Abstract: PUB054
Disparities in Digital Health Literacy: Artificial Intelligence (AI)-Driven Evaluation of AKI Information Across Top-Ranked Websites
Session Information
Category: Artificial Intelligence, Digital Health, and Data Science
- 300 Artificial Intelligence, Digital Health, and Data Science
Authors
- Balakrishnan, Suryanarayanan, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Thongprayoon, Charat, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Kashani, Kianoush, Mayo Clinic Minnesota, Rochester, Minnesota, United States
- Cheungpasitporn, Wisit, Mayo Clinic Minnesota, Rochester, Minnesota, United States
Background
Acute kidney injury (AKI) is a major contributor to morbidity and progression to chronic kidney disease (CKD) and end-stage renal disease (ESRD). As patients increasingly turn to the internet for medical guidance, evaluating the accuracy, readability, and consistency of online AKI content is essential to improving public health education and health literacy. This updated cross-sectional study uses artificial intelligence (AI) to reanalyze the top Google search results related to AKI.
Methods
In April 2025, the first ten websites generated from a Google search for “acute kidney injury” were reviewed. Each site was assessed across six clinical domains: definition, symptoms, causes, risk factors, diagnosis, and treatment. Advanced natural language processing models (ChatGPT-4o and Google’s consensus summarizer) were utilized to evaluate sentiment polarity, consensus consistency, readability (Flesch-Kincaid grade level), and inferred public engagement based on content authority and audience targeting.
Results
40% of websites used a recovery-focused, upbeat tone, while 60% maintained a neutral clinical tone. Readability levels varied significantly, with some websites written at an 8th–9th grade level (e.g., MedlinePlus, National Kidney Foundation) and others at a college level (e.g., Wikipedia, academic institutions). High inferred engagement was found in established sources such as Mayo and Cleveland Clinics. While core information on AKI symptoms and treatments was generally consistent across platforms, substantial differences emerged in the discussion of etiologies and risk factors. Some content included complex medical classifications that may hinder comprehension among general readers.
Conclusion
Significant variability persists in the tone, complexity, and accessibility of online AKI information. Applying advanced computational tools to evaluate and guide the development of digital health content may improve public understanding, support earlier recognition of AKI, and help reduce disparities in kidney health outcomes.