Disparities in seizure outcomes revealed by large language models.

We used large language models (LLMs) and natural language processing to extract seizure outcomes from clinical note text.We found no evidence of intrinsic bias in the LLM algorithm, in that it performed similarly across all demographic groups.Using LLM-extracted seizure outcomes, female sex, public insurance, and lower income zip- codes were associated with higher likelihood of seizures at each visit.Black race was associated with higher likelihood of seizures in univariable but not multivariable analysis.These findings highlight the critical need to reduce disparities in the care of people with epilepsy.