Bolstering gun injury surveillance accuracy using capture–recapture methods
Using a single source of data, such as police records, or combining data from multiple sources results in an undercount of gun-related injuries. To improve gun-related injury surveillance accuracy by using capture–recapture methods, data were culled from law enforcement, emergency departments, emergency medical services, media, and medical examiner records. The data overlap was operationalized using capture–recapture to generate estimates of uncounted gun incidents. Dependencies between data sources were controlled using log-linear modeling for accurate estimates. New Haven, Connecticut. The study population included subjects injuried/killed from a gun projectile. Incidence was measured using capture–recapture. 49 gun injuries occurred within the defined geography. No single source recorded more than 43 gun-related injuries/deaths. Log-linear modeling estimated the actual number of injuries to be 49.1 (95% CI 49–49.9). Capture–recapture may be less useful in large metropolitan areas that cross state geographical boundaries because of how government agency data are aggregated within each state. No single data source achieves complete gun-related case ascertainment. Log-linear and capture–recapture methods significantly improve gun-related injury estimates.
KeywordsGun violence Firearm violence Gun injury surveillance Gun case definition
Compliance with ethical standards
Conflict of interest
Lori Ann Post, Zev Balsen, Richard Spano, and Federico E. Vaca declare that they have no conflict of interest.
Human and animal rights and Informed consent
All procedures performed in studies involving human participants were in accordance with the ethical standards of the institutional and/or national research committee and with the 1964 Helsinki declaration and its later amendments or comparable ethical standards. For this type of study formal consent is not required.
This study relied on secondary data analysis and thus informed consent was unnecessary. All identifiers were deleted from database and data only are presented in the aggregate. There were no human participants or animals. This study was reviewed and approved by Northwestern and Yale Universities, and University of New Haven IRBs as it was exempt. This study was unfunded.
- CDC. (2018). WONDER: Underlying Cause of Death. Underlying Cause of Death, 1999-2017. Retrieved from https://wonder.cdc.gov/controller/datarequest/D76
- CDC. (2019). WISQARS nonfatal Injury Reports. Nonfatal Injury Reports, 2000–2016. Retrieved from https://webappa.cdc.gov/sasweb/ncipc/nfirates.html
- Duquet, N., & Alstein, M. V. (2015). Firearms and violent deaths in Europe. Retrieved from Europe: https://www.vlaamsvredesinstituut.eu/sites/vlaamsvredesinstituut.eu/files/files/reports/firearms_and_violent_deaths_in_europe_web.pdf
- Garcia, M. (2011). Creating a healthy and safe city: The impact of violence in New Haven. Retrieved from New Haven, CT:Google Scholar
- Miller, T. R., Lawrence, B. A., Carlson, N. N., Hendrie, D., Randall, S., Rockett, I. R., et al. (2016). Perils of police action: a cautionary tale from US data sets. Injury Prevention, injuryprev-2016-042023.Google Scholar
- National Research Council. (2005). Firearms and violence: A critical review. Washington, DC: The National Academies Press.Google Scholar
- Rivest, L., & Baillargeon, S. (2014). Rcapture: Loglinear models for capture–recapture experiments. R package version 1.4-2.Google Scholar
- Rothman, E. F., Stoklosa, H., Baldwin, S. B., Chisolm-Straker, M., Kato Price, R., Atkinson, H. G., et al. (2017). Public health research priorities to address US human trafficking: American Public Health Association.Google Scholar
- Straus, M. A. (2017). Physical violence in American families: Incidence rates, causes, and trends. In Abused and battered (pp. 17–34). Routledge.Google Scholar
- US Cenus Bureau. (2019). American Community Survey. Retrieved from https://factfinder.census.gov/faces/nav/jsf/pages/index.xhtml