Artificial intelligence (AI) in hematology and oncology fellowship (HOF) training: A multicenter survey of education, attitudes, and clinical use.

Abstract

9001 Background: A prior national survey of U.S. HOF curricula demonstrated substantial heterogeneity and limited protected time for didactics beyond traditional lecture-based education 1 . More recently, artificial intelligence (AI), including large language models and ambient listening tools, has become increasingly integrated into trainee education and clinical practice 2-5 . We conducted a multi-center survey to assess the use of AI among HOFs. Methods: Hematology/oncology (H/O) fellows were recruited via email by program leadership to complete an anonymous survey adapted from our prior study, with added questions on AI education, attitudes, and clinical use. Responses were collected via REDCap and summarized using descriptive statistics. Results: A total of 118 H/O fellows responded from 18 of 30 invited U.S. HOF programs (60%), primarily from academic centers (94%), with a near-uniform distribution across fellowship training years. Most fellows (74%) reported using AI tools such as ChatGPT or OpenEvidence. Commonly used resources included NCCN guidelines (92%), UpToDate (86%), faculty lecture slides (70%), primary journals (65%), podcasts (58%), textbooks (29%), and social media (20%). Only 8% reported receiving AI education during HOF training. Most fellows felt AI was useful for medical education (93%) and were confident using AI tools for learning (74%). The majority anticipated increasing AI use over the next 5 years (92%) and expressed interest in AI training during fellowship (82%). Fellows most commonly used large language model–based AI tools to clarify difficult concepts (86%), summarize journal articles (83%), and learn about emerging research (75%), whereas fewer reported use for question generation (31%) or patient case simulations (29%). AI-assisted documentation was the most commonly used clinical AI application (51%). Reported barriers to AI use included (in order of highest concern): uncertainty regarding accuracy, lack of formal training, data privacy concerns, and unclear ethical or institutional guidelines. Conclusions: AI tools are widely used and perceived as useful for clinical and educational purposes by current H/O fellows, yet formal training during fellowship remains limited. These findings highlight an unmet need for structured education on effective, safe, and ethical AI use, with opportunities for multi-institutional collaboration to achieve scalable impact aligned with contemporary oncology practice.