S869 Can Artificial Intelligence Help Improve Medical Literacy in Patients With Educational Disparities? A Pilot Study Assessing the Role of Chat GPT in Colonoscopy (CS) and Esophagogastroduodenoscopy (EGD) Patient Education

Abstract

Introduction: Around 15 million colonoscopies (CS) and 6-7 million esophagogastroduodenoscopy (EGDs) are performed annually in the United States. The prevalence of a clear understanding of CS and EGD among patients can vary significantly based on several factors, including communication, educational resources, and individual patient characteristics like level of education. Educational disparities exist, with 37% population having no more than a high school diploma and only 14% having advanced education. Previous studies showed that simplified instructions improved bowel preparation quality, and cancellation rates were reduced to half. However, there are no studies customizing these instructions per patient literacy level. With an aim to enhance awareness and comprehension, we aimed to explore the potential role of Chat Generative Pre-Trained Transformer (ChatGPT) to determine how it could assist in simplifying text for improved understanding. Methods: We used the American College of Gastroenterology patient education resources on CS and EGD, for their overview, indications, sedation, preparation, procedure, risks and alternative tests. We asked ChatGPT to modify the text so that someone who is at a 6th-grade level or below can understand it. To assess the education level of a specific text we compared the original and modified texts for their Flesch-Kincaid Grade Level (FKGL), Flesch Reading Ease Score (FRES), Reading Level, Average Words per Sentence, and Average Syllables per Word. We analyzed the statistical significance using a paired t-test. Results: The average original FKGL before and after modification by ChatGPT was: CS (12.38 and 7.12) and EGD (6.48 and 5.20). Using ChatGPT there was a significant change in FKGL, FRESe Score (FRES), Reading Level, Average Syllables per Word, and Total Words (Table 1). Noticeably, there was a significant reduction in text for the topic of bowel preparation for CS eliminating many medical terms compared to other topics. Conclusion: Our study showed that ChatGPT can make medical information about CS and EGD easier to understand and more usable to people with varying literacy levels. The modifications made by ChatGPT were statistically significant (P-value < 0.05), indicating successful text simplification. This demonstrates that AI can effectively convey information and tailor it to specific audiences. This pilot study would drive further research to address patient education despite the educational, gender, racial, ethnic, socioeconomic and regional disparities. Table 1. - Comparison of Patient Education Text before (Pre) and after (Post) ChatGPT's modification for Colonoscopy and Esophagogastroduodenoscopy using text assessment scores Sub-topic FKGL (Pre) FKGL (Post) FRES (Pre) FRES (Post) RL (Pre) RL (Post) Avg. W/S (Pre) Avg. W/S (Post) Avg. S/W (Pre) Avg. S/W (Post) Avg. Sentences (Pre) Avg. Sentences (Post) Avg. Words (Pre) Avg. Words (Post) Colonoscopy Overview 11.9 8.6 39.8 52.2 College 10th & 12 grade 14.5 10.7 1.8 1.7 31 23 448 247 Indications 13.9 7.5 38.6 66 College 8th & 9th grade 24.1 13.7 1.7 1.5 11 9 265 123 Alternative Screening Methods 12.1 6.9 32.4 62 College 8th & 9th grade 13.5 9.3 1.9 1.6 11 7 149 65 Bowel Preparation 10.5 6 47.3 69.8 College 8th & 9th grade 15.5 10 1.7 1.5 27 5 418 50 Complications 13.5 6.6 28.8 62.9 College 8th & 9th grade 17 8.4 1.9 1.6 7 9 119 76 p value (Paired t-test) 0.001* 0.002* NA 0.007* 0.004* 0.175 0.042* Esophagogastroduodenoscopy Overview 8.9 7.4 56.9 66.2 10th & 12 grade 8th & 9th grade 14.4 13.5 1.6 1.5 14 13 202 175 Indications 9.2 5.3 40 71.6 College 7th grade 6 8.2 1.9 1.5 26 11 155 90 Sedation 6.1 4.8 69.6 78.2 8th & 9th grade 7th grade 10.2 10 1.5 1.4 10 9 102 90 Preparation 8.8 5.5 51.9 76.6 10th & 12 grade 7th grade 11 11.6 1.7 1.4 8 7 88 81 Before Procedure 7.1 4.2 67 85.4 8th & 9th grade 6th grade 12.7 11.3 1.5 1.3 6 6 76 68 During Procedure 6.4 5.3 74.3 76.9 7th grade 7th grade 13.9 11.3 1.4 1.4 7 6 97 68 After Procedure 7 5.8 67.2 75.6 8th & 9th grade 7th grade 12.5 12.6 1.5 1.4 6 5 75 63 Complications 4.3 5.2 74.3 71.9 7th grade 7th grade 5.6 7.9 1.5 1.5 11 8 62 63 Alternative Screening Methods 7.6 5.5 65.6 81.8 8th & 9th grade 6th grade 14.1 14.8 1.5 1.3 11 6 155 89 p value (Paired t-test) 0.005* 0.007* NA 0.872 0.008* 0.082 0.017* Footnotes: (1) Abbreviations: Pre (Text before ChatGPT's modification), Post (Text after ChatGPT's modification), FKGL (Flesh-Kincaid Grade Level), FRES (Flesch Reading Ease Score), RL (Reading Level), Avg. (Average), W/S (Words per Sentence), S/W (Syllables per Word), NA (Not Applicable); (2) Bold* P-values are statistically significant (P<0.05)

Ähnliche Arbeiten