Capturing Signs and Events Related to Catheters in Clinical Text

Abstract

Annotated clinical corpora are necessary to extract information from clinical text for answering clinical questions. However, publicly available annotated clinical corpora are limited because of privacy issues, ethical concerns, and resource requirements for curating annotated corpora. When available, they are annotated for specific purposes and might lack the annotations required to answer clinical research questions. These challenges open up the opportunity to identify considerations and develop an annotated clinical corpus to answer a clinical research question.

To narrow the scope of capturing clinical concepts and knowledge within clinical text, this work focuses on a use case. The use case is capturing signs and events related to catheters from clinical adverse event notes for reducing sepsis and infection rates. This work addresses four research questions: RQ1: What methods utilize clinical text to reduce sepsis and infections? RQ2: What characteristics of catheter-related signs and events can be captured from clinical text? RQ3: How can an incremental annotation-based method be developed to extract information about catheter-related signs and events from clinical text? RQ4: How can clinical knowledge about catheter-related signs and events be captured?

Addressing the research questions resulted in four publications. The contributions are identifying research gaps, developing an annotated corpus, and developing a corresponding knowledge model. Results from this work are being extended in ongoing research.