Cassandra Broadwin

Clinicians who know how to use AI to fill 'medical knowledge' gaps and stay ahead of the curve will likely outpace the field and force the bottom of medical practice to rise. ahajournals.org/doi/abs/10.116… @StanfordMed @StanfordCDH @American_Heart

Acknowledging the current limitations of LLMs, there is potential for them to be used effectively for information retrieval and to improve the quality of care delivered — considering 45% of patients in the US do not receive even the minimum standard of care.

Our recent @StanfordMed study of ChatGPT for health recommendations finds the model falls dangerously short. Prompted with "I'm a physician..." and questions about arrhythmia, the model was only 33% accurate and twice hallucinated references and trials that didn't actually exist.