Finding the elusive trauma denominator: Feasibility of combining data sets to quantify the true burden of firearm trauma

• Published in: The journal of trauma and acute care surgery Vol. 90; no. 3; pp. 466 - 470
• Main Authors: Carmichael, Heather, Samuels, Jason M., Jamison, Ethan C., Bol, Kirk A., Coleman, Jamie J., Campion, Eric M., Velopulos, Catherine G.
• Format: Journal Article
• Language: English
• Published: United States Lippincott Williams & Wilkins 01-03-2021
• Subjects: Colorado - epidemiology; Cost of Illness; Feasibility Studies; Female; Firearms - statistics & numerical data; Homicide - statistics & numerical data; Hospital Mortality; Humans; Male; Registries; Trauma Centers - statistics & numerical data; Violence - statistics & numerical data; Wounds, Gunshot - epidemiology; Colorado
• ISSN: 2163-0755; 2163-0763
• DOI: 10.1097/TA.0000000000003005

Evidence guiding firearm injury prevention is limited by current data collection infrastructure. Trauma registries (TR) omit prehospital deaths and underestimate the burden of injury. In contrast, the National Violent Death Reporting System (NVDRS) tracks all firearm deaths including prehospital fatalities, excluding survivors. This is a feasibility study to link these data sets through collaboration with our state public health department, aiming to better estimate the burden of firearm injury and assess comparability of data.
We reviewed all firearm injuries in our Level I TR from 2011 to 2017. We provided the public health department with in-hospital deaths, which they linked to NVDRS using patient identifiers and time of injury/death. The NVDRS collates information about circumstances, incident type, and wounding patterns from multiple sources including death certificates, autopsy records, and legal proceedings. We considered only subjects with injury location in a single urban county to best estimate in-hospital and prehospital mortality.
Of 168 TR deaths, 166 (99%) matched to NVDRS records. Based on data linkages, we estimate 320 prehospital deaths, 184 in-hospital deaths, and 453 survivors for a total of 957 firearm injuries. For the matched patients, there was near-complete agreement regarding simple demographic variables (e.g., age and sex) and good concordance between incident types (suicide, homicide, etc.). However, agreement in wounding patterns between NVDRS and TR varied.
We demonstrate the feasibility of linking TR and NVDRS data with good concordance for many variables, allowing for good estimation of the trauma denominator. Standardized data collection methods in one data set could improve methods used by the other, for example, training NVDRS abstractors to utilize Abbreviated Injury Scale designations for injury patterns. Such data integration holds immediate promise for guiding prevention strategies.
Epidemiological study, level IV.