METHODS: Chat-GPT 3.5 was prompted to generate 10 unpublished, innovative research topics on breast reconstruction surgery, followed by 10 additional subtopics. Results were filtered for systematic reviews in PubMed, and novel ideas were identified. To evaluate Chat-GPT\'s power in generating improved responses, two additional searches were conducted using search terms generated by Chat-GPT.
RESULTS: Chat-GPT produced 83 novel ideas, leading to an accuracy rate of 83%. There was a wide range of novel ideas produced among topics such as transgender women, generating 10 ideas, whereas acellular dermal matrix (ADM) generated five ideas. Chat-GPT increased the total number of manuscripts generated by a factor of 2.3, 3.9, and 4.0 in the first, second, and third trials, respectively. While the search results were accurate to our manual searches (95.2% accuracy), the greater number of manuscripts potentially diluted the quality of articles, resulting in fewer novel systematic review ideas.
CONCLUSIONS: Chat-GPT proves valuable in identifying gaps in the literature and offering insights into areas lacking research in breast reconstruction surgery. While it displays high sensitivity, refining its specificity is imperative. Prudent practice involves evaluating accomplished work and conducting a comprehensive review of all components involved.
方法:系统提示Chat-GPT3.5生成10个未发布的,关于乳房重建手术的创新研究课题,其次是10个额外的子主题。在PubMed中过滤结果以进行系统评价,并确定了新的想法。要评估Chat-GPT在生成改进响应方面的能力,使用Chat-GPT生成的搜索词进行了另外两次搜索.
结果:Chat-GPT产生了83个新颖的想法,准确率为83%。在变性女性等主题中产生了广泛的新颖想法,产生10个想法,而无细胞真皮基质(ADM)产生了五个想法。Chat-GPT增加了产生的手稿总数,其中第一个增加了2.3、3.9和4.0倍,第二,和第三次审判,分别。虽然搜索结果对我们的手动搜索是准确的(准确率为95.2%),更多的手稿可能会稀释文章的质量,导致较少新颖的系统审查思路。
结论:Chat-GPT在发现文献空白和提供对乳房重建手术缺乏研究领域的见解方面被证明是有价值的。虽然它显示高灵敏度,完善其特殊性势在必行。谨慎的做法包括评估已完成的工作并对所有涉及的组件进行全面审查。