Large Language Models and Primary Care: A Scoping Review.

Abstract

Large Language Models (LLMs) offer transformative potential for primary care but carry risks regarding bias and reliability. This scoping review synthesizes evidence on LLM applications in primary care adhering to PRISMA-ScR guidelines. We searched 10 databases for original research and assessed risk of bias using PROBAST. Of 28 included studies, most originated from high-income countries, with significant underrepresentation of low-and-middle-income regions. Common applications included information extraction (25%) and predictive modeling (14%). While performance was generally high (F1-scores 0.70-0.95) for structured tasks, interpretive tasks showed greater variability. Crucially, only 18% of studies had a low overall risk of bias, with 39% exhibiting high risk. While LLMs demonstrate utility in administrative and clinical tasks, the current landscape is geographically skewed and methodologically limited, requiring rigorous, inclusive evaluation frameworks before widespread adoption.