Back

Textual Triage: Assessing GPT4 for Classification of Free-Text Medication-Related Messages for Hypertension Management

Batugo, A.; Hwang, S.; Davoudi, A.; Luong, T.; Lee, N.; Mowery, D.

2024-09-24 health informatics
10.1101/2024.09.23.24314207 medRxiv
Show abstract

Patient-generated free-text messages are a well-recognized source of clinical burden and burnout for clinicians. Machine learning approaches such as Large Language Models (LLMs) may be applied to alleviate this burden by automatically triaging and classifying messages, but their performance in this domain has not been fully characterized. In this study, we analyzed the effectiveness of GPT4 for classifying patient and provider messages for hypertension management through prompt engineering, comparing its performance to an alternative unsupervised generative statistical approach. The results of this study suggest GPT is promising for classification of medical-related messages even with very few guiding examples.

Matching journals

The top 4 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.