Evaluating the Effectiveness of Generative AI for the Creation of Patient Education Materials on Coronary Heart Disease: A Comparative Study
Bibliographic record
Abstract
BACKGROUND: Generative Artificial Intelligence (AI) has shown great potential in various fields, including healthcare. However, its application in developing patient education materials(PEMs), particularly those with coronary heart disease (CHD), remains underexplored. Traditional methods for creating these materials are time-consuming and lack personalization, which limits their effectiveness. OBJECTIVE: This study aims to explore the effectiveness of Generative AI tools(ChatGPT and DeepSeek) in generating PEMs for CHD patients and to compare them with materials developed by a professional medical team. METHODS: In February 2025, PEMs for CHD patients were developed using a framework designed by a professional medical team. Structured prompts were used to generate materials through two Generative AI models-ChatGPT-4o and DeepSeek R1. These AI-generated materials were compared with those created by the medical team in terms of development time, readability, understandability, actionability, and accuracy. RESULTS: The total time for manual preparation was 14 hours, while ChatGPT and DeepSeek consumed 0.62 and 0.78 hours, respectively. Regarding readability, the frequency of difficult words was more variable in Manually Written and ChatGPT materials, while DeepSeek showed more consistency. The proportion of simple sentences was highest in DeepSeek, followed by ChatGPT, with complete separation between Manually Written and ChatGPT (δ = 1). Content word frequency was highest in Manually Written, while ChatGPT had the lowest but most stable values. Personal pronouns were most frequent in Manually Written, with high variability, and least in DeepSeek, which was stable.All three groups had similar readability levels, reached Chinese elementary school-level readability for simple sentence proportion and personal pronouns, with high school-level difficulty words and content word frequency. The understandability and actionability scores were above 70, with ChatGPT being more stable in understandability, and DeepSeek in actionability. No significant differences were found between groups..In terms of accuracy, inter-group comparisons showed significant differences (H = 7.27, P = .03), but no significant differences in multiple comparisons. The direct comparison between ChatGPT and DeepSeek showed a negligible effect size (δ = .02), with no significant difference (Z = -.06, P = .96) . Four out of eight experts noted accuracy issues in the AI-generated materials. CONCLUSIONS: Generative AI significantly improved the efficiency of developing PEMs for CHD patients. The materials generated by ChatGPT-4o and DeepSeek R1 were comparable to the professionally written ones in terms of readability, understandability, and actionability. However, improvements in reducing difficult words and increasing content word frequency are needed to enhance readability. The accuracy of AI-generated materials still poses concerns, including potential AI "hallucinations" and requires review by healthcare professionals. Generative AI holds considerable potential in generating PEMs, and future research should assess its applicability and effectiveness in real-world patient and family contexts.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame machine prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. The Gemma side is a direct model label for every work in the frame, read from the title-only record. The Codex side is a classifier learned from the 10,348 direct Codex labels and calibrated to design-weighted sample rates; fields without enough sample support carry no Codex call. Candidate is the union of the two sides; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels.
Distilled classifier scores by category (both heads)
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.021 | 0.088 |
| Meta-epidemiology (narrow) | 0.001 | 0.000 |
| Meta-epidemiology (broad) | 0.001 | 0.001 |
| Bibliometrics | 0.003 | 0.001 |
| Science and technology studies | 0.001 | 0.001 |
| Scholarly communication | 0.002 | 0.002 |
| Open science | 0.002 | 0.002 |
| Research integrity | 0.001 | 0.001 |
| Insufficient payload (model declined to judge) | 0.003 | 0.001 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; a candidate call from one source (direct Gemma or distilled Codex), not a consensus.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".