Is It Time to Incorporate Large Language Models into EHRs?

Article discusses use of ChatGPT-4 in medicine, highlights benefits, risks and limitations. Halamka and Cerrato warn against reliance on internet-trained models for diagnostics and treatment due to biases and misinformation. Concerns include lack of medically trained staff in fine tuning and transparency in source assembly. Despite risks, some advocate LLM use in writing EHR notes. Nayak and colleagues favor "prompt engineering", claim close AI developer-clinician collaboration needed for output accuracy. Preiksaitis argues against LLM use in medical documentation due to potential threats to authenticity and bias propagation. Generative AI experiments aim to create skeleton notes for human augmentation/editing to reduce documentation time, increase time for patient care.

Opens on healthsystemcio.com