Biostatistics Weekly Seminar

Medication information is important to determine dosing in drug-related studies, such as pharmacokinetic, pharmacodynamic, or pharmacogenomic analyses. Natural language processing (NLP) systems have been developed to address this task. These are often more general-purpose, intended to extract all drugs simultaneously and with various intended uses, e.g., medication reconciliation. Existing systems may struggle to optimize performance with respect to a specific drug(s) of interest, which is critical for obtaining high-quality research datasets from electronic health records.

In this talk I will present medExtractR, an NLP system we built using R to extract medication information from free-text clinical notes. Users can easily customize this system through specifying function arguments, providing custom dictionaries, or even editing the publicly available R source code. Unlike existing systems, medExtractR is intended to be used for a specific drug(s) of interest, allowing researchers to tailor the system to a particular drug, study site, or patient cohort.

I will discuss how we developed medExtractR on notes from the Synthetic Derivative and compare its performance to three existing medication NLP systems: MedXN, MedEx, and CLAMP. I will also demonstrate how we applied and evaluated medExtractR in clinical notes from intensive care unit admissions in the MIMIC-III Clinical Care Database.

MRBIII, Room 1220
2 October 2019
1:30pm