P3.C050: An Artificial Intelligence Curriculum for Pediatric Trainees and Faculty- Where do you start?

Background: The swift adoption of Artificial Intelligence (AI) in healthcare, and specifically in pediatrics underscores the need for an AI curriculum tailored for pediatric trainees and faculty. AI applications, including diagnostic and patient management tools, are reshaping patient care, necessitating proficiency in both AI theory and practice among pediatricians. However, an educational gap exists in current pediatric training regarding AI. Developing a structured AI curriculum is crucial for enhancing clinical decision-making and equipping the pediatric workforce to handle AI's ethical, legal, and social implications in the care of children.

Methods: To understand the educational needs and preferences regarding AI among pediatric professionals, a structured 32-question survey was created and electronically distributed via the Qualtrics platform, after IRB approval. The single-site survey targeted pediatric faculty and trainees at a large children's hospital, gathering information on roles, AI knowledge, past AI usage, motivations for learning AI (Table 1), acquisition of specific AI skills (Table 2), and educational format preferences. Descriptive statistics and Chi-Square test to compare groups were performed using SPSS software. Survey feedback is being used to develop a tailored AI curriculum ensuring its applicability across various pediatric subspecialties and experience levels.

Results: In this needs assessment survey with a 27% response rate (171/636), the majority of respondents were attending physicians (69%), with residents and fellows in smaller proportions. Participants hailed from diverse specialties, notably pediatric hospital medicine, and pediatric emergency medicine. Despite 70% not having used AI tools in their practice or studies, those who did referred to specific applications (a particular large language model; an ambient recording software for patient encounters), along with various diagnostic support systems. 36% identified themselves as 'very interested' in learning more about generative AI. This interest level was higher among attendings vs trainees (p=0.01); yet, there were no differences in self-reported AI knowledge or opinion about its’ importance amongst groups, with 69% overall viewing generative AI as crucial for the future of medicine. Several themes were generated describing respondent’s motivation for learning AI. Preferred learning formats include online courses, hands-on training, and workshops, with a preference for sessions lasting 30-45 minutes. Key challenges in integrating AI included steep learning curves, confidentiality concerns, accuracy of AI outputs, and the potential for perpetuating existing biases.

Conclusion: These results underscore the necessity for an AI curriculum that is accessible, practically focused, addresses ethical considerations, and is responsive to the concerns and time constraints of pediatric professionals at an academic institution. We will obtain feedback during the various planned monthly sessions which aim to disseminate the knowledge of applying generative AI to various aspects of pediatric clinical care, research, education, advocacy, and practice management. There is a potential for widespread adoption of this curriculum across various pediatric institutions.