Abstract: Although unstructured clinical text data are harder to analyze, they can often contain significant information that can expose new domains for analysis. Using an open-source Python package, we tested simple text classification methods on clinical text from the emergency department and admit notes to classify social features, specifically housing stability, of acute care patients. We found that these methods can be easily implemented and generalized while yielding statistically meaningful results.

Describe the new knowledge and additional skills the participant will gain after attending your presentation.: Attendees will understand our preliminary findings of extracting social features from unstructured clinical text notes and how open-source NLP tools can be easily implemented and useful in a clinical setting.

Authors:

Andrew Teng (Presenter)
University of Washington

Adam Wilcox, University of Washington