Review ChatGPT Utility in Healthcare Education, Research, and Practice: Systematic Review on the Promising Perspectives and Valid Concerns
综述ChatGPT在医疗教育、研究和实践中的应用：前景与隐忧的系统性评价
Malik Sallam $^{1,2(\oplus)}$
Malik Sallam $^{1,2(\oplus)}$
Citation: Sallam, M. ChatGPT Utility in Healthcare Education, Research, and Practice: Systematic Review on the Promising Perspectives and Valid Concerns. Healthcare 2023, 11, 887. https://doi.org/10.3390/ healthcare 11060887
引用：Sallam, M. ChatGPT在医疗教育、研究和实践中的应用价值：关于前景与隐忧的系统性综述。Healthcare 2023, 11, 887. https://doi.org/10.3390/healthcare11060887
Academic Editor: Daniele Giansanti
学术编辑：Daniele Giansanti
Copyright: $\circledcirc$ 2023 by the author. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license ( creative commons.org/licenses/by/ 4.0/).
版权：$\circledcirc$ 2023 作者所有。获许可方 MDPI，瑞士巴塞尔。本文是一篇开放获取文章，依据知识共享署名许可协议 (CC BY) (https://creativecommons.org/licenses/by/4.0/) 的条款和条件发布。
Abstract: ChatGPT is an artificial intelligence (AI)-based conversational large language model (LLM). The potential applications of LLMs in health care education, research, and practice could be promising if the associated valid concerns are pro actively examined and addressed. The current systematic review aimed to investigate the utility of ChatGPT in health care education, research, and practice and to highlight its potential limitations. Using the PRIMSA guidelines, a systematic search was conducted to retrieve English records in PubMed/MEDLINE and Google Scholar (published research or preprints) that examined ChatGPT in the context of health care education, research, or practice. A total of 60 records were eligible for inclusion. Benefits of ChatGPT were cited in 51/60 $(85.0%)$ ) records and included: (1) improved scientific writing and enhancing research equity and versatility; (2) utility in health care research (efficient analysis of datasets, code generation, literature reviews, saving time to focus on experimental design, and drug discovery and development); (3) benefits in health care practice (streamlining the workflow, cost saving, documentation, personalized medicine, and improved health literacy); and (4) benefits in health care education including improved personalized learning and the focus on critical thinking and problem-based learning. Concerns regarding ChatGPT use were stated in 58/60 $(96.7%)$ ) records including ethical, copyright, transparency, and legal issues, the risk of bias, plagiarism, lack of originality, inaccurate content with risk of hallucination, limited knowledge, incorrect citations, cyber security issues, and risk of infodemics. The promising applications of ChatGPT can induce paradigm shifts in health care education, research, and practice. However, the embrace of this AI chatbot should be conducted with extreme caution considering its potential limitations. As it currently stands, ChatGPT does not qualify to be listed as an author in scientific articles unless the ICMJE/COPE guidelines are revised or amended. An initiative involving all stakeholders in health care education, research, and practice is urgently needed. This will help to set a code of ethics to guide the responsible use of ChatGPT among other LLMs in health care and academia.
摘要：ChatGPT是一种基于人工智能(AI)的对话式大语言模型(LLM)。若能主动审视并解决相关合理关切，大语言模型在医疗教育、研究和实践中的潜在应用前景广阔。本系统综述旨在探讨ChatGPT在医疗教育、研究和实践中的效用，并强调其潜在局限性。根据PRIMSA指南，我们系统检索了PubMed/MEDLINE和Google Scholar中探讨ChatGPT在医疗教育、研究或实践背景下应用的英文文献(已发表研究或预印本)。共纳入60篇文献。51/60 $(85.0%)$ 篇文献指出ChatGPT的优势包括：(1)提升科学写作能力，促进科研公平性与多样性；(2)在医疗研究中的实用性(高效分析数据集、代码生成、文献综述、节省时间以专注实验设计、药物发现与开发)；(3)在医疗实践中的益处(优化工作流程、节约成本、文档处理、个性化医疗及提升健康素养)；(4)在医疗教育中的优势包括促进个性化学习、培养批判性思维和基于问题的学习。58/60 $(96.7%)$ 篇文献提出ChatGPT的使用顾虑，涉及伦理、版权、透明度、法律问题、偏见风险、剽窃、缺乏原创性、存在幻觉风险的不准确内容、知识有限、错误引用、网络安全问题及信息流行病风险。ChatGPT的广阔应用可能引发医疗教育、研究和实践的范式转变。然而，鉴于其潜在局限性，采用这款AI聊天机器人需极度谨慎。就目前而言，除非ICMJE/COPE指南修订，否则ChatGPT不具备被列为科学论文作者的资格。亟需发起一项涵盖医疗教育、研究和实践所有利益相关者的行动，以制定伦理准则，指导医疗界和学术界负责任地使用ChatGPT等大语言模型。
Keywords: machine learning; digital health; artificial intelligence; healthcare; ethics
关键词：机器学习；数字健康；人工智能；医疗保健；伦理
1. Introduction
1. 引言
Artificial intelligence (AI) can be defined as the multidisciplinary approach of computer science and linguistics that aspires to create machines capable of performing tasks that normally require human intelligence [1]. These tasks include the ability to learn, adapt, rationalize, understand, and to fathom abstract concepts as well as the reactivity to complex human attributes such as attention, emotion, creativity, etc. [2].
人工智能 (AI) 可定义为计算机科学与语言学的多学科交叉领域，旨在创造能够执行通常需要人类智能的任务的机器 [1]。这些任务包括学习、适应、推理、理解抽象概念的能力，以及对注意力、情感、创造力等复杂人类特征的响应能力 [2]。
The history of AI as a scientific discipline can be traced back to the mid-XX century at the Dartmouth Summer Research Project on AI [3]. This was followed by the development of machine learning (ML) algorithms that allow decision-making or predictions based on the patterns in large datasets [4]. Subsequently, the development of neural networks (brainmimicking algorithms), genetic algorithms (finding optimal solutions for complex problems by application of evolutionary principles), and other advanced techniques followed [5].
人工智能作为一门科学学科的历史可以追溯到20世纪中叶的达特茅斯夏季人工智能研究项目 [3]。随后出现了机器学习 (ML) 算法的发展，这些算法能够基于大数据集中的模式进行决策或预测 [4]。接着发展出了神经网络 (模拟大脑的算法)、遗传算法 (通过应用进化原理寻找复杂问题的最优解) 以及其他先进技术 [5]。
Launched in November 2022, “ChatGPT” is an AI-based large language model (LLM) trained on massive text datasets in multiple languages with the ability to generate humanlike responses to text input [6]. Developed by OpenAI (OpenAI, L.L.C., San Francisco, CA, USA), ChatGPT etymology is related to being a chatbot (a program able to understand and generate responses using a text-based interface) and is based on the generative pre-trained transformer (GPT) architecture [6,7]. The GPT architecture utilizes a neural network to process natural language, thus generating responses based on the context of input text [7]. The superiority of ChatGPT compared to its GPT-based predecessors can be linked to its ability to respond to multiple languages generating refined and highly sophisticated responses based on advanced modeling [6,7].
2022年11月推出的"ChatGPT"是一种基于AI的大语言模型(LLM)，通过多语言海量文本数据集训练而成，能够针对文本输入生成类人响应[6]。由OpenAI(OpenAI, L.L.C., 美国旧金山)开发的ChatGPT，其名称源于chatbot(能够通过文本界面理解并生成响应的程序)，并基于生成式预训练Transformer(GPT)架构[6,7]。GPT架构利用神经网络处理自然语言，从而根据输入文本的上下文生成响应[7]。ChatGPT相较于前代GPT模型的优势在于：通过先进建模技术，能够以多语言生成精炼且高度复杂的响应[6,7]。
In the scientific community and academia, ChatGPT has received mixed responses reflecting the history of controversy regarding the benefits vs. risks of advanced AI technologies [8–10]. On one hand, ChatGPT, among other LLMs, can be beneficial in conversational and writing tasks, assisting to increase the efficiency and accuracy of the required output [11]. On the other hand, concerns have been raised in relation to possible bias based on the datasets used in ChatGPT training, which can limit its capabilities and could result in factual inaccuracies, but alarmingly appear to be scientifically plausible (a phenomenon termed hallucination) [11]. Additionally, security concerns and the potential of cyberattacks with the spread of misinformation utilizing LLMs should also be considered [11].
在科学界和学术界，ChatGPT 引发的反响褒贬不一，这反映了关于先进 AI (Artificial Intelligence) 技术利弊的长期争议 [8–10]。一方面，ChatGPT 等大语言模型能在对话和写作任务中提升效率与输出准确性 [11]；另一方面，人们担忧其训练数据可能导致的偏见——这些偏见不仅会限制模型能力，还可能产生看似科学合理实则失实的内容（这种现象称为幻觉）[11]。此外，大语言模型传播错误信息引发的安全问题及网络攻击风险也值得警惕 [11]。
The innate resistance of the human mind to any change is a well-described phenomenon and can be understandable from evolutionary and social psychology perspectives [12]. Therefore, the concerns and debate that arose instantaneously following the widespread release of ChatGPT appear to be understandable. The attention that ChatGPT received involved several disciplines. In education, for example, ChatGPT release could mark the end of essays as assignments [13]. In health care practice and academic writing, factual inaccuracies, ethical issues, and the fear of misuse including the spread of misinformation should be considered [14–16].
人类思维对任何变化的固有抗拒是一种已被充分描述的现象，从进化心理学和社会心理学角度可以理解 [12]。因此，ChatGPT广泛发布后立即引发的担忧和辩论似乎也在情理之中。ChatGPT受到的关注涉及多个学科领域。例如在教育界，ChatGPT的发布可能标志着论文作业时代的终结 [13]；在医疗实践和学术写作领域，则需要考量事实性错误、伦理问题以及对滥用（包括虚假信息传播）的担忧 [14-16]。
The versatility of human intelligence (HI) compared to AI is related to its biological evolutionary history, adaptability, creativity, the ability of emotional intelligence, and the ability to understand complex abstract concepts [2]. However, HI-AI cooperation can be beneficial if an accurate and reliable output of AI is ensured. The promising utility of AI in health care has been outlined previously with possible benefits in personalized medicine, drug discovery, and the analysis of large datasets aside from the potential applications to improve diagnosis and clinical decisions [17,18]. Additionally, the utility of AI chatbots in health care education is an interesting area to probe. This is related to the massive information and various concepts that health care students are required to grasp [19]. However, all of these applications should be considered cautiously considering the valid concerns, risks, and categorical failures experienced and cited in the context of LLM applications. Specifically, Borji comprehensively highlighted the caveats of ChatGPT use that included, but were not limited to, the generation of inaccurate content, the risk of bias and discrimination, lack of transparency and reliability, cyber security concerns, ethical consequences, and societal implications [20].
人类智能 (HI) 相比人工智能的多样性源于其生物进化史、适应性、创造力、情商能力及理解复杂抽象概念的能力 [2]。然而，若能确保AI输出的准确性和可靠性，HI与AI的合作将大有裨益。此前已有研究概述了AI在医疗保健领域的潜在应用价值，包括个性化医疗、药物研发、大规模数据分析，以及改善诊断和临床决策的可能性 [17,18]。此外，AI聊天机器人在医疗教育中的应用也是一个值得探索的领域，这与医学生需要掌握海量信息和多样概念密切相关 [19]。但考虑到大语言模型应用中实际存在的隐患、风险及已被引证的典型失败案例，所有这些应用都应谨慎评估。Borji特别全面指出了ChatGPT的使用注意事项，包括但不限于：生成不准确内容、偏见与歧视风险、缺乏透明度和可靠性、网络安全问题、伦理影响及社会后果 [20]。
Therefore, the aim of the current review was to explore the future perspectives of ChatGPT as a prime example of LLMs in health care education, academic/scientific writing, health care research, and health care practice based on the existing evidence. Importantly, the current review objectives extended to involve the identification of potential limitations and concerns that could be associated with the application of ChatGPT in the aforementioned areas in health care settings.
因此，本次综述旨在基于现有证据，探讨以ChatGPT为代表的大语言模型(LLM)在医疗教育、学术/科研写作、医疗研究及临床实践中的未来前景。值得注意的是，本次综述目标还延伸至识别ChatGPT在上述医疗领域应用中可能存在的局限性与潜在问题。
2. Materials and Methods
2. 材料与方法
2.1. Search Strategy and Inclusion Criteria
2.1. 搜索策略与纳入标准
The current systematic review was conducted according to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMSA) guidelines [21]. The information sources included PubMed/MEDLINE and Google Scholar.
本次系统综述按照系统综述和荟萃分析优先报告条目(PRIMSA)指南[21]开展。信息来源包括PubMed/MEDLINE和Google Scholar。
The eligibility criteria involved any type of published scientific research or preprints (article, review, communication, editorial, opinion, etc.) addressing ChatGPT that fell under the following categories: (1) health care practice/research; (2) health care education; and (3) academic writing.
入选标准包括涉及ChatGPT的任何类型的已发表科学研究或预印本(文章、综述、通讯、社论、观点等)，且属于以下类别之一：(1) 医疗实践/研究；(2) 医学教育；(3) 学术写作。
The exclusion criteria included: (1) non-English records; (2) records addressing ChatGPT in subjects other than those mentioned in the eligibility criteria; and (3) articles from non-academic sources (e.g., newspapers, internet websites, magazines, etc.).
排除标准包括：(1) 非英文记录；(2) 记录涉及ChatGPT的主题不符合入选标准；(3) 非学术来源的文章(如报纸、网站、杂志等)。
The exact PubMed/MEDLINE search strategy, which concluded on 16 February 2023, was as follows: (ChatGPT) AND ((“2022/11/ $'30^{\prime\prime}$ [Date–Publication]: $\mathrm{'}3000^{\prime\prime}$ [Date–Publication])), which yielded 42 records.
截至2023年2月16日的PubMed/MEDLINE精确检索策略如下：(ChatGPT) AND (("2022/11/30"[Date-Publication] : "3000"[Date-Publication]))，共获得42条记录。
The search on Google Scholar was conducted using Publish or Perish (Version 8) [22]. The search term was “ChatGPT” for the years 2022–2023, and the Google Scholar search yielded 238 records and concluded on 16 February 2023.
在Google Scholar上的搜索使用了Publish or Perish (版本8) [22]。搜索关键词为"ChatGPT"，时间范围为2022年至2023年，Google Scholar搜索共获得238条记录，搜索结束于2023年2月16日。
2.2. Summary of the Record Screening Approach
2.2. 文献筛选方法概述
The records retrieved following the PubMed/MEDLINE and Google Scholar searches were imported to EndNote v.20 for Windows (Thomson Research Soft, Stanford, CA, USA), which yielded a total of 280 records.
通过PubMed/MEDLINE和Google Scholar检索获取的记录被导入至EndNote v.20 for Windows (Thomson Research Soft, Stanford, CA, USA)，共获得280条记录。
Next, screening of the title/abstract was conducted for each record with the exclusion of duplicate records $(n=40)$ , followed by the exclusion of records published in languages other than English $(n=32)$ ). Additionally, the records that fell outside the scope of the review (records that examined ChatGPT in a context outside health care education, health care practice, or scientific research/academic writing) were excluded $(n=80)$ ). Moreover, the records published in non-academic sources (e.g., newspapers, magazines, Internet websites, blogs, etc.) were excluded $(n=18)$ .
接下来，对每条记录进行标题/摘要筛选，排除重复记录 $(n=40)$ ，随后剔除非英语发表的记录 $(n=32)$ 。此外，还排除了超出综述范围的记录（即在医疗教育、医疗实践或科学研究/学术写作之外研究ChatGPT的记录） $(n=80)$ 。同时，剔除非学术来源（如报纸、杂志、网站、博客等）发表的记录 $(n=18)$ 。
Afterward, full screening of the remaining records $(n=110)$ ) was carried out with the exclusion of an additional 41 records that fell outside the scope of the current review. An additional nine records were excluded due to the inability to access the full text of these records, being subscription-based. This yielded a total of 60 records eligible for inclusion in the current review.
随后对剩余记录 $(n=110)$ 进行了全面筛选，排除了另外41条不属于当前综述范围的记录。由于无法获取基于订阅的全文，又排除了9条记录。最终共有60条记录符合纳入当前综述的条件。
2.3. Summary of the Descriptive Search for ChatGPT Benefits and Risks in the Included Records
2.3. 纳入记录中ChatGPT益处与风险的描述性检索总结
Each of the included records was searched specifically for the following: (1) type of record (preprint, published research article, opinion, commentary, editorial, review, etc.); (2) the listed benefits/applications of ChatGPT in health care education, health care practice, or scientific research/academic writing; (3) the listed risks/concerns of ChatGPT in health care education, health care practice, or scientific research/academic writing; and (4) the main conclusions and recommendations regarding ChatGPT in health care education, health care practice, or scientific research/academic writing.
对每份纳入文献的检索具体包括以下方面：(1) 文献类型 (预印本、已发表研究论文、观点、评论、社论、综述等)；(2) ChatGPT在医疗教育、医疗实践或科研/学术写作中列出的优势/应用；(3) ChatGPT在医疗教育、医疗实践或科研/学术写作中列出的风险/隐患；(4) 关于ChatGPT在医疗教育、医疗实践或科研/学术写作中的主要结论与建议。
Categorization of the benefits/applications of ChatGPT was as follows: (1) educational benefits in health care education (e.g., generation of realistic and variable clinical vignettes, customized clinical cases with immediate feedback based on the student’s needs, enhanced communications skills); (2) benefits in academic/scientific writing (e.g., text generation, sum mari z ation, translation, and literature review in scientific research); (3) benefits in scientific research (e.g., efficient analysis of large datasets, drug discovery, identification of potential drug targets, generation of codes in scientific research); (4) benefits in health care practice (e.g., improvements in personalized medicine, diagnosis, treatment, lifestyle recommendations based on personalized traits, documentation/generation of reports); and (5) being a freely available package.
ChatGPT的优势/应用分类如下：(1) 医疗教育领域优势（如生成真实多样的临床案例、根据学生需求提供定制化临床病例及即时反馈、提升沟通技巧）；(2) 学术/科研写作优势（如文本生成、摘要总结、翻译及科研文献综述）；(3) 科研领域优势（如高效分析大型数据集、药物发现、潜在药物靶点识别、科研代码生成）；(4) 医疗实践优势（如改进个性化医疗、诊断治疗、基于个性化特征的生活方式建议、文档/报告生成）；(5) 作为免费开源工具。
Categorization of the risks/concerns of ChatGPT was as follows: (1) ethical issues (e.g., risk of bias, discrimination based on the quality of training data, plagiarism); (2) hallucination (the generation of scientifically incorrect content that sounds plausible); (3) transparency issues (black box application); (4) risk of declining need for human expertise with subsequent psycho logic, economic and social issues; (5) over-detailed, redundant, excessive content; (6) concerns about data privacy for medical information; (7) risk of declining clinical skills, critical thinking and problem-solving abilities; (8) legal isues (e.g., copyright issues, authorship status); (9) interpret ability issues; (10) referencing issues; (11) risk of academic fraud in research; (12) incorrect content; and (13) infodemic risk.
ChatGPT的风险/担忧分类如下：(1) 伦理问题（如偏见风险、基于训练数据质量的歧视、剽窃）；(2) 幻觉（生成听起来合理但科学上不正确的内容）；(3) 透明度问题（黑箱应用）；(4) 人类专业知识需求下降带来的心理、经济和社会问题风险；(5) 内容过度详细、冗余或过量；(6) 医疗信息数据隐私担忧；(7) 临床技能、批判性思维和解决问题能力下降风险；(8) 法律问题（如版权问题、作者身份认定）；(9) 可解释性问题；(10) 参考文献问题；(11) 学术研究造假风险；(12) 内容错误；(13) 信息疫情风险。
3. Results
3. 结果
A total of 280 records were identified, and following the full screening process, a total of 60 records were eligible to be included in the review. The PRISMA flowchart of the record selection process is shown in Figure 1.
共识别出280条记录，经过全面筛选后，最终有60条记录符合纳入综述的条件。记录筛选过程的PRISMA流程图如图1所示。
Figure 1. Flowchart of the record selection process based on the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRIMSA) guidelines.
图 1: 基于系统评价与荟萃分析优先报告条目(PRIMSA)指南的记录筛选流程图。
3.1. Summary of the ChatGPT Benefits and Limitations/Concerns in Health Care
3.1. ChatGPT在医疗保健领域的优势与局限性/关注点总结
Summaries of the main conclusions of the included studies regarding ChatGPT utility in academic writing, health care education, and health care practice/research are provided in Table 1 for the records comprising editorials/letters to the editors, in Table 2 for the records comprising research articles, commentaries, news articles, perspectives, case studies, brief reports, communications, opinions, or recommendations, and in Table 3 for the records representing preprints
关于ChatGPT在学术写作、医疗教育和医疗实践/研究中效用的主要结论摘要如下：包含社论/致编辑信函的记录见表1，包含研究论文、评论、新闻文章、观点、案例研究、简报、通讯、意见或建议的记录见表2，预印本记录见表3
Table 1. A summary of the main conclusions of the included records comprising editorials/letters to the editors.
| Author(s) [Record] | Design, Aims | Applications,Benefits | Risks,Concerns,Limitations | Suggested Action, Conclusions |
| Chen [23] | Editorial on ChatGPT applications in scientific writing | ChatGPT helps to overcome language barriers promoting equity in research | Ethical concerns (ghostwriting); doubtful accuracy; citation problems | Embrace this innovation with an open mind; authors should have proper knowledge on how to exploit AI6 tools |
| Thorp [24] | Editorial: "ChatGPT is not an author" | Content is not original; incorrect answers that sound plausible; issues of referencing; risk of plagiarism | Revise assignments in education In Science journals, the use of ChatGPT is considered as a scientific misconduct | |
| Kitamura [25] | Editorial on ChatGPT and the future of medical writing | Improved efficiency in medical writing; translation purposes | Ethical concerns, plagiarism; lack of originality; inaccurate content; risk of bias | "AI in the Loop: Humans in Charge" |
| Lubowitz [26] | Editorial, ChatGPT impact on medical literature | Inaccurate content; risk of bias; spread of misinformation and disinformation; lack of references; redundancy in text | Authors should notuse ChatGPT to compose any part of scientific submission; however, it can be used under careful human supervision to ensure the integrity and originality of the scientific work | |
| Nature [27] | Nature editorial on the rules of ChatGPT use to ensure transparent science | ChatGPT can help to summarize research papers; to generate helpful computer codes | Ethical issues; transparency issues | LLM 7 tools will be accepted as authors;if LLM tools are to be used, it should be documented in the methods or acknowledgements; advocate for transparency in methods, and integrity and truth from researchers |
| Moons and Van Bulck [28] | Editorial on ChatGPT potential in cardiovascular nursing practice and research | ChatGPT can help to summarize a large text; it can facilitate the work of researchers; it can help in | Information accuracy issues; the limited knowledge up to the year 2021; limited capacity | ChatGPT can be a valuable tool in health care practice |
| Cahan and Treutlein [29] | Editorial reporting a conversation with ChatGPT on stem cell research | data collection ChatGPT helps to save time | Repetition; several ChatGPT responses lacked depth and insight; lack of references | ChatGPT helped to write an editorial saving valuable time |
| Ahn [30] | A letter to the editor reporting a conversation of ChatGPT regarding CPR 1 | Personalized interaction; quick response time; it can help to provide easily accessible and understandable information | Inaccurate information might be generated with possible serious medical consequences | Explore the potential utility of ChatGPT to provide information and education on CPR |
| Gunawan [31] | An editorial reporting a conversationwith ChatGPT regarding the future | regarding CPR to the general public ChatGPT can help to increase efficiency; it helps to reduce errors in | Lack of emotional and personal support | ChatGPT can provide valuable perspectives in health care |
| D'Amico et al. [32] | of nursing An editorial reporting a conversation of ChatGPT regarding incorporating Chatbots into neurosurgical | care delivery ChatGPT can help to provide timely and accurate information for the patients about their treatment | Possibility of inaccurate information; privacy concerns; ethical issues; legal issues; | Neurosurgery practice can be leading in utilizing ChatGPT into patient care and research |
| Fijacko et al. [33] | practice and research A letter to the editor to report the accuracy of ChatGPT responses with regards to life support exam | andcare ChatGPT provides relevant and accurate responses on occasions | risk of bias Referencing issues; over-detailed responses | ChatGPT did not pass any of the exams; however, it can be a powerful self-learning tool to prepare for the life |
| Mbakwe et al. [34] | questions by the AHA 2 An editorial on ChatGPT ability to pass the USMLE 3 | Risk of bias; lack of thoughtful reasoning | support exams ChatGPT passing the USMLE revealed the deficiencies in medical education and assessment; there is a need to reevaluate the current medical students' training and educational tools |
表 1: 包含社论/致编辑信的主要结论摘要
| 作者 [文献] | 设计与目标 | 应用与优势 | 风险、担忧与局限 | 建议行动与结论 |
|---|---|---|---|---|
| Chen [23] | 关于ChatGPT在科学写作中应用的社论 | ChatGPT有助于克服语言障碍，促进研究公平性 | 伦理担忧(代笔)；准确性存疑；引用问题 | 以开放心态拥抱创新；作者应掌握正确使用AI工具的知识 |
| Thorp [24] | 社论："ChatGPT不能成为作者" | 内容非原创；看似合理实则错误的答案；参考文献问题；抄袭风险 | 修订教育评估方式；《Science》期刊将使用ChatGPT视为学术不端行为 | |
| Kitamura [25] | 关于ChatGPT与医学写作未来的社论 | 提高医学写作效率；翻译用途 | 伦理问题，抄袭；缺乏原创性；内容不准确；偏见风险 | "人在主导，AI辅助" |
| Lubowitz [26] | 社论，ChatGPT对医学文献的影响 | 内容不准确；偏见风险；错误与虚假信息传播；缺乏参考文献；文本冗余 | 作者不应使用ChatGPT撰写科学投稿的任何部分；但在严格人工监督下可使用，以确保科研工作的完整性与原创性 | |
| Nature [27] | 《自然》关于ChatGPT使用规则以确保科学透明性的社论 | ChatGPT可帮助总结研究论文；生成实用代码 | 伦理问题；透明度问题 | 大语言模型工具可作为作者被接受；若使用大语言模型工具，应在方法或致谢部分说明；倡导方法透明及研究者的诚信与真实 |
| Moons和Van Bulck [28] | 关于ChatGPT在心血管护理实践与研究潜力的社论 | ChatGPT可帮助总结长文本；促进研究者工作；辅助 | 信息准确性问题；知识截止至2021年；能力有限 | ChatGPT可成为医疗实践中的有价值工具 |
| Cahan和Treutlein [29] | 报告与ChatGPT关于干细胞研究对话的社论 | 数据收集 ChatGPT帮助节省时间 | 重复；部分回答缺乏深度与洞察力；缺少参考文献 | ChatGPT帮助撰写社论，节省宝贵时间 |
| Ahn [30] | 致编辑信，报告与ChatGPT关于CPR的对话 | 个性化互动；响应迅速；可提供易获取且易懂的信息 | 可能生成不准确信息，导致严重医疗后果 | 探索ChatGPT在提供CPR信息与教育方面的潜在用途 |
| Gunawan [31] | 报告与ChatGPT关于护理未来对话的社论 | 关于向公众普及CPR ChatGPT可提高效率；帮助减少 | 缺乏情感与个人支持 | ChatGPT可为医疗保健提供有价值视角 |
| D'Amico等 [32] | 报告将Chatbot整合到神经外科实践与研究的社论 | 护理服务 ChatGPT可帮助为患者提供及时准确的治疗信息 | 信息不准确可能性；隐私担忧；伦理问题；法律问题 | 神经外科实践可引领将ChatGPT应用于患者护理与研究 |
| Fijacko等 [33] | 关于ChatGPT对生命支持考试回答准确性的致编辑信 | 护理 ChatGPT有时能提供相关准确回答 | 偏见风险 参考文献问题；回答过于详细 | ChatGPT未通过任何考试；但可作为备考生命支持考试的有力自学工具 |
| Mbakwe等 [34] | 关于ChatGPT通过USMLE能力的社论 | 偏见风险；缺乏深思熟虑的推理 | ChatGPT通过USMLE暴露了医学教育与评估的缺陷；需要重新评估现有医学生培训与教育工具 |
Table 1. Cont.
| Author(s) [Record] | Design, Aims | Applications,Benefits | Risks, Concerns, Limitations | Suggested Action, Conclusions |
| Huh [35] | An editorial of JEEHP 4 policy towards ChatGPT use | Reponses were not accurate in someareas | JEEHP will not accept ChatGPT as an author; however, ChatGPT content can be used if properly cited and documented | |
| O'Connor * [36] | An editorial writtenwith ChatGPT assistance on ChatGPT in nursing education | ChatGPT can help to provide a personalized learning experience ChatGPT can help in the | Risk of plagiarism; biased or misleading results Risk of hallucination | Advocate ethical and responsible use of ChatGPT; improve assessment in nursing education |
| Shen et al. [37] | An editorial on ChatGPT strengths and limitations | generation of medical reports; providing summaries ofmedical records; drafting letters to the insurance provider; improving the interpretability of CAD 5 systems | (inaccurateinformation that sounds plausible scientifically); the need to carefully construct ChatGPT prompts; possible inaccurate or incomplete results; dependence on the training data; risk of bias; risk of research fraud | Careful use of ChatGPT is needed to exploit its powerful applications |
| Gordijn and Have [38] | Editorial on the revolutionary nature of ChatGPT | Risk of factual inaccuracies; risk of plagiarism; risk of fraud; copyright infringements possibility | In the near future, LLM can write papers with the ability to pass peer review; therefore, the scientific community should be prepared to address this serious issue | |
| Mijwil et al. [39] | An editorial on the role of cybersecurity in the protection of medical information | Versatility; efficiency; high quality of text generated; cost saving; innovation potential; improved decision making; improved diagnostics; predictive modeling; improved personalized medicine; streamline clinical workflow increasing efficiency; remote monitoring | Data security issues | The role of cybersecurity to protect medical information should be emphasized |
| TheLancet Digital Health [40] | An editorial on the strengths and limitations of ChatGPT | ChatGPT can help to improve language and readability | Over-detailed content;incorrect or biased content; potential to generate harmful errors; risk of spread of misinformation; risk of plagiarism; issues with | Widespread use of ChatGPT is inevitable; proper documentationofChatGPTuse is needed; ChatGPT should not belisted or cited as an author |
| Aljanabi et al. [41] | An editorial on the possibilities provided by ChatGPT | ChatGPT can help in academic writing; helpful in code generation | integrity of scientific records Inaccurate content including inabilityhandlemathematical calculations to reliably | or co-author ChatGPT will receive a growing interest in the scientific community |
1 CPR: cardiopulmonary resuscitation; 2 AHA: American Heart Association; 3 USMLE: United States Medical Licensing Examination; 4 JEEHP: Journal of Educational Evaluation for Health Professions; 5 CAD: computeraided diagnosis; 6 AI: artificial intelligence; 7 LLM: large-scale language model. * ChatGPT generative pre-trained transformer was listed as an author.
表 1. （续）
| 作者 [记录] | 设计与目标 | 应用与优势 | 风险、问题与局限性 | 建议行动与结论 |
|---|---|---|---|---|
| Huh [35] | JEEHP 4 关于 ChatGPT 使用政策的社论 | 在某些领域回应不准确 | JEEHP 不接受 ChatGPT 作为作者；但如果正确引用和记录，可以使用 ChatGPT 生成的内容 | |
| O'Connor * [36] | 一篇关于 ChatGPT 在护理教育中应用的社论，由 ChatGPT 辅助撰写 | ChatGPT 可提供个性化学习体验；有助于生成医疗报告、病历摘要、保险信函草拟，并提升 CAD 5 系统的可解释性 | 剽窃风险；结果可能存在偏见或误导；幻觉风险（看似科学但实际不准确的信息） | 倡导 ChatGPT 的伦理与负责任使用；改进护理教育评估 |
| Shen 等 [37] | 关于 ChatGPT 优势与局限的社论 | 需谨慎构建提示词；结果可能不准确或不完整；依赖训练数据；存在偏见风险；研究欺诈风险 | 需谨慎使用 ChatGPT 以发挥其强大应用价值 | |
| Gordijn 和 Have [38] | 关于 ChatGPT 革命性本质的社论 | 事实错误风险；剽窃风险；欺诈风险；可能侵犯版权 | 短期内大语言模型可撰写通过同行评审的论文，科学界应为此严峻问题做好准备 | |
| Mijwil 等 [39] | 关于网络安全在医疗信息保护中作用的社论 | 多功能性；高效性；生成文本质量高；节约成本；创新潜力；改进决策；优化诊断；预测建模；提升个性化医疗；简化临床工作流程；远程监测 | 数据安全问题 | 应强调网络安全对医疗信息的保护作用 |
| TheLancet Digital Health [40] | 关于 ChatGPT 优势与局限的社论 | ChatGPT 可提升语言表达与可读性 | 内容过度详细；存在错误或偏见内容；可能生成有害错误；传播错误信息风险；剽窃风险；科学记录完整性问题 | ChatGPT 的广泛使用不可避免；需规范使用记录；不应将 ChatGPT 列为作者或引用 |
| Aljanabi 等 [41] | 关于 ChatGPT 可能性的社论 | ChatGPT 有助于学术写作；辅助代码生成 | 内容不准确（包括无法可靠处理数学计算） | ChatGPT 将日益受到科学界关注 |
1 CPR: 心肺复苏；2 AHA: 美国心脏协会；3 USMLE: 美国医师执照考试；4 JEEHP: 健康专业教育评价杂志；5 CAD: 计算机辅助诊断；6 AI: 人工智能；7 LLM: 大语言模型。* ChatGPT 生成式预训练 Transformer 被列为作者。
Table 2. A summary of the main conclusions of the included records comprising research articles, commentaries, news articles, perspectives, case studies, brief reports, communications, opinions, or recommendations.
| Author(s) [Record] | Design, Aims | Applications,Benefits | Risks,Concerns,Limitations | Suggested Action,Conclusions |
| Stokel- Walker [13] | News explainer | Well-organized content withdecentreferences;a free package Original, precise, and | Imminentendofconventional educationalassessment; concernsregardingthe effecton humanknowledge and ability Failuretofollow the | Theneedtoreviseeducational assessmenttoolstoprioritize criticalthinking orreasoning |
| Kumar [42] | Briefreport;assessment of ChatGPTforacademic writinginbiomedicine | accurate responses with systematic approach; helpful for training and to improving topic clarity; efficiency in time; promotingmotivation towrite | instructionscorrectlyon occasions;failuretocite referencesin-text;inaccurate references;lack of practical examples;lackofpersonal experience highlights; superficialresponses | ChatGPT can help in improving academic writing skills; advocateforuniversaldesignfor learning; proper use of ChatGPT under academicmentoring |
表 2: 纳入记录(包括研究论文、评论、新闻文章、观点、案例研究、简报、通讯、意见或建议)的主要结论汇总
| 作者 [记录] | 设计与目标 | 应用与优势 | 风险/顾虑/局限性 | 建议措施与结论 |
|---|---|---|---|---|
| Stokel-Walker [13] | 新闻解读 | 结构清晰的内容与可靠参考文献；免费资源包；原创、精确且 | 传统教育评估即将终结；对人类知识与能力影响的担忧 | 需要改革教育评估工具以优先考察批判性思维与推理能力 |
| Kumar [42] | 简报；评估ChatGPT在生物医学学术写作中的应用 | 系统方法生成准确回答；有助于培训与提升主题明确性；节省时间；提升写作动机 | 偶尔无法正确遵循指令；未标注文中引用；参考文献不准确；缺乏实际案例；缺少个人经验描述；回答流于表面 | ChatGPT可帮助提升学术写作能力；倡导通用学习设计；建议在学术指导下合理使用ChatGPT |
Table 2. Cont.
| Author(s) [Record] | Design, Aims | Applications, Benefits | Risks, Concerns, Limitations | Suggested Action, Conclusions ChatGPT does not meet ICMJE 4 |
| Zielinski et al. [43] | WAME1 recommendations on ChatGPT | ChatGPT can be a useful tool for researchers | Risk of incorrect or non-sensical answers; restricted knowledge to the period before 2021; lack of legal personality; risk of plagiarism | criteria and cannot belisted as an author; authors should be transparent regarding ChatGPT use and take responsibility for its content; editors need appropriate detection tools for ChatGPT-generated content |
| Biswas [44] | A perspective record on the future of medical writing in light of ChatGPT | Improved efficiency in medical writing | Suboptimal understanding of the medicalfield;ethical concerns; risk of bias; legal issues;transparencyissues | A powerful tool in the medical field;however, several limitations of ChatGPTshould be considered |
| Stokel- Walker [45] | A news article on the view of ChatGPT as an author | ChatGPT can help to | Risk of plagiarism; lack of accountability; concerns about misuse in the academia | ChatGPT should not be considered as an author ChatGPT ban will not work; develop rules for accountability, |
| van Dis et al. [46] | A commentary on the priorities for ChatGPT research | accelerate innovation; to increase efficiency in publication time; it can make science more equitable and increase the diversity of scientific perspectives; more free time for experimental designs; it could optimize academic training | Compromised research quality; transparency issues; risk of spread of misinformation; inaccuracies in content, risk of bias and plagiarism; ethical concerns; possible future monopoly; lack of transparency | integrity, transparency and honesty; carefully consider which academic skills remain essential to researchers; widen the debateinthe academia; an initiative is needed to address the development and responsible use of LLM 5 for research |
| Lund and Wang [47] | News article on ChatGPT impact in academia | Useful for literature review; can help in data analysis; can help in translation | Ethical concerns,issues about data privacy and security; risk of bias; transparency issues | ChatGPT has the potential to advance academia; consider how to use ChatGPT responsibly and ethically Robust AI authorship guidelines |
| Liebrenz et al. [48] | A commentary on the ethical issues of ChatGPT use in medical publishing | ChatGPT can help to overcome language barriers | Ethical issues (copyright, attribution, plagiarism, and authorship); inequalities in scholarly publishing; risk of spread of misinformation; inaccurate content | are needed in scholarly publishing; COPE AI6 should be followed;AI cannot belisted as an author and it must be properly acknowledged upon its use |
| Manohar and Prasad [49] | A case study written with ChatGPT assistance | ChatGPT helped to generate a clear, comprehensible text | Lack of scientific accuracy and reliability; citation inaccuracies | ChatGPT use is discouraged due to risk of false information and non-existent citations; it can be misleading in health care practice |
| Akhter and Cooper [50] | A case study written with ChatGPT assistance | ChatGPT helped to provide a relevant general introductory summary | Inability to access relevant literature; the limited knowledge up to 2021; citation inaccuracies; limited ability to critically discuss results | Currently, ChatGPT cannot replaceindependentliterature reviews in scientific research |
| Holzinger et al. [51] | An article on AI2/ ChatGPT use in biotechnology | Biomedical image analysis; diagnostics and disease prediction; personalized medicine; drug discovery and development | Ethical and legal issues; limited data availability to train the models; the issue of reproducibility of the runs | The scientists aspire for fairness, open science, and open data |
| Mann [52] | A perspective on ChatGPT intranslationalresearch | Efficiency in writing; analysis oflarge datasets (e.g., electronic health records or genomic data); predict risk factors for disease; predict disease outcomes | Compromised quality of data available; inability to understand the complexity of biologic systems | ChatGPT use in scientific and medical journals is inevitable in near future |
| Patel and Lam [53] | A commentary on ChatGPT utility in documentation of discharge summary | ChatGPT can help to reduce the burden of discharge summaries providing high-quality and efficient output | Data governance issues; risk of depersonalization of care; risk of incorrect or inadequate information | Proactive adoption of ChatGPT is needed to limit any possible future issues and limitations |
表 2. （续）
| 作者 [记录] | 设计与目标 | 应用与优势 | 风险、担忧与局限性 | 建议行动与结论 ChatGPT 不符合 ICMJE 4 项标准 |
|---|---|---|---|---|
| Zielinski 等 [43] | WAME1 关于 ChatGPT 的建议 | ChatGPT 可作为研究人员的实用工具 | 存在错误或无意义答案的风险；知识仅限于 2021 年前；缺乏法律人格；抄袭风险 | 不能列为作者；作者应透明披露 ChatGPT 使用情况并对其内容负责；编辑需配备检测 ChatGPT 生成内容的工具 |
| Biswas [44] | 关于 ChatGPT 对医学写作未来影响的观点记录 | 提升医学写作效率 | 对医学领域理解不足；伦理问题；偏见风险；法律问题；透明度问题 | 医学领域的强大工具，但需考虑 ChatGPT 的多项局限性 |
| Stokel-Walker [45] | 关于将 ChatGPT 视为作者的新闻文章 | ChatGPT 可协助 | 抄袭风险；缺乏问责；学术界滥用担忧 | ChatGPT 不应被视为作者；禁令无效，需制定问责规则 |
| van Dis 等 [46] | 关于 ChatGPT 研究优先事项的评论 | 加速创新；缩短发表时间；促进科学公平性与视角多样性；为实验设计腾出时间；优化学术培训 | 研究质量受损；透明度问题；错误信息传播风险；内容不准确、偏见与抄袭；伦理问题；未来垄断可能；缺乏透明度 | 需确保诚信、透明与诚实；审慎评估研究者必备技能；扩大学术界讨论；需发起倡议推动大语言模型 (LLM) 5 的负责任研究应用 |
| Lund 和 Wang [47] | 关于 ChatGPT 对学术界影响的新闻文章 | 助力文献综述；辅助数据分析；协助翻译 | 伦理问题；数据隐私与安全隐患；偏见风险；透明度问题 | ChatGPT 有推动学术进步的潜力；需考虑负责任与伦理的使用方式；需健全 AI 作者指南 |
| Liebrenz 等 [48] | 关于医学出版中 ChatGPT 使用伦理问题的评论 | ChatGPT 可帮助克服语言障碍 | 伦理问题（版权、归属、抄袭与作者身份）；学术出版不平等；错误信息传播风险；内容不准确 | 学术出版需遵循 COPE AI6 准则；AI 不能列为作者且使用时应明确标注 |
| Manohar 和 Prasad [49] | 借助 ChatGPT 撰写的案例研究 | ChatGPT 帮助生成清晰易懂的文本 | 缺乏科学准确性与可靠性；引用不准确 | 因虚假信息与虚构引用风险不推荐使用；在医疗实践中可能产生误导 |
| Akhter 和 Cooper [50] | 借助 ChatGPT 撰写的案例研究 | ChatGPT 协助提供相关通用性介绍摘要 | 无法获取相关文献；知识限于 2021 年前；引用不准确；结果批判性讨论能力有限 | 目前 ChatGPT 无法替代科研中的独立文献综述 |
| Holzinger 等 [51] | 关于 AI2/ChatGPT 在生物技术中应用的文章 | 生物医学图像分析；诊断与疾病预测；个性化医疗；药物研发 | 伦理与法律问题；模型训练数据有限；实验可复现性问题 | 科学家追求公平性、开放科学与开放数据 |
| Mann [52] | 关于 ChatGPT 在转化研究中应用的观点 | 提升写作效率；分析大型数据集（如电子健康记录或基因组数据）；预测疾病风险因素；预测疾病结果 | 数据质量受损；无法理解生物系统复杂性 | ChatGPT 在科学与医学期刊中的应用在未来不可避免 |
| Patel 和 Lam [53] | 关于 ChatGPT 在出院小结记录中效用的评论 | ChatGPT 可减轻出院小结负担，提供高效优质输出 | 数据治理问题；诊疗去个性化风险；信息错误或不完整风险 | 需主动采用 ChatGPT 以规避未来潜在问题与限制 |
Table 2. Cont.
| Author(s) | |||||
| [Record] Zhavoronkov * [54] | Design, Aims A perspective reporting a conversationwith ChatGPT about rapamycin use from | Applications, Benefits ChatGPT provided correct summary of rapamycin side effects;it referred to theneed to consult a health | Risks, Concerns, Limitations | Suggested Action, Conclusions Demonstration of ChatGPT's potential to generate complex philosophical arguments | |
| Hallsworth et al. [55] | a philosophical perspective A comprehensive opinion article submitted before ChatGPT launching on the value of | care provider based on the specific situation It can help to circumvent language barriers; it can robustly help to process massive data in short time; it can stimulate creativity | Ethical issues; legal responsibility issues; lack of empathy and personal | Despite the AI potential in science, there is an intrinsic value of human engagement in the scientific process which | |
| Stokel- Walker and Van | theory-based research Naturenewsfeature article on ChatGPT implications | by humans if "AI in the Loop: Humans in Charge" is applied More productivity among researchers | communication; lack oftransparency Problems in reliability and factual inaccuracies; misleading information that seem plausible | cannot be replaced by AI contribution "AI in the Loop: Humans in Charge" should be used; ChatGPT widespread use in the | |
| Noorden [14] Huh [56] | in science A study to compare ChatGPT performance on a parasitology exam to the performance of Korean | ChatGPT performance will improve by deep learning | (hallucination); over-detailed content; risk of bias; ethical issues; copyright issues ChatGPT performance was lower compared to medical students; plausible explanations for incorrect | nearfuture would beinevitable ChatGPT performance will continue to improve, and health care educators/students are advised to incorporate this tool | |
| Khan et al. [57] | medical students A communication on ChatGPT use in medical education and clinical management | ChatGPT can help in automated scoring; assistance in teaching; improved personalized learning; assistance in research; generation of clinical vignettes; rapid access to information; translation; documentation in clinical practice; support | answers (hallucination) Lack of human-like understanding; the limited knowledge up to 2021 | into the educational process ChatGPT is helpful in medical education, research, and in clinical practice; however, the human capabilities are still needed | |
| Gilson et al. [58] | An article on the performance of ChatGPT on USMLE3 | in clinical decisions; personalized medicine Ability to understand context and to complete a coherent and relevant conversation in the medical field; can be used as an | The limited knowledge up to 2021 | ChatGPT passes the USMLE with performance at a 3rd year medical student level; can help to facilitate learning as a virtual medical tutor | |
| Kung et al. [59] | An article showing the ChatGPT raised accuracy which enabled passing the USMLE | adjunct in group learning Accuracy withhigh concordance and insight; it can facilitate patient communication; improved personalized medicine | ChatGPT has a promising potential in medical education; future studies are recommended to consider non-biased approach withquantitativenatural language processing and text mining tools such as word | ||
| Marchandot et al. [60] | A commentary on ChatGPT use in academic writing | ChatGPT can assist in literature review saving time; the ability to summarize papers; the ability to improve language | Risk of inaccurate content;risk of bias; ChatGPT may lead to decreased critical thinking and creativity in science; ethical concerns; risk of plagiarism | network analysis ChatGPT can be listed as an author based on its significant contribution | |
1 WAME: World Association of Medical Editors; 2 AI: artificial intelligence; 3 USMLE: United States Medical Licensing Examination; 4 ICMJE: International Committee of Medical Journal Editors; 5 LLM: large-scale language model; 6 COPE AI in decision making: Committee on Publication Ethics, Artificial Intelligence (AI) in decision making, available from: https://publication ethics.org/node/50766, accessed on 18 February 2023. * ChatGPT generative pre-trained transformer was listed as an author.
表 2. 续
| 作者 | 设计与目标 | 应用与优势 | 风险、担忧与局限 | 建议行动与结论 |
|---|---|---|---|---|
| Zhavoronkov * [54] | 从哲学视角报道与ChatGPT关于雷帕霉素应用的对话 | ChatGPT正确总结了雷帕霉素副作用；建议根据具体情况咨询医疗专业人员 | 展示了ChatGPT生成复杂哲学论证的潜力 | |
| Hallsworth等 [55] | 在ChatGPT发布前提交的关于理论研究的综合观点文章 | 可帮助克服语言障碍；快速处理海量数据；激发创造力 | 伦理问题；法律责任；缺乏同理心与个人沟通；透明度不足 | 尽管AI在科学中有潜力，但人类参与科研过程的内在价值无法被AI取代 |
| Stokel-Walker与Van Noorden [14] | Nature新闻特写关于ChatGPT对科学的影响 | 若采用"AI在环：人类主导"模式可提升研究人员生产力 | 可靠性与事实准确性存疑；看似合理但误导的信息(幻觉)；内容过度详细；偏见风险；伦理问题；版权问题 | "AI在环：人类主导"应被采用；ChatGPT在近未来的广泛使用不可避免 |
| Huh [56] | 比较ChatGPT与韩国医学生在寄生虫学考试表现的研究 | ChatGPT表现将通过深度学习提升 | 表现低于医学生；错误答案的合理解释(幻觉) | ChatGPT表现将持续改进，建议医学教育者/学生将该工具纳入教学 |
| Khan等 [57] | 关于ChatGPT在医学教育与临床管理中应用的通讯 | 可辅助自动评分、教学、个性化学习、研究、临床案例生成、快速获取信息、翻译、临床记录、临床决策支持、个性化医疗 | 缺乏类人理解；知识截止至2021年 | ChatGPT有助于医学教育、研究和临床实践，但仍需人类能力 |
| Gilson等 [58] | 关于ChatGPT在美国医师执照考试(USMLE)表现的文章 | 能理解医学领域语境并完成连贯对话；可作为小组学习辅助工具 | 知识截止至2021年 | ChatGPT以三年级医学生水平通过USMLE；可作为虚拟医学导师促进学习 |
| Kung等 [59] | 展示ChatGPT提高准确率从而通过USMLE的文章 | 高一致性准确率与洞察力；改善医患沟通；促进个性化医疗 | ChatGPT在医学教育中潜力巨大；建议未来研究采用无偏见的定量自然语言处理与文本挖掘工具 | |
| Marchandot等 [60] | 关于ChatGPT在学术写作中应用的评论 | 可辅助文献综述节省时间；总结论文能力；改进语言能力 | 内容不准确风险；偏见风险；可能降低科研批判思维与创造力；伦理担忧；抄袭风险 | 基于显著贡献，ChatGPT可被列为作者 |
1 WAME：世界医学编辑协会；2 AI：人工智能；3 USMLE：美国医师执照考试；4 ICMJE：国际医学期刊编辑委员会；5 LLM：大语言模型；6 COPE AI决策：出版伦理委员会人工智能决策，来源：https://publicationethics.org/node/50766，访问于2023年2月18日。* ChatGPT生成式预训练Transformer被列为作者。
Table 3. A summary of the main conclusions of the included records representing preprints.
| Author(s) | |||||
| [Record] Wang et al. [61] | Design, Aims An arXiv preprint 1; investigating ChatGPT effectiveness to generate Boolean queries for | Applications, Benefits Higher precision compared to the current automatic | Risks, Concerns, Limitations Non-suitability for high-recall retrieval; many incorrect MeSH 11 terms; variability in query | Suggested Action, Conclusions A promising tool for research | |
| systematic literature reviews An arXiv preprint; to | query formulation methods | effectiveness across multiple requests; a black-box application Problems in spatial, temporal, physical, psychological and logical reasoning; limited capability to calculate mathematical expressions; factual errors; risk of bias and discrimination; difficulty in using idioms; lack of real | Implementation ofresponsible use and precautions; proper | ||
| Borji [20] | highlight the limitations of ChatGPT | Extremely helpful in scientific writing | emotions and thoughts; no perspective for the subject; over-detailed; lacks human-like divergences; lack of transparency and reliability; security concerns with vulnerability to data poisoning; violation of data privacy; plagiarism; impact on the environment and climate; ethical andsocial consequences | monitoring; transparent communication; regular inspection for biases, misinformation, among other harmful purposes (e.g., identity theft) | |
| Cotton et al. [62] | An EdArXiv 2 preprint on the academic integrity in ChatGPT era A bioRxiv 3 preprint | Risk of plagiarism; academic dishonesty | Careful thinking of educational assessment tools | ||
| Gao et al. [63] | comparing the scientific abstracts generated by ChatGPT to original abstracts | A tool to decrease the burden of writing and formatting; it can help to overcome language barriers ChatGPT can help to | Misuse to falsify research; risk of bias | The use of ChatGPTinscientific writing or assistance should be clearly disclosed and documented | |
| Polonsky and Rotman [64] | An SSRN 4 preprint on listing ChatGPT as an author | accelerate the research process; it can help to increase accuracy and precision | Intellectual property issues if financial gains are expected | AI 12 can be listed as an author in some instances | |
| Aczel and Wagenmak- ers [65] | A PsyArXiv 5 preprint as a guide of transparent ChatGPT use in scientific writing | Issues of originality, transparency issues | There is a need to provide sufficient information on ChatGPT use, with accreditation and verification of its use Carefully weigh ChatGPT | ||
| De Angelis et al. [66] | An SSRN preprint discussing the concerns of an AI-driven infodemic | ChatGPT can support and expedite academic research | Generation ofmisinformation and the risk of subsequent infodemics; falsified or fake research; ethical concerns | possible benefits with its possible risks; there is a need to establish ethical guidelines for ChatGPT use; a science-driven debateis neededto address ChatGPT utility | |
| Benoit [67] | A medRxiv 6 preprint on the generation, revision, and evaluation of clinical vignettes as a tool in health education using ChatGPT | Consistency, rapidity and flexibility of text and style; ability to generate plagiarism-free text | Clinical vignettes' ownership issues; inaccurate or non-existent references Risk of bias or inaccuracies; | ChatGPT can allow for improved medical education and better patient communication | |
| Sharma and Thakur [68] | A ChemRxiv 7 preprint on ChatGPT possible use in drug discovery | ChatGPT can help to identify and validate new drug targets; to design new drugs; to optimize drug properties; to assess toxicity; and to generate drug-related reports | inability to understand the complexity of biologic systems; transparency issues;lack of experimental validation; limited interpretability; limited handling of uncertainty; ethical issues | ChatGPT can be a powerful and promising tool in drug discovery; however, its accompanying ethical issues should be addressed | |
表 3: 收录预印本记录的主要结论摘要
| 作者 | 设计与目标 | 应用与优势 | 风险、问题与局限性 | 建议行动与结论 |
|---|---|---|---|---|
| Wang 等 [61] | 一项 arXiv 预印本；研究 ChatGPT 为系统文献综述生成布尔查询的有效性 | 相比当前自动查询生成方法具有更高精度 | 不适用于高召回检索；许多错误的 MeSH 术语；查询效果随请求变化；黑箱应用问题 | 研究中的潜力工具；需负责任使用并采取预防措施 |
| Borji [20] | 强调 ChatGPT 的局限性 | 对科学写作极有帮助 | 空间、时间、物理、心理和逻辑推理问题；数学计算能力有限；事实错误；偏见与歧视风险；难以使用习语；缺乏真实情感与思考；透明性与可靠性不足；数据中毒漏洞的安全问题；数据隐私侵犯；剽窃；对环境和气候的影响；伦理与社会后果 | 需透明沟通；定期检查偏见、错误信息及其他有害用途（如身份盗用） |
| Cotton 等 [62] | 一项 EdArXiv 预印本，探讨 ChatGPT 时代的学术诚信 | 剽窃风险；学术不端 | 需谨慎设计教育评估工具 | |
| Gao 等 [63] | 一项 bioRxiv 预印本，比较 ChatGPT 生成的科学摘要与原始摘要 | 减轻写作与格式负担的工具；可帮助克服语言障碍 | 滥用以伪造研究；偏见风险 | 在科学写作或辅助中使用 ChatGPT 应明确披露并记录 |
| Polonsky 和 Rotman [64] | 一项 SSRN 预印本，讨论将 ChatGPT 列为作者 | 加速研究进程；提高准确性与精确度 | 若涉及经济利益则存在知识产权问题 | 某些情况下可将 AI 列为作者 |
| Aczel 和 Wagenmakers [65] | 一项 PsyArXiv 预印本，作为科学写作中透明使用 ChatGPT 的指南 | 原创性问题；透明度问题 | 需提供 ChatGPT 使用的充分信息，并验证其使用 | |
| De Angelis 等 [66] | 一项 SSRN 预印本，讨论 AI 驱动信息疫情的担忧 | ChatGPT 可支持并加速学术研究 | 生成错误信息及后续信息疫情风险；伪造研究；伦理问题 | 需权衡利弊；制定 ChatGPT 使用的伦理指南；开展科学驱动的讨论 |
| Benoit [67] | 一项 medRxiv 预印本，探讨 ChatGPT 作为健康教育工具生成、修订和评估临床案例 | 文本与风格的一致性、快速性和灵活性；可生成无剽窃文本 | 临床案例所有权问题；不准确或不存在的参考文献 | 可改善医学教育与患者沟通 |
| Sharma 和 Thakur [68] | 一项 ChemRxiv 预印本，探讨 ChatGPT 在药物发现中的潜在用途 | 帮助识别和验证新药物靶点；设计新药；优化药物性质；评估毒性；生成药物相关报告 | 无法理解生物系统复杂性；透明度问题；缺乏实验验证；可解释性有限；不确定性处理能力有限；伦理问题 | ChatGPT 是药物发现中强大且有潜力的工具，但需解决其伦理问题 |
Table 3. Cont.
| Author(s) [Record] | Design, Aims | Applications, Benefits | Risks,Concerns,Limitations Lack of references; alignment | Suggested Action, Conclusions Using ChatGPT for radiologic |
| Rao et al. [69] | A medRxiv preprint on the usefulness of ChatGPT in radiologic decision making | ChatGPT showed moderate accuracy to determine appropriate imaging steps in breast cancer screening and evaluation of breast pain | with userintent;inaccurate information; over-detailed; recommending imaging in futile situations; providing rationale for incorrect imaging decisions; the black box nature with lack of transparency | decision making is feasible, potentially improving the clinical workflow and responsible use of radiology services |
| Antaki et al. [70] | A medRxiv preprint assessing ChatGPT's ability to answer a diverse MCQ 8 exam in ophthalmology | ChatGPT currently performs at the level of an average first-year ophthalmology resident | Inability to process images; risk of bias; dependence on training dataset quality | There is a potential of ChatGPT use in ophthalmology; however, its applications should be carefully addressed Expression of knowledge can be |
| Aydin and Karaarslan [71] | An SSRN preprint on the use of ChatGPT to conduct a literature review on digital twin in health care | Low risk of plagiarism; accelerated literature review; more free time for researchers ChatGPT can provide | Lack of originality | accelerated using ChatGPT; further work will use ChatGPT in citation analysis to assess the attitude towards the findings |
| Sanmarchi et al. [72] | A medRxiv preprint evaluating ChatGPTvalue in an epidemiologic study following the STROBE 9 recommendations | appropriate responses if proper constructs are developed; more free time for researchers to focus on experimental phase Generation of rapid and accurate responses; easily | Risk of bias in the training data; risk of devaluation of human expertise; risk of scientific fraud; legal issues; reproducibility issues | Despite ChatGPT possible value, the research premise and originality will remain the function of human brain |
| Duong and Solomon [73] | A medRxiv preprint evaluating ChatGPT versus human responses to questions on genetics | accessibleinformationfor the patients with genetic disease and theirfamilies;it can help can health professionals in the diagnosis and treatment of genetic diseases; it could make genetic information widely available and help non-experts to understand suchinformation | Plausible explanations for incorrect answers (hallucination); reproducibility issues | Thevalue of ChatGPTwill increase in research and clinical settings |
| Yeo et al. [74] | A medRxiv preprint evaluating ChatGPT responses to questions on cirrhosis and hepatocellular carcinoma | Improved health literacy with better patient outcome; free availability; increased efficiency among health providers; emulation of empathetic responses | Non-comprehensive responses; the limited knowledge up to 2021; responses can be limited and not tailored to specific country or region; legal issues | ChatGPT may serve as a useful aid for patients besides the standard of care; future studies on ChatGPT utility arerecommended |
| Basic et al. [75] | An arXiv preprint on the performance of ChatGPT in essay writing compared to masters forensic students in Croatia | Risk of plagiarism; lack of originality; ChatGPT use did not accelerate essay writing | The concerns in the academia towards ChatGPT are not totally justified; ChatGPT text detectors can fail | |
| Hisan and Amri [76] | An RG 10 preprint on ChatGPT use medical education | Generation ofeducational content; useful to learn languages | Ethical concerns; scientific fraud (papermills); inaccurate responses; declining quality of educations with the issues of cheating | Appropriate medical exam design is needed, especially for practical skills |
| Jeblick et al. [77] | An arXiv preprint on ChatGPTutility to simplify and summarize radiology reports | Generation of medical information relevant for the patients;moving towards patient-centered care; cost efficiency | Bias and fairness issues; misinterpretation of medical terms; imprecise responses; odd language; hallucination (plausible yet inaccurate response); unspecific location of injury/disease | Demonstration of the ability of ChatGPT simplified radiology reports; however, the limitations should be considered. Improvements of patient-centered care in radiology could be achieved via ChatGPT use |
| Nisar and Aslam [78] | An SSRN preprint on the assessmentof ChatGPT usefulness to study pharmacology | Good accuracy | Contentwasnotsufficientfor research purposes | ChatGPT can be a helpful self-learning tool |
表 3. 续表
| 作者 [记录] | 设计、目标 | 应用、优势 | 风险、担忧、局限性 | 建议行动、结论 |
|---|---|---|---|---|
| Rao 等 [69] | 一篇 medRxiv 预印本，探讨 ChatGPT 在放射科决策中的实用性 | ChatGPT 在乳腺癌筛查和乳房疼痛评估中表现出中等准确度，能确定适当的成像步骤 | 与用户意图不一致；信息不准确；过于详细；在无效情况下推荐成像；为错误的成像决策提供理由；黑箱性质缺乏透明度 | 使用 ChatGPT 进行放射科决策是可行的，可能改善临床工作流程并促进放射服务的负责任使用 |
| Antaki 等 [70] | 一篇 medRxiv 预印本，评估 ChatGPT 回答眼科多样化多选题考试的能力 | ChatGPT 目前的表现相当于眼科第一年住院医师的平均水平 | 无法处理图像；存在偏见风险；依赖训练数据集质量 | ChatGPT 在眼科领域有潜在应用价值，但其应用需谨慎对待 |
| Aydin 和 Karaarslan [71] | 一篇 SSRN 预印本，探讨使用 ChatGPT 进行医疗保健数字孪生文献综述 | 抄袭风险低；加速文献综述；为研究人员提供更多自由时间 | 缺乏原创性 | 使用 ChatGPT 可加速知识表达；后续工作将利用 ChatGPT 进行引文分析以评估对研究结果的态度 |
| Sanmarchi 等 [72] | 一篇 medRxiv 预印本，遵循 STROBE 建议评估 ChatGPT 在流行病学研究中的价值 | 如果构建得当，能提供恰当回应；为研究人员节省更多时间专注于实验阶段 | 训练数据存在偏见风险；贬低人类专业知识的风险；科学欺诈风险；法律问题；可重复性问题 | 尽管 ChatGPT 可能有价值，但研究前提和原创性仍取决于人脑 |
| Duong 和 Solomon [73] | 一篇 medRxiv 预印本，评估 ChatGPT 与人类回答遗传学问题的对比 | 为遗传病患者及其家属提供可访问信息；可帮助医疗专业人员诊断和治疗遗传病；使遗传信息广泛传播并帮助非专业人士理解 | 对错误答案的合理解释（幻觉）；可重复性问题 | ChatGPT 在研究和临床环境中的价值将增加 |
| Yeo 等 [74] | 一篇 medRxiv 预印本，评估 ChatGPT 对肝硬化和肝细胞癌问题的回答 | 提高健康素养，改善患者预后；免费可用；提高医疗提供者效率；模拟共情回应 | 回答不全面；知识仅限于 2021 年前；回答可能有限且未针对特定国家或地区定制；法律问题 | ChatGPT 可作为患者标准护理的有用辅助工具；建议未来研究 ChatGPT 的实用性 |
| Basic 等 [75] | 一篇 arXiv 预印本，比较 ChatGPT 与克罗地亚法医学硕士生在论文写作中的表现 | 抄袭风险；缺乏原创性；使用 ChatGPT 并未加速论文写作 | 学术界对 ChatGPT 的担忧并非完全合理；ChatGPT 文本检测器可能失效 | |
| Hisan 和 Amri [76] | 一篇 RG 预印本，探讨 ChatGPT 在医学教育中的应用 | 生成教育内容；有助于语言学习 | 伦理问题；科学欺诈（论文工厂）；不准确回应；教育质量下降与作弊问题 | 需要设计适当的医学考试，尤其是实践技能考核 |
| Jeblick 等 [77] | 一篇 arXiv 预印本，探讨 ChatGPT 简化和总结放射学报告的实用性 | 生成与患者相关的医疗信息；迈向以患者为中心的护理；成本效益 | 偏见和公平问题；医学术语误解；回应不精确；语言奇怪；幻觉（看似合理但不准确的回应）；损伤/疾病位置不具体 | 展示了 ChatGPT 简化放射学报告的能力，但应考虑其局限性。通过使用 ChatGPT 可实现放射科以患者为中心的护理改进 |
| Nisar 和 Aslam [78] | 一篇 SSRN 预印本，评估 ChatGPT 在药理学学习中的实用性 | 准确性良好 | 内容不足以用于研究目的 | ChatGPT 可作为一种有用的自学工具 |
Table 3. Cont.
| Author(s) [Record] | Design, Aims | Applications,Benefits | Risks,Concerns,Limitations | SuggestedAction,Conclusions |
| Lin [79] | A PsyArXiv preprint to describeChatGPT'sutility inacademiceducation | Versatility | Hallucination(inaccurate informationthatsounds scientifically plausible); fraudulentresearch;riskof plagiarism; copyright issues | ChatGPThasatransformative long-term potential; embrace ChatGPTanduseittoaugment human capabilities; however, adequate guidelines and codes of conduct are urgently needed |
表 3. (续)
| 作者 [记录] | 设计目标 | 应用与优势 | 风险/顾虑/局限性 | 建议措施与结论 |
|---|---|---|---|---|
| Lin [79] | 一篇描述ChatGPT在学术教育中实用性的PsyArXiv预印本 | 多功能性 | 幻觉(听起来科学合理但不准确的信息)、学术欺诈、剽窃风险、版权问题 | ChatGPT具有变革性的长期潜力，应拥抱该技术并用于增强人类能力，但亟需制定适当的指导方针和行为准则 |
1 arXiv: A free distribution service and an open-access archive for scholarly articles in the fields of physics, mathematics, computer science, quantitative biology, quantitative finance, statistics, electrical engineering and systems science, and economics, materials on arXiv are not peer-reviewed by arXiv, available from: https://arxiv.org/, accessed on 18 February 2023; 2 EdArXiv: A preprint server for the education research community, available from: https://edarxiv.org/, accessed on 19 February 2023; 3 bioRxiv: A free online archive and distribution service for unpublished preprints in the life sciences, available from: https://www.biorxiv.org/, accessed on 19 February 2023; 4 SSRN: Social Science Research Network repository for preprints, available from: https://www.ssrn.com/index.cfm/en/, accessed on 19 February 2023; 5 PsyArXiv: Psychology archive for preprints, available from: https://psyarxiv.com/, accessed on 18 February 2023; 6 medRxiv: Free online archive and distribution server for complete but unpublished manuscripts (preprints) in the medical, clinical, and related health sciences, available from: https://www.medrxiv.org/, accessed on 18 February 2023; 7 ChemRxiv is a free submission, distribution, and archive service for unpublished preprints in chemistry and related areas, available from: https://chemrxiv.org/engage/chemrxiv/public-dashboard, accessed on 18 February 2023; 8 MCQ: Multiple choice exam; 9 STROBE: Strengthening the reporting of observational studies in epidemiology; 10 RG: Research Gate: A commercial social networking site for scientists and researchers, available from: https://www.research gate.net/about, accessed on 19 February 2023; 11 MeSH: Medical Subject Headings; 12 AI: Artificial intelligence.
1 arXiv: 物理学、数学、计算机科学、定量生物学、定量金融学、统计学、电气工程与系统科学以及经济学领域的免费学术文章分发服务和开放获取存档平台，arXiv上的材料未经arXiv同行评审，访问地址: https://arxiv.org/，访问日期: 2023年2月18日；
2 EdArXiv: 教育研究社区的预印本服务器，访问地址: https://edarxiv.org/，访问日期: 2023年2月19日；
3 bioRxiv: 生命科学领域未发表预印本的免费在线存档和分发服务，访问地址: https://www.biorxiv.org/，访问日期: 2023年2月19日；
4 SSRN: 社会科学研究网络的预印本存储库，访问地址: https://www.ssrn.com/index.cfm/en/，访问日期: 2023年2月19日；
5 PsyArXiv: 心理学领域的预印本存档，访问地址: https://psyarxiv.com/，访问日期: 2023年2月18日；
6 medRxiv: 医学、临床及相关健康科学领域已完成但未发表手稿(预印本)的免费在线存档和分发服务器，访问地址: https://www.medrxiv.org/，访问日期: 2023年2月18日；
7 ChemRxiv: 化学及相关领域未发表预印本的免费提交、分发和存档服务，访问地址: https://chemrxiv.org/engage/chemrxiv/public-dashboard，访问日期: 2023年2月18日；
8 MCQ: 多选题考试；
9 STROBE: 加强流行病学观察性研究报告；
10 RG: Research Gate: 面向科学家和研究人员的商业社交网站，访问地址: https://www.researchgate.net/about，访问日期: 2023年2月19日；
11 MeSH: 医学主题词表；
12 AI: 人工智能。
3.2. Characteristics of the Included Records
3.2. 纳入记录的特征
A summary of the record types included in the current review is shown in Figure 2.
当前综述中包含的记录类型汇总如图 2 所示。
Figure 2. Summary of the types of included records $(n=60)$ 0. Preprints (not peer reviewed) are highlighted in grey while published records are highlighted in blue.
图 2: 纳入记录类型汇总 $(n=60)$ 0. 预印本(未经同行评审)以灰色高亮显示，已发表记录以蓝色高亮显示。
One-third of the included records were preprints $(n=20)$ ), with the most common preprint server being medRxiv $(n=6,30.0%)$ , followed by SSRN and arXiv $(n=4,20.0%)$ for each. Editorials/letters to editors were the second most common type of included records $(n=19,31.7%)$ .
收录的记录中有三分之一是预印本$(n=20)$)，最常见的预印本服务器是medRxiv$(n=6,30.0%)$，其次是SSRN和arXiv$(n=4,20.0%)$。社论/致编辑信是第二常见的收录记录类型$(n=19,31.7%)$。
3.3. Benefits and Possible Applications of ChatGPT in Health Care Education, Practice, and Research Based on the Included Records
3.3. ChatGPT在医疗教育、实践及研究中的优势与潜在应用（基于纳入文献）
The benefits of ChatGPT were most frequently cited in the context of academic/scientific writing, which was mentioned in 31 records $(51.7%)$ . Examples included efficiency and versatility in writing with text of high quality, improved language, readability, and translation promoting research equity, and accelerated literature review. Benefits in scientific research followed, which was mentioned in 20 records $(33.3%)$ . Examples included the ability to analyze massive data including electronic health records or genomic data, the availability of more free time for the focus on experimental design, and drug design and discovery. Benefits in health care practice was mentioned by 14 records $(23.3%)$ , with examples including personalized medicine, prediction of disease risk and outcome, streamlining the clinical workflow, improved diagnostics, documentation, cost saving, and improved health literacy. Educational benefits in health care disciplines were mentioned in seven records $(11.7%)$ with examples including the generation of accurate and versatile clinical vignettes, improved personalized learning experience, and being an adjunct in group learning. Being a free package was mentioned as a benefit in two records $(3.3%)_ {\cdot}$ Figure 3).
ChatGPT的优势最常被提及的领域是学术/科学写作，共有31条记录（51.7%）提到这一点。具体优势包括高效、多功能的优质文本撰写能力，改进语言表达、可读性和翻译水平以促进研究公平性，以及加速文献综述过程。其次是科研方面的优势，20条记录（33.3%）提到，例如分析电子健康记录或基因组数据等海量信息的能力，为实验设计、药物研发腾出更多自由时间等。14条记录（23.3%）提及医疗实践中的优势，包括个性化医疗、疾病风险与预后预测、优化临床工作流程、提升诊断与记录效率、降低成本以及改善健康素养等。7条记录（11.7%）提到其在医疗教育领域的价值，例如生成精准多样的临床案例、提升个性化学习体验、作为小组学习辅助工具等。另有2条记录（3.3%）指出其免费使用的优势（图3）。
Benefits/applications of ChatGPT in healthcare settings
医疗环境中ChatGPT的优势与应用
Figure 3. Summary of benefits/applications of ChatGPT in health care education, research, and practice based on the included records.
图 3: 基于纳入文献总结的ChatGPT在医疗教育、研究和实践中的优势/应用
3.4. Risks and Concerns toward ChatGPT in Health Care Education, Practice, and Research Based on the Included Records
3.4. 基于纳入记录的ChatGPT在医疗教育、实践及研究中的风险与担忧
Ethical concerns were commonly mentioned by 33 records $(55.0%)$ , especially in the context of risk of bias (mentioned by 18 records, $30.0%$ ) and plagiarism (mentioned by 14 records, $23.3%$ ) among data privacy and security issues.
33份记录 $(55.0%)$ 普遍提及伦理问题，尤其集中在数据隐私与安全领域的偏见风险（18份记录提及，$30.0%$）和剽窃行为（14份记录提及，$23.3%$）。
Other concerns involved the risk of incorrect/inaccurate information, which was mentioned by 20 records $(33.3%)$ ; citation/reference inaccuracy or inadequate referencing, which was mentioned by 10 records $(16.7%)$ : transparency issues, which was mentioned by 10 records $(16.7%)$ ; legal issues were mentioned in seven records $(11.7%)$ ; restricted knowledge before 2021 was mentioned by six records $(10.0%)$ ; risk of misinformation spread was mentioned by five records $(8.3%)$ ; over-detailed content was mentioned in five records $(8.3%)$ ; copyright issues were mentioned in four records $(6.7%)$ ; and the lack of originality was mentioned by four records $(6.7%$ , Figure 4).
其他关注点包括：错误/不准确信息的风险（20条记录提及，占33.3%）；引用/参考文献不准确或不足（10条记录提及，占16.7%）；透明度问题（10条记录提及，占16.7%）；法律问题（7条记录提及，占11.7%）；2021年前知识受限（6条记录提及，占10.0%）；错误信息传播风险（5条记录提及，占8.3%）；内容过于详细（5条记录提及，占8.3%）；版权问题（4条记录提及，占6.7%）；以及缺乏原创性（4条记录提及，占6.7%，图4）。
Risks/concerns of ChatGPT in healthcare settings
医疗环境中ChatGPT的风险与担忧
Figure 4. Summary of risks/concerns of ChatGPT use in health care education, research, and practice based on the included records.
图 4: 基于纳入文献总结的ChatGPT在医疗教育、研究和实践中使用的风险/问题。
4. Discussion
4. 讨论
The far-reaching consequences of ChatGPT among other LLMs can be described as a paradigm shift in academia and health care practice [16]. The discussion of its potential benefits, future perspectives, and importantly, its limitations, appear timely and relevant [80].
ChatGPT等大语言模型的深远影响可视为学术界和医疗实践领域的一次范式转变[16]。探讨其潜在优势、未来前景以及(更重要的是)局限性显得及时且必要[80]。
Therefore, the current review aimed to highlight these issues based on the current evidence. The following common themes emerged from the available literature.
因此，本次综述旨在基于现有证据重点探讨这些问题。现有文献中浮现出以下共同主题。
4.1. Benefits of ChatGPT in Scientific Research
4.1. ChatGPT 在科研中的优势
ChatGPT, as an example of other LLMs, can be described as a promising or even a revolutionary tool for scientific research in both academic writing and in the research process itself. Specifically, ChatGPT was listed in several sources as an efficient and promising tool for conducting comprehensive literature reviews and generating computer codes, thereby saving time for the research steps that require more efforts from human intelligence (e.g., the focus on experimental design) [14,27,28,41,44,46,47,60,71,72]. Additionally, ChatGPT can be helpful in generating queries for comprehensive systematic review with high precision, as shown by Wang et al., despite the authors highlighting the transparency issues and un suitability for high-recall retrieval [61]. Moreover, the utility of ChatGPT extends to involve an improvement in language and a better ability to express and communicate research ideas and results, ultimately speeding up the publication process with the faster availability of research results [23,25,29,40,48,63,66]. This is particularly relevant for researchers who are non-native English speakers [23,25,63]. Such a practice can be acceptable considering the already existent English editing services provided by several academic publishers. Subsequently, this can help to promote equity and diversity in research [46,55].
以ChatGPT为代表的大语言模型，可被视为学术写作和研究过程中极具潜力甚至革命性的科研工具。具体而言，多篇文献指出ChatGPT能高效完成文献综述和代码生成工作，从而为需要人类投入更多智慧的科研环节（如实验设计）节省时间 [14,27,28,41,44,46,47,60,71,72]。Wang等学者研究表明，尽管存在透明度问题且不适用于高召回率检索，ChatGPT仍能生成高精度的系统综述查询语句 [61]。此外，ChatGPT还能提升语言表达水平，帮助研究者更清晰地传达研究思路与成果，最终通过加速成果发布来缩短出版周期 [23,25,29,40,48,63,66]，这对非英语母语研究者尤为有益 [23,25,63]。考虑到现有学术出版机构提供的英文润色服务，此类应用具有合理性，并有助于促进科研公平性与多样性 [46,55]。
4.2. Limitations of ChatGPT Use in Scientific Research
4.2. ChatGPT在科研应用中的局限性
On the other hand, the use of ChatGPT in academic writing and scientific research should be conducted in light of several limitations that could compromise the quality of research as follows. First, superficial, inaccurate, or incorrect content was frequently cited as a shortcoming of ChatGPT use in scientific writing [14,28,29,40,60]. The ethical issues including the risk of bias based on training datasets and plagiarism were also frequently mentioned, aside from the lack of transparency regarding content generation, which justifies the description of ChatGPT, on occasions, as a black box technology [14,25,26,40,44–48,55,60,63,65,72]. Importantly, the concept of ChatGPT hallucination could be risky if the generated content is not thoroughly evaluated by researchers and health providers with proper expertise [37,56,73,77,79]. This comes in light of the ability of ChatGPT to generate incorrect content that appears plausible from a scientific point of view [81].
另一方面，在学术写作和科学研究中使用ChatGPT时，需注意以下可能影响研究质量的局限性。首先，ChatGPT生成的内容常被指出存在肤浅、不准确或错误的问题[14,28,29,40,60]。除内容生成缺乏透明度外，伦理问题也频繁被提及，包括基于训练数据的偏见风险和剽窃问题，这使得ChatGPT有时被称为黑箱技术[14,25,26,40,44–48,55,60,63,65,72]。值得注意的是，若研究人员和医疗专业人员未以专业能力充分评估生成内容，ChatGPT的幻觉概念可能带来风险[37,56,73,77,79]。这源于ChatGPT能够生成看似科学合理实则错误的內容[81]。
Second, several records mentioned the current problems regarding citation inaccuracies, insufficient references, and ChatGPT referencing to non-existent sources [23,26]. This was clearly shown in two recently published case studies with ChatGPT use in a journal contest [29,49,50]. These case studies discouraged the use of ChatGPT, citing the lack of scientific accuracy, the limited updated knowledge, and the lack of ability to critically discuss the results [38,49,50]. Therefore, the ChatGPT generated content, albeit efficient, should be meticulously examined prior to its inclusion in any research manuscripts or proposals for research grants.
其次，多份记录指出当前存在引用不准确、参考文献不足以及ChatGPT引用虚构来源的问题 [23,26]。近期两项在期刊竞赛中使用ChatGPT的案例研究 [29,49,50] 清晰地展现了这一现象。这些案例研究反对使用ChatGPT，理由是缺乏科学准确性、知识更新有限以及无法批判性讨论结果 [38,49,50]。因此，尽管ChatGPT生成内容效率高，但在将其纳入研究手稿或科研经费申请提案前，仍需仔细核查。
Third, the generation of non-original, over-detailed, or excessive content can be an additional burden for researchers who should carefully supervise the ChatGPT-generated content [14,24–26,65,71]. This can be addressed by supplying ChatGPT with proper prompts (text input), since varying responses might be generated based on the exact approach of prompt construction [72,82].
第三，生成非原创、过于详细或过多的内容可能会给需要仔细监督ChatGPT生成内容的研究人员带来额外负担[14,24–26,65,71]。这一问题可以通过为ChatGPT提供适当的提示(文本输入)来解决，因为根据提示构建的具体方法可能会产生不同的响应[72,82]。
Fourth, as it currently stands, the knowledge of ChatGPT is limited to the period prior to 2021 based on the datasets used in ChatGPT training [6]. Thus, ChatGPT cannot be used currently as a reliable updated source of literature review [83]. Nevertheless, ChatGPT can be used as a motivation to organize the literature in a decent format, if supplemented by reliable and up-to-date references [28,74].
第四，就目前而言，ChatGPT的知识受限于其训练数据集 [6] ，仅覆盖2021年前的信息。因此，ChatGPT目前无法作为可靠的文献综述更新来源 [83] 。但若辅以可靠的最新参考文献 [28,74] ，它仍可作为文献整理的有效激发工具。
Fifth, the risk of research fraud (e.g., ghostwriting, falsified or fake research) involving ChatGPT should be considered seriously [23,37,38,66,72,79] as well as the risk of generating mis- or disinformation with the subsequent possibility of infodemics [26,46,48,66].
第五，需要认真考虑涉及ChatGPT的学术欺诈风险（如代笔、伪造或虚假研究）[23,37,38,66,72,79]，以及生成错误或虚假信息进而引发信息疫情的可能性[26,46,48,66]。
Sixth, legal issues in relation to ChatGPT use were also raised by several records including copyright issues [14,38,44,55,79]. Finally, the practice of listing ChatGPT as an author does not appear to be acceptable based on the current ICMJE and COPE guidelines for determining authorship, as illustrated by Zielinski et al. and Liebrenz et al. [43,48]. This comes in light of the fact that authorship entails legal obligations that are not met by ChatGPT [43,48]. However, other researchers have suggested the possibility of ChatGPT inclusion as an author in some specified instances [60,64].
第六，多份记录提出了与使用ChatGPT相关的法律问题，包括版权问题[14,38,44,55,79]。最后，根据当前ICMJE和COPE关于作者身份认定的指南，将ChatGPT列为作者的做法似乎不可接受，如Zielinski等人和Liebrenz等人所述[43,48]。这是因为作者身份涉及ChatGPT无法履行的法律义务[43,48]。然而，其他研究人员提出了在某些特定情况下将ChatGPT列为作者的可能性[60,64]。
A few instances were encountered in this review, where ChatGPT was listed as an author that can point to the initial perplexity of a few publishers regarding the role of LLM including ChatGPT in research [36,54]. The disapproval of including ChatGPT or any other LLM in the list of authors was clearly explained in Science, Nature, and the Lancet editorials, which referred to such practice as scientific misconduct, and this view was echoed by many scientists [24,27,35,40,45]. In the case of ChatGPT use in the research process, several records advocated the need for the proper and concise disclosure and documentation of ChatGPT or LLM use in the methodology or acknowledgement sections [35,63,65]. A noteworthy and comprehensive record by Borji can be used as a categorical guide for the issues and concerns of ChatGPT use, especially in the context of scientific writing [20].
在本综述中，我们遇到了一些将ChatGPT列为作者的情况，这反映出部分出版商最初对大语言模型(包括ChatGPT)在研究中角色的困惑[36,54]。《科学》《自然》和《柳叶刀》的社论明确指出，将ChatGPT或其他大语言模型列为作者属于学术不端行为，这一观点得到了许多科学家的响应[24,27,35,40,45]。关于在研究过程中使用ChatGPT的问题，多篇文献主张应在方法学或致谢部分对ChatGPT或大语言模型的使用进行恰当而简洁的披露和记录[35,63,65]。Borji提出的一个值得注意且全面的记录，可作为ChatGPT使用问题(特别是在科学写作背景下)的分类指南[20]。
4.3. Benefits of ChatGPT in Health Care Practice
4.3. ChatGPT在医疗实践中的优势
From the health care practice perspective, the current review showed a careful excitement vibe regarding ChatGPT applications. The ability of ChatGPT to help in streamlining the clinical workflow appears promising, with possible cost savings and increased efficiency in health care delivery [31,37,39,77]. This was illustrated recently by Patel and Lam, highlighting the ability of ChatGPT to produce efficient discharge summaries, which can be valuable to reduce the burden of documentation in health care [53]. Additionally, ChatGPT, among other LLMs, can have a transforming potential in health care practice via enhancing diagnostics, prediction of disease risk and outcome, and drug discovery among other areas in translational research [51,52,68]. Moreover, ChatGPT showed moderate accuracy in determining the imaging steps needed in breast cancer screening and in the evaluation of breast pain, which can be a promising application in decision making in radiology [69]. ChatGPT in health care settings also has the prospects of refining personalized medicine and the ability to improve health literacy by providing easily accessible and understandable health information to the general public [30,32,59,73,74]. This utility was demonstrated by ChatGPT responses, highlighting the need to consult health care providers among other reliable sources on specific situations [16,54].
从医疗实践角度来看，当前综述显示出对ChatGPT应用持谨慎乐观态度。ChatGPT在简化临床工作流程方面展现出良好前景，可能降低医疗成本并提高服务效率[31,37,39,77]。Patel和Lam近期研究证实，ChatGPT能高效生成出院小结，有助于减轻医疗文书负担[53]。此外，ChatGPT等大语言模型通过增强诊断能力、疾病风险与预后预测、药物研发等转化研究领域，可能为医疗实践带来变革[51,52,68]。在乳腺癌筛查影像学步骤和乳房疼痛评估方面，ChatGPT表现出中等准确度，有望辅助放射科决策[69]。ChatGPT在医疗场景中还具有优化个性化医疗的潜力，并能通过向公众提供易获取、易理解的健康信息来提升健康素养[30,32,59,73,74]。其响应内容表明，在特定情况下仍需咨询医疗专业人员等可靠信源[16,54]。
4.4. Concerns Regarding ChatGPT Use in Health Care Practice
4.4. 医疗实践中使用ChatGPT的相关顾虑
On the other hand, several concerns regarding ChatGPT use in health care settings were raised. Ethical issues including the risk of bias and transparency issues appeared as recurring major concerns [51,68,69,77]. Additionally, the generation of inaccurate content can have severe negative consequences in health care; therefore, this valid concern should be cautiously considered in health care practice [30,32,53,84]. This concern also extends to involve the ability of ChatGPT to provide justification for incorrect decisions [69].
另一方面，关于ChatGPT在医疗保健领域应用的若干担忧也被提出。伦理问题(包括偏见风险和透明度问题)成为反复出现的主要关切点[51,68,69,77]。此外，生成不准确内容可能对医疗保健造成严重的负面影响，因此这一合理担忧应在医疗实践中被谨慎考量[30,32,53,84]。该担忧还延伸至ChatGPT为错误决策提供合理化解释的能力问题[69]。
Other ChatGPT limitations including the issues of interpret ability, reproducibility, and the handling of uncertainty were also raised, which can have harmful consequences in health care settings including health care research [68,72,73]. In the area of personalized medicine, the lack of transparency and unclear information regarding the sources of data used for ChatGPT training are important issues in health care settings considering the variability observed among different populations in several health-related traits [69]. The issue of reproducibility between the ChatGPT prompt runs is of particular importance, which can be a major limitation in health care practice [51].
ChatGPT的其他局限性还包括可解释性、可重复性以及不确定性处理等问题，这些问题可能对医疗保健环境（包括医疗研究）产生有害影响[68,72,73]。在个性化医疗领域，考虑到不同人群在多项健康相关特征中观察到的差异性，ChatGPT训练数据来源缺乏透明度和信息不明确是医疗保健环境中的重要问题[69]。ChatGPT提示运行间的可重复性问题尤为关键，这可能是医疗实践中的主要限制因素[51]。
Medico-legal and accountability issues in the case of medical errors caused by ChatGPT application should be carefully considered [44]. Importantly, the current LLMs including ChatGPT are unable to comprehend the complexity of biologic systems, which is an important concept needed in health care decisions and research [52,68]. The concerns regarding data governance, health care cyber security, and data privacy should draw specific attention in the discussion regarding the utility of LLMs in health care [32,39,53].
在ChatGPT应用导致医疗错误的情况下，应审慎考虑法医学与责任归属问题[44]。值得注意的是，当前包括ChatGPT在内的大语言模型尚无法理解生物系统的复杂性，而这是医疗决策和研究所需的重要概念[52,68]。关于数据治理、医疗网络安全和数据隐私的担忧，应在大语言模型医疗应用讨论中获得特别关注[32,39,53]。
Other issues accompanying ChatGPT applications in health care include the lack of personal and emotional perspectives needed in health care delivery and research [30,55]. However, ChatGPT emulation of empathetic responses was reported in a preprint in the context of hepatic disease [74]. Additionally, the issue of devaluing the function of the human brain should not be overlooked; therefore, stressing the indispensable human role in health care practice and research is important to address any psycho logic, economic, and social consequences that could accompany the application of LLM tools in health care settings [72].
ChatGPT在医疗保健应用中伴随的其他问题包括缺乏医疗服务和研究所需的个人及情感视角 [30,55]。然而，一份关于肝病领域的预印本报告指出，ChatGPT能够模拟共情反应 [74]。此外，不应忽视对人类大脑功能贬值的担忧，因此强调人类在医疗实践和研究中的不可替代性至关重要，以应对大语言模型工具在医疗场景应用可能带来的心理、经济和社会影响 [72]。
4.5. Benefits and Concerns Regarding ChatGPT Use in Health Care Education
4.5. ChatGPT在医疗教育中的应用优势与隐忧
In the area of health care education, ChatGPT appears to have a massive transformative potential. The need to rethink and revise the current assessment tools in health care education comes in light of ChatGPT’s ability to pass reputable exams (e.g., USMLE) and possibility of ChatGPT misuse, which would result in academic dishonesty [24,34,58,59,62,76,85–87].
在医疗保健教育领域，ChatGPT展现出巨大的变革潜力。鉴于ChatGPT能够通过权威考试(如USMLE)及其可能被滥用导致学术不端[24,34,58,59,62,76,85-87]，当前亟需重新思考并修订医疗保健教育的评估工具。
Specifically, in ophthalmology examination, Antaki et al. showed that ChatGPT currently performed at the level of an average first-year resident [70]. Such a result highlights the need to focus on questions involving the assessment of critical and problem-based thinking [34]. Additionally, the utility of ChatGPT in health care education can involve tailoring education based on the needs of the student with immediate feedback [46]. Inte rest ingly, a recent preprint by Benoit showed the promising potential of ChatGPT in rapidly crafting consistent realistic clinical vignettes of variable complexities that can be a valuable educational source with lower costs [67]. Thus, ChatGPT can be useful in health care education including enhanced communication skills given proper academic mentoring [42,57,67]. However, the copyright issues should be taken into account regarding the ChatGPT-generated clinical vignettes, aside from the issue of inaccurate references [67].
具体而言，在眼科检查中，Antaki等人研究表明ChatGPT目前表现相当于普通一年级住院医师水平 [70]。这一结果凸显了需要重点关注涉及批判性思维和基于问题思考能力评估的题目 [34]。此外，ChatGPT在医疗教育中的应用可包括根据学生需求提供即时反馈的个性化教学 [46]。值得注意的是，Benoit最近的预印本研究表明，ChatGPT在快速构建复杂度可调且连贯逼真的临床案例方面展现出巨大潜力，这些案例能成为低成本的高价值教学资源 [67]。因此，在适当的学术指导下，ChatGPT可有效提升包括沟通技巧在内的医疗教育水平 [42,57,67]。但需注意的是，除了参考文献不准确的问题外，ChatGPT生成的临床案例还涉及版权问题 [67]。
Additionally, ChatGPT availability can be considered as a motivation in health care education based on the personalized interaction it provides, enabling powerful self-learning as well as its utility as an adjunct in group learning [30,33,36,57,58].
此外，ChatGPT的可及性可被视为医疗教育中的一项激励因素，因其能提供个性化互动，既能支持高效自主学习，也可作为小组学习的辅助工具 [30,33,36,57,58]。
Other limitations of ChatGPT use in health care education include the concern regarding the quality of training datasets that could result in biased content and inaccurate information limited to the period prior to the year 2021. Additionally, other concerns include the current inability of ChatGPT to handle images as well as its low performance in some topics (e.g., failure to pass a paras it ology exam for Korean medical students), and the issue of possible plagiarism [33,56–58,70,75]. Despite ChatGPT versatility in the context of academic education [79], the content of ChatGPT in research assignments was discouraged, being currently insufficient, biased, or misleading [36,78].
ChatGPT在医疗教育应用中的其他局限还包括：其训练数据集质量可能引发偏见内容，且信息准确性受限于2021年前的数据。此外，当前版本无法处理图像，在特定领域表现欠佳（例如未能通过韩国医学生寄生虫学考试），并存在潜在剽窃风险 [33,56–58,70,75]。尽管ChatGPT在学术教育中展现出多样性 [79]，但研究任务中生成的内容因存在不充分、偏见或误导性而被谨慎对待 [36,78]。
4.6. Future Perspectives
4.6. 未来展望
As stated comprehensively in a commentary by van Dis et al., there is an urgent need to develop guidelines for ChatGPT use in scientific research, taking into account the issues of accountability, integrity, transparency, and honesty [46,88]. Thus, the application of ChatGPT to advance academia and health care should be carried out ethically and responsibly, taking into account the potential risks and concerns it entails [47,89].
正如van Dis等人在一篇评论中全面指出的那样，亟需制定ChatGPT在科学研究中的使用指南，同时考虑到责任、诚信、透明和诚实等问题[46,88]。因此，在推动学术界和医疗保健领域应用ChatGPT时，应以道德和负责任的方式进行，并充分考虑其潜在风险和隐患[47,89]。
More studies are needed to evaluate the content of LLMs including its potential impact to advance academia and science with a particular focus on health care settings [90]. In academic writing, a question arises as to whether authors would prefer an AI-editor and an AIreviewer considering the previous flaws in the editorial and peer review processes [91–93]. A similar question would also arise in health care settings involving the personal preference of emotional support from health care providers, rather than the potential efficiency of AI-based systems.
需要更多研究来评估大语言模型(LLM)的内容，包括其对推动学术界和科学发展的潜在影响，尤其要关注医疗保健领域[90]。在学术写作中，考虑到编辑和同行评审过程中存在的既往缺陷[91-93]，作者是否会倾向于选择AI编辑和AI评审员成为一个值得探讨的问题。类似问题也会出现在医疗保健场景中，这涉及患者对医护人员情感支持的个人偏好，而非基于AI系统的潜在效率。
In health care education, more studies are needed to evaluate the potential impact of ChatGPT on the quality and efficiency of both educational content and assessment tools. ChatGPT utility to help in refining communication skills among health care students is another aspect that should be further explored as well as the applications of LLMs in the better achievement of the intended learning outcomes through personalized and instantaneous feedback for the students.
在医疗保健教育领域，需要更多研究来评估ChatGPT对教育内容和评估工具质量与效率的潜在影响。ChatGPT在提升医学生沟通技巧方面的效用，以及大语言模型通过为学生提供个性化即时反馈来更好实现预期学习成果的应用，都是值得进一步探索的方向。
4.7. Strengths and Limitations
4.7. 优势与局限性
The current review represents the first rapid and concise overview of ChatGPT utility in health care education, research, and practice. However, the results of the current review should be viewed carefully in light of several shortcomings that include: (1) the quality of the included records can be variable, compromising the general iz ability of the results; (2) the exclusion of non-English records might have resulted in selection bias; (3) the exclusion of several records that could not be accessed could have resulted in missing relevant data despite being small in number; (4) the inclusion of preprints that have not been peer reviewed but might also compromise the general iz ability of the results; (5) the swift growth of literature addressing ChatGPT applications/risks mandate the need for further studies and reviews considering that the search in this review was concluded on 16 February 2023; and (6) this systematic review was based on the screening and interpretation of a single author, which may limit the interpret ability of the results; therefore, future systematic reviews should consider collaborative work to improve the quality and credibility of the review results.
当前综述首次快速、简明地概述了ChatGPT在医疗教育、研究和实践中的应用价值。然而，鉴于以下局限性，需谨慎看待本次综述结果：(1) 纳入文献质量参差不齐，可能影响结果普适性；(2) 排除非英语文献可能导致选择偏倚；(3) 少量无法获取的文献虽占比小，但可能遗漏相关数据；(4) 包含未经同行评审的预印本可能削弱结果普适性；(5) 鉴于检索截止于2023年2月16日，ChatGPT应用/风险相关文献快速增长，需后续研究及综述补充；(6) 本系统综述仅由单一作者筛选和解读，可能限制结果的可解释性，未来系统综述应考虑协作研究以提升结果质量与可信度。
5. Conclusions
5. 结论
The imminent dominant use of LLM technology including the widespread use of ChatGPT in health care education, research, and practice is inevitable. Considering the valid concerns raised regarding its potential misuse, appropriate guidelines and regulations are urgently needed with the engagement of all stakeholders involved to ensure the safe and responsible use of ChatGPT powers. The proactive embrace of LLM technologies with careful consideration of the possible ethical and legal issues can limit the potential future complications. If properly implemented, ChatGPT, among other LLMs, have the potential to expedite innovation in health care and can aid in promoting equity and diversity in research by overcoming language barriers. Therefore, a science-driven debate regarding the pros and cons of ChatGPT is strongly recommended and its possible benefits should be weighed with the possible risks of misleading results and fraudulent research [94].
大语言模型(LLM)技术即将在医疗教育、研究和实践中占据主导地位，包括ChatGPT的广泛应用已不可避免。鉴于其潜在滥用引发的合理担忧，迫切需要所有相关利益方共同参与制定适当的指导方针和法规，以确保安全负责任地使用ChatGPT的强大功能。在谨慎考量可能涉及的伦理与法律问题前提下，主动接纳大语言模型技术能够限制未来潜在的复杂问题。若实施得当，ChatGPT等大语言模型有望加速医疗创新，并通过消除语言障碍促进研究领域的公平性与多样性。因此，我们强烈建议开展以科学为基础的辩论来权衡ChatGPT的利弊，在评估其可能益处时需同步考量误导性结果和学术不端等潜在风险[94]。
Based on the available evidence, health care professionals could be described as carefully enthusiastic regarding the huge potential of ChatGPT among other LLMs in clinical decision-making and optimizing the clinical workflow. “ChatGPT in the Loop: Humans in Charge” can be the proper motto to follow based on the intrinsic value of human knowledge and expertise in health care research and practice [14,25,55]. An inspiring example of this motto could be drawn based on the relationship between the human character Cooper and the robotic character TARS from Christopher Nolan’s movie Interstellar [95].
根据现有证据，医疗专业人士对ChatGPT等大语言模型(LLM)在临床决策和优化工作流程中的巨大潜力表现出谨慎的乐观态度。"人类主导的ChatGPT协同模式"应成为基本准则，这源于人类知识与专业经验在医疗研究和实践中的核心价值[14,25,55]。Christopher Nolan电影《星际穿越》中人类角色Cooper与机器人TARS的协作关系，恰好为这一准则提供了生动例证[95]。
However, before its widespread adoption, the impact of ChatGPT from the health care perspective in a real-world setting should be conducted (e.g., using a risk-based approach) [96]. Based on the title of an important perspective article “AI in the hands of imperfect users” by Kostick-Quenet and Gerke [96], the real-world impact of ChatGPT among other LLMs should be properly evaluated to prevent any negative impact of its potential misuse. The same innovative and revolutionary tool can be severely deleterious if used improperly. An example to illustrate such severe negative consequences of ChatGPT misuse can be based on Formula 1 racing, as follows. In the 2004 Formula 1 season, the Ferrari F2004 (the highly successful Formula 1 racing car) broke several Formula 1 records in the hands of Michael Schumacher, one of the most successful Formula 1 drivers of all time. However, in my own hands —as a humble researcher without expertise in Formula 1 driving— the same highly successful car would only break walls and be damaged beyond repair.
然而，在广泛采用之前，应从医疗保健角度在现实环境中评估ChatGPT的影响（例如采用基于风险的方法）[96]。根据Kostick-Quenet和Gerke的重要观点文章《不完美用户手中的AI》[96]，应对ChatGPT等大语言模型在现实世界的影响进行恰当评估，以防止其潜在滥用带来的负面影响。同样的创新革命性工具若使用不当，可能造成严重危害。以一级方程式赛车为例说明ChatGPT滥用的严重后果：2004赛季中，法拉利F2004（这款极其成功的F1赛车）在史上最杰出车手之一Michael Schumacher的驾驭下屡破纪录。但若由我这样毫无F1驾驶专长的普通研究者操控，这台顶级赛车只会撞毁围墙，导致无法修复的损毁。
Funding: This research received no external funding.
资助：本研究未获得外部资助。
Institutional Review Board Statement: Not applicable.
机构审查委员会声明：不适用。
Informed Consent Statement: Not applicable.
知情同意声明：不适用。
Data Availability Statement: Data supporting this systematic review are available in the original publications, reports, and preprints that were cited in the reference section. In addition, the analyzed data that were used during the current systematic review are available from the author on reasonable request.
数据可用性声明：本系统综述所支持的数据可在参考文献部分引用的原始出版物、报告和预印本中获取。此外，当前系统综述期间使用的分析数据可根据合理要求向作者索取。
Acknowledgments: I am sincerely grateful to the reviewers for their time and effort in reviewing the manuscript, which provided insightful and valuable comments that helped to improve the quality of the final manuscript to a great degree.
致谢：衷心感谢审稿人花费时间和精力审阅稿件，他们提出的深刻而有价值的意见极大提升了最终稿件的质量。
Conflicts of Interest: The author declares no conflict of interest.
利益冲突: 作者声明无利益冲突。
