New research suggests that the Gun Violence Archive (GVA), an independent data collection group, can provide valuable insights into community firearm violence in large cities. Although the GVA’s data collection methods have been questioned, a study comparing their data to police department records showed promising results. The GVA’s dataset offers timeliness, geographic coverage, and spatial resolution, but its bias toward certain types of shootings limits its use for detailed epidemiological studies. Healthcare stakeholders are utilizing data-driven approaches and exploring the role of artificial intelligence in firearm violence prevention.
Recent research highlights the potential benefits of utilizing independently collected gun violence data for studying community firearm violence in major cities. A study published in JAMA Network Open emphasizes that the Gun Violence Archive (GVA), an independent organization dedicated to data collection, can serve as a valuable, albeit limited, resource for researchers investigating community firearm violence in large cities.
The study’s authors acknowledge that firearm injuries are a significant public health issue in the United States, yet there is currently no single validated national data source available for studying this phenomenon. The GVA has compiled a dataset of firearm violence events using public records, media reports, and location information similar to that employed by police departments. This comprehensive dataset offers valuable insights into gun violence across the country.
While the GVA has the potential to expand the scope, timeliness, and flexibility of firearm injury research, concerns have been raised regarding the validity of its data collection methods. To assess the GVA’s viability as an epidemiological data source for community firearm violence, including firearm homicides and nonfatal shootings resulting from interpersonal violence, the research team conducted a cross-sectional observational study.
The study compared community firearm violence data from the GVA with publicly available police department data, which are the standard source of data for this type of research. Data were collected between January 1, 2015, and December 31, 2020, from four cities with populations exceeding 300,000 people, according to the 2020 US Census: Philadelphia, Pennsylvania; New York, New York; Chicago, Illinois; and Cincinnati, Ohio.
The researchers matched community firearm violence events from the GVA with police department shooting events based on location and date. They calculated the sensitivity and positive predictive value for each dataset to assess performance.
During the study period, the GVA documented 26,679 shooting events, while police department data documented 32,588 events. Both datasets reported a total of 29,791 individuals involved in shootings. Among them, 87.3 percent were involved in shootings with multiple injured individuals, and 69.7 percent were involved in nonfatal shootings.
The researchers discovered that the GVA had failed to capture information on 9,643 individuals involved in shootings, with 82.1 percent of these cases involving a single injured individual and 62.6 percent being nonfatal. The data exhibited a “systematic missingness,” with shootings involving women or multiple individuals less likely to be included in the GVA database.
Overall, the GVA demonstrated a sensitivity of 81.1, which steadily improved over the six-year study period. The dataset exhibited a positive predictive value of 99.0 percent.
Based on these findings, the researchers concluded that the GVA can be utilized in specific research endeavors that require its unique features, such as timeliness, geographic coverage, and spatial resolution. However, they cautioned against using the GVA for detailed examinations of epidemiological trends due to its bias toward shootings involving women, children, and multiple individuals.
To combat gun violence in the United States, healthcare stakeholders are employing data-driven studies as one approach, with some health systems exploring the integration of gun safety conversations into patient-provider communication, considering crime and violence as social determinants of health (SDOH).
Furthermore, researchers are investigating the potential of artificial intelligence (AI) to inform firearm violence prevention strategies. In a previous study, a research team demonstrated that machine learning (ML) models trained using administrative data on handgun transactions can accurately predict the risk of firearm suicide. These insights could contribute to the development of targeted suicide prevention interventions.
Using such ML models yielded moderately informative predictions, and the analysis identified fifteen variables deemed crucial for forecasting firearm suicide risk, including purchaser age, race, ethnicity, and handgun category.
