Background The majority of gun violence in the United States does not result in physical injury and therefore cannot be completely measured using hospital data. To measure the full scope of gun violence, the nation’s crime reporting systems that collect police reports of crimes committed with a firearm are vital. However, crime data reporting conventions may underestimate gun violence in the U.S. This paper compares crime data sources to assess underestimation of gun violence. Findings The Federal Bureau of Investigation’s Summary Reporting System (SRS) and National Incident Based Reporting System (NIBRS) measures of gun violence were compared in 2019 for states comprehensively reporting data to both systems. Gun violence is underestimated in the SRS compared to NIBRS. Within the sample, 18.8% more aggravated assaults with a firearm are recorded and 2.1% more robberies with a firearm are recorded in NIBRS. The proportion of assaults and robberies committed with a firearm measured in both sources did not differ. If the additional gun violence events recorded in the NIBRS sample are consistent with national crime reporting, the number of additional gun violence events per year captured using NIBRS totals approximately 65,071 additional events, or an additional 178 gun violence events per day. Of the additional gun violence events, approximately 31% are due to omitted crime categories, with the remaining variation driven mostly by aggravated assaults with a firearm. Conclusions Police data are important data sources for estimating the full scope of gun violence. Comparisons between police data sources suggest that the proportion of crimes committed with a firearm is unchanged. Due to crime reporting conventions, however, the number of gun violence events may be substantially understated. Despite advantages in measuring gun violence, agency participation in NIBRS is alarmingly low and jeopardizes accurate and reliable national crime data.
IMPORTANCE Nonfatal gunshot injuries are the most common firearm injury, but where they frequently occur remains unclear owing to data limitations. Natural language processing can be applied to medical text narratives of gunshot injury records to classify injury location and inform prevention efforts. OBJECTIVE To examine the performance of natural language processing (NLP) and machine learning models to predict nonfatal gunshot injury locations and generate new national estimates of the locations in which these injuries occur.
Accurate data on the nature of firearm injuries are essential for crafting effective policies for prevention but are currently lacking. It has been established that medical record coders often misclassify assault cases as "unintentional," with the result that publicly available statistics on nonfatal firearms injuries are heavily biased with respect to the distribution of intents. The study by Miller et al 1 investigates causes of misclassification, using patient case-level data from 3 level I US trauma centers. The authors found that 28% of assaults (234 of 837) were misclassified as accidents by medical record coders. Almost half (114) of these errors involved cases in which the medical record included a description of the circumstances that unmistakably indicated an assault; in the other cases, "assault" was the only reasonable supposition (eg, if the patient sustained multiple gunshot injuries).Although it is now normal for medical records coding to include an external cause of injury, it remains true that the primary purpose of coding is for billing and that payments are not affected by the choice of external-cause code. Hence, there is no financial incentive for providers to code the external cause accurately. In the 3 trauma centers in the study by Miller et al, 1 the trauma registrars, using the same medical-record information as the medical-record coders, accurately coded intent of firearms injuries, with no bias against assault. It appears, then, that medical record coders could do much better.Since 2010, more than 800 journal articles have examined firearm injuries using hospital data sources according to a Google Scholar search performed on October 25, 2022, using the following terms: National Emergency Department Sample or emergency department data or State Emergency
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.