Abstract: The Smoking History And Pack-year Extraction System (SHAPES) is a rules-based Natural Language Processing systems developed to extract quantitative tobacco exposure from clinical notes. The system was validated on 4040 notes and identified 17 of 22 patients eligible for lung cancer screening, F-measure 0.82. SHAPES identified all 35 patients eligible for abdominal aortic aneurysm screening, F-measure 0.85. SHAPES performs well for pragmatic problems such as identifying patients for lung cancer or abdominal aortic aneurysm screening.

Learning Objective 1: Participants will learn the additional benefits of quantitative smoking exposure data versus smoking status only.


Travis Osterman (Presenter)
Vanderbilt University

Julie Wu, Vanderbilt University
Dara Mize, Vanderbilt University
Wei-Qi Wei, Vanderbilt University
Joshua Denny, Vanderbilt University

Presentation Materials: