LLM Use to Identify Adolescent Patient Portal Account Access by Guardians
JAMA Network Open
|
Summary
A study published in JAMA Network Open assessed the effectiveness of using a large language model (LLM) to identify guardian authorship of messages in adolescent patient portals. Given HIPAA and minor consent laws, confidentiality is essential, but a significant number of adolescent portal accounts are accessed by guardians. The research tested GPT-4 on identifying such guardian-authored messages from a dataset and found high sensitivity (98.1–98.3%) and specificity (88.4–88.9%), outperforming prior natural language processing algorithms. This LLM-based approach could improve the detection of unauthorized access and enhance patient confidentiality in electronic health records.
