Assessing the Gun Violence Archive as an Epidemiologic Data Source for Community Firearm Violence in 4 US Cities

Abstract

ImportanceFirearm injury is a major public health burden in the US, and yet there is no single, validated national data source to study community firearm violence, including firearm homicide and nonfatal shootings that result from interpersonal violence.ObjectiveTo assess the validity of the Gun Violence Archive as a source of data on events of community firearm violence and to examine the characteristics of individuals injured in shootings.Design, Setting, and ParticipantsThis cross-sectional observational study compared data on community firearm violence from the Gun Violence Archive with publicly available police department data, which were assumed to be the reference standard, between January 1, 2015, and December 31, 2020. Cities included in the study (Philadelphia, Pennsylvania; New York, New York; Chicago, Illinois; and Cincinnati, Ohio) had a population of greater than 300 000 people according to the 2020 US Census and had publicly available shooting data from the city police department. A large city was defined as having a population greater than or equal to 500 000 (ie, Philadelphia, New York City, and Chicago). Data analysis was performed in December 2022.Main Outcomes and MeasuresEvents of community firearm violence from the Gun Violence Archive were matched to police department shootings by date and location. The sensitivity and positive predictive value of the data were calculated (0.9-1.0, excellent; 0.8-0.9, good; 0.7-0.8, fair; 0.6-0.7, poor; and <0.6, failed).ResultsA total of 26 679 and 32 588 shooting events were documented in the Gun Violence Archive and the police department databases, respectively, during the study period. The overall sensitivity of the Gun Violence Archive over the 6-year period was 81.1%, and the positive predictive value was 99.0%. The sensitivity steadily improved over time. Shootings involving multiple individuals and those involving women and children were less likely to be missing from the Gun Violence Archive, suggesting a systematic missingness.Conclusions and RelevanceThese findings support the use of the Gun Violence Archive in large cities for research requiring its unique advantages (ie, spatial resolution, timeliness, and geographic coverage), albeit with caution regarding a more granular examination of epidemiology given its apparent bias toward shootings involving multiple persons and those involving women and children.

Examining media reports of pediatric unintentional firearm injury deaths for prevention messaging concerning secured storage of firearms: U.S., 2021–2022

“…In particular, fatal shootings and incidents involving children were found to have a high capture rate when compared to police data. However, those findings were based on 4 urban areas; therefore, results may differ for other geographic areas (Gobaud et al 2023 ). The findings of this report are subject to at least 6 limitations.…”

“…Third, since the GVA is not 100% complete in the reporting of unintentional firearm injury deaths, we could not establish a true mortality rate. Fourth, the GVA capture rate may be impacted by external influences on news media coverage (Gobaud et al 2023 ). Factors such as politics, advocacy, and news media goals may increase or decrease reporting in shooting events.…”

Domestic firearm violence against women (2018–2021)

“…Additionally, though there have been few comprehensive studies conducted to confirm the reliability, reproducibility, or validity of the GVA database, a recent study by Gobaud et al revealed that the GVA is generally an appropriate model for firearm violence research. Despite these limitations, the GVA is one of the most comprehensive databases on this topic that has been utilized in numerous studies and by news sources [ 3 , [27] , [28] , [29] , [30] ]. Furthermore, the methods of collection by the GVA have changed over time and vary depending on the location (e.g., increased coverage of urban compared to rural locations).…”

Evaluating Firearm Violence After New Jersey’s Cash Bail Reform

“…GVA data are typically used to measure intentional, interpersonal firearm violence, because intentional self-harm incidents (ie, suicides and suicide attempts) are omitted and unintentional shootings are a small proportion of total incidents. Validation work has found good sensitivity (81%) and excellent specificity (99%) when comparing the GVA with incident-level firearm violence data compiled by local police agencies, which are considered the criterion standard . We geocoded the GVA location data to US counties and analyzed a combined total count of fatal and nonfatal shootings from 2014 to 2019.…”

“…Validation work has found good sensitivity (81%) and excellent specificity (99%) when comparing the GVA with incident-level firearm violence data compiled by local police agencies, which are considered the criterion standard. 34 We geocoded the GVA location data to US counties and analyzed a combined total count of fatal and nonfatal shootings from 2014 to 2019. For both outcomes (NCHS firearm-related mortality and GVA combined fatal and nonfatal shootings), we estimated a quarterly, county-level rate using the annual county population provided by the US Bureau of the Census County Population Estimates (2014-2019).…”