Investigation of Racial Bias in an Epilepsy Machine Learning Algorithm Trained on Physician Notes

Abstract

Purpose: There are 470,000 children with epilepsy in the United States, of which 20-40% are drug-resistant and may be candidates for resective epilepsy surgery. Surgical treatment has drastically better outcomes than continuing pharmacotherapy alone. Racial disparities in the utilization of epilepsy surgery are well documented. It is unknown if a machine learning (ML) algorithm trained on physician notes would produce biased recommendations for epilepsy presurgical evaluations. Our objective was to assess the impact of patient demographics on ML recommendations for epilepsy surgical candidacy. Methods: …