Generative AI for ECG Interpretation Education: Impact on Nursing, Student Performance, and AI Model Accuracy

BACKGROUND: Electrocardiogram (ECG) interpretation is a critical yet challenging skill for nurses. Generative artificial intelligence (AI) offers potential for personalized, adaptive learning. PURPOSE: The aim was to evaluate the effectiveness, accuracy, and cost-efficiency of AI models in nursing ECG education. METHODS: A 2-part study compared 4 AI models (GoodNurse, ChatGPT-5, Claude Sonnet 4, Microsoft Copilot) on an 88-item ECG exam and assessed cost-effectiveness. GoodNurse was then integrated into a 4-credit ECG course; AI usage, satisfaction, and grades were analyzed. RESULTS: Accuracy varied (P <.01): GoodNurse 85.3%, ChatGPT-5 83.1%, Copilot 80.9%, Sonnet 79.1%. GoodNurse had the fewest waveform errors and the best cost-per-accuracy ($8.06 per 1% gain). In the course, 43% of students used GoodNurse, achieving higher grades (95.1 ± 2.5%) than nonusers (88.8 ± 5.9%; P =.0048), with a $5.80 per 1% grade improvement. CONCLUSION: Domain-specific AI, such as GoodNurse, enhances ECG learning, diagnostic accuracy, and cost-efficiency, supporting its integration into nursing education.