Evaluating Artificial Intelligence-generated Patient Education Guides on Cardiomyopathies: A Cross-sectional Comparison of ChatGPT and DeepSeek Artificial Intelligence.

Abstract

Introduction: Patient education is essential in the management of cardiomyopathies, including dilated, restrictive, and hypertrophic subtypes, which often involve complex diagnostic and treatment pathways. Traditional educational resources may not address varying health literacy levels. Recent advancements in artificial intelligence (AI), particularly large language models such as ChatGPT and DeepSeek AI, offer new avenues for generating accessible and scalable patient education materials. This study evaluates the performance of these tools in that context.

Methodology: This cross-sectional study involved generating educational brochures for three cardiomyopathies using ChatGPT and DeepSeek AI. The responses were evaluated using the Flesch-Kincaid Grade Level and reading ease (for readability), QuillBot similarity checker (for originality), and the modified DISCERN instrument (for reliability). Statistical analysis included the Shapiro-Wilk test for normality, Levene's test for equality of variances, and independent samples t-tests. Pearson correlation was also performed between readability and reliability scores.

Results: There were no statistically significant differences between the two AI tools across all assessed parameters. ChatGPT generally produced content with higher grade levels and longer text, while DeepSeek AI demonstrated higher originality in restrictive cardiomyopathy and superior readability in hypertrophic cardiomyopathy. Both tools achieved moderate DISCERN scores, with DeepSeek AI slightly outperforming ChatGPT in dilated cardiomyopathy.

Conclusion: This study found no significant difference in ease score, reliability score, and grade scores between patient education brochures on cardiomyopathies generated by ChatGPT and DeepSeek AI.