Objective:

To evaluate the performance and limitations of widely used LLMs in generating patient education texts for cardiac myxoma, focusing on specific challenges such as readability and actionability.

Approach:

Key Findings:

• Significant heterogeneity in information quality, understandability, and readability across generated texts, highlighting the need for improved educational resources.
• Clinical actionability of the texts was generally low, indicating a gap in practical guidance for patients.
• Longer texts correlated positively with higher informational quality scores, suggesting a trade-off between length and clarity.
• Reading difficulty indices were negatively correlated with actionability, emphasizing the challenge of creating accessible materials.
• Three text phenotypes identified: moderate-quality/low-readability, high-quality/high-actionability, and low-quality/easy-to-read, which can inform future content development.

Interpretation:

The application of LLMs in patient education for cardiac myxoma is limited by reading burden and insufficient behavioral guidance, which may hinder patient understanding and engagement.

Limitations:

• Only a small subset of outputs approached a favorable quality–actionability profile, raising concerns about the overall utility of LLM-generated texts.
• The study did not directly elicit real-world patient information needs, which may limit the relevance of the generated materials.

Conclusion:

Current LLM-generated educational materials for cardiac myxoma require improvements in readability and actionability.

Attribution Notice

This content is an AI-generated, fully rewritten summary based on a published scholarly article. It does not reproduce the original text and is not a substitute for the original publication. Readers are encouraged to consult the source for full context, data, and methodology.