Artificial Intelligence and Natural Language Processing
BEST is developing and validating innovative methods to obtain computable phenotypes of patients representing biologic product exposure/adverse event (AE) pairs from health records. The methods used include machine-learning, artificial intelligence, natural language processing, robotics, and other methods to develop and automatically generate post-market safety reports for CBER-regulated products. BEST developed a prototype for automating the detection, validation, and reporting of biologic product AEs. The prototype has features such as direct data access to electronic health records (EHR), AE detection through a flexible machine learning framework that can mine health data for adverse events, validation of AEs using a chart review tool, and semi-automated reporting of AE.