背景:TUSOMONRSTEM-IMPRESS计划看到了一个未来,每个人都有机会和资源成为医生,无论种族如何,性别,性取向,社会经济地位,或共享学习环境中的任何其他差异。我们努力通过鼓励所有大学新生和大二学生做出进入医疗领域的决定来实现这一愿景;用包括临床轮换在内的全面计划来充实他们,研究项目,MCAT准备工作,面试提示,和小组会议;并通过建立职业指导与他们接触。计划价值:Equity-IMPRESS为那些原本无法获得机会和接触医疗保健职业的人提供服务。多样性-印象认识到包括种族在内的医学多样性的重要性,性别,性取向,社会经济地位,意识形态,或任何其他差异。Inclusion-IMPRESS重视每个观点的独特性,经验,以及它们对医学领域更具包容性的环境的共同贡献。Culture-IMPRESS促进了医学文化,反映了新奥尔良及周边地区文化的丰富性和活力。目的是评估由Chatbot生成预训练变压器版本4(ChatGPT-4)在与结肠和直肠手术有关的科学文献中生成的参考文献的准确性。
方法:选择了与结肠和直肠手术有关的十个经常研究的关键词:结肠癌,直肠癌,肛门癌,全新辅助治疗,憩室炎,低位前切除术,经肛门微创手术,回肠袋肛门吻合术,腹部手术切除,和痔疮切除术.ChatGPT-4被提示为所有关键词搜索最具代表性的引文。在这之后,两名独立的评估员精心检查了每个关键要素的结果,在与Scopus交叉引用时,对任何字段中没有差异的引用授予完全准确性,Google,和PubMed数据库。来自ChatGPT-4的参考文献经过了彻底的审查过程,其中包括仔细审查文章标题等关键要素,作者,日记帐名称,出版年份,和数字对象标识符(DOI)。
结果:生成的100个参考中有41个是完全准确的;但是,但没有包括DOI。在67个参考文献中观察到部分准确性,可以通过标题和日记来识别。性能因特定关键字而异;例如,结肠癌和直肠癌的参考文献通过标题和杂志100%可识别,但没有一个术语在所有类别中都有100%的准确性。值得注意的是,生成的参考文献没有正确列出所有作者。在短时间内进行,在此期间ChatGPT4正在迅速发展并更新其知识库。
结论:虽然ChatGPT-4提供了对其前身的改进,并显示了在学术文献中使用的潜力,其不同类别的表现不一致,缺少DOIs,作者列表中的违规行为引起了人们对其在结肠和直肠手术研究领域的应用准备的担忧。
BACKGROUND: The TUSOM ONR STEM-IMPRESS Program sees a future where everyone is given opportunities and resources to become a physician regardless of race, gender, sexual orientation, social economic status, or any other differences in the shared learning environment. We strive to achieve this vision by EMPOWERING all college freshmen and sophomores to make the decision to enter the medical field; ENRICHING them with a well-rounded program that includes clinical rotations, research projects, MCAT preparations, interview tips, and panel sessions; and ENGAGING with them by establishing career-long mentorship. Program values: Equity-IMPRESS offers services to those who would otherwise not have access to opportunities and exposure to healthcare careers. Diversity-IMPRESS recognizes the importance of diversity in medicine including race, gender, sexual orientation, social economic status, ideology, or any other differences. Inclusion-IMPRESS values the uniqueness of each individual viewpoint, experience, and their combined contribution to a more inclusive environment in medicine. Culture-IMPRESS promotes a culture in medicine that reflects the richness and vitality in the culture of New Orleans and the surrounding area. The objective is to assess the precision of references generated by Chatbot Generative Pretrained Transformer version 4 (ChatGPT-4) in scientific literature pertaining to colon and rectal surgery.
METHODS: Ten frequently studied keywords pertaining to colon and rectal surgery were chosen: colon cancer, rectal cancer, anal cancer, total neoadjuvant therapy, diverticulitis, low anterior resection, transanal minimally invasive surgery, ileal pouch anal anastomosis, abdominoperineal resection, and hemorrhoidectomy. ChatGPT-4 was prompted to search for the most representative citations for all keywords. After this, two separate evaluators meticulously examined the outcomes each key element, awarding full accuracy to generated citations in which there was no discrepancies in any of the fields when cross-referenced with the Scopus, Google, and PubMed databases. References from ChatGPT-4 underwent a thorough review process, which involved careful examination of key elements such as the article title, authors, journal name, publication year, and Digital Object Identifier (DOI).
RESULTS: Forty-one of the 100 references generated by were fully accurate; however, but none included a DOI. Partial accuracy was observed in 67 of the references, which were identifiable by title and journal. Performance varied across specific keywords; for example, references for colon and rectal cancer were 100% identifiable by title and journal, but no term had 100% accuracy across all categories. Notably, none of the generated references correctly listed all authors. Conducted within a short timeframe during which ChatGPT4 is rapidly evolving and updating its knowledge base.
CONCLUSIONS: While ChatGPT-4 offers improvements over its predecessors and shows potential for use in academic literature, its inconsistent performance across categories, lack of DOIs, and irregularities in authorship listings raise concerns about its readiness for application in the field of colon and rectal surgery research.