Epidemiologic Trends in Fatal and Nonfatal Firearm Injuries in the US, 2009-2017

Key Points

Question  What is the incidence of fatal and nonfatal firearm injury in the US, including self-harm, assault, and unintentional injury?

Findings  In this cross-sectional study of data from nationwide US databases, from 2009 to 2017, there was an annual average of 85 694 emergency department visits for nonfatal firearm injury and 34 538 deaths from firearm injury. Although suicide accounted for 61% of deaths, self-harm accounted for only 3% of nonfatal injuries; assaults accounted for 35% of deaths and 41% of nonfatal injuries, and unintentional injuries accounted for 2% of deaths and 51% of nonfatal injuries.

Meaning  In the US, nonfatal firearm injuries are more than twice as prevalent as deaths from firearm injury; the most common injuries are self-harm among deaths, unintentional injuries among nonfatal injuries, and assaults overall.

Importance  Firearm injury research in the US has focused on fatal injuries. The incidence and epidemiologic factors associated with nonfatal firearm injuries are less understood.

Objective  To evaluate estimates of incidence and trends over time of fatal and nonfatal firearm injuries.

Design, Setting, and Participants  A cross-sectional, ecologic study was conducted using data throughout the US from 2009 to 2017. Data on fatal injuries from the Centers for Disease Control and Prevention were combined with national data on emergency department visits for nonfatal firearm injury from the Nationwide Emergency Department (ED) sample. Data analysis was conducted from August 2019 to September 2020.

Exposures  Firearm injuries identified with International Classification of Diseases external cause of injury codes and categorized by intent of injury, age group, and urban-rural location.

Main Outcomes and Measures  Incidence, case fatality rate, and trends over time of firearm injury according to intent, age group, and urban-rural location.

Results  From 2009 to 2017, there was a mean of 85 694 ED visits for nonfatal firearm injury and 34 538 deaths each year. An annual mean of 26 445 deaths (76.6%) occurred outside of the hospital. Assault was the most common overall mechanism (38.9%), followed by unintentional injuries (36.9%) and intentional self-harm (19.6%). Self-harm, which accounted for 21 128 deaths (61.2%), had the highest case fatality rate (89.4%; 95% CI, 88.5%-90.4%), followed by assault (25.9%; 95% CI, 23.7%-28.6%) and legal intervention (23.4%; 95% CI, 21.6%-25.5%). Unintentional injuries were the most common nonfatal injuries (43 729 [51.0%]) and had the lowest case fatality rate (1.2%; 95% CI, 1.1%-1.3%). Self-harm deaths, 87.8% of which occurred outside the hospital, increased in all age groups in both rural and urban areas during the study period and were most common among people aged 55 years and older. The rate of fatal assault injuries was higher in urban than in rural areas (16.6 vs 9.0 per 100 000 per year) and highest among people aged 15 to 34 years (38.6 per 100 000 per year). Rates of unintentional injury were higher in rural than in urban areas (18.5 per 100 000 vs 12.4 per 100 000).

Conclusions and Relevance  In this cross-sectional study, suicide appears to be the most common cause of firearm injury death in the US, and most people who die from suicide never reach the hospital. These findings suggest that assaults and unintentional injuries account for most nonfatal and overall firearm injuries and for most of the injuries that are treated in hospitals.

Firearm injury is a major public health challenge in the US, causing 39 773 deaths in 2017.¹ Data to plan, implement, and evaluate public health interventions have gaps.² Most studies of firearm injury focus on deaths, not the overall incidence of firearm injury, but most injured individuals survive. Factors ranging from firearm caliber³ and type⁴ to number and location of wounds⁵ and access to care⁶ can determine the outcome of injury.

There is no comprehensive, national data source for firearm injuries. Fatal injury data are compiled from death certificate data made available by the Centers for Disease Control and Prevention (CDC)⁷ and the National Violent Death Reporting System.⁸ Although firearm mortality decreased in the 1990s and stabilized in the early 2000s, firearm suicides began to increase in 2006, and firearm homicides began to increase in 2014.⁹ The CDC collects nonfatal injury data,¹⁰ but estimates are based on a sample of less than 100 of the 5000 US emergency departments (EDs); analyses have found these data to be unreliable.^11,12 Studies of nonfatal injuries have often used trauma registries, but registries include only patients treated at designated trauma centers and miss up to 1 in 3 firearm injuries.¹³ The National Crime Victimization Survey provides information about assaults but does not include unintentional or self-harm injuries.¹⁴

Health care administrative data can bridge some of these gaps. Since 2006, the Agency for Healthcare Research and Quality Healthcare Cost and Utilization Program has released the Nationwide ED Sample (NEDS), a stratified sample of 900 to 1000 EDs across the US weighted to provide national estimates.¹⁵ Gani et al¹⁶ used the NEDS to estimate approximately 80 000 ED visits and 25 000 hospitalizations for firearm injury annually, from 2006 to 2014. These estimates, however, did not include most deaths, which occur out-of-hospital.⁷ To provide a fuller picture of the population-level incidence of fatal and nonfatal firearm injuries, we combined CDC and NEDS data for firearm injury reported from 2009 to 2017. Data analysis was conducted from August 2019 to September 2020.

Fatal injury data were obtained from the CDC Wide-Ranging Online Data for Epidemiologic Research Underlying Cause of Death database.⁷ International Statistical Classification of Diseases and Related Health Problems, Tenth Revision (ICD-10) external-cause-of-injury codes were used to classify firearm deaths as unintentional (W32, W33, W34), self-harm (X72, X73, X74), assault (X93, X94, X95), legal intervention (Y35.0), or undetermined intent (Y22, Y23, Y24). Sequelae of injury were excluded. The classification legal intervention refers to injuries inflicted by law enforcement officers pursuing their duties.¹⁷

We derived data on nonfatal injury from the NEDS. The NEDS consists of complete discharge records for visits originating in an ED and includes approximately 20% of US hospitals drawn from participating states.¹⁸ Firearm injuries were identified using International Classification of Diseases, Ninth Revision (ICD-9) external-cause-of-injury codes. Data from 2009 through September 2015 used ICD-9 codes (unintentional injury: E922.0-E922.9, self-harm: E955.0-E955.4, assault: E965.0-E965.4, legal intervention: E970, and undetermined: E985.0-E985.4).¹⁹ Data from October 2015 through 2017 used ICD-10 codes. The CDC does not classify the death of any individual younger than 10 years as a suicide.²⁰ For consistency, injuries coded as self-harm in the NEDS for patients younger than 10 years were recoded as unintentional. This study was considered exempt from review by the University of Pennsylvania Institutional Review Board because of use of deidentified data. This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline for cross-sectional studies.

We tabulated injuries and deaths by intent, sex, age group, and urban-rural location according to the National Center for Health Statistics' urban-rural classification scheme for counties. The 2006 classification was used for 2009-2012 and the 2013 revision for 2013-2017.²¹ Counties were classified as urban if they were located in metropolitan areas with a population greater than or equal to 250 000. Patients who died in the hospital were excluded from the NEDS sample to avoid double-counting. Deaths reported by the CDC were characterized as in-hospital or out-of-hospital. NEDS sample weights were used to create national estimates of injury incidence according to Healthcare Research and Quality Healthcare Cost and Utilization Program–stratified sampling methods.²² Case fatality rates were calculated as the percentage of injuries that were fatal for each subcategory. Trends over time were assessed using the nonparametric test for trend, an extension of the Wilcoxon rank-sum test for ordered groups.²³ We tabulated characteristics of patients and injuries presenting to EDs. Categories of surgical procedures were derived from ICD-9 and ICD-10 procedure codes using the Clinical Classification Software categories provided by the Healthcare Research and Quality Healthcare Cost and Utilization Program.²⁴ We calculated abbreviated injury scores, which are assigned by body region (range, 0 [no injury] to 6 [most severe injury])²⁵ and injury severity scores (range, 0-75), which are calculated by summing the squares of the highest abbreviated injury scores code in the 3 most injured injury severity scores body regions,²⁶ from diagnosis codes using the ICD Programs for Injury Classification.^27,28 Patients transferred from the ED are at risk for double counting if they were transferred to another ED that was sampled for the NEDS or if they died after transfer. We performed a sensitivity analysis excluding these patients. The threshold of significance was set at α = .05 using 2-sided, unpaired testing. Data analysis was conducted with Stata, version 14 (StataCorp LLC).

Quiz Ref IDBetween 2009 and 2017, there was a mean of 120 232 firearm injuries in the US each year, including 34 538 deaths (range, 31 347-39 773) and 85 694 ED visits for nonfatal injury (range, 74 728-100 720).Quiz Ref ID Assault injuries were most common, with 14.80 cases per 100 000 people per year (38.9% of total injuries), followed by unintentional injuries (13.99 per 100 000 [36.9%]) and self-harm injuries (7.47 per 100 000 [19.6%]) (Table 1). Each year, an average of 26 772 injured people (22.3%) died before reaching the hospital. Of all deaths, 77.5% occurred outside of the hospital. Of people who reached the ED, 6632 died each year (7.3%). Annual counts of nonfatal injury stratified by age, intent, and urban-rural location are reported in eTable 1 in the Supplement.

Quiz Ref IDSelf-harm injuries caused 21 128 deaths (61.2%) and 2501 nonfatal injuries (2.9%) per year. The case fatality rate was 89.4% (95% CI, 88.5%-90.4%), which was the highest of any intent, with no significant change over time (eFigure in the Supplement). The case fatality rate for assault was 25.9% (95% CI, 23.7%-28.6%), with 12 156 deaths and 34 687 nonfatal injuries per year and no significant change over time. Legal intervention accounted for a mean of 453 deaths and 1481 nonfatal injuries per year, for a case fatality rate of 23.4% (95% CI, 21.6%-25.5%). The case fatality rate increased from 15.8% in 2009 to 33.3% in 2017 (P = .006). There were 526 deaths due to unintentional injury and 43 729 nonfatal injuries each year for a case fatality rate of 1.2% (95% CI, 1.1%-1.3%). Between 2009 and 2017, the case fatality rate decreased from 1.3% to 0.9% (P = .02).

Case fatality rates varied between urban and rural areas and by age (Figure 1). In general, firearm injuries at older ages more often led to death. For self-harm, the lowest case fatality rate was 78.2% among rural children younger than 15 years, and the highest was 94.4% in urban adults aged 55 years or older. For assault, the lowest case fatality rate was 22.5% among individuals aged 15 to 34 years in urban areas and the highest was 55.8% among those aged 55 years or older in rural areas.

Rates of self-harm injury and death were higher in rural areas (11.2 per 100 000 people per year) than in urban areas (6.3 per 100 000) and were highest in people aged 55 years or older (11.3 per 100 000). Rates of death from self-harm increased from 6.1 to 7.3 per 100 000 per year from 2009 to 2017 (P = .004 for trend) with no significant change in the rate of nonfatal injuries (Figure 2; eTable 2 in the Supplement). Deaths from self-harm increased in every age group in both urban and rural areas (eTable 3 in the Supplement). Characteristics of individuals with fatal firearm injuries are reported in Table 2.

The rate of fatal assault injuries was higher in urban than in rural areas (16.6 vs 9.0 per 100 000 per year) and highest among people aged 15 to 34 years (38.6 per 100 000 per year) (Figure 1; eTable 2 in the Supplement). Although there was no significant change in overall assault rates, nonfatal assault rates increased by 9.3% in rural areas from 6.2 to 6.8 per 100 000 (P = .05) (eTable 3 in the Supplement). This increase was significant in individuals aged less than 15, 35 to 54, and greater than or equal to 55 years. Fatal assaults also increased significantly among individuals aged 35 to 54 years in both rural and urban areas.

Rates of unintentional injury were higher in rural than in urban areas (18.5 per 100 000 vs 12.4 per 100 000), and the highest rates were among people aged 15 to 34 years (Figure 1). Quiz Ref IDUnintentional injuries were the most common type among children younger than 15 years, with an incidence of 11.3 per 100 000 per year. The rate of fatal unintentional injuries in that age cohort decreased slightly, with no significant change in overall nonfatal injury rates. In both urban and rural areas, there were significant decreases in nonfatal, unintentional injury in children younger than 15 years but significant increases in nonfatal, unintentional injury among adults aged 35 years or older (eTable 3 in the Supplement).

Rates of legal intervention injury were similar in urban and rural locations, with 0.1 deaths and 0.5 nonfatal injuries per 100 000 per year. Fatal legal intervention injuries increased from 0.1 to 0.2 per 100 000 (P = .01), but nonfatal injuries decreased from 0.6 to 0.4 (P = .008). In rural areas, deaths increased and nonfatal injuries decreased; in urban areas deaths remained stable and nonfatal injuries decreased (eTable 3 in the Supplement). Given the rarity of legal intervention injuries, trends over time were not analyzed by age group.

Patient and injury characteristics of individuals treated in the ED or hospital for nonfatal injuries are reported in Table 3. More than 3 in 4 patients were male across intents. Median age was 25 years (interquartile range, 19-35 years). Injuries were minor (injury severity score <9) in 71.9% of patients. Patients with self-harm injuries had the most severe injuries and the highest incidence of injuries to the head and neck. Extremity injuries were most common in assault and unintentional injury. Three in 10 patients were admitted to the hospital, and the median length of stay was 4 days (interquartile range, 2-9 days). A mean of 17 257 patients (18.7%) each year underwent major operations for firearm injury (eTable 4 in the Supplement). Results for a sensitivity analysis excluding patients who were transferred to another ED or who died after transfer were similar.

By integrating death certificate data with what is, to our knowledge, the largest-available sample of ED cases, this study provides a detailed examination of firearm injury in the US. Between 2009 and 2017, there were approximately 120 232 people injured by firearms per year, or 329 every day. For every person who died from a firearm injury, there were another 2 people who survived, but rates and likelihood of survival varied widely by intent. Although self-harm accounted for most firearm fatalities, when nonfatal injuries were taken into account, assaults were the most common injury type (38.9%), followed by unintentional injuries (36.9%).

Deaths due to self-harm increased in urban and rural areas and for all age groups, consistent with prior evidence.²⁹ Suicide, a form of “death of despair,” has increased in incidence among people with limited resources in the US as they age and as access to firearms has grown.³⁰ Individuals who attempt suicide with a firearm are more likely to die than those who use any other means.³¹ Therefore, access to firearms increases the risk of completed suicide, reducing the opportunity for mental health treatment.³²

Nearly 3 in 4 people who experienced firearm assault survived. Consistent with findings for homicide,³³ assault rates were highest in urban areas and among people aged 15 to 34 years. However, the proportion of assault injuries resulting in death was higher in rural areas and among older age groups. The reasons for this disparity should be further explored but may relate to factors such as firearm caliber, access to trauma care, and context of assault.^3-5,34

Unintentional injuries accounted for half of all nonfatal injuries. Unintentional injuries were the most common type among children aged 0 to 14 years in both rural and urban areas, and 1 in 4 deaths from unintentional injury were among children.³⁵ Injury prevention strategies aimed at unintentional injury include safe storage programs, such as gun lock distribution,³⁶ and child access prevention laws that attribute liability to the owners of firearms accessible to or used by children.^37,38 Data on fatal injuries provide an incomplete picture of the extent of unintentional firearm injuries among children, limiting the ability to determine the results of these interventions.

Firearm injuries and death from legal intervention are a major public concern.³⁹ Data on these injuries are lacking.^40,41 Vital statistics, including the data sources used in this study, underrepresent legal intervention deaths because they are coded as homicides or as justifiable homicides by law enforcement officers.⁴⁰ We identified similar rates of death from legal intervention in urban and rural areas, consistent with recent work using the Fatal Force Database collected by the Washington Post.⁴² In urban areas, legal intervention injuries decreased between 2009 and 2017, while rates of fatal injuries remained stable. The reasons for this difference are uncertain and require further study.

Our findings also highlight the need for dedicated data collection about firearm injuries at the national level to provide detail on victims, context, shooters, and weapons, as well as the need for increased research funding.^43-46 The expansion of the National Violent Death Reporting System to all 50 states may add insights, but only for fatalities. Innovative use of law enforcement data can contribute to the context.⁴⁷ For example, in a 2017 study, Hipple and Magee³⁴ found that motive was a key factor in the lethality of injuries, with drug-related shootings having the highest risk of death, followed by retaliation, domestic incidents, and arguments. Homicides in rural areas more often involved intimate partners or family members than those in urban areas, and shootings that occur indoors in a confined area may be more likely to result in death.⁴⁸

This study has limitations. The NEDS does not include data on race or ethnicity; thus, we could not study how racial and ethnic disparities in firearm injury are reflected in nonfatal injury. The highest rate of firearm homicide occurs among Black adolescent boys and young men.⁴⁹ The highest rates of firearm suicide are among older White men and Native Americans.⁵⁰ To understand the relationship of nonfatal firearm injuries to economic deprivation, structural racism, and other social forces, detailed understanding of racial and ethnic disparities is essential. We also lacked data on nonfatal firearm injuries treated in medical settings other than an ED or hospital and on injuries that were not treated. Although the incidence of such injuries is uncertain, excluding them increases the apparent case fatality rates. Our findings are also limited by inaccuracies in hospital coding. We assumed that the diagnosis of firearm injury is largely correct and that the frequency of coding errors was stable over time. Quiz Ref IDAssignment of intentionality for nonfatal injuries, however, may rely on patient report or physician documentation.⁵¹ In one study, 2 in 5 deaths coded as unintentional firearm injury should have been classified as suicide or homicide, but the reverse was also true, leading to approximately correct total counts.⁵² In addition, we excluded in-hospital deaths from NEDS data to avoid double counting patients. Although likely rare, it is still possible that patients who died after discharge were counted twice.

Suicide appears to be the most common cause of firearm injury death in the US, and most people who die from suicide never reach the hospital. Assault and unintentional injuries account for most nonfatal and overall firearm injuries. Comprehensive data including nonfatal injuries are needed to develop effective public health strategies for prevention and treatment.

Accepted for Publication: September 24, 2020.

Published Online: December 7, 2020. doi:10.1001/jamainternmed.2020.6696

Corresponding Author: Elinore J. Kaufman, MD, MSHP, Division of Traumatology, Surgical Critical Care, and Emergency Surgery, University of Pennsylvania Perelman School of Medicine, 51 N 39th St, Medical Office Building, Ste 120, Philadelphia, PA 19104 (elinore.kaufman@pennmedicine.upenn.edu).

Author Contributions: Dr Kaufman had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: Kaufman, Wiebe, Morrison, Seamon, Delgado.

Acquisition, analysis, or interpretation of data: Kaufman, Xiong, Delgado.

Drafting of the manuscript: Kaufman, Wiebe, Xiong, Morrison, Delgado.

Critical revision of the manuscript for important intellectual content: Kaufman, Wiebe, Morrison, Seamon, Delgado.

Statistical analysis: Kaufman, Xiong, Morrison.

Obtained funding: Delgado.

Administrative, technical, or material support: Kaufman, Delgado.

Supervision: Kaufman, Wiebe, Delgado.

Conflict of Interest Disclosures: Dr Delgado reported receiving grants from the National Institutes of Health and from the Abramson Family Foundation during the conduct of the study. No other disclosures were reported.

Funding/Support: Dr Delgado was supported by the National Institute of Child Health and Human Development (grant K23HD090272001) and by a philanthropic grant from the Abramson Family Foundation.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Meeting Presentation: This study was presented at the 15th Annual Academic Surgical Congress; February 5, 2020; Orlando, Florida.