P3A054: Translating Pediatric Disease Terminology into Plain Language with Prompt Learning: Leveraging AI and Machine Learning

Background: Pediatric health literacy is vital to achieving optimal health outcomes in children, as poor health literacy may lead to misunderstandings about health conditions, medication errors, and inadequate self-health management. Consequently, many children and adolescents turn to the internet to gather information. Current resources often use language that is too technical for children to understand and there is a need for innovative and accessible tools that can help children understand their health conditions better.

Project Description: We developed an application leveraging OpenAI's application programming interface to create a search engine utilizing a generative pre-trained transformer model. This model has been optimized to recognize the user's disease, grade level, and native language as critical tokens within the input sequence. This model has set specific parameters utilizing the Agency for Healthcare Research and Quality Health Literacy Universal Precautions Toolkit to ensure safe use of more sensitive topics.

Discussion: In our pediatric health literacy tool, users are able to enter their grade level and the condition they are interested in learning about; the application subsequently generates a response tailored to the patient’s grade level. The uses are not limited to diseases: the algorithm also responds to queries such as “my knee hurts from basketball” with age appropriate responses. The algorithm also provides viable answers to mental health related and “what should I do” questions; though, many limitations remain.

Conclusion: The application has the potential to help improve pediatric health literacy by providing accessible and easy-to-understand health information, but the new technology must be well researched before implementation.