文献速递小儿耳鼻咽喉

儿童耳鼻咽喉科信息可读性:人工智能生成内容与谷歌搜索结果的比较

Readability of Pediatric Otolaryngology Information: Comparing AI-Generated Content With Google Search Results.
PubMed 文献 · Otolaryngol Head Neck Surg · 2025-08-26

本研究为横断面可读性分析,比较ChatGPT4o生成与谷歌搜索获取的10种儿童耳鼻咽喉疾病患者教育材料的可读性。ChatGPT4o生成内容在FKGL、Gunning-Fog、ARI和SMOG评分上显著更高、FRES评分更低(P<.05),提示其更复杂,差异在变应性鼻炎和外耳道炎等简单疾病中尤为明显。

要点

  • ChatGPT4o生成的儿童耳鼻咽喉科患者教育材料在多项可读性指标上比谷歌搜索材料更难读(P<.05)。
  • 差异在变应性鼻炎、外耳道炎等较简单疾病中最为明显。
  • 喉软化、唇腭裂等复杂主题的可读性评分在两种来源间无显著差异(P>.05)。

核心结论

  • ChatGPT4o生成的患者教育材料总体上比谷歌搜索内容更难读,在较简单疾病中尤其明显;作者认为人工智能生成材料可能需要进一步改进以在不影响准确性的前提下提高可及性。

研究设计与样本

  • 横断面可读性分析,覆盖10种儿童耳鼻咽喉疾病,教育文章来源于谷歌搜索或ChatGPT4o生成,文本经标准化处理后用6种经验证的可读性指标评估。

主要发现

  • ChatGPT4o生成内容相比谷歌来源材料在FKGL、Gunning-Fog、ARI和SMOG评分上显著更高、FRES评分显著更低(P<.05),提示更复杂。
  • 在变应性鼻炎和外耳道炎等较简单疾病中,两种来源的可读性差异最为明显。
  • 在喉软化和唇腭裂等较复杂主题中,两种来源的可读性评分无显著差异(P>.05)。

关键数值与推荐

  • 可读性评估使用6种指标:Flesch-Kincaid Grade Level(FKGL)、Flesch Reading Ease Score(FRES)、Gunning-Fog Index、Simple Measure of Gobbledygook(SMOG)、Coleman-Liau Index和Automated Readability Index(ARI)。
  • 统计比较采用配对t检验或Wilcoxon符号秩检验。

临床怎么用

  • 原文未据此给出推荐。

局限性

  • 原文未报告局限性。

以上解读依据论文摘要(全文未开放获取或许可不允许翻译转载)。

原文

OBJECTIVE: This study evaluates and compares the readability of pediatric otolaryngology patient education materials generated by ChatGPT4o and those retrieved from Google searches. The goal is to determine whether artificial intelligence (AI)-generated content improves accessibility compared to institutionally affiliated online resources. STUDY DESIGN: Cross-sectional readability analysis. SETTING: Online educational materials focused on pediatric otolaryngology topics. METHODS: Educational articles covering 10 pediatric otolaryngology conditions were sourced either via Google search or generated using ChatGPT4o. All texts were standardized by removing extraneous formatting. Readability was assessed using six validated metrics: Flesch-Kincaid Grade Level (FKGL), Flesch Reading Ease Score (FRES), Gunning-Fog Index, Simple Measure of Gobbledygook (SMOG), Coleman-Liau Index, and Automated Readability Index (ARI). Statistical comparisons were performed using paired t tests or Wilcoxon signed-rank tests to evaluate differences in scores between sources. RESULTS: ChatGPT4o-generated content demonstrated significantly higher FKGL, Gunning-Fog, ARI, and SMOG scores and lower FRES scores compared to Google-sourced materials, indicating greater complexity (P < .05). These differences were most pronounced for simpler conditions such as allergic rhinitis and otitis externa. For more complex topics like laryngomalacia and cleft lip and palate, readability scores were not significantly different between the two sources (P > .05). CONCLUSION: ChatGPT4o-generated patient education materials are generally more difficult to read than Google-sourced content, especially for less complex conditions. Given the importance of readability in patient education, AI-generated materials may require further refinement to improve accessibility without compromising accuracy. Enhancing clarity could increase the utility of AI tools for educating parents and caregivers in pediatric otolaryngology.

来源:Otolaryngol Head Neck Surg,查看原文。版权归原作者所有。

引用

  1. Carnino JM, Rampam S, Puyo EM, et al. Readability of Pediatric Otolaryngology Information: Comparing AI-Generated Content With Google Search Results. Otolaryngol Head Neck Surg. 2025. PMID: 40856230. doi:10.1002/ohn.70011

供执业医师工作参考,不替代临床判断;数值与分级以指南原文为准。