- Email: [email protected]
Trends in Nonfatal and Fatal Firearm-Related Injury Rates in the United States, 1985–1995
INJURY PREVENTION/ORIGINAL CONTRIBUTION
Trends in Nonfatal and Fatal Firearm-Related Injury Rates in the United States, 1985–1995 From the National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA. Received for publication June 5, 1997. Revisions received November 7, 1997, and February 3 and March 11, 1998. Accepted for publication March 18, 1998.
Darci Cherry, MPH Joseph L Annest, PhD James A Mercy, PhD Marcie-jo Kresnow, MS Daniel A Pollock, MD
Copyright © 1998 by the American College of Emergency Physicians.
See related editorials, p 77 and p 79. Study objective: To characterize trends in annual estimates of nonfatal firearm-related injuries treated in US hospital emergency departments and to compare trends in quarterly rates of such injuries with those of firearm-related fatalities in the US population. Methods: Data on nonfatal firearm-related injuries were obtained from the National Electronic Injury Surveillance System (NEISS) by review of medical records for June 1, 1992, through May 31, 1995. Data on firearm-related fatalities were obtained from the National Vital Statistics System for January 1, 1985, through December 31, 1995. NEISS comprises 91 hospitals that represent a stratified probability sample of all hospitals in the United States and its territories that have at least six beds and provide 24-hour emergency service. The main outcome measures were numbers, percentages, and quarterly population rates for nonfatal and fatal firearm-related injuries. Results: An estimated 288,538 nonfatal firearm-related injuries (95% confidence interval [CI], 169,776 to 407,300) were treated in EDs during the 3-year study period. The annual number of nonfatal firearm-related injuries increased from 99,025 for June 1992 through May 1993 (95% CI, 58,266 to 139,784) to 101,669 for June 1993 through May 1994 (95% CI, 59,822 to 143,516), then decreased to 87,844 for June 1994 through May 1995 (95% CI, 51,687 to 124,001). Before the third quarter of 1993, quarterly nonfatal and fatal firearm-related injury rates in the total US population and quarterly nonfatal firearm assaultive injury and firearm homicide rates for males aged 15 to 24 years were observed to be on the rise. Since then, these rates have significantly declined. Conclusion: Analysis of national trends indicates that nonfatal and fatal firearm-related injuries are declining in the United States, although the rate of firearm-related deaths remains high, especially among males aged 15 to 24 years, in relation to other leading causes of injury death. An assessment of factors responsi-
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 1
FIREARM-RELATED INJURY RATES Cherry et al
ble for the decline in firearm-related injuries is needed to design further prevention efforts. [Cherry D, Annest JL, Mercy JA, Kresnow M-J, Pollock DA: Trends in nonfatal and fatal firearm-related injury rates in the United States, 1985–1995. Ann Emerg Med July 1998;32:51-59.]
terize nonfatal firearm-related injuries over time by selected characteristics and to compare trends in quarterly nonfatal firearm-related injury rates with trends in firearm-related death rates.
METHODS INTRODUCTION
In 1993, firearm-related injuries ranked second to motor vehicle-related injuries as the leading cause of injury death in the United States, taking 39,595 lives.1,2 From 1968 to 1993, the number of motor vehicle-related deaths steadily declined, whereas the number of firearm-related deaths increased.1,3 If those trends continue, firearm-related injuries are predicted to replace motor-vehicle crashes as the leading cause of injury death in the United States by the year 2001 (based on an updated regression analysis3 that includes mortality data for 1968 through 1995). By 1993, this crossover had already occurred in 10 states and the District of Columbia, according to mortality data from the National Vital Statistics System.1 Although historical data on national firearm-related fatalities are well documented3-8 and trend data on nonfatal firearm-related injuries sustained by crime victims and by police officers are available,9 annual trends in national estimates of nonfatal firearm-related injuries treated in US hospital emergency departments have not been described. Nationally representative data for nonfatal firearm-related injuries treated in hospital EDs have only recently become available. Studies9,10 suggest that nonfatal firearm-related injuries treated in hospital EDs outnumber fatal firearmrelated injuries by about 3 to 1. In firearm-related fatalities, suicide is the most common cause of death,1 but most nonfatal firearm-related injuries result from interpersonal violence.10 Furthermore, 4% of firearm-related deaths are unintentional,1 whereas 20% of nonfatal firearm-related injuries treated in hospital EDs result from unintentional shootings.10,11 National data on nonfatal firearm-related injuries are now available to begin assessing whether these epidemiologic patterns are changing over time. The National Electronic Injury Surveillance System (NEISS) of the US Consumer Product Safety Commission (CPSC) provides the first statistically valid and nationally representative data for describing temporal trends in nonfatal firearm-related injuries treated in US hospital EDs.12 It also provides a potential model for state and local efforts to improve completeness and quality of cause-specific ED injury data. In this report, we used NEISS data to charac-
5 2
Data were collected for all gun-related injuries treated in NEISS hospital EDs during the 3-year study period (June 1, 1992, through May 31, 1995). Study data were obtained through an ongoing agreement between the Centers for Disease Control and Prevention (CDC) and CPSC. NEISS comprises 91 hospitals selected as a stratified probability sample of all hospitals in the United States and its territories that have at least six beds and provide 24hour emergency care. Hospital size, specified by the total number of ED visits each year, defined the four strata from which the sample was selected (small, medium, large, and very large). Urban, suburban, and rural hospitals from all regions of the United States were included in the sample.13 Detailed descriptions of sampling methods and sensitivity of the system can be found elsewhere.10,12 CPSC personnel reviewed ED records and logs daily to identify all persons with gun-related injuries who were initially treated at NEISS hospitals. Gun-related injuries were defined as those cases in which a gun was involved in the incident (eg, penetrating gunshot wounds, BB or pellet gun injuries, powder burns, pistol whipping), regardless of intent of the injury.14 However, only those patients who had a penetrating injury or gunshot wound from a weapon that uses a powder charge to fire a projectile and who were discharged from the ED alive were included. No adjustments were made to account for patients who died after leaving the ED, because previous studies have suggested that most firearm-related deaths occur before hospitalization.10,15,16 Cases from each hospital were assigned a sample weight based on the inverse of the probability of selection. National estimates of nonfatal firearm-related injuries were produced by summing the sample weights.13 The 95% confidence intervals (CIs) of these estimates were calculated based on generalized standard errors, which were computed with the use of a formula that accounts for the stratified sample design and sample weights.10,13 From the medical records for each case, data were abstracted to describe the demographic characteristics of the patient, mode of transport to the hospital, injury diagnosis, body part affected, disposition at discharge from the ED, type of injury, type of firearm used, victim-offender relationship, and other variables of interest.14 Brief narrative descriptions of the nature and
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
Concerning the classification of nonfatal firearm-related injuries by type (or intent) of injury, the number of cases reported as “unknown” decreased by 45% over the 3-year study period, from 15.8% in study year 1 to 8.6% in study year 3. This substantial improvement in classification was the result of workshops and visits to NEISS hospitals by CDC investigators to inform on-site NEISS coders and health care providers about the need for better documentation in the medical record—and in particular, better description of the nature and circumstances of the injury incident. To investigate trends in nonfatal firearm-related injuries and injury rates by type of injury, cases with “unknown” type of injury were allocated to one of three categories: assault/ legal intervention, unintentional injury, or suicide attempt. More specifically, “unknown” cases for each quarter of each year were allocated to these three categories based on the weighted distribution of cases with known injury type
circumstances of the injury incident also were recorded. We examined annual trends for nonfatal firearm-related injuries by sex-and-age category, race/ethnicity, and type of injury category (assault/legal intervention, unintentional, and suicide attempt). National vital statistics data for 1985 through 1995 from the National Center for Health Statistics were used to compute quarterly rates of fatal firearm-related injuries for trend analysis and comparison with national estimates of quarterly rates of nonfatal firearm-related injuries obtained from NEISS data. Mortality data were collected from death certificates filed in all 50 states and the District of Columbia.17 The 95% CIs for fatal firearm-related injury rates were based on calculations of the standard error that account for random variation in the number of deaths each year.18 Population data for the calculation of rates were obtained from the US Bureau of the Census.
Table.
Annual estimates of numbers and percentages of persons with nonfatal firearm-related injuries treated in US hospital EDs, overall and by selected characteristics, United States, June 1992 through May 1995* June 1992–May 1993 Characteristic Overall Sex/Age‡ Male total 0–14 15–24 25–34 35–44 45+ Not stated Female total 0–14 15–24 25–34 35–44 45+ Not stated Race/Ethnicity Black White, non-Hispanic Hispanic Other Not stated Type of injuryII Assault/Legal intervention Unintentional Suicide attempt
June 1993–May 1994
No.
%†
99,025
100.0
86,000 2,926 38,371 24,017 13,226 7,213 247§ 13,008 842§ 5,011 3,403 2,285 1,467§ 0
100.0 3.4 44.6 27.9 15.4 8.4 .3 100.0 6.5 38.5 26.2 17.6 11.3 —
47,663 30,444 11,105 3,294 6,519 70,139 22,582 6,304
No.
June 1994–May 1995 %
No.
%
100.0
87,844
100.0
89,542 2,599 42,539 23,019 12,707 7,962 716§ 12,127 687§ 4,416 3,409 2,109 1,360§ 146§
100.0 2.9 47.5 25.7 14.2 8.9 .8 100.0 5.7 36.4 28.1 17.4 11.2 1.2
77,526 2,506 36,403 21,207 10,523 6,802 85§ 10,318 532§ 3,480 3,094 1,898§ 1,314§ 0
100.0 3.2 47.0 27.4 13.6 8.8 .1 100.0 5.2 33.7 30.0 18.4 12.7 —
48.1 30.7 11.2 3.3 6.6
49,485 28,725 14,485 4,206 4,768
48.7 28.3 14.3 4.1 4.7
39,624 28,247 12,986 3,386 3,601
45.1 32.2 14.8 3.9 4.1
70.8 22.8 6.4
75,923 18,331 7,415
74.7 18.0 7.3
65,583 16,091 6,170
74.7 18.3 7.0
101,669
*From
the Centers for Disease Control and Prevention Firearm Injury Surveillance Study. percentages may not total 100% because of rounding. study year 1, sex of the injured person was not stated for 1 case (sample weight=17). §Estimate based on fewer than 50 cases. IIUnknowns are allocated in these distributions; see text for details. †Column ‡In
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 3
FIREARM-RELATED INJURY RATES Cherry et al
for the respective quarters of study year 3. These cases were allocated for the total population and for the male population aged 15 to 24 years. Nonfatal firearm assault and legal intervention cases were combined for comparison with firearm homicide and legal intervention deaths, because the latter are traditionally reported together.17 Legal intervention cases are defined as firearm-related injuries purposefully inflicted by law enforcement officers in the line of duty. Approximately 1% to 2% of all nonfatal and fatal firearm-related injuries are classified as legal intervention.1,10 To examine differences in the classification of nonfatal firearm-related injuries by selected characteristics (Table), the null hypothesis of no difference across years in the percentage distribution for each characteristic was tested by χ2 analysis.19 For trend analysis, quarterly nonfatal and fatal firearmrelated injury rates were calculated using slightly different quarter definitions. Since the NEISS study began on June 1, it was necessary to compute quarterly nonfatal firearmrelated injury rates for a staggered set of months (DecemberFebruary, March-May, June-August, September-November), instead of the conventional set of months used to define quarters for the NCHS mortality data (January-March, AprilJune, July-September, October-December). Trend analysis of quarterly nonfatal and fatal firearmrelated injury rates was accomplished with SAS software20 running piecewise regression models21 to account for ob-
R E S U LT S
A total of 9,401 cases of nonfatal firearm-related injury were recorded in the NEISS system from June 1, 1992, through
24
33
18
22
12
∆
o
44
11
∆
o
Nonfatal Predicted Nonfatal Observed Fatal Predicted Fatal Observed
6
Fatal Firearm Injury Rate per 100,000
Nonfatal and fatal firearmrelated injury rates by quarter, United States, 1985–1995.
Nonfatal Firearm Injury Rate per 100,000
Figure 1.
served changes in the linear relation of the rates over time. A preliminary review of scatter plots of observed nonfatal firearm-related injury rates by quarter suggested two possible distinct slopes (cutpoint at third-quarter 1993 [3Q 1993]), whereas death rate plots suggested three possible distinct slopes (cutpoints at 1Q 1988 and 3Q 1993). Modeling of rates in segments over time yielded an improved model fit in comparison with a simple linear regression model. Improved fit was determined by comparing the mean square errors and the adjusted R2 values of the hierarchical models and by testing the significance of adding cutpoints to the models. Statistical testing for discontinuous piecewise regression models was also conducted to determine whether the rate changed significantly at the start of each new time period. No significant changes were observed. Therefore, regression lines for nonfatal firearmrelated injury rates are presented for two time periods (3Q 1992 to 3Q 1993 and 3Q 1993 to 2Q 1995), and regression lines for fatal firearm-related injury rates are presented for three time periods (1Q 1985 to 1Q 1988, 1Q 1988 to 3Q 1993, and 3Q 1993 to 4Q 1995). For each model, statistical tests were conducted to determine whether the slope of the regression line was significantly different from zero.
0
0 85–1 86–1 87–1 88–1 89–1 90–1 91–1 92–1 93–1 94–1 95–1 Year–Quarter
5 4
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
May 31, 1995. Summing of the sample weights of these cases produced a national estimate of 288,538 (95% CI, 169,776 to 407,300) nonfatal firearm-related injuries treated in US hospital EDs during the 3-year study period. The overall annual rate of nonfatal firearm-related injury increased slightly between the period from June 1992 through May 1993 (38.8/100,000 [95% CI, 22.8 to 54.8]) and the later period from June 1993 through May 1994 (39.4/100,000 [95% CI, 23.2 to 55.6]). The rate then decreased to 33.7/ 100,000 (95% CI, 19.8 to 47.6) for the period from June 1994 through May 1995. The number of nonfatal firearmrelated injuries also dropped between the early period, June 1993 through May 1994 (101,669 cases [95% CI, 59,822 to 143,516]), and the later period, June 1994 through May 1995 (87,844 cases [95% CI, 51,687 to 124,001]), a 13.6% decline. The number of nonfatal firearm-related injuries decreased between the period June 1993 through May 1994 and the period June 1994 through May 1995 across all ages for both sexes, all racial/ethnic groups, and all types of injuries (Table). Also, χ2 analysis indicated that the percentage distributions for race/ethnicity categories (P=.28) and for age groups within sex categories (males, P=.55; females, P=.77) were not significantly different across the 3 years of study. For type of injury, the percentage of firearm injuries associated with assault/legal intervention was significantly lower for year 1 than for years 2 and 3 (70.8% versus 74.8%; P=.03); however, the difference was relatively small.
240
60
180
45
120
30
∆
o
Nonfatal Predicted Nonfatal Observed Fatal Predicted Fatal Observed
o
∆
15
60
0
Fatal Firearm Homicide Rate per 100,000
Nonfatal firearm assault and firearm homicide rates by quarter for males aged 15 to 24 years, United States, 1985–1995.
Nonfatal Firearm Assault Rate per 100,000
Figure 2.
Approximately 73.4% of all nonfatal firearm-related injuries were classified as assault/legal intervention; in comparison, 46.9%1 of fatal firearm-related injuries were firearm homicide/legal intervention. Suicide attempts made up the smallest proportion of nonfatal firearm-related injuries (6.9%), whereas suicide accounted for the largest proportion of firearm-related deaths (47.8%).1 In contrast, unintentional firearm-related injuries comprised 19.8% of all nonfatal cases and 3.8%1 of all deaths. Young males aged 15 to 24 years are a high-risk group for both nonfatal and fatal firearm-related injuries. Males in this age group accounted for 117,313 cases (95% CI, 69,027 to 165,599), or 40.7% of nonfatal firearm-related injuries, with assault/legal intervention making up the largest proportion (79.5%). In the third year of study, the rate of nonfatal firearm assault/legal intervention injury for males in this age group was 156.4/100,000 (95% CI, 92.0 to 220.8), 6.2 times the rate of nonfatal firearm assault/legal intervention injury for the entire US population (25.1/100,000 [95% CI, 14.8 to 35.4]). Similarly, in 1994 the homicide/ legal intervention rate for men aged 15 to 24 years (32.8/ 100,000 [95% CI, 31.9 to 33.7]) was five times the homicide/legal intervention rate for the entire US population (6.6/100,000 [95% CI, 6.5 to 6.7]). The results of regression analysis of trends in the quarterly rates for the total US population indicated a decline in rates of firearm-related injury and death after 3Q 1993 (Figure 1). From 1985 to 3Q 1993, the rate of firearm-
0 85–1 86–1 87–1 88–1 89–1 90–1 91–1 92–1 93–1 94–1 95–1 Year–Quarter
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 5
FIREARM-RELATED INJURY RATES Cherry et al
related death increased continuously, with a significant (P<.01) and sharper increase beginning in 1988. NEISS nonfatal firearm-related injury data became available in 3Q 1992, and from that time until 3Q 1993 the rate of nonfatal firearm-related injury also increased, although the increase was not statistically significant. After peaking in 3Q 1993, both nonfatal (P=.03) and fatal (P<.001) firearmrelated injury rates decreased significantly. Based on predicted values from the regression analysis, both the nonfatal and fatal firearm-related injury rates decreased between 3Q 1993 and 2Q 1995, by 23.0% and 10.3%, respectively. There were also peaks in both fatal and nonfatal firearm assault/legal intervention rates for the high-risk group, men aged 15 to 24 years, during 3Q 1993 (Figure 2). The firearm homicide/legal intervention rates increased significantly between 1985 and 3Q 1993 (P<.01), with most of the rise occurring between 1988 and 3Q 1993 (P<.001). An increase in the rate of nonfatal firearm assault/legal intervention injuries between the 3Q 1992 and 3Q 1993 was not statistically significant. After 3Q 1993, both the nonfatal firearm assault/legal intervention rate (P=.03) and the firearm homicide/legal intervention rate (P<.001) for men aged 15 to 24 years decreased significantly. Based on predicted values from the regression analysis, from 3Q 1993 through 2Q 1995 the nonfatal and fatal firearm assaultive/ legal intervention injury rates for males aged 15 to 24 years decreased by 27.6% and 16.5%, respectively. DISCUSSION
In this report, we present annual estimates of nonfatal firearm-related injuries treated in US EDs for the 3-year study period, June 1992 through May 1995. We also examine and compare trends in the quarterly rates of nonfatal and fatal firearm-related injury in the total US population and in the US male population aged 15 to 24 years. Our data provide further evidence that firearm-related injuries and deaths are beginning to decline in the United States.6,22 The annual number of nonfatal firearm-related injuries dropped in the third year of this study. This decrease was observed across all age groups for both sexes, all racial/ ethnic groups, and all types of injury (ie, assault/legal intervention, unintentional injury, and suicide attempt). In addition, quarterly rates of nonfatal and fatal firearm-related injury decreased significantly after peaking in 3Q 1993. These declines were observed for the firearm-related injury rate in the entire US population and for both nonfatal and fatal firearm assaultive/legal intervention injury rates in the highest risk group, males aged 15 to 24 years. These are noteworthy and encouraging trends, but it is important to
5 6
keep them in perspective. Although nonfatal and fatal firearm-related injuries appear to be declining, the rate of firearm-related death still remains high compared with other leading causes of injury death in the United States. After 3Q 1993, nonfatal firearm-related injury rates declined more sharply than fatal firearm-related injury rates. Over all firearm-related injuries, the estimated case-fatality rate increased from 27.9% in 3Q 1993 to 31.6% in 2Q 1995. Similarly, for firearm assaultive injuries among males aged 15 to 24 years, the estimated case-fatality rate increased from 15.9% in 3Q 1993 to 17.5% in 2Q 1995. These findings indicate that the lethality of firearm injuries may be increasing, which may be partially explained by an increase in the use of higher-powered, semiautomatic handguns in assaultive violence over the past few years.23-25 The overall trend in quarterly fatal and nonfatal firearmrelated injury rates is predominantly driven by changes in fatal and nonfatal firearm assaultive injuries. Firearm-related suicide and unintentional death rates have remained virtually unchanged since the early 1980s.26,27 The observed fluctuations in annual national estimates of nonfatal firearmrelated suicide attempts over the 3-year NEISS Firearm Injury Study period may be caused by sampling variation, because these estimates are based on approximately 150 firearm injury cases. On the other hand, the 29% decline in the annual estimate of the number of nonfatal unintentional firearm-related injuries is intriguing. Preliminary analysis of quarterly rates over the 3-year study period indicated that this decline was gradual and linear with some seasonal variation. This trend will be examined in more detail as additional data on nonfatal firearm-related injury, obtained through ongoing data collection from NEISS, become available. This study indicates that NEISS provides a sensitive system for capturing nonfatal firearm-related injury cases and can be useful for observing trends in nonfatal firearm-related injury rates across time. A previously conducted sensitivity study12 found that approximately 92% of gunshot wound cases treated in US EDs are detected by NEISS. In our study, the percentage distributions of nonfatal firearm injuries for sex by age, race/ethnicity, and type of injury categories were similar across the 3 years of study. Therefore, changes in rates of nonfatal firearm-related injuries over the 3-year study period were not explained by these factors. There are several limitations in the use of NEISS data for analysis of quarterly trends in nonfatal firearm-related injuries. First, additional training after the first year of data collection improved the documentation of injury circumstances by ED health care providers and the narrative descriptions of injury incidents recorded by NEISS personnel. By
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
the third year of data collection, the type of injury was recorded as “not stated” for about 9% of cases, compared with 16% in the first year. This substantial improvement in classification could have affected trend analysis. To overcome this limitation, cases classified as “unknown” for type of injury were allocated to the three other categories (i.e., assault/legal intervention, unintentional injury, and suicide attempt) based on the distribution of these three types of injury in the third year of study. Second, three years of NEISS data are currently available for investigating trends. This provides only 12 data points for analysis of quarterly rates of nonfatal firearm-related injury. However, even with a limited number of data points, we were able to detect statistically significant changes in nonfatal firearm-related injury rates. These early findings are a promising demonstration of the future utility of NEISS as a system for collecting and tracking nonfatal firearmrelated injury data. Although the sample design of NEISS only accommodates the calculation of national estimates,12 temporal trends in nonfatal firearm injuries can be investigated for selected population groups defined by age, sex, race/ethnicity, and type (or intent) of injury. Third, data on type of firearm used, victim-offender relationship, and locale where the injury occurred were obtained on only about one half of all firearm-related injury cases. NEISS hospital ED staff were asked to record pertinent information about the injury incident, but the medical records of firearm-related injury cases often did not contain descriptions of these characteristics. Attending physicians and nurses often are not able to obtain this information from patients, especially those who are severely injured. Major improvement of data on these factors would require additional resources for follow-up interviews with law enforcement officers and for data linkage with police crime and crime laboratory reports. Systems are being developed to link outcomes (eg, fatal or nonfatal firearm-related assaultive injury) with data on guns used in crimes, victim-offender relationship, and other circumstances of the firearm-related injury incident using information from the national, state, and local criminal justice systems.23-25 The methodology developed for these systems could be adapted for use in supplementing the NEISS data on injury circumstances. Finally, there are major factors associated with the NEISS sample design12 that could affect national estimates of nonfatal firearm-related injury over the long term. About every 8 to 10 years, the sampling frame of NEISS is updated to account for changes in US hospitals with EDs that have opened or closed their doors for business. As a result, some NEISS hospitals are dropped and others are added to reflect the new sampling frame. Therefore, changes over time in
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
the number, size (number of annual ED visits), and geographic distribution of US hospitals with EDs and in the level of care provided could influence the sample of hospitals drawn for inclusion in NEISS. In turn, this could influence the number and distribution of reported nonfatal firearm-related injuries. CPSC is currently updating the sampling frame and NEISS hospital sample, which will provide an opportunity to investigate these issues in the future. Our finding that rates of firearm-related assaultive/legal intervention injury and death have declined is consistent with other recent reports of declines in rates of homicide, firearm-related death, and criminal victimization between 1993 and 1995 nationally.28,29 The cause of the drops in rates and numbers of nonfatal and fatal firearm-related injuries is difficult to determine, especially in light of the many factors that may affect such injuries.22,30-43 Numerous explanations for these declines have been proposed, including aggressive policing, decline and stabilization of the crack cocaine trade, aging of the population, improved economic conditions that have resulted in reduced unemployment and income inequality, social and violence prevention programs for youth, and longer prison sentences, which have kept some violent offenders off the streets. Definitive explanations for the declines in rates and the dramatic increases in youth violence which preceded these declines have not been identified. Further research is needed to identify the most plausible explanations. For example, if recruitment of youth into the distribution of crack cocaine increased their access to guns and accelerated more widespread use of guns by a much wider group of children and adolescents,44 then perhaps recent declines could be explained by the waning of crack cocaine markets. This possibility could be explored by examining the temporal association between the introduction and decline of crack cocaine markets and patterns in homicide and firearm injury across urban areas. A more recent idea is that trends in homicide and firearm injury may be a tipping-point phenomenon.45 This theory suggests that there may be threshold points at which time social and psychological conditions lead to sharp increases or declines in violent behavior in communities akin to epidemics of infectious disease. One hypothesis consistent with this theory is that increases or decreases in firearm-related assaults are closely related to the level of fear of victimization and reactive gun-carrying in a community.46 Although NEISS and the national vital statistics system are relatively inexpensive sources of basic information about nonfatal and fatal firearm-related injuries, more detailed population-based data are needed to evaluate the effects of firearm-related interventions and policies, clarify the influ-
5 7
FIREARM-RELATED INJURY RATES Cherry et al
ence of social and behavioral factors, and determine possible explanations for the recent downward trends in rates of firearm-related injury and death. Data collection methods used in NEISS can serve as a model for local ED surveillance. National estimates from NEISS can provide a basis of comparison to gauge the relative magnitude of nonfatal firearm-related injury problems in specific localities. A national firearm-related injury surveillance system is necessary for the collection of standardized data describing circumstances and place of injury, demographic and socioeconomic characteristics of victim and offender, specific information about the firearm, behavioral risk factors for injury, severity of injury, extent of hospital and rehabilitation services, cost of medical care, and specific health outcomes of injury. Without comprehensive firearm-related injury data, the public health impacts of such injuries and the factors contributing to them cannot be completely assessed or understood.47 When linked with police department and criminal justice data, ED-based data can provide crucial information about the nature and scope of intentional and unintentional firearm-related injuries. Emergency physicians are in a unique position to ensure that firearm-related injury data are collected in a complete and accurate manner. Obviously, the highest priority must be placed on ascertaining the anatomic and physiologic severity of the injury. Still, identifying and recording injury intent can yield direct clinical benefits and can serve a variety of additional purposes that are valuable for individual patient care and public health. In the aftermath of immediate care, knowing that an injury was inflicted intentionally can help focus clinical attention on the psychological impact of interpersonal violence and the need for interventions aimed at preventing recurrent injury. 48,49 Prompt reporting of an assaultive injury also helps safeguard society by facilitating investigation by law enforcement agencies.50 Further, when ED data describing injury intentionality are collected by public health surveillance systems, these data can help characterize the magnitude and distribution of injury risks in the community and guide population-based prevention efforts.51 If clinical ED data describing firearm-related injuries are to be aggregated for public health surveillance, a concerted effort is needed to ensure that these data are consistent and complete. Uniformity in data element definitions, coding conventions, and other specifications can help maximize the yield of data generated during episodes of care.52 New initiatives are needed to develop or enhance ED-based surveillance systems at the state and local levels to complement existing national systems. Workshops like those we
5 8
used to improve ED data quality at NEISS hospitals can exert positive effects on recordkeeping practices in other EDs.
REFERENCES 1. Centers for Disease Control and Prevention: Injury Mortality: National Summary of Injury Mortality Data, 1987–1993. Atlanta: US Department of Health and Human Services, Public Health Service, CDC, 1996. 2. Centers for Disease Control and Prevention: 1993: 10 Leading Causes of Death. Atlanta: US Department of Health and Human Services, Public Health Service, CDC, 1996. 3. Centers for Disease Control and Prevention: Deaths resulting from firearm- and motor-vehiclerelated injuries—United States, 1968–1991. MMWR Morb Mortal Wkly Rep 1994;43:37-42. 4. Fingerhut LA, Jones C, Makuc DM: Firearm and motor vehicle injury mortality: Variations by state, race, and ethnicity—United States, 1990–1991, in Advance Data From Vital and Health Statistics, vol 242. Hyattsville, MD: National Center for Health Statistics, 1994:1-12. 5. Centers for Disease Control and Prevention: Firearm-related years of potential life lost before age 65 years—United States, 1980–1991. MMWR Morb Mortal Wkly Rep 1994;43:609-611. 6. Centers for Disease Control and Prevention: Trends in rates of homicide—United States, 1985–1994. MMWR Morb Mortal Wkly Rep 1996;45:460-464. 7. Fingerhut LA: Firearm mortality among children, youth, and young adults 1–34 years of age, trends and current status: United States, 1985–1990, in Advance Data from Vital and Health Statistics, vol 231. Hyattsville, MD: National Center for Health Statistics, 1993:1-17. 8. Wintemute GJ: Firearms as a cause of death in the United States, 1920–1982. J Trauma 1987;27:532-536. 9. Zawitz MW: Firearm Injury from Crime. NCJ-160093. Washington, DC: US Department of Justice, Bureau of Justice Statistics, 1996. 10. Annest JL, Mercy JA, Gibson DR, et al: National estimates of nonfatal firearm-related injuries: Beyond the tip of the iceberg. JAMA 1995;273:1749-1754. 11. Sinauer N, Annest JL, Mercy JA: Unintentional, nonfatal firearm-related injuries: A preventable public health burden. JAMA 1996;275:1740-1743. 12. Davis Y, Annest JL, Powell KE, et al: An evaluation of the National Electronic Injury Surveillance System for use in monitoring nonfatal firearm injuries and obtaining national estimates. J Safety Research 1996;27:83-91. 13. US Consumer Product Safety Commission: National Electronic Injury Surveillance System (NEISS) Sample Design and Implementation. Washington, DC: US Consumer Product Safety Commission, 1994. 14. US Consumer Product Safety Commission: National Electronic Injury Surveillance System (NEISS) Firearm Injury Special Study Instructions Manual. Washington, DC: US Consumer Product Safety Commission, October 1993. 15. Lee RK, Waxweiler RJ, Dobbins JG, et al: Incidence rates of firearm injuries in Galveston, Texas, 1979–1981. Am J Epidemiol 1991;134:511-521. 16. Centers for Disease Control and Prevention: Firearm-related deaths and hospitalizations— Wisconsin, 1994. MMWR Morb Mortal Wkly Rep 1996;45:757-760. 17. Gardner P, Hudson BL: Advance report of final mortality statistics, 1993. Monthly Vital Statistics Report, vol 44, no. 7, suppl. Hyattsville, MD: National Center for Health Statistics, 1996. 18. Singh GK, Mathews TJ, Clarke SC, et al: Annual summary of births, marriages, divorces, and deaths: United States, 1994. Monthly Vital Statistics Report, vol 43, no. 13. Hyattsville, MD: National Center for Health Statistics, 1995. 19. Shah BV, Barnwell BG, Bieler GS: SUDAAN, Software for Analysis of Correlated Data, Release 6.40: User’s Manual. Research Triangle, NC: Research Triangle Institute, 1995. 20. SAS Institute Inc: SAS/STAT User’s Guide, Version 6, ed 4, vol 2. Cary, NC: SAS Institute Inc, 1989:1350-1456. 21. Neter J, Wasserman W, Kutner M: Applied Linear Regression Models, ed 2. Homewood, IL: Irwin, 1989:370-374. 22. Taylor BM: National Crime Victimization Survey: Changes in Criminal Victimization, 1994–1995. NCJ-162032:2. Washington, DC: US Department of Justice, Bureau of Justice Statistics, April 1997.
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
23. Zawitz MW: Guns Used in Crime. NCJ-148201. Washington, DC: US Department of Justice, Bureau of Justice Statistics, July 1995. 24. Barber CW, Ozonoff VV, Schuster M, et al: When bullets don’t kill: A new surveillance system targets firearm injuries. Public Health Rep 1996;111:483-493. 25. Hargarten SW, Karlson TA, O’Brien M, et al: Characteristics of firearms involved in fatalities. JAMA 1996;275:42-45. 26. Kachur PS, Potter LB, James SP, et al: Suicide in the United States, 1980–1992. Violence Surveillance Summary Series, No. 1. Atlanta: Centers for Disease Control and Prevention, National Center for Injury Prevention and Control, 1995;25-31. 27. National Safety Council: Accident Facts, 1995 Edition. Library of Congress Catalog Card Number: 91-60648. Itasca, IL: National Safety Council, 1995;42-45. 28. Taylor BM: Changes in Criminal Victimization, 1994–1995. NCJ-162032. Washington, DC: US Department of Justice, Bureau of Justice Statistics, April 1997.
52. National Center for Injury Prevention and Control: Data Elements for Emergency Department Systems, Release 1.0. Atlanta: Centers for Disease Control and Prevention, 1997.
We would like to express our thanks and appreciation to Art McDonald, director, and Eileen Kessler, statistician and project officer, Division of Hazard and Injury Data Systems, and other staff of the US Consumer Product Safety Commission for their diligence in providing high-quality surveillance data on nonfatal firearm-related injuries using the NEISS. We also thank Mr Steve James for his assistance in data preparation and Dr George Ryan for statistical consultation at the National Center for Injury Prevention and Control.
Reprint no. 47/1/90751 Address for reprints: Joseph L Annest, PhD
29. Singh GK, Kochanek KD, Macdorman MF: Advance report of final mortality statistics, 1994. DHHS Publication No. (PHS) 96-1120. Monthly Vital Statistics Report, vol 45, no. 3, suppl. Hyattsville, MD: National Center for Health Statistics, 1996:22-57.
Director, Office of Statistics and Programming
30. Mullins RJ, Veum-Stone J, Helfand M, et al: Outcome of hospitalized injured patients after institution of a trauma system in an urban area. JAMA 1994;271:1919-1924.
Centers for Disease Control and Prevention
31. Shackford SR, Hollingworth-Fridlund P, Cooper GF, et al: The effect of regionalization upon the quality of trauma care as assessed by concurrent audit before and after institution of a trauma system: A preliminary report. J Trauma 1986;26:812-820.
Atlanta, GA 30341-3724
32. Hammond WR, Yung BR: Preventing violence in at-risk African-American youth. J Health Care Poor Underserved 1991;2:359-373.
National Center for Injury Prevention and Control 4770 Buford Hwy, NE, MS/K59 770-488-4804 Fax 770-488-1665
33. Tolan P, Guerra N: What Works in Reducing Adolescent Violence: An Empirical Review of the Field. Boulder, CO: University of Colorado, Boulder, Institute for Behavioral Sciences, Center for the Study and Prevention of Violence, 1994. 34. Zigler E, Taussig C, Black K: Early childhood intervention: A promising preventative for juvenile delinquency. Am Psychol 1992;47:997-1006. 35. Loftin C, McDowall D, Wiersema B, et al: Effects of restrictive licensing of handguns on homicide and suicide in the District of Columbia. N Engl J Med 1991;325:1615-1620. 36. Sloan JH, Kellermann AL, Reay DT, et al: Handgun regulations, crime, assaults, and homicide: A tale of two cities. N Engl J Med 1988;319:1256-1262. 37. Wintemute GJ: The relationship between firearm design and firearm violence: Handguns in the 1990s. JAMA 1996;275:1749-1753. 38. Kellermann AL, Rivara FP, Somes G, et al: Suicide in the home in relation to gun ownership. N Engl J Med 1992;327:467-472. 39. Kellermann AL, Rivara FP, Rushforth NB, et al: Gun ownership as a risk factor for homicide in the home. N Engl J Med 1993;329:1084-1091. 40. Brent DA, Perper JA, Allman CJ, et al: The presence and accessibility of firearms in the homes of adolescent suicides: A case control study. JAMA 1991;266:2989-2995. 41. Hutson RH, Anglin D, Kyriacou DN, et al: The epidemic of gang-related homicides in Los Angeles County from 1979 through 1994. JAMA 1995;274:1031-1036. 42. McGonigal MD, Cole J, Schwab CW, et al: Urban firearm deaths: A five-year perspective. J Trauma 1993;35:532-537. 43. Centerwall BS: Race, socioeconomic status, and domestic homicide. JAMA 1995;273:1755-1758. 44. Blumstein A: Youth violence, guns, and the illicit-drug industry. Journal of Criminal Law and Criminology 1995;86:10-36. 45. Gladwell M: The tipping point. The New Yorker. June 3, 1996:32-38. 46. Kennedy DM: Can we keep guns away from kids? The American Prospect Summer 1994:74-80. 47. Teret SP: The firearm injury reporting system revisited. JAMA 1996;275:70. 48. Feliciano DV: Patterns of injury, in Feliciano DV, Moore EE, Mattox KL (eds): Trauma, ed 3. Stamford, CT: Appleton & Lange, 1996:85-103. 49. Bell CC, Jenkins EJ, Kpo W, et al: Response of emergency rooms to victims of violence. Hosp Commun Psychiatry 1994;45:142-146. 50. Wetli CV: Forensic issues, in Ivatury RR, Cayten CG (eds): The Textbook of Penetrating Trauma. Baltimore, MD: Williams & Wilkins, 1996:1084-1097. 51. Garrison HG, Runyan CW, Tintinalli JE, et al: Emergency department surveillance: An examination of issues and a proposal for a national strategy. Ann Emerg Med 1994;24:849-856.
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 9
Trends in Nonfatal and Fatal Firearm-Related Injury Rates in the United States, 1985–1995 From the National Center for Injury Prevention and Control, Centers for Disease Control and Prevention, Atlanta, GA. Received for publication June 5, 1997. Revisions received November 7, 1997, and February 3 and March 11, 1998. Accepted for publication March 18, 1998.
Darci Cherry, MPH Joseph L Annest, PhD James A Mercy, PhD Marcie-jo Kresnow, MS Daniel A Pollock, MD
Copyright © 1998 by the American College of Emergency Physicians.
See related editorials, p 77 and p 79. Study objective: To characterize trends in annual estimates of nonfatal firearm-related injuries treated in US hospital emergency departments and to compare trends in quarterly rates of such injuries with those of firearm-related fatalities in the US population. Methods: Data on nonfatal firearm-related injuries were obtained from the National Electronic Injury Surveillance System (NEISS) by review of medical records for June 1, 1992, through May 31, 1995. Data on firearm-related fatalities were obtained from the National Vital Statistics System for January 1, 1985, through December 31, 1995. NEISS comprises 91 hospitals that represent a stratified probability sample of all hospitals in the United States and its territories that have at least six beds and provide 24-hour emergency service. The main outcome measures were numbers, percentages, and quarterly population rates for nonfatal and fatal firearm-related injuries. Results: An estimated 288,538 nonfatal firearm-related injuries (95% confidence interval [CI], 169,776 to 407,300) were treated in EDs during the 3-year study period. The annual number of nonfatal firearm-related injuries increased from 99,025 for June 1992 through May 1993 (95% CI, 58,266 to 139,784) to 101,669 for June 1993 through May 1994 (95% CI, 59,822 to 143,516), then decreased to 87,844 for June 1994 through May 1995 (95% CI, 51,687 to 124,001). Before the third quarter of 1993, quarterly nonfatal and fatal firearm-related injury rates in the total US population and quarterly nonfatal firearm assaultive injury and firearm homicide rates for males aged 15 to 24 years were observed to be on the rise. Since then, these rates have significantly declined. Conclusion: Analysis of national trends indicates that nonfatal and fatal firearm-related injuries are declining in the United States, although the rate of firearm-related deaths remains high, especially among males aged 15 to 24 years, in relation to other leading causes of injury death. An assessment of factors responsi-
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 1
FIREARM-RELATED INJURY RATES Cherry et al
ble for the decline in firearm-related injuries is needed to design further prevention efforts. [Cherry D, Annest JL, Mercy JA, Kresnow M-J, Pollock DA: Trends in nonfatal and fatal firearm-related injury rates in the United States, 1985–1995. Ann Emerg Med July 1998;32:51-59.]
terize nonfatal firearm-related injuries over time by selected characteristics and to compare trends in quarterly nonfatal firearm-related injury rates with trends in firearm-related death rates.
METHODS INTRODUCTION
In 1993, firearm-related injuries ranked second to motor vehicle-related injuries as the leading cause of injury death in the United States, taking 39,595 lives.1,2 From 1968 to 1993, the number of motor vehicle-related deaths steadily declined, whereas the number of firearm-related deaths increased.1,3 If those trends continue, firearm-related injuries are predicted to replace motor-vehicle crashes as the leading cause of injury death in the United States by the year 2001 (based on an updated regression analysis3 that includes mortality data for 1968 through 1995). By 1993, this crossover had already occurred in 10 states and the District of Columbia, according to mortality data from the National Vital Statistics System.1 Although historical data on national firearm-related fatalities are well documented3-8 and trend data on nonfatal firearm-related injuries sustained by crime victims and by police officers are available,9 annual trends in national estimates of nonfatal firearm-related injuries treated in US hospital emergency departments have not been described. Nationally representative data for nonfatal firearm-related injuries treated in hospital EDs have only recently become available. Studies9,10 suggest that nonfatal firearm-related injuries treated in hospital EDs outnumber fatal firearmrelated injuries by about 3 to 1. In firearm-related fatalities, suicide is the most common cause of death,1 but most nonfatal firearm-related injuries result from interpersonal violence.10 Furthermore, 4% of firearm-related deaths are unintentional,1 whereas 20% of nonfatal firearm-related injuries treated in hospital EDs result from unintentional shootings.10,11 National data on nonfatal firearm-related injuries are now available to begin assessing whether these epidemiologic patterns are changing over time. The National Electronic Injury Surveillance System (NEISS) of the US Consumer Product Safety Commission (CPSC) provides the first statistically valid and nationally representative data for describing temporal trends in nonfatal firearm-related injuries treated in US hospital EDs.12 It also provides a potential model for state and local efforts to improve completeness and quality of cause-specific ED injury data. In this report, we used NEISS data to charac-
5 2
Data were collected for all gun-related injuries treated in NEISS hospital EDs during the 3-year study period (June 1, 1992, through May 31, 1995). Study data were obtained through an ongoing agreement between the Centers for Disease Control and Prevention (CDC) and CPSC. NEISS comprises 91 hospitals selected as a stratified probability sample of all hospitals in the United States and its territories that have at least six beds and provide 24hour emergency care. Hospital size, specified by the total number of ED visits each year, defined the four strata from which the sample was selected (small, medium, large, and very large). Urban, suburban, and rural hospitals from all regions of the United States were included in the sample.13 Detailed descriptions of sampling methods and sensitivity of the system can be found elsewhere.10,12 CPSC personnel reviewed ED records and logs daily to identify all persons with gun-related injuries who were initially treated at NEISS hospitals. Gun-related injuries were defined as those cases in which a gun was involved in the incident (eg, penetrating gunshot wounds, BB or pellet gun injuries, powder burns, pistol whipping), regardless of intent of the injury.14 However, only those patients who had a penetrating injury or gunshot wound from a weapon that uses a powder charge to fire a projectile and who were discharged from the ED alive were included. No adjustments were made to account for patients who died after leaving the ED, because previous studies have suggested that most firearm-related deaths occur before hospitalization.10,15,16 Cases from each hospital were assigned a sample weight based on the inverse of the probability of selection. National estimates of nonfatal firearm-related injuries were produced by summing the sample weights.13 The 95% confidence intervals (CIs) of these estimates were calculated based on generalized standard errors, which were computed with the use of a formula that accounts for the stratified sample design and sample weights.10,13 From the medical records for each case, data were abstracted to describe the demographic characteristics of the patient, mode of transport to the hospital, injury diagnosis, body part affected, disposition at discharge from the ED, type of injury, type of firearm used, victim-offender relationship, and other variables of interest.14 Brief narrative descriptions of the nature and
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
Concerning the classification of nonfatal firearm-related injuries by type (or intent) of injury, the number of cases reported as “unknown” decreased by 45% over the 3-year study period, from 15.8% in study year 1 to 8.6% in study year 3. This substantial improvement in classification was the result of workshops and visits to NEISS hospitals by CDC investigators to inform on-site NEISS coders and health care providers about the need for better documentation in the medical record—and in particular, better description of the nature and circumstances of the injury incident. To investigate trends in nonfatal firearm-related injuries and injury rates by type of injury, cases with “unknown” type of injury were allocated to one of three categories: assault/ legal intervention, unintentional injury, or suicide attempt. More specifically, “unknown” cases for each quarter of each year were allocated to these three categories based on the weighted distribution of cases with known injury type
circumstances of the injury incident also were recorded. We examined annual trends for nonfatal firearm-related injuries by sex-and-age category, race/ethnicity, and type of injury category (assault/legal intervention, unintentional, and suicide attempt). National vital statistics data for 1985 through 1995 from the National Center for Health Statistics were used to compute quarterly rates of fatal firearm-related injuries for trend analysis and comparison with national estimates of quarterly rates of nonfatal firearm-related injuries obtained from NEISS data. Mortality data were collected from death certificates filed in all 50 states and the District of Columbia.17 The 95% CIs for fatal firearm-related injury rates were based on calculations of the standard error that account for random variation in the number of deaths each year.18 Population data for the calculation of rates were obtained from the US Bureau of the Census.
Table.
Annual estimates of numbers and percentages of persons with nonfatal firearm-related injuries treated in US hospital EDs, overall and by selected characteristics, United States, June 1992 through May 1995* June 1992–May 1993 Characteristic Overall Sex/Age‡ Male total 0–14 15–24 25–34 35–44 45+ Not stated Female total 0–14 15–24 25–34 35–44 45+ Not stated Race/Ethnicity Black White, non-Hispanic Hispanic Other Not stated Type of injuryII Assault/Legal intervention Unintentional Suicide attempt
June 1993–May 1994
No.
%†
99,025
100.0
86,000 2,926 38,371 24,017 13,226 7,213 247§ 13,008 842§ 5,011 3,403 2,285 1,467§ 0
100.0 3.4 44.6 27.9 15.4 8.4 .3 100.0 6.5 38.5 26.2 17.6 11.3 —
47,663 30,444 11,105 3,294 6,519 70,139 22,582 6,304
No.
June 1994–May 1995 %
No.
%
100.0
87,844
100.0
89,542 2,599 42,539 23,019 12,707 7,962 716§ 12,127 687§ 4,416 3,409 2,109 1,360§ 146§
100.0 2.9 47.5 25.7 14.2 8.9 .8 100.0 5.7 36.4 28.1 17.4 11.2 1.2
77,526 2,506 36,403 21,207 10,523 6,802 85§ 10,318 532§ 3,480 3,094 1,898§ 1,314§ 0
100.0 3.2 47.0 27.4 13.6 8.8 .1 100.0 5.2 33.7 30.0 18.4 12.7 —
48.1 30.7 11.2 3.3 6.6
49,485 28,725 14,485 4,206 4,768
48.7 28.3 14.3 4.1 4.7
39,624 28,247 12,986 3,386 3,601
45.1 32.2 14.8 3.9 4.1
70.8 22.8 6.4
75,923 18,331 7,415
74.7 18.0 7.3
65,583 16,091 6,170
74.7 18.3 7.0
101,669
*From
the Centers for Disease Control and Prevention Firearm Injury Surveillance Study. percentages may not total 100% because of rounding. study year 1, sex of the injured person was not stated for 1 case (sample weight=17). §Estimate based on fewer than 50 cases. IIUnknowns are allocated in these distributions; see text for details. †Column ‡In
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 3
FIREARM-RELATED INJURY RATES Cherry et al
for the respective quarters of study year 3. These cases were allocated for the total population and for the male population aged 15 to 24 years. Nonfatal firearm assault and legal intervention cases were combined for comparison with firearm homicide and legal intervention deaths, because the latter are traditionally reported together.17 Legal intervention cases are defined as firearm-related injuries purposefully inflicted by law enforcement officers in the line of duty. Approximately 1% to 2% of all nonfatal and fatal firearm-related injuries are classified as legal intervention.1,10 To examine differences in the classification of nonfatal firearm-related injuries by selected characteristics (Table), the null hypothesis of no difference across years in the percentage distribution for each characteristic was tested by χ2 analysis.19 For trend analysis, quarterly nonfatal and fatal firearmrelated injury rates were calculated using slightly different quarter definitions. Since the NEISS study began on June 1, it was necessary to compute quarterly nonfatal firearmrelated injury rates for a staggered set of months (DecemberFebruary, March-May, June-August, September-November), instead of the conventional set of months used to define quarters for the NCHS mortality data (January-March, AprilJune, July-September, October-December). Trend analysis of quarterly nonfatal and fatal firearmrelated injury rates was accomplished with SAS software20 running piecewise regression models21 to account for ob-
R E S U LT S
A total of 9,401 cases of nonfatal firearm-related injury were recorded in the NEISS system from June 1, 1992, through
24
33
18
22
12
∆
o
44
11
∆
o
Nonfatal Predicted Nonfatal Observed Fatal Predicted Fatal Observed
6
Fatal Firearm Injury Rate per 100,000
Nonfatal and fatal firearmrelated injury rates by quarter, United States, 1985–1995.
Nonfatal Firearm Injury Rate per 100,000
Figure 1.
served changes in the linear relation of the rates over time. A preliminary review of scatter plots of observed nonfatal firearm-related injury rates by quarter suggested two possible distinct slopes (cutpoint at third-quarter 1993 [3Q 1993]), whereas death rate plots suggested three possible distinct slopes (cutpoints at 1Q 1988 and 3Q 1993). Modeling of rates in segments over time yielded an improved model fit in comparison with a simple linear regression model. Improved fit was determined by comparing the mean square errors and the adjusted R2 values of the hierarchical models and by testing the significance of adding cutpoints to the models. Statistical testing for discontinuous piecewise regression models was also conducted to determine whether the rate changed significantly at the start of each new time period. No significant changes were observed. Therefore, regression lines for nonfatal firearmrelated injury rates are presented for two time periods (3Q 1992 to 3Q 1993 and 3Q 1993 to 2Q 1995), and regression lines for fatal firearm-related injury rates are presented for three time periods (1Q 1985 to 1Q 1988, 1Q 1988 to 3Q 1993, and 3Q 1993 to 4Q 1995). For each model, statistical tests were conducted to determine whether the slope of the regression line was significantly different from zero.
0
0 85–1 86–1 87–1 88–1 89–1 90–1 91–1 92–1 93–1 94–1 95–1 Year–Quarter
5 4
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
May 31, 1995. Summing of the sample weights of these cases produced a national estimate of 288,538 (95% CI, 169,776 to 407,300) nonfatal firearm-related injuries treated in US hospital EDs during the 3-year study period. The overall annual rate of nonfatal firearm-related injury increased slightly between the period from June 1992 through May 1993 (38.8/100,000 [95% CI, 22.8 to 54.8]) and the later period from June 1993 through May 1994 (39.4/100,000 [95% CI, 23.2 to 55.6]). The rate then decreased to 33.7/ 100,000 (95% CI, 19.8 to 47.6) for the period from June 1994 through May 1995. The number of nonfatal firearmrelated injuries also dropped between the early period, June 1993 through May 1994 (101,669 cases [95% CI, 59,822 to 143,516]), and the later period, June 1994 through May 1995 (87,844 cases [95% CI, 51,687 to 124,001]), a 13.6% decline. The number of nonfatal firearm-related injuries decreased between the period June 1993 through May 1994 and the period June 1994 through May 1995 across all ages for both sexes, all racial/ethnic groups, and all types of injuries (Table). Also, χ2 analysis indicated that the percentage distributions for race/ethnicity categories (P=.28) and for age groups within sex categories (males, P=.55; females, P=.77) were not significantly different across the 3 years of study. For type of injury, the percentage of firearm injuries associated with assault/legal intervention was significantly lower for year 1 than for years 2 and 3 (70.8% versus 74.8%; P=.03); however, the difference was relatively small.
240
60
180
45
120
30
∆
o
Nonfatal Predicted Nonfatal Observed Fatal Predicted Fatal Observed
o
∆
15
60
0
Fatal Firearm Homicide Rate per 100,000
Nonfatal firearm assault and firearm homicide rates by quarter for males aged 15 to 24 years, United States, 1985–1995.
Nonfatal Firearm Assault Rate per 100,000
Figure 2.
Approximately 73.4% of all nonfatal firearm-related injuries were classified as assault/legal intervention; in comparison, 46.9%1 of fatal firearm-related injuries were firearm homicide/legal intervention. Suicide attempts made up the smallest proportion of nonfatal firearm-related injuries (6.9%), whereas suicide accounted for the largest proportion of firearm-related deaths (47.8%).1 In contrast, unintentional firearm-related injuries comprised 19.8% of all nonfatal cases and 3.8%1 of all deaths. Young males aged 15 to 24 years are a high-risk group for both nonfatal and fatal firearm-related injuries. Males in this age group accounted for 117,313 cases (95% CI, 69,027 to 165,599), or 40.7% of nonfatal firearm-related injuries, with assault/legal intervention making up the largest proportion (79.5%). In the third year of study, the rate of nonfatal firearm assault/legal intervention injury for males in this age group was 156.4/100,000 (95% CI, 92.0 to 220.8), 6.2 times the rate of nonfatal firearm assault/legal intervention injury for the entire US population (25.1/100,000 [95% CI, 14.8 to 35.4]). Similarly, in 1994 the homicide/ legal intervention rate for men aged 15 to 24 years (32.8/ 100,000 [95% CI, 31.9 to 33.7]) was five times the homicide/legal intervention rate for the entire US population (6.6/100,000 [95% CI, 6.5 to 6.7]). The results of regression analysis of trends in the quarterly rates for the total US population indicated a decline in rates of firearm-related injury and death after 3Q 1993 (Figure 1). From 1985 to 3Q 1993, the rate of firearm-
0 85–1 86–1 87–1 88–1 89–1 90–1 91–1 92–1 93–1 94–1 95–1 Year–Quarter
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 5
FIREARM-RELATED INJURY RATES Cherry et al
related death increased continuously, with a significant (P<.01) and sharper increase beginning in 1988. NEISS nonfatal firearm-related injury data became available in 3Q 1992, and from that time until 3Q 1993 the rate of nonfatal firearm-related injury also increased, although the increase was not statistically significant. After peaking in 3Q 1993, both nonfatal (P=.03) and fatal (P<.001) firearmrelated injury rates decreased significantly. Based on predicted values from the regression analysis, both the nonfatal and fatal firearm-related injury rates decreased between 3Q 1993 and 2Q 1995, by 23.0% and 10.3%, respectively. There were also peaks in both fatal and nonfatal firearm assault/legal intervention rates for the high-risk group, men aged 15 to 24 years, during 3Q 1993 (Figure 2). The firearm homicide/legal intervention rates increased significantly between 1985 and 3Q 1993 (P<.01), with most of the rise occurring between 1988 and 3Q 1993 (P<.001). An increase in the rate of nonfatal firearm assault/legal intervention injuries between the 3Q 1992 and 3Q 1993 was not statistically significant. After 3Q 1993, both the nonfatal firearm assault/legal intervention rate (P=.03) and the firearm homicide/legal intervention rate (P<.001) for men aged 15 to 24 years decreased significantly. Based on predicted values from the regression analysis, from 3Q 1993 through 2Q 1995 the nonfatal and fatal firearm assaultive/ legal intervention injury rates for males aged 15 to 24 years decreased by 27.6% and 16.5%, respectively. DISCUSSION
In this report, we present annual estimates of nonfatal firearm-related injuries treated in US EDs for the 3-year study period, June 1992 through May 1995. We also examine and compare trends in the quarterly rates of nonfatal and fatal firearm-related injury in the total US population and in the US male population aged 15 to 24 years. Our data provide further evidence that firearm-related injuries and deaths are beginning to decline in the United States.6,22 The annual number of nonfatal firearm-related injuries dropped in the third year of this study. This decrease was observed across all age groups for both sexes, all racial/ ethnic groups, and all types of injury (ie, assault/legal intervention, unintentional injury, and suicide attempt). In addition, quarterly rates of nonfatal and fatal firearm-related injury decreased significantly after peaking in 3Q 1993. These declines were observed for the firearm-related injury rate in the entire US population and for both nonfatal and fatal firearm assaultive/legal intervention injury rates in the highest risk group, males aged 15 to 24 years. These are noteworthy and encouraging trends, but it is important to
5 6
keep them in perspective. Although nonfatal and fatal firearm-related injuries appear to be declining, the rate of firearm-related death still remains high compared with other leading causes of injury death in the United States. After 3Q 1993, nonfatal firearm-related injury rates declined more sharply than fatal firearm-related injury rates. Over all firearm-related injuries, the estimated case-fatality rate increased from 27.9% in 3Q 1993 to 31.6% in 2Q 1995. Similarly, for firearm assaultive injuries among males aged 15 to 24 years, the estimated case-fatality rate increased from 15.9% in 3Q 1993 to 17.5% in 2Q 1995. These findings indicate that the lethality of firearm injuries may be increasing, which may be partially explained by an increase in the use of higher-powered, semiautomatic handguns in assaultive violence over the past few years.23-25 The overall trend in quarterly fatal and nonfatal firearmrelated injury rates is predominantly driven by changes in fatal and nonfatal firearm assaultive injuries. Firearm-related suicide and unintentional death rates have remained virtually unchanged since the early 1980s.26,27 The observed fluctuations in annual national estimates of nonfatal firearmrelated suicide attempts over the 3-year NEISS Firearm Injury Study period may be caused by sampling variation, because these estimates are based on approximately 150 firearm injury cases. On the other hand, the 29% decline in the annual estimate of the number of nonfatal unintentional firearm-related injuries is intriguing. Preliminary analysis of quarterly rates over the 3-year study period indicated that this decline was gradual and linear with some seasonal variation. This trend will be examined in more detail as additional data on nonfatal firearm-related injury, obtained through ongoing data collection from NEISS, become available. This study indicates that NEISS provides a sensitive system for capturing nonfatal firearm-related injury cases and can be useful for observing trends in nonfatal firearm-related injury rates across time. A previously conducted sensitivity study12 found that approximately 92% of gunshot wound cases treated in US EDs are detected by NEISS. In our study, the percentage distributions of nonfatal firearm injuries for sex by age, race/ethnicity, and type of injury categories were similar across the 3 years of study. Therefore, changes in rates of nonfatal firearm-related injuries over the 3-year study period were not explained by these factors. There are several limitations in the use of NEISS data for analysis of quarterly trends in nonfatal firearm-related injuries. First, additional training after the first year of data collection improved the documentation of injury circumstances by ED health care providers and the narrative descriptions of injury incidents recorded by NEISS personnel. By
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
the third year of data collection, the type of injury was recorded as “not stated” for about 9% of cases, compared with 16% in the first year. This substantial improvement in classification could have affected trend analysis. To overcome this limitation, cases classified as “unknown” for type of injury were allocated to the three other categories (i.e., assault/legal intervention, unintentional injury, and suicide attempt) based on the distribution of these three types of injury in the third year of study. Second, three years of NEISS data are currently available for investigating trends. This provides only 12 data points for analysis of quarterly rates of nonfatal firearm-related injury. However, even with a limited number of data points, we were able to detect statistically significant changes in nonfatal firearm-related injury rates. These early findings are a promising demonstration of the future utility of NEISS as a system for collecting and tracking nonfatal firearmrelated injury data. Although the sample design of NEISS only accommodates the calculation of national estimates,12 temporal trends in nonfatal firearm injuries can be investigated for selected population groups defined by age, sex, race/ethnicity, and type (or intent) of injury. Third, data on type of firearm used, victim-offender relationship, and locale where the injury occurred were obtained on only about one half of all firearm-related injury cases. NEISS hospital ED staff were asked to record pertinent information about the injury incident, but the medical records of firearm-related injury cases often did not contain descriptions of these characteristics. Attending physicians and nurses often are not able to obtain this information from patients, especially those who are severely injured. Major improvement of data on these factors would require additional resources for follow-up interviews with law enforcement officers and for data linkage with police crime and crime laboratory reports. Systems are being developed to link outcomes (eg, fatal or nonfatal firearm-related assaultive injury) with data on guns used in crimes, victim-offender relationship, and other circumstances of the firearm-related injury incident using information from the national, state, and local criminal justice systems.23-25 The methodology developed for these systems could be adapted for use in supplementing the NEISS data on injury circumstances. Finally, there are major factors associated with the NEISS sample design12 that could affect national estimates of nonfatal firearm-related injury over the long term. About every 8 to 10 years, the sampling frame of NEISS is updated to account for changes in US hospitals with EDs that have opened or closed their doors for business. As a result, some NEISS hospitals are dropped and others are added to reflect the new sampling frame. Therefore, changes over time in
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
the number, size (number of annual ED visits), and geographic distribution of US hospitals with EDs and in the level of care provided could influence the sample of hospitals drawn for inclusion in NEISS. In turn, this could influence the number and distribution of reported nonfatal firearm-related injuries. CPSC is currently updating the sampling frame and NEISS hospital sample, which will provide an opportunity to investigate these issues in the future. Our finding that rates of firearm-related assaultive/legal intervention injury and death have declined is consistent with other recent reports of declines in rates of homicide, firearm-related death, and criminal victimization between 1993 and 1995 nationally.28,29 The cause of the drops in rates and numbers of nonfatal and fatal firearm-related injuries is difficult to determine, especially in light of the many factors that may affect such injuries.22,30-43 Numerous explanations for these declines have been proposed, including aggressive policing, decline and stabilization of the crack cocaine trade, aging of the population, improved economic conditions that have resulted in reduced unemployment and income inequality, social and violence prevention programs for youth, and longer prison sentences, which have kept some violent offenders off the streets. Definitive explanations for the declines in rates and the dramatic increases in youth violence which preceded these declines have not been identified. Further research is needed to identify the most plausible explanations. For example, if recruitment of youth into the distribution of crack cocaine increased their access to guns and accelerated more widespread use of guns by a much wider group of children and adolescents,44 then perhaps recent declines could be explained by the waning of crack cocaine markets. This possibility could be explored by examining the temporal association between the introduction and decline of crack cocaine markets and patterns in homicide and firearm injury across urban areas. A more recent idea is that trends in homicide and firearm injury may be a tipping-point phenomenon.45 This theory suggests that there may be threshold points at which time social and psychological conditions lead to sharp increases or declines in violent behavior in communities akin to epidemics of infectious disease. One hypothesis consistent with this theory is that increases or decreases in firearm-related assaults are closely related to the level of fear of victimization and reactive gun-carrying in a community.46 Although NEISS and the national vital statistics system are relatively inexpensive sources of basic information about nonfatal and fatal firearm-related injuries, more detailed population-based data are needed to evaluate the effects of firearm-related interventions and policies, clarify the influ-
5 7
FIREARM-RELATED INJURY RATES Cherry et al
ence of social and behavioral factors, and determine possible explanations for the recent downward trends in rates of firearm-related injury and death. Data collection methods used in NEISS can serve as a model for local ED surveillance. National estimates from NEISS can provide a basis of comparison to gauge the relative magnitude of nonfatal firearm-related injury problems in specific localities. A national firearm-related injury surveillance system is necessary for the collection of standardized data describing circumstances and place of injury, demographic and socioeconomic characteristics of victim and offender, specific information about the firearm, behavioral risk factors for injury, severity of injury, extent of hospital and rehabilitation services, cost of medical care, and specific health outcomes of injury. Without comprehensive firearm-related injury data, the public health impacts of such injuries and the factors contributing to them cannot be completely assessed or understood.47 When linked with police department and criminal justice data, ED-based data can provide crucial information about the nature and scope of intentional and unintentional firearm-related injuries. Emergency physicians are in a unique position to ensure that firearm-related injury data are collected in a complete and accurate manner. Obviously, the highest priority must be placed on ascertaining the anatomic and physiologic severity of the injury. Still, identifying and recording injury intent can yield direct clinical benefits and can serve a variety of additional purposes that are valuable for individual patient care and public health. In the aftermath of immediate care, knowing that an injury was inflicted intentionally can help focus clinical attention on the psychological impact of interpersonal violence and the need for interventions aimed at preventing recurrent injury. 48,49 Prompt reporting of an assaultive injury also helps safeguard society by facilitating investigation by law enforcement agencies.50 Further, when ED data describing injury intentionality are collected by public health surveillance systems, these data can help characterize the magnitude and distribution of injury risks in the community and guide population-based prevention efforts.51 If clinical ED data describing firearm-related injuries are to be aggregated for public health surveillance, a concerted effort is needed to ensure that these data are consistent and complete. Uniformity in data element definitions, coding conventions, and other specifications can help maximize the yield of data generated during episodes of care.52 New initiatives are needed to develop or enhance ED-based surveillance systems at the state and local levels to complement existing national systems. Workshops like those we
5 8
used to improve ED data quality at NEISS hospitals can exert positive effects on recordkeeping practices in other EDs.
REFERENCES 1. Centers for Disease Control and Prevention: Injury Mortality: National Summary of Injury Mortality Data, 1987–1993. Atlanta: US Department of Health and Human Services, Public Health Service, CDC, 1996. 2. Centers for Disease Control and Prevention: 1993: 10 Leading Causes of Death. Atlanta: US Department of Health and Human Services, Public Health Service, CDC, 1996. 3. Centers for Disease Control and Prevention: Deaths resulting from firearm- and motor-vehiclerelated injuries—United States, 1968–1991. MMWR Morb Mortal Wkly Rep 1994;43:37-42. 4. Fingerhut LA, Jones C, Makuc DM: Firearm and motor vehicle injury mortality: Variations by state, race, and ethnicity—United States, 1990–1991, in Advance Data From Vital and Health Statistics, vol 242. Hyattsville, MD: National Center for Health Statistics, 1994:1-12. 5. Centers for Disease Control and Prevention: Firearm-related years of potential life lost before age 65 years—United States, 1980–1991. MMWR Morb Mortal Wkly Rep 1994;43:609-611. 6. Centers for Disease Control and Prevention: Trends in rates of homicide—United States, 1985–1994. MMWR Morb Mortal Wkly Rep 1996;45:460-464. 7. Fingerhut LA: Firearm mortality among children, youth, and young adults 1–34 years of age, trends and current status: United States, 1985–1990, in Advance Data from Vital and Health Statistics, vol 231. Hyattsville, MD: National Center for Health Statistics, 1993:1-17. 8. Wintemute GJ: Firearms as a cause of death in the United States, 1920–1982. J Trauma 1987;27:532-536. 9. Zawitz MW: Firearm Injury from Crime. NCJ-160093. Washington, DC: US Department of Justice, Bureau of Justice Statistics, 1996. 10. Annest JL, Mercy JA, Gibson DR, et al: National estimates of nonfatal firearm-related injuries: Beyond the tip of the iceberg. JAMA 1995;273:1749-1754. 11. Sinauer N, Annest JL, Mercy JA: Unintentional, nonfatal firearm-related injuries: A preventable public health burden. JAMA 1996;275:1740-1743. 12. Davis Y, Annest JL, Powell KE, et al: An evaluation of the National Electronic Injury Surveillance System for use in monitoring nonfatal firearm injuries and obtaining national estimates. J Safety Research 1996;27:83-91. 13. US Consumer Product Safety Commission: National Electronic Injury Surveillance System (NEISS) Sample Design and Implementation. Washington, DC: US Consumer Product Safety Commission, 1994. 14. US Consumer Product Safety Commission: National Electronic Injury Surveillance System (NEISS) Firearm Injury Special Study Instructions Manual. Washington, DC: US Consumer Product Safety Commission, October 1993. 15. Lee RK, Waxweiler RJ, Dobbins JG, et al: Incidence rates of firearm injuries in Galveston, Texas, 1979–1981. Am J Epidemiol 1991;134:511-521. 16. Centers for Disease Control and Prevention: Firearm-related deaths and hospitalizations— Wisconsin, 1994. MMWR Morb Mortal Wkly Rep 1996;45:757-760. 17. Gardner P, Hudson BL: Advance report of final mortality statistics, 1993. Monthly Vital Statistics Report, vol 44, no. 7, suppl. Hyattsville, MD: National Center for Health Statistics, 1996. 18. Singh GK, Mathews TJ, Clarke SC, et al: Annual summary of births, marriages, divorces, and deaths: United States, 1994. Monthly Vital Statistics Report, vol 43, no. 13. Hyattsville, MD: National Center for Health Statistics, 1995. 19. Shah BV, Barnwell BG, Bieler GS: SUDAAN, Software for Analysis of Correlated Data, Release 6.40: User’s Manual. Research Triangle, NC: Research Triangle Institute, 1995. 20. SAS Institute Inc: SAS/STAT User’s Guide, Version 6, ed 4, vol 2. Cary, NC: SAS Institute Inc, 1989:1350-1456. 21. Neter J, Wasserman W, Kutner M: Applied Linear Regression Models, ed 2. Homewood, IL: Irwin, 1989:370-374. 22. Taylor BM: National Crime Victimization Survey: Changes in Criminal Victimization, 1994–1995. NCJ-162032:2. Washington, DC: US Department of Justice, Bureau of Justice Statistics, April 1997.
ANNALS OF EMERGENCY MEDICINE
32:1 JULY 1998
FIREARM-RELATED INJURY RATES Cherry et al
23. Zawitz MW: Guns Used in Crime. NCJ-148201. Washington, DC: US Department of Justice, Bureau of Justice Statistics, July 1995. 24. Barber CW, Ozonoff VV, Schuster M, et al: When bullets don’t kill: A new surveillance system targets firearm injuries. Public Health Rep 1996;111:483-493. 25. Hargarten SW, Karlson TA, O’Brien M, et al: Characteristics of firearms involved in fatalities. JAMA 1996;275:42-45. 26. Kachur PS, Potter LB, James SP, et al: Suicide in the United States, 1980–1992. Violence Surveillance Summary Series, No. 1. Atlanta: Centers for Disease Control and Prevention, National Center for Injury Prevention and Control, 1995;25-31. 27. National Safety Council: Accident Facts, 1995 Edition. Library of Congress Catalog Card Number: 91-60648. Itasca, IL: National Safety Council, 1995;42-45. 28. Taylor BM: Changes in Criminal Victimization, 1994–1995. NCJ-162032. Washington, DC: US Department of Justice, Bureau of Justice Statistics, April 1997.
52. National Center for Injury Prevention and Control: Data Elements for Emergency Department Systems, Release 1.0. Atlanta: Centers for Disease Control and Prevention, 1997.
We would like to express our thanks and appreciation to Art McDonald, director, and Eileen Kessler, statistician and project officer, Division of Hazard and Injury Data Systems, and other staff of the US Consumer Product Safety Commission for their diligence in providing high-quality surveillance data on nonfatal firearm-related injuries using the NEISS. We also thank Mr Steve James for his assistance in data preparation and Dr George Ryan for statistical consultation at the National Center for Injury Prevention and Control.
Reprint no. 47/1/90751 Address for reprints: Joseph L Annest, PhD
29. Singh GK, Kochanek KD, Macdorman MF: Advance report of final mortality statistics, 1994. DHHS Publication No. (PHS) 96-1120. Monthly Vital Statistics Report, vol 45, no. 3, suppl. Hyattsville, MD: National Center for Health Statistics, 1996:22-57.
Director, Office of Statistics and Programming
30. Mullins RJ, Veum-Stone J, Helfand M, et al: Outcome of hospitalized injured patients after institution of a trauma system in an urban area. JAMA 1994;271:1919-1924.
Centers for Disease Control and Prevention
31. Shackford SR, Hollingworth-Fridlund P, Cooper GF, et al: The effect of regionalization upon the quality of trauma care as assessed by concurrent audit before and after institution of a trauma system: A preliminary report. J Trauma 1986;26:812-820.
Atlanta, GA 30341-3724
32. Hammond WR, Yung BR: Preventing violence in at-risk African-American youth. J Health Care Poor Underserved 1991;2:359-373.
National Center for Injury Prevention and Control 4770 Buford Hwy, NE, MS/K59 770-488-4804 Fax 770-488-1665
33. Tolan P, Guerra N: What Works in Reducing Adolescent Violence: An Empirical Review of the Field. Boulder, CO: University of Colorado, Boulder, Institute for Behavioral Sciences, Center for the Study and Prevention of Violence, 1994. 34. Zigler E, Taussig C, Black K: Early childhood intervention: A promising preventative for juvenile delinquency. Am Psychol 1992;47:997-1006. 35. Loftin C, McDowall D, Wiersema B, et al: Effects of restrictive licensing of handguns on homicide and suicide in the District of Columbia. N Engl J Med 1991;325:1615-1620. 36. Sloan JH, Kellermann AL, Reay DT, et al: Handgun regulations, crime, assaults, and homicide: A tale of two cities. N Engl J Med 1988;319:1256-1262. 37. Wintemute GJ: The relationship between firearm design and firearm violence: Handguns in the 1990s. JAMA 1996;275:1749-1753. 38. Kellermann AL, Rivara FP, Somes G, et al: Suicide in the home in relation to gun ownership. N Engl J Med 1992;327:467-472. 39. Kellermann AL, Rivara FP, Rushforth NB, et al: Gun ownership as a risk factor for homicide in the home. N Engl J Med 1993;329:1084-1091. 40. Brent DA, Perper JA, Allman CJ, et al: The presence and accessibility of firearms in the homes of adolescent suicides: A case control study. JAMA 1991;266:2989-2995. 41. Hutson RH, Anglin D, Kyriacou DN, et al: The epidemic of gang-related homicides in Los Angeles County from 1979 through 1994. JAMA 1995;274:1031-1036. 42. McGonigal MD, Cole J, Schwab CW, et al: Urban firearm deaths: A five-year perspective. J Trauma 1993;35:532-537. 43. Centerwall BS: Race, socioeconomic status, and domestic homicide. JAMA 1995;273:1755-1758. 44. Blumstein A: Youth violence, guns, and the illicit-drug industry. Journal of Criminal Law and Criminology 1995;86:10-36. 45. Gladwell M: The tipping point. The New Yorker. June 3, 1996:32-38. 46. Kennedy DM: Can we keep guns away from kids? The American Prospect Summer 1994:74-80. 47. Teret SP: The firearm injury reporting system revisited. JAMA 1996;275:70. 48. Feliciano DV: Patterns of injury, in Feliciano DV, Moore EE, Mattox KL (eds): Trauma, ed 3. Stamford, CT: Appleton & Lange, 1996:85-103. 49. Bell CC, Jenkins EJ, Kpo W, et al: Response of emergency rooms to victims of violence. Hosp Commun Psychiatry 1994;45:142-146. 50. Wetli CV: Forensic issues, in Ivatury RR, Cayten CG (eds): The Textbook of Penetrating Trauma. Baltimore, MD: Williams & Wilkins, 1996:1084-1097. 51. Garrison HG, Runyan CW, Tintinalli JE, et al: Emergency department surveillance: An examination of issues and a proposal for a national strategy. Ann Emerg Med 1994;24:849-856.
JULY 1998
32:1
ANNALS OF EMERGENCY MEDICINE
5 9