{"title":"Reliability and Quality of AI-Generated Nursing Care Plans: A 2-Year Follow-Up Comparing GPT-3.5 and GPT-5.","authors":"Funda Çam, Ayşe Dost, Mahmut Dağcl","doi":"10.1097/CIN.0000000000001627","DOIUrl":null,"url":null,"abstract":"<p><p>Advancements in large language models have increased the use of AI-generated content in nursing education; however, concerns remain regarding the reliability and quality of such outputs. This study aimed to evaluate the reliability and quality of nursing care plans generated by ChatGPT-5 and to compare them with those previously produced using ChatGPT-3.5. This methodological follow-up study used a descriptive and comparative design. Forty nursing diagnoses previously analyzed in 2023 were re-evaluated using ChatGPT-5 to ensure methodological continuity. Generated care plans were assessed using a descriptive criteria form and the DISCERN instrument. Paired comparisons were conducted using the Wilcoxon signed-rank test. ChatGPT-5-generated care plans showed significantly higher numbers of total and accessible references and higher DISCERN reliability and total scores than those reported in the 2023 evaluation of ChatGPT-3.5. No significant differences were observed in care-related information quality or overall quality scores; however, category-based analyses showed a shift toward higher quality levels. ChatGPT-5-generated nursing care plans showed higher reference accessibility and reliability scores than those reported in the 2023 evaluation of ChatGPT-3.5 using the same nursing diagnoses, although human oversight remains essential, particularly regarding reference accuracy.</p>","PeriodicalId":520598,"journal":{"name":"Computers, informatics, nursing : CIN","volume":" ","pages":""},"PeriodicalIF":1.9000,"publicationDate":"2026-09-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Computers, informatics, nursing : CIN","FirstCategoryId":"1085","ListUrlMain":"https://doi.org/10.1097/CIN.0000000000001627","RegionNum":0,"RegionCategory":null,"ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"","JCRName":"","Score":null,"Total":0}
引用次数: 0
Abstract
Advancements in large language models have increased the use of AI-generated content in nursing education; however, concerns remain regarding the reliability and quality of such outputs. This study aimed to evaluate the reliability and quality of nursing care plans generated by ChatGPT-5 and to compare them with those previously produced using ChatGPT-3.5. This methodological follow-up study used a descriptive and comparative design. Forty nursing diagnoses previously analyzed in 2023 were re-evaluated using ChatGPT-5 to ensure methodological continuity. Generated care plans were assessed using a descriptive criteria form and the DISCERN instrument. Paired comparisons were conducted using the Wilcoxon signed-rank test. ChatGPT-5-generated care plans showed significantly higher numbers of total and accessible references and higher DISCERN reliability and total scores than those reported in the 2023 evaluation of ChatGPT-3.5. No significant differences were observed in care-related information quality or overall quality scores; however, category-based analyses showed a shift toward higher quality levels. ChatGPT-5-generated nursing care plans showed higher reference accessibility and reliability scores than those reported in the 2023 evaluation of ChatGPT-3.5 using the same nursing diagnoses, although human oversight remains essential, particularly regarding reference accuracy.
求助内容：
应助结果提醒方式：
京公网安备 11010802042870号
