Authors: Brenda Y. Miao, Christopher YK Williams, Ebenezer Chinedu-Eneh, Travis Zack, Emily Alsentzer, Atul J. Butte, Irene Y. Chen

Published on: February 06, 2024

Impact Score: 8.12

Arxiv code: Arxiv:2402.03597

Summary

• What is new: The novel use of GPT-4 to identify reasons for switching contraceptives from unstructured clinical notes, outperforming older models with higher accuracy and providing demographic-specific insights.
• Why this is important: Difficulty in extracting factors driving contraceptive switching from unstructured clinical notes.
• What the research proposes: Employing GPT-4’s zero-shot capabilities to accurately extract reasons for contraceptive selection and switching.
• Results: GPT-4 achieved microF1 scores of 0.849 and 0.881 for starting and stopping contraceptive extraction, respectively, with a human evaluation accuracy of 91.4%.

Technical Details

Technological frameworks used: GPT-4 via HIPAA-compliant Microsoft Azure API

Models used: Baseline BERT-based models, unsupervised topic modeling approaches

Data used: UCSF Information Commons clinical notes dataset

Potential Impact

Healthcare providers, insurance companies, pharmaceutical firms, and reproductive health-focused organizations could benefit from or be disrupted by these insights.

Want to implement this idea in a business?

We have generated a startup concept here: MediMatchRX.