. 2023 Apr 19;21(1):269.
649
650 117. Zielinski C, Winker M, Aggarwal R, et al. Chatbots, ChatGPT, and Scholarly Manuscripts: WAME Recommendations on ChatGPT and
651 Chatbots in Relation to Scholarly Publications. Open Access Macedonian Journal of Medical Sciences 2023;11:83-86. DOI:
652 10.3889/oamjms.2023.11502.
653
654 118. Zimmerman A. A Ghostwriter for the Masses: ChatGPT and the Future of Writing. Ann Surg Oncol. 2023 Jun;30(6):3170-3173.
655
656
657
658
659
660
661
662
663
664
665
666
667
 86

668
669
670
671
672
673
674 Table-2: Summaries of full text of 118 selected article were uploaded in pdf format on ChatPDF, subsequently 12
675 questions were asked. Questions and respective answers given by ChatPDF are listed in this table.
676
677
Number Question Answer by ChatPDF Summary
1 What are potential ChatGPT can be used in medical writing to produce coherent and ChatGPT has
role of ChatGPT in readable content. It can also be used in scientific research to generate applications in
medical writing creative ideas, write essays and homework assignments, and even medical writing,
and research? contribute as an author on research papers. Additionally, ChatGPT has scientific
the potential to automate medical tasks and enhance learning, research, and
interpretation, and recall of medical information. However, concerns automation of
persist about its accuracy, potential biases in the training data, privacy medical tasks, but
concerns, issues related to authorship in academic papers, and ethical concerns remain
considerations of its use. Despite these challenges, AI tools like regarding
ChatGPT are expected to be widely adopted in the medical field in the accuracy, biases,
future. privacy,
authorship, and
ethics. Adoption
in the medical
field is
anticipated.
2 What could be the ChatGPT can assist with patient inquiries, help with writing medical notes ChatGPT offers
role of ChatGPT in and discharge summaries, aid in decision-making and treatment assistance in
clinical practice? planning, and support medical educators in various tasks. In clinical patient inquiries,
practice, ChatGPT could potentially be used as a valuable tool for medical note
personalized learning and encouraging a shift towards critical thinking writing, decision-
 87

and problem-based learning. However, it is important to note that there making, and
are challenges associated with the use of ChatGPT in clinical practice, medical
such as the need for access to current information, potential biases in the education, but
training data, privacy concerns, issues related to authorship in academic challenges like
papers, and ethical considerations of its use. Therefore, regulations and biases, privacy,
control mechanisms should be established to address ethical utilization of authorship, and
ChatGPT and similar tools in medicine. ethics need to be
addressed for
ethical utilization
in medicine.
3 What are ethical Paper writing entails several ethical concerns, including the risks of bias, Ethical concerns
issues associated plagiarism, copyright infringement, transparency issues, legal in paper writing
with paper implications, lack of originality, incorrect responses, limited knowledge, with ChatGPT
writing? and inaccurate citations. These concerns arise from the potential for involve bias,
ChatGPT to generate fabricated and unreliable research outputs, as well plagiarism,
as the lack of transparency and accountability in its responses. transparency,
Addressing these ethical issues is crucial to uphold research integrity and legal issues, and
ensure the reliability and trustworthiness of the generated results. lack of
accountability,
which need
addressing for
reliable and
trustworthy
research.
4 Can CHATGPT be The question of whether ChatGPT can be considered an author is a Including
an author? subject of controversy and may pose challenges in a legal context. The ChatGPT as an
traditional authorship criteria, such as those outlined by the International author is
Committee of Medical Journal Editors (ICMJE), require individuals to controversial due
contribute substantially to various aspects of the research process, to its inability to
including conception, design, analysis, interpretation, and manuscript meet authorship
preparation. As an AI language model, ChatGPT cannot independently criteria, but it can
fulfill these criteria. be used as a tool
 88

to assist human
However, ChatGPT can serve as a valuable tool to support human authors. Decisions
authors in writing papers and generating ideas. Its contributions can be should be made
acknowledged in the paper, but it may not meet the requirements for on a case-by-case
authorship in the traditional sense. basis, adhering to
ethical guidelines.
Determining whether or not to include ChatGPT as an author should be
approached on a case-by-case basis and aligned with established ethical
guidelines for scientific research. It is essential to consider the specific
contributions made by ChatGPT and ensure that human authors take
responsibility for the substantial aspects of the research process as
outlined by the ICMJE guidelines.
5 Can CHATGPT ChatGPT can assist in the editing process, including language editing, ChatGPT can aid
write text in good and summarizing the text to compose a suitable abstract. However, the in the editing
English and free quality of the AI-generated content needs careful evaluation as it is not a process, but
of plagiarism? substitute for human expertise, creativity, and critical thinking. While human expertise,
ChatGPT can rapidly comprehend information and connect evidence, it creativity, and
lacks the nuance, style, or originality that a human author can provide. critical thinking
Additionally, there is a risk of plagiarism if ChatGPT is not properly are essential.
trained or if it draws on sources without proper attribution. Therefore, it is Careful evaluation
important to carefully evaluate and edit any content generated by and editing are
ChatGPT to ensure that it meets high standards of quality and originality. needed to ensure
quality, originality,
and avoid
plagiarism.
6 Role of CHATGPT While there is limited specific information on the role of ChatGPT in While there is no
so far in neurological disorders related to clinical practice and research, it is worth specific
neurological noting that the integration of artificial intelligence, natural language information on
disorders related processing, and GPT technologies in psychiatry shows promise in ChatGPT's role in
clinical practice transforming the diagnosis, treatment, and monitoring of psychiatric neurological
and research. disorders. These technologies have the potential to assist psychiatrists in disorders, AI
making more accurate diagnoses, developing personalized treatment technologies
 89

plans, enhancing the accuracy and efficiency of clinical trials, and show promise in
increasing access to mental health care through the use of chatbots and revolutionizing
virtual assistants. Although there may not be explicit examples or studies psychiatric
on ChatGPT's application in neurological disorders, similar applications diagnosis,
could potentially be developed in the future. treatment, and
accessibility to
mental health
care.
7 Effectiveness and Upon review, I couldn't find specific information on the effectiveness and There is no
efficiency of Chat efficiency of ChatGPT in medical research and clinical settings. The specific
GPT in medical provided pages discuss the potential applications of AI and natural information on
research and language processing in medicine, including psychiatry, but do not offer ChatGPT's
clinical settings specific studies or data on the use of ChatGPT in these settings. effectiveness in
medical research
It is important to emphasize that while AI tools like ChatGPT show and clinical
promise in improving medical research and clinical practice, they should settings,
not be seen as a substitute for human expertise. Instead, they should be highlighting the
utilized alongside established best practices and ethical guidelines. need for further
Further research is necessary to assess the effectiveness and efficiency research and
of ChatGPT specifically in medical research and clinical contexts. consideration of
ethical guidelines.
8 Potential benefits ChatGPT presents potential benefits and limitations in medical research ChatGPT has
and limitations of and clinical applications. Some benefits include its assistance in patient potential benefits
Chat GPT in questionnaires, interviews, and epidemiological research by streamlining in mental health
medical research data collection, recording, and analysis. The integration of artificial care, data
and clinical intelligence, natural language processing, and GPT technologies in collection, and
applications psychiatry holds promise for transforming psychiatric diagnosis, manuscript
treatment, and monitoring. It can aid in accurate diagnoses, personalized checks but has
treatment design, improved clinical trial efficiency, and increased limitations in
accessibility to mental health care through chatbots and virtual providing
assistants. ChatGPT can also be useful for tasks like proofreading and comprehensive
manuscript checks. diagnoses and
 90

replacing human
However, there are limitations to consider. ChatGPT cannot update its expertise,
training data in real-time and may provide general or vague answers in creativity, and
medical-related conversations. Its diagnoses may not be comprehensive critical thinking.
or sufficient, indicating that it cannot independently handle the Careful evaluation
complexities of clinical practice. It cannot replace the specialized and ethical
knowledge, innovative ideas, and opinions that human input provides. guidelines are
Ethical guidelines and further exploration of research ethics are necessary.
necessary for the responsible use of AI in health research. The quality of
AI-generated content requires careful evaluation as it cannot substitute
for human expertise, creativity, and critical thinking. While ChatGPT can
comprehend information and connect evidence rapidly, it lacks the
nuance and originality that human authors provide. Moreover, there is a
risk of plagiarism if ChatGPT is not appropriately trained or if it fails to
provide proper source attribution.
9 The ethical There are ethical implications of using ChatGPT in medical research and Using ChatGPT in
implications of clinical practice. - Almost 97% of the records expressed concerns or medical research
using Chat GPT in possible risks associated with ChatGPT. Ethical issues, such as risk of and clinical
medical research bias, plagiarism, copyright issues, transparency issues, legal concerns, practice raises
and clinical lack of originality, incorrect responses, limited knowledge, and inaccurate ethical concerns
practice citations were the most common. The quality of the AI-generated content such as bias,
needs careful evaluation as it is not a substitute for human expertise, plagiarism, and
creativity, and critical thinking. While ChatGPT can rapidly comprehend limited human-like
information and connect evidence, it lacks the nuance or originality that a qualities.
human author can provide. Additionally, there is a risk of plagiarism if Adherence to
ChatGPT is not properly trained or if it draws on sources without proper research ethics
attribution. Further exploration of research ethics and the moral use of AI and careful
in health research is required to establish guidelines. It is important to evaluation are
consider these ethical implications when using ChatGPT in medical crucial.
research and clinical practice. Researchers and practitioners should
ensure that they are using AI tools like ChatGPT in an ethical manner
that respects patient privacy and autonomy while also adhering to
 91

established best practices for research ethics. Additionally, they should

carefully evaluate any content generated by ChatGPT to ensure that it
meets high standards of quality and originality.
10 Identify the gaps There are gaps in the current research on ChatGPT and areas for further Gaps in current
in the current investigation include: 1. The effectiveness and efficiency of ChatGPT in research on
research on Chat medical research and clinical settings: While there is potential for ChatGPT include
GPT and suggest ChatGPT to be useful in these settings, there is a lack of specific studies investigating its
areas for further or data on its effectiveness and efficiency. 2. The ethical implications of effectiveness,
investigation. using ChatGPT in medical research and clinical practice: While there are ethical
concerns about ethical issues associated with ChatGPT, further implications,
exploration of research ethics and the moral use of AI in health research limitations,
is required to establish guidelines. 3. The limitations of ChatGPT: While content quality,
ChatGPT has potential benefits, it also has limitations such as its inability and role in
to update its training data in real-time and provide comprehensive neurological
diagnoses. Further investigation into these limitations could help improve disorders in
the technology. 4. The quality of AI-generated content: There is a need medical research
for careful evaluation of the quality of AI-generated content as it lacks the and clinical
nuance or originality that a human author can provide. Further practice.
investigation into how to improve the quality of AI-generated content
could help make it more useful in medical research and clinical practice.
5. The role of ChatGPT in neurological disorders related clinical practice
and research: While there is discussion about the potential applications
of AI and natural language processing in medicine, including psychiatry,
there is no specific information on the role of ChatGPT in neurological
disorders related clinical practice and research. Further investigation into
these areas could help improve our understanding of how to best use
ChatGPT in medical research and clinical practice while also addressing
ethical concerns associated with its use.
11 Provide insights There are potential future applications of ChatGPT in medical research Potential future
into the potential and clinical practice, including: 1. Clinical decision support: ChatGPT applications of
future applications could assist in clinical decision-making by providing evidence-based ChatGPT in
of Chat GPT in recommendations and identifying patterns and correlations in large medical research
 92

medical research datasets. 2. Clinical trial recruitment: ChatGPT could help identify eligible and clinical
and clinical patients for clinical trials by analyzing electronic health records and other practice include
practice data sources. 3. Clinical data management: ChatGPT could help manage clinical decision
large amounts of clinical data by extracting relevant information from support, trial
electronic health records and other sources. 4. Research support: recruitment, data
ChatGPT could assist researchers in identifying relevant literature, management,
generating hypotheses, and analyzing data. 5. Patient education: research support,
ChatGPT could provide personalized education to patients about their and patient
conditions, treatments, and medications. 6. Automation of tasks: education. Ethical
ChatGPT could automate tasks such as data analysis and image considerations
interpretation, freeing up time for healthcare professionals to focus on and effective
patient care. 7. Drug discovery: ChatGPT could assist in drug discovery utilization are
by recognizing, classifying, and describing chemical formulas or crucial.
molecular structures. 8. Disease prediction, diagnosis, and assessment
of therapeutic targets: ChatGPT could help predict diseases based on
patient data, diagnose diseases based on symptoms or test results, and
assess therapeutic targets for drug development. Overall, the potential
future applications of ChatGPT in medical research and clinical practice
are vast. However, it is important to address ethical concerns associated
with its use while also ensuring that it is used effectively to improve
patient outcomes.
12 Recommendations 1. Researchers should be transparent about the use of ChatGPT in Researchers,
for researchers, scientific publications and ensure that the generated text is accurate and clinicians, and
clinicians, and reliable. policymakers
policymakers on 2. Clinicians should use ChatGPT with caution and recognize its should ensure
the use of limitations in providing reliable information. It can be useful for tasks like transparency,
ChatGPT in proofreading and manuscript checks, but cannot replace the specialized evaluate
medical research knowledge, innovative ideas, and opinions that human input provides. limitations,
and clinical 3. Policymakers should establish regulations and control mechanisms to establish
practice address ethical utilization of ChatGPT and similar tools in medicine. This regulations, and
includes guidelines for transparency in disclosing the use of AI tools like address ethical
ChatGPT in scientific publications. concerns related
 93

4. Researchers should evaluate and monitor the use of ChatGPT to to ChatGPT in

ensure that it is being used ethically and responsibly in medical medical
publishing and research. applications.
5. Clinicians should be aware of the potential biases in the training data
used to develop ChatGPT and should take steps to mitigate these biases
when using the tool.
6. Policymakers should consider the potential impact of ChatGPT on
clinical and translational medicine and establish guidelines for its use in
these areas.
7. Researchers, clinicians, and policymakers should work together to
address the ethical dilemmas associated with the use of ChatGPT in
medical research and clinical practice.
678