Evaluation of Prompts to Simplify Cardiovascular Disease Information Generated Using a Large Language Model:...

Area of Science:

• Cardiology
• Artificial Intelligence
• Health Communication

Background:

• Cardiovascular disease prevention relies on accessible information.
• Large language models like GPT-4 are increasingly used for health information generation.
• Evaluating AI-generated health content is crucial for patient safety and education.

Purpose of the Study:

• To evaluate the completeness, readability, and syntactic complexity of cardiovascular disease prevention information generated by GPT-4.
• To assess the impact of different prompt types on the quality of AI-generated health content.

Main Methods:

• A cross-sectional study design was employed.
• GPT-4 generated responses to four distinct prompts related to cardiovascular disease prevention.
• Information completeness, readability scores (e.g., Flesch-Kincaid), and syntactic complexity metrics were analyzed.

Main Results:

• GPT-4 produced generally complete and readable information on cardiovascular disease prevention.
• Syntactic complexity varied depending on the prompt, indicating potential areas for refinement.
• The AI demonstrated a capacity for generating understandable health advice.

Conclusions:

• GPT-4 shows promise in generating accessible cardiovascular disease prevention information.
• Prompt engineering can influence the quality and complexity of AI-generated health content.
• Further research is needed to optimize AI for nuanced health communication.