Determining the Consistency Between Nurses and Artificial Intelligence (ChatGPT-5) in Delivering Scenario-Based Discharge Education to Coronary Artery Bypass Graft Patients: A Methodological Study

Обзор

This methodological study aims to determine the level of agreement between nurses and an artificial intelligence system (ChatGPT-4.0) in providing scenario-based discharge education for patients who have undergone coronary artery bypass graft (CABG) surgery. Thirty standardized patient scenarios representing different demographic, clinical, and psychosocial characteristics will be used. For each scenario, both expert nurses and ChatGPT-4.0 will prepare discharge education content based on six main domains and twenty-four subtopics identified from the literature and clinical guidelines. The educational materials will be independently evaluated by two blinded reviewers in terms of content accuracy, completeness, scientific consistency, and clarity of language. Agreement between nurses and AI-generated content will be analyzed using Cohen's Kappa coefficient and Fisher's Exact Test. The findings are expected to provide evidence for the reliability and applicability of AI-assisted discharge education systems in cardiac surgery nursing practice.

Подробное описание

This methodological study aims to determine the agreement between expert nurses and an artificial intelligence (AI) system (ChatGPT-5) in providing scenario-based discharge education for patients who have undergone coronary artery bypass graft (CABG) surgery. The purpose of the study is to evaluate whether ChatGPT-5 can generate discharge education content that is comparable in accuracy, completeness, and clinical appropriateness to that prepared by experienced cardiovascular surgery nurses.

Thirty standardized patient scenarios will be developed to represent a wide range of CABG cases with diverse demographic, socioeconomic, psychosocial, and clinical characteristics. Each scenario will simulate realistic postoperative conditions, including potential complications (e.g., delirium, wound infection, bleeding, arrhythmia), comorbidities (e.g., diabetes, hypertension, COPD), and psychosocial variables such as anxiety level, family structure, and social support. All scenarios will be reviewed and validated by a multidisciplinary expert panel including cardiovascular surgeons and academic nurse specialists to ensure clinical realism and content validity.

Discharge education will be structured around six main domains and twenty-four subtopics derived from national and international guidelines and evidence-based literature. These domains include: (1) medical management and follow-up, (2) daily life and functional recovery, (3) psychosocial and social support, (4) risk factors and preventive health, (5) quality of life and specific conditions, and (6) religious practices. For each scenario, both expert nurses and ChatGPT-5 will independently prepare written discharge education materials using this standardized framework.

The educational materials will be anonymized and evaluated by two blinded reviewers in terms of scientific accuracy, content completeness, linguistic clarity, and alignment with clinical standards. In case of disagreement, a third independent reviewer will provide a final decision to ensure objectivity. Statistical analyses will include Cohen's Kappa coefficient to measure inter-rater agreement and Fisher's Exact Test for categorical comparisons. Diagnostic performance measures such as sensitivity, specificity, positive predictive value (PPV), negative predictive value (NPV), and F1 score will also be computed.

Data will be analyzed using SPSS v25 (IBM Corp., Armonk, NY, USA). Descriptive statistics (frequencies, percentages, means, and standard deviations) will be reported to summarize the characteristics of the scenarios and evaluations. Agreement levels will be interpreted according to Landis and Koch's classification. A p-value of \<0.05 will be considered statistically significant.

The findings of this study are expected to provide evidence regarding the reliability, validity, and usability of ChatGPT-5 as an innovative and supportive tool for preparing individualized discharge education materials in cardiovascular surgery nursing. Results may contribute to developing new technology-assisted educational models that can reduce nurse workload, improve the standardization of discharge education, and enhance patient understanding and satisfaction in the postoperative period.

Первичные конечные точки

• Agreement Between Nurse- and ChatGPT-5-Generated Discharge Education Content [Срок оценки: During data collection (expected within 8 months after study start).]
• Agreement Between Nurse- and ChatGPT-5-Generated Discharge Education Content [Срок оценки: During data collection (expected within 12 months after study start).]

Критерии участия

Критерии включения

• Patient scenarios representing individuals who have undergone coronary artery bypass graft (CABG) surgery.
• Scenarios that include demographic, socioeconomic, clinical, and psychosocial information consistent with current literature and clinical guidelines.
• Scenarios describing patients who underwent median sternotomy and on-pump CABG procedure.
• Scenarios that include relevant postoperative complications (e.g., delirium, bleeding, wound infection, arrhythmia) and comorbidities (e.g., diabetes, hypertension, COPD).
• Scenarios that enable both nurse and ChatGPT-5 to prepare discharge education materials under the same standardized framework.
• Scenarios reviewed and validated by cardiovascular surgery experts and nurse academicians for content validity.

Критерии исключения

• Patient scenarios not related to coronary artery bypass graft (CABG) surgery.
• Scenarios lacking sufficient demographic, clinical, or psychosocial information to prepare individualized discharge education.
• Scenarios that do not follow the standardized structure of six main domains and twenty-four subtopics.
• Scenarios with inconsistent or contradictory medical data (e.g., incompatible diagnosis and treatment details).
• Scenarios not validated by the expert review panel for clinical accuracy and content validity.
• Scenarios that do not allow comparison between nurse-generated and ChatGPT-5-generated discharge education materials.

Критерии приведены из реестра в оригинале (на английском). Окончательную оценку соответствия проводит исследовательский центр.

Здоровые добровольцы: Да

Дизайн исследования

Модель наблюдения

Другое

Центры проведения

Turkey (Türkiye) · 1 центр

• Hasan Kalyoncu University Faculty of Nursing — Gaziantep

Публикации

• Thompson R, Traylor D, Halguist E. Evaluating the clinical accuracy of ChatGPT-generated patient instructions: a review study. Digital Health. 2025;11:2055207625123456.
• Su H, Vrdoljak J, Busch M, et al. Artificial intelligence in patient discharge education: improving readability and patient understanding. Journal of Medical Internet Research. 2025;27:e45612.
• Rushton M, Hemmings L, Marsh L, et al. Enhancing recovery after cardiac surgery: discharge education and follow-up. European Journal of Cardiovascular Nursing. 2017;16(2):114-123. doi:10.1177/1474515116643395.
• Akbari M, Celik S. The effect of discharge training on stress, anxiety, and pain in patients after coronary artery bypass graft surgery. Journal of Perioperative Nursing. 2015;28(3):165-172.
• Fredericks S, Guruge S, Sidani S, Wan T. Postoperative patient education: a systematic review. Clin Nurs Res. 2010 May;19(2):144-64. doi: 10.1177/1054773810365994. PMID 20435785

Идентификаторы

NCT: NCT07263724 · CABG-AI-EDU-PHASE1-2025

Первоисточники (государственные реестры)

• ClinicalTrials.gov (NIH, США)

Открыть это исследование на ClinicalTrials.gov ↗