Detection of potential patient harms from clinical notes in electronic healthcare records: The Shakespeare Method
Many methods for finding adverse events in the text of healthcare records rely on predefining potential adverse events before searching for prespecified words and phrases or manual labeling (standardization) by investigators.
We developed the Shakespeare Method to identify potential adverse events, even if unknown or unattributed, without any pre-specifications or standardization of notes from electronic healthcare records.
The method was inspired by word-frequency analysis methods used to uncover the true authorship of disputed works credited to William Shakespeare. In this talk I will cover the details of the Shakespeare Method where we use a combination of filtering, classification, and topic modeling to identify documents from a specific group or time of interest, to review for potential adverse events.
Summer Rankin
Senior Lead Data Scientist at Booz Allen Hamilton
Dr. Rankin is a senior data scientist in the Strategic Innovation Group at Booz Allen Hamilton in Honolulu Hawaii. She leads projects that involve a range of machine learning techniques including deep learning, natural language processing, anomaly detection, and performance measurement.
She serves as an artificial intelligence subject matter expert for Indo-Pacific defense and health projects with recent publications on JAUST (a cognitive search tool) and the results of Project Shakespeare under the direction of Dr. Roselie Bright, an Epidemiologist at the FDA.
She holds a Ph.D. in Complex Systems and Brain Sciences and completed a postdoctoral fellowship (5 years) with Charles Limb, MD at Johns Hopkins School of Medicine where she studied musical creativity, auditory perception (fMRI), and human coordination with 1/f-type (fractal) auditory signals. She has multiple peer-reviewed publications, public software releases, and conference presentations in the fields of AI, data science, and neuroscience.
Presented by
