We developed and evaluated a natural language processing (NLP) approach for identifying intimate partner violence (IPV) from clinical narratives. 191,728 clinical notes from 12,276 adult female patients with emergency department visits were used as a training dataset to iteratively develop a lexicon of IPV, customized negations, and NLP algorithms. Through manual review of clinical notes for 312 NLP identified IPV patients (2.5%) in the training dataset, our NLP pipelines yielded 93.3% precision.

Learning Objective: Natural language processing apporach extracting the evidence of intimate partner violence from clinical notes

Authors:

Vivienne Zhu (Presenter)
Medical University of South Caarolina

Christine Carr, Medical University of South Caarolina
Kit Simpson, Medical University of South Caarolina
Roberta Dawson, Medical University of South Caarolina
Alyssa Rheingold, Medical University of South Caarolina
Leslie Lenert, Medical University of South Caarolina

Presentation Materials:

• Abstract/Manuscript - 1