Department of Urology, David Geffen School of Medicine, University of California, Los Angeles, Los Angeles, California.
Department of Urology, Vattikuti Urology Institute, Henry Ford Hospital, Detroit, Michigan.
J Ren Nutr. 2024 Mar;34(2):170-176. doi: 10.1053/j.jrn.2023.10.002. Epub 2023 Oct 13.
The American Medical Association recommends health information to be written at a 6th grade level reading level. Our aim was to determine whether Artificial Intelligence can outperform the existing health information on kidney stone prevention and treatment.
The top 50 search results for "Kidney Stone Prevention" and "Kidney Stone Treatment" on Google, Bing, and Yahoo were selected. Duplicate webpages, advertisements, pages intended for health professionals such as science articles, links to videos, paid subscription pages, and links nonrelated to kidney stone prevention and/or treatment were excluded. Included pages were categorized into academic, hospital-affiliated, commercial, nonprofit foundations, and other. Quality and readability of webpages were evaluated using validated tools, and the reading level was descriptively compared with ChatGPT generated health information on kidney stone prevention and treatment.
50 webpages on kidney stone prevention and 49 on stone treatment were included in this study. The reading level was determined to equate to that of a 10th to 12th grade student. Quality was measured as "fair" with no pages scoring "excellent" and only 20% receiving a "good" quality. There was no significant difference between pages from academic, hospital-affiliated, commercial, and nonprofit foundation publications. The text generated by ChatGPT was considerably easier to understand with readability levels measured as low as 5th grade.
The language used in existing information on kidney stone disease is of subpar quality and too complex to understand. Machine learning tools could aid in generating information that is comprehensible by the public.
美国医学协会建议将健康信息编写为六年级阅读水平。我们的目的是确定人工智能是否能超越现有的肾结石预防和治疗健康信息。
在谷歌、必应和雅虎上搜索“肾结石预防”和“肾结石治疗”的前 50 个搜索结果。排除重复网页、广告、针对健康专业人士（如科学文章）的网页、链接到视频、付费订阅页面以及与肾结石预防和/或治疗无关的链接。收录的网页分为学术、医院附属、商业、非营利基金会和其他。使用经过验证的工具评估网页的质量和可读性，并描述性地比较 ChatGPT 生成的肾结石预防和治疗健康信息的阅读水平。
本研究共纳入 50 篇肾结石预防网页和 49 篇结石治疗网页。阅读水平相当于 10 至 12 年级学生的水平。质量被评为“一般”，没有一个网页评为“优秀”，只有 20%的网页评为“良好”。学术、医院附属、商业和非营利基金会出版物的网页之间没有显著差异。ChatGPT 生成的文本理解起来要容易得多，可读性水平低至 5 年级。
现有的肾结石疾病信息质量较差，语言过于复杂，难以理解。机器学习工具可以帮助生成公众能够理解的信息。
