Back to the Future Again: Relinking NC-Vdrs Data for Future Linkage Projects

Tuesday, June 6, 2017: 2:20 PM
440, Boise Centre
Scott Proescholdbell , North Carolina Department of Health and Human Services, Raleigh, NC
Shana M Geary , North Carolina Department of Health and Human Services, Raleigh, NC
Mary E Cox , North Carolina Department of Health and Human Services, Raleigh, NC
Tammy Norwood , North Carolina Department of Health and Human Services, Raleigh, NC

BACKGROUND: More than 42 states and jurisdictions are now using the Center for Disease Control and Prevention National Violent Death Reporting System (CDC NVDRS) web-based system to collect additional data on deaths due to violence. Since CDC NVDRS transitioned to the web-based system in 2014 with de-identifiable data, some states have found methods to re-link de-identified NVDRS state data back to other data sources for additional analyses to better understand violent deaths. NC has started to develop linkage protocols to make the data more useable for public health purposes.

METHODS: The North Carolina Violent Death Reporting System (NC-VDRS) creates a linkage key for probabilistic linkage using several existing semi-unique variables. Other data sources, such as vital records, can then be linked back to the NC-VDRS data to provide limited personally identifiable information (PII) so that additional linkages are possible. A step-wise progression is developed to check and confirm matches between the NC-VDRS and vital records data. An unmatched file can then be separated out to view in greater detail.

RESULTS: In one small project, we were able to test a number of ways to create the best matching ID across NC-VDRS and vital records data sources. The most successful matching ID resulted in nearly 18,553 of 18,778 cases from each source were successfully linked with a 98.8% match rate. The linked data will then be used to link to other data sets. For example, we currently have a graduate student who is linking these cases to court files to better understand restraining orders and their effectiveness in preventing intimate partner violence deaths.

CONCLUSIONS: Data linkage is an increasingly important skill for epidemiologists and data analysts to have in their toolbox. Most states have disparate data systems specific to key health outcomes. Developing linkage tools and processes is one way to demonstrate the value added in linking data. In this project, we re-linked NC-VDRS data back to the original death certificate data to enable additional data linkages using limited PII.