Related Experiment Video
Updated: May 22, 2026

Lumped-Parameter and Finite Element Modeling of Heart Failure with Preserved Ejection Fraction
Published on: February 13, 2021
An analysis of artificial intelligence heart failure discharge instructions compared to routinely referenced
Evan Derector1, Daniel Ricketti2, Anupam A Kumar3
1Cooper Medical School of Rowan University, Camden, NJ, USA.
Insights
Artificial intelligence (AI) generated heart failure (HF) discharge instructions offer high-quality information but demand college-level literacy. Current AI tools need refinement for better patient understanding and accessibility.
Area of Science:
- Medical Informatics
- Patient Education
- Health Literacy
Background:
- Heart failure (HF) is a prevalent global condition with significant healthcare costs and high hospital readmission rates.
- Improving the post-discharge process is crucial for reducing complications and enhancing patient outcomes.
Purpose of the Study:
- To compare the quality, readability, and health literacy demands of AI-generated discharge instructions against established patient education materials.
- To identify patient-centered discharge materials with optimal readability and literacy requirements.
Main Methods:
- A blinded survey assessed discharge instructions from WebMD, Lexicomp®, and ChatGPT-4o, evaluated by 15 cardiovascular disease fellows.
- Quality was measured using modified DISCERN criteria and Global Quality Scale (GQS).
- Readability was analyzed using Flesch Reading Ease Score (FRES), Flesch-Kincaid Reading Grade Level (FKGL), and Gunning-Fog Index (GFI).
Main Results:
- No statistical differences in quality metrics were found between AI-generated and traditional materials.
- ChatGPT-4o instructions required college-level literacy, significantly higher than WebMD and Lexicomp® (8th-9th grade).
- AI-generated text relied on abstract cognitive tasks, unlike the direct commands in WebMD and Lexicomp®, increasing the literacy burden.
Conclusions:
- All evaluated sources provided high-quality HF discharge information, but AI materials presented a higher literacy barrier.
- There is a critical need to enhance the readability of AI-generated patient education materials.
- Future research should focus on prompt engineering to favor concrete examples, making AI tools more accessible to patients with lower health literacy.
Background:
Heart failure (HF) affects 1-3% of adults globally with high hospital re-admission rates and healthcare costs. Research aims to improve the discharge process to reduce post-discharge complications.
Objectives:
This study compares the quality, readability, and health literacy requirements of artificial intelligence (AI) generated discharge instructions with commonly accessed patient education materials.
Methods:
A blinded survey of 15 cardiovascular disease fellows evaluated HF discharge instructions generated from WebMD, Lexicomp®, and ChatGPT-4o. Quality was assessed using a modified DISCERN criteria and the Global Quality Scale (GQS) while readability was analyzed using the Flesch Reading Ease Score (FRES), Flesch-Kincaid Reading Grade Level (FKGL), and the Gunning-Fog Index (GFI). Statistical analysis identified the highest quality patient-centered discharge materials and reading level requirements.
Results:
Quality metrics showed no statistical differences. However, ChatGPT-4o was rated as "good" while WebMD and Lexicomp® were rated as "excellent." Importantly, expert reviewers identified no clinical inaccuracies in the AI-generated text. Readability analysis found ChatGPT-4o-generated instructions required college-level literacy, significantly higher than WebMD and Lexicomp® (8th-9th grade). Qualitative comparison revealed that WebMD and Lexicomp® utilized direct, behavioral commands, whereas AI-generated text relied on abstract cognitive tasks, contributing to the higher literacy burden.
Conclusion:
While all sources provided high-quality information, AI-generated materials required significantly higher literacy levels. This study highlights the need for improved readability in patient education materials. Future study should analyze prompt engineering strategies that prioritize concrete examples over abstract concepts, ensuring that high-quality AI tools are accessible to patients with lower health literacy.
Related Concept Videos
Heart Failure VII: Nursing Interventions
Heart Failure VI: Adjunct Therapies
Heart Failure IV: Classification and Diagnostic Evaluation
Heart Failure I: Introduction
Heart Failure V: Medical Management
Heart Failure II: Pathophysiology