A Simple Free-Text Method for Extracting Semi-Structured Data from Electronic Health Records


The Epic Electronic Health Record (EHR) is regularly used in healthcare settings. A recent study featuring PISC Senior Scholar Alexis M. Zebrowski, PhD, MPH, and External Advisory Board Member Brendan Carr, MD, MS, evaluates the use of a free-text-like method to obtain data from the semi-structured flowsheet fields of the EHR. The team demonstrates the free-text-like method through a use case analyzing mortality in emergency department triage. Models were trained from 2014-2017 using gradient boosting and logistic regression. Demographic and clinical data from 2014-2017 was retrieved from the EHR. Results indicate a 2% mortality rate of ED patients within 48 hours of admittance. All aspects of the study have an area under the curve of at least 0.95. The study effectively models how “a free-text-like approach can be useful for extracting knowledge from large amounts of semi-structured EHR data.”


Go to Article

Back to News