August 16, 2024JAMA Network OpenOpen Access

Enhancing Postmarketing Surveillance of Medical Products With Large Language Models

View Full Paper

Ask AI

Bookmark

Authors

Discussion

Review highlights opportunities for large language models in adverse event detection, indicating false positive risks require prompt engineering.

Large language models may uncover novel adverse event signals from unstructured electronic health records, enhancing postmarketing surveillance across diverse clinical datasets.
Assessment of surveillance methodology evaluates probabilistic phenotyping and natural language processing to extract essential confounding factors from narrative clinical text.
Highlights that prompt engineering and equitable infrastructure are essential to minimize nonfactual statements and avoid costly false positive signals during regulatory reviews.

Matheny et al. (2024) studied this question.

View Full Paper

Ask AI

Bookmark

Review highlights opportunities for large language models in adverse event detection, indicating false positive risks require prompt engineering.