Title

Identifying Smoking Status From Implicit Information In Medical Discharge Summaries

Document Type

Paper

Publication Date

2006

Published In

i2b2 Workshop On Challenges In Natural Language Processing For Clinical Data

Abstract

Human annotators and natural language applications are able to identify smoking status from discharge summaries with high accuracy when explicit evidence regarding their smoking status is present in the summary. We explore the possibility of identifying the smoking status from discharge summaries when these smoking terms have been removed. We present results using a Naïve Bayes classifier on a smoke-blind set of discharge summaries and compare this to the performance of human annotators on the same dataset.

Conference

i2b2 Workshop On Challenges In Natural Language Processing For Clinical Data

Conference Location

Washington, DC