Accuracy of GPT vs Google Translate for Translation of Patient-Specific Emergency Discharge Instructions

Poster Number: P71
Presentation Time: 05:00 PM - 06:30 PM

Abstract Keywords: Health Equity, Diversity, Equity, Inclusion, Accessibility, and Health Equity, Delivering Health Information and Knowledge to the Public
Primary Track: Applications
Programmatic Theme: Clinical Informatics

We aimed to compare translation accuracy of Google Translate (GT) versus OpenAI GPT. Using 50 real emergency department patient discharge instructions, we evaluated the accuracy of GT vs GPT for translation into Chinese, Spanish, and Russian. For Chinese translations (all analyses will be complete by summer 2024), GPT was significantly more accurate than GT (94.6% vs 89.5%, p=0.018). Grammar anomalies and sentences providing return precautions were significantly associated with inaccurate translations by GT.

Speaker(s):
Elaine Khoong, MD, MS
University of California San Francisco