Data Scientist
Dept of Veterans Affairs; University of Utah
Disclosure(s): No financial relationships to disclose
Kelly Peterson is a Data Scientist in the Dept of Veterans Affairs and the Division of Epidemiology at the University of Utah. I specialize in Natural Language Processing (NLP) and Machine Learning (ML) as a Data Scientist and as a Health Sciences Researcher. The common theme of my work is leveraging models to consume large quantities of data to identify common factors, as it is prohibitive for humans to read so many documents. Historically, I have focused on developing novel algorithms to identify risk factors in clinical text as it is difficult to anticipate what risk factors may be present in a population. In this situation, I utilized topic modeling techniques to aid in uncovering the significant themes, terms and phrases which may be present across thousands or millions of documents. I’ve been employing such topic models to more rapidly and thoroughly understand both biomedical literature and clinical documentation. Advancements in natural language processing permit a “distant reading” of such quantities such that we can more rapidly explore their contents. Recently, I utilized NLP and ML to rapidly identify positive COVID patients for accurate reporting and infection control in which this work is consumed on a daily basis by VA Operations and Whitehouse Advisory Board. Recent achievements using these models include clinical text processing include a machine learning model which extracts positively affirmed mentions of Travel History. This model was trained as a research project with the University of Utah and is now deployed in an operational capacity where it runs on thousands of documents nightly to generate travel history profiles given the 1.6+ million notes it has processed. These accomplishments coupled with decades of experience in software development demonstrate a record of timely and successful technical contributions to the field of NLP, ML and ultimately improved care for patients.