Development of a NLP Pipeline for MRSA Surveillance Among Veterans Receiving Care at VA Facilities Project ID 689-13-1-1455-0013

Project: Research project

Project Details

Description

Introduction Dr. Matthew Scotch is a Natural Language Processing (NLP) expert with formal training in biomedical informatics, and public health and will assist with the VA Connecticut Healthcare Systems work on the Consortium for Healthcare Informatics Research (CHIR). The work will entail developing an NLP pipeline for Methicillin-resistant Staphylococcus Aureus (MRSA) surveillance among Veterans receiving care at VA facilities. Background Methicillin-resistant Staphylococcus Aureus (MRSA) is responsible for an increasing number of community and healthcare associated infections in the US. In 2006-2007, the Veterans Health Administration (VHA) began a national initiative to decrease the impact of MRSA on Veterans health. Evidence in the medical literature has suggested that programs utilizing such active surveillance effectively decrease the impact of MRSA. Part of the VAs initiative mandates testing of all hospitalized patients for MRSA on admission, when transferred between acute care units and on hospital discharge. While the use of active surveillance remains controversial, the VA has made a substantial commitment to its use as a key factor in decreasing the impact of MRSA on Veterans health. However, problems exist for the adoption of this strategy due to the additional workload required to collate, analyze and effectively respond to the large amount of complex data. While testing individuals to identify colonization may control the impact of MRSA in the health care setting, this activity is labor intensive, prone to error and consumes significant resources. VistA, the VHAs electronic medical record (EMR), provides a unique opportunity for the rapid development and implementation of EMR-based informatics tools to support the VA MRSA initiative. Automatic methods to extract this information are not straightforward as much of the data in VistaA is hard to interpret given its variation including structure vs. unstructured format. Natural Language Processing (NLP) is a scientific domain that focuses on extracting and unlocking information from free text. NLP has been used in medical informatics for several years now for problems associated with health services research, hospital surveillance, and patient outcomes. NLP tackles the problem that structured data, including administrative and other coded data sources, do not adequately capture the clinically- and epidemiologically-relevant information needed for performing MRSA surveillance or for monitoring the impact of infection control programs. NLP can also consider the use of the semi-structured data that is characteristic of microbiology laboratory reports. Scope A Natural Language Processing pipeline that can process annotated MRSA notes and identify concepts and relationships between those concepts. The pipeline needs to link to a machine learning classifier in order to identify status of MRSA including whether the patient has a urinary track infection (UTI) or a blood stream infection (BSI) related to the MRSA.
StatusFinished
Effective start/end date4/15/134/15/15

Funding

  • VA: Boston Healthcare System: $80,000.00

Fingerprint Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.