SHM Converge Scientific Abstract ePoster Gallery-Generative Artificial Intelligence Use Among Internal Medicine Residents

Pdf Summary

This survey studied how internal medicine residents in North Carolina are using large language models (LLMs), such as ChatGPT, and what they think about them. A total of 152 residents responded (29% response rate) across five programs.<br /><br />Key findings:<br />- 45% reported using an LLM at least once.<br />- 26% said they used LLMs for professional purposes.<br />- Among professional users, use was split between clinical work (12.5%), non-clinical work (35%), and both (52.5%).<br />- In clinical settings, most users reported only occasional use, with weekly or daily use uncommon.<br />- In personal life, LLMs were used more often than in professional settings.<br /><br />Residents had significant concerns about LLMs:<br />- 96% mentioned at least one risk or limitation.<br />- The most common concerns were inaccuracies/hallucinations, legal and ethical issues, data privacy, fake references, plagiarism, bias, and low-quality output.<br /><br />Despite these concerns, interest in learning more was high:<br />- 65% wanted more education about LLMs, and another 30% answered “maybe.”<br />- About half thought residency programs should include a formal LLM curriculum, with 42% selecting “maybe.”<br />- Preferred educational approaches included experiential learning, didactics, and interactive modules.<br />- No respondents reported having received formal training on LLMs.<br /><br />Overall, the study suggests that internal medicine residents are already experimenting with generative AI in both personal and professional contexts. They recognize important risks, but also see a growing role for LLMs in medicine and want more guidance, training, and clear practice standards from educators and administrators.

Asset Subtitle

Aaron Fried

Meta Tag

Author List Aaron Fried, Aimee Zaas, Debra Bynum, Donna Williams, Emily Mullen, Spencer Dorn, William Leland

Category Research

Concept Large Language Model

Concept Internal Medicine Resident

Concept Generative AI

Concept Medical Training

Concept Clinical Practice

Distinguished Non-Finalist

Presenter Organization University of North Carolina at Chapel Hill

Presenting Author Aaron Fried

Track Technology in Hospital Medicine

Keywords

large language models

ChatGPT

internal medicine residents

North Carolina

survey study

generative AI

clinical use

professional use

data privacy

medical education

Large Language Model

Internal Medicine Resident

Generative AI

Medical Training

Clinical Practice