Cargando…

SAT119 Artificial Intelligence Mediated Assistants: Will They Educate The Public About Health Disparities In Diabetes?

Disclosure: W. Zhao: None. G. Wu: None. E. Cheng: None. J. Lee: None. A. Wong: None. Background: In 2022, 37 million US adults had diabetes and 96 million had pre-diabetes. Diabetes and prediabetes affect 49.3% population in America. In the diabetic population, 12.1% are non-Hispanic blacks, 11.8% o...

Descripción completa

Detalles Bibliográficos
Autores principales: Zhao, Weichen, Wu, Gloria, Cheng, Emily, Lee, Jay Yong, Wong, Adrial
Formato: Online Artículo Texto
Lenguaje:English
Publicado: Oxford University Press 2023
Materias:
Acceso en línea:https://www.ncbi.nlm.nih.gov/pmc/articles/PMC10554334/
http://dx.doi.org/10.1210/jendso/bvad114.984
Descripción
Sumario:Disclosure: W. Zhao: None. G. Wu: None. E. Cheng: None. J. Lee: None. A. Wong: None. Background: In 2022, 37 million US adults had diabetes and 96 million had pre-diabetes. Diabetes and prediabetes affect 49.3% population in America. In the diabetic population, 12.1% are non-Hispanic blacks, 11.8% of Hispanics, and 9.5% of Asian Americans. 54% of Americans read at or below the sixth-grade level. ChatGPT(TM,) launched 11/30/22, has the newest Artificial Intelligence (AI) software versus Google Assistant(TM), launched 05/2016. Purpose: Can ChatGPT use its natural language software to educate the lay public about diabetes and health disparities? Methods: Evaluate the readability of responses after five questions were posed to ChatGPT(TM) and Google Assistant(TM)Questions: 1. Do African Americans have a higher risk of complications with Type 2 Diabetes? 2. Do Hispanic Americans have a higher risk of complications with Type 2 diabetes? 3. Do Asian Americans have a higher risk of complications with Type 2 diabetes? 4. Do women get Type 2 diabetes more than men? 5. What is pre-diabetes vs diabetes? Word count and Dale-Chall score assessed the readability level of both ChatGPT and Google Assistant responses. The Dale-Chall readability formula is calculated based on the percentage of difficult words and average sentence length in words. The U.S. Department of Education used the Dale-Chall formula to adjust the text to grade-level readability. Result: Word count: CHAT:Q1=104; Q2=135; Q3=136; Q4=133; Q5=232 vs GOOGLE: Q1=38, Q2=49, Q3=16; Q4=50; Q5=42. CHAT Word Count Avg=148, SD=48.81; GOOGLE Word Count Avg=39, SD=13.7; P-value=0.003908 Dale-Chall: CHAT: Q1=8.4; Q2=8.6; Q3=9.8, Q4=8.6, Q5=8.7 vs. GOOGLE: Q1=6.7, Q2=9, Q3=9.4, Q4=6.3, Q5=9.1. CHAT Dale-Chall Score Avg=8.82, SD=0.56; GOOGLE Dale-Chall Score Avg=8.1, SD=1.47; P-value=0.1306. ChatGPT has nearly the same grade level of comprehension as Google Assistant, including statistics about health disparities. However, ChatGPT was trained using information and transforming it into paragraphs, and google assistant extracts the data from websites and the variability is higher with a higher standard deviation than ChatGPT. Conclusion: ChatGPT with its artificial intelligence software is able to provide answers to our patients. While the answers may require a high school reading level and beyond, this is an excellent beginning for health education. Presentation: Saturday, June 17, 2023