Attitudes of physicians towards large language models in healthcare: insights from a stratified random survey in China

The integration of large language models (LLMs) in healthcare has the potential to streamline physician workflows and enhance clinical efficiency. Understanding physicians’ attitudes toward LLMs is crucial for their development and integration into healthcare. However, research in this area remains limited. This study surveyed physicians across three major hospitals in Chongqing, China, from October to December 2023, using a stratified random sampling approach. The survey consisted of a 20-item questionnaire covering demographics, perception of LLMs, the demand for LLM applications in medical practice, and factors influencing their acceptance. Statistical analysis was performed using a Generalized Linear Mixed Model to identify factors associated with positive attitudes toward LLMs. Of the 419 respondents, 211 (50.3%) held a positive view of LLMs in healthcare. Physicians with prior experience using LLMs were more likely to have positive attitudes (Odds Ratio 1.58, 95% Confidence Interval: 1.05–2.37). Significant demand for LLMs was observed, particularly for diagnostic assistance (73.7%), literature searches (70.6%), and data analysis (69.2%). However, concerns about the maturity of the technology (69.0%), data privacy and security (58.5%), and the risk of LLMs generating erroneous content (58%) were prevalent. This study reveals a cautiously optimistic stance among physicians from three university-affiliated hospitals in Chongqing, China towards LLMs, highlighting both the growing demand for and reservations about these technologies and providing valuable insights for their guided application in healthcare settings.