Hematology-Oncology Fellows' Use of Artificial Intelligence: A Multicenter Educational Practice and Needs Assessment Survey.

Abstract

Artificial Intelligence (AI) is rapidly being incorporated within healthcare, including medical education. Hematology-oncology (HO) is a complex field and AI is of great value in education and practice. We explored HO fellow's confidence and concerns with use of AI, AI training that HO fellows currently receive, and interest in future AI training. Multi-institutional survey administered to fellows in six adult HO programs. The survey was distributed electronically over one month with weekly reminders. Results were collected in RedCap and analyzed using R. 36 of 153 potential participants responded (23.5% response rate). Almost all (35, 97%) reported using AI. Fellows who reported using AI for work used it primarily for patient care (28, 82%) and/or research (29, 85%). Most were "somewhat confident" in the accuracy of AI-generated clinical information (25, 74%) and found AI "somewhat reliable" for clinical decision support (25, 74%). Respondents cited concerns about the use of AI in clinical practice, most often reliability of clinical data (32, 94%). Only one respondent (2.8%) received formal education on AI during fellowship and over who received informal training only reported its quality as "fair" or "poor" with 29 of these remaining 35 (83%) respondents interested in formal training. 31 fellows (86%) reported thinking AI would be important in their future career plans and 17 (49%) felt AI training should be required for HO fellows. While almost all HO fellows use AI, formal education in AI is rare in HO fellowship. Most fellows expressed interested in such training and believe that AI will be important in their careers, but also expressed concerns about use of AI, especially reliability of AI-generated clinical data. We plan to create formal curricula specific to HO training.