American Journal of Epidemiology Advance Access published October 22, 2013 American Journal of Epidemiology © The Author 2013. Published by Oxford University Press on behalf of the Johns Hopkins Bloomberg School of Public Health. All rights reserved. For permissions, please e-mail:
[email protected].
DOI: 10.1093/aje/kwt255
Original Contribution Firearm-related Hospitalizations and In-Hospital Mortality in the United States, 2000–2010
Bindu Kalesan*, Clare French, Jeffrey A. Fagan, Dennis L. Fowler, and Sandro Galea
Initially submitted July 18, 2013; accepted for publication September 25, 2013.
Most firearm-related injuries are nonfatal and require hospitalization. Using data on 3,257,720 hospitalizations from the National Hospital Discharge Survey (2000–2010), we determined overall and cause-, gender-, and race-specific trends in firearm-related hospitalization (FRH) and determinants of in-hospital firearm mortality. Types of FRH evaluated, according to International Classification of Diseases, Ninth Revision, Clinical Modification, E-diagnostic codes, were accident (codes E922.0–E922.3, E922.8, and E922.9), assault (codes E965.0–E965.4), attempted suicide (codes E955.0–E955.4), legal intervention (code E970), undetermined intent (codes E985.0– E985.3), and war (code E991). A moderate reduction in FRH rates was observed from 2000 to 2011: from 62 FRHs per 100,000 hospitalizations to 57 per 100,000 (P-trend = 0.0016). The majority of FRHs were due to assault (P-trend = 0.19) or accident (P-trend = 0.32) and showed no significant reduction in rates over time, whereas rates for 14% of all FRHs—those due to attempted suicide (P-trend = 0.002) and undetermined intent (P-trend = 0.0029)— declined moderately. Moderate declines were observed among both blacks (from 213.1 FRHs per 100,000 hospitalizations to 164.4 per 100,000; P-trend = 0.049) and whites (from 38.4 FRHs per 100,000 hospitalizations to 32.2 per 100,000; P-trend = 0.031). The decline was significant only among men (effect size = 0.9, P-trend = 0.004). In conclusion, the reduction in FRH was driven by a reduction in self-inflicted and undetermined injuries. FRH rates were 6-fold greater among blacks than among whites and 14-fold greater in men than in women throughout the period. firearms; hospitalization; injury nonfatal; trends; violence
Abbreviations: CI, confidence interval; FRH, firearm-related hospitalization; NHDS, National Hospital Discharge Survey; OR, odds ratio; SD, standard deviation.
are nearly 40 times as many nonfatal firearm-related events as there are firearm-related deaths annually (2). In 2011, 478,400 fatal and nonfatal firearm incidents occurred in the United States, the vast majority of which were nonfatal (97.4%) (2). Approximately three-fourths of nonfatal firearm injuries required medical attention, and 80% of persons requiring medical care were hospitalized. Therefore, the majority of the victims of firearm incidents survive and live wounded, compromised lives with injuries that complicate their day-to-day activities (6). Until now, the focus of gun control policies has narrowly been informed by firearm fatalities, which is merely the “tip of the iceberg” of this problem. Inclusion of nonfatal firearm incidents in national discussions
Numbers of fatal and nonfatal firearm injuries in the United States declined dramatically during the 1990s but have remained relatively constant in the intervening decade (1–3). At the same time, the US firearm-related death rate remains twice than that of the next-highest country, Northern Ireland, 3 times that of Canada, and greater than that of Brazil, which tops the list among upper-middle-income countries (4, 5). In the wake of several high-profile incidents of gun-related violence, particularly the December 2012 massacre in Newtown, Connecticut, a public discussion now rages about US gun policies. Although much of the public discussion around firearms in the United States has focused on fatal firearm injuries, there 1
Downloaded from http://aje.oxfordjournals.org/ at Columbia University Libraries on October 23, 2013
* Correspondence to Dr. Bindu Kalesan, Department of Epidemiology, Mailman School of Public Health, Columbia University, 722 West 168th Street, New York, NY 10032 (e-mail:
[email protected]).
2 Kalesan et al.
regarding gun policy can help contribute to a comprehensive discussion of gun policy in the United States. Therefore, we were interested in documenting trends in firearm-related hospitalizations (FRHs) from 2000 to 2010. Our central aim was to document overall trends in FRH and rates of cause-, gender- and race-specific FRHs. Moreover, we aimed to determine any differential in the risk of in-hospital mortality due to FRHs according to demographic characteristics. MATERIALS AND METHODS Data source
The National Center for Health Statistics, part of the Centers for Disease Control and Prevention, conducted the National Hospital Discharge Survey (NHDS) annually from 1965 to 2010 (http://www.cdc.gov/nchs/nhds.htm). Demographic and medical information on inpatients discharged from nonfederal short-stay hospitals was collected from a sample of hospitalizations using a national probability selection of hospitals in the 50 states and the District of Columbia (6). The NHDS included general hospitals or children’s general hospitals with an average length of stay of fewer than 30 days for all patients. Federal, military, and Department of Veterans Affairs hospitals were excluded, as were hospital units of institutions and hospitals with fewer than 6 beds staffed for patient use. In 1988, the NHDS was redesigned, and the sample was selected from a frame of short-stay hospitals listed in the 1987 SMG Hospital Market Database (SMG Hospital Marketing Group, Inc., Chicago, Illinois). Details about the redesigned survey methodology are available elsewhere (7). From 1995 to 2007, the sample included 501–525 hospitals. From 2008 to 2010, the NHDS used a half-sample of 239 hospitals. The Research Ethics Review Board of the National Center for Health Statistics approved data collection for the NHDS. Analysis of deidentified data from the survey is exempt from federal regulations for the protection of human research participants. Study population
Publicly available NHDS data sets from 2000–2010 were used for this analysis (n = 3,257,720). Each record pertained to 1 hospitalization, and estimates from the NHDS are for hospitalizations, not persons; thus, a person may have had multiple hospitalizations. Variables and definitions
Information on FRH was derived from E codes (15 diagnostic codes) of the International Classification of Diseases, Ninth Revision, Clinical Modification, as represented in the NHDS. The first-listed code is the principal diagnosis; since E codes are by definition a secondary diagnosis, we defined FRH from any listed E-codes. Types of FRH, as defined by cause, were: accident (codes E922.0–E922.3, E922.8, and E922.9), assault (codes E965.0–E965.4), attempted suicide or self-inflicted injury (codes E955.0–E955.4), legal intervention (code E970), undetermined intent (codes E985.0–
E985.3), and war (code E991). There were no hospitalizations in the NHDS due to legal intervention or war. All FRHs were defined as due to accident, assault, attempted suicide, or undetermined intent. Demographic variables included age, gender, race (white, black, or other), current marital status (married or not married), and geographic region (4 US Census Bureau regions: Northeast, Midwest, South, and West). Hospital characteristics used in the analysis were hospital ownership ( proprietary, government, or nonprofit), hospital size (number of beds: 6–99, 100–199, 200–299, 300–499, or ≥500), and payment source (government, private, or other). In-hospital mortality was also included in this analysis. Statistical analysis
We combined data from cross-sectional surveys carried out between 2000 and 2010 to obtain more precise estimates and to conduct subgroup analyses (8). Rates of overall FRH and of each type of FRH, per 100,000 hospitalizations, were calculated after weighting with analytical weights in each of the 11 data sets to reflect all US hospital discharges. Rates for each survey year were calculated and then analyzed using random-effects meta-analysis to estimate a pooled rate for 2000– 2010. Since the data were from repeated cross-sectional surveys and not panel data, meta-analysis was used to obtain pooled estimates and to assess trends across the survey years. Heterogeneity among survey years was examined using the I 2 statistic. The I 2 statistic ranges from 0% to 100% and describes the percentage of total variation across studies that is attributable to heterogeneity between survey years rather than chance (9, 10). Meta-regression was used to determine annual changes in rates and standard deviations, assuming linear trends. We calculated effect size or Cohen’s D (annual change in rate/standard deviation) to assess the magnitude of the reduction (11, 12). An effect size greater than 0.5 was considered to be an effect of moderate magnitude. Trends in rates of hospitalization from 2000 to 2010 were calculated by trend test (χ2 statistic for trend), using the weighted frequencies (13, 14). We used z tests to compare rates between groups. To examine the associations of overall FRH and each type of FRH with sociodemographic and hospital characteristics, we calculated odds ratios and 95% confidence intervals using bivariate logistic regression weighted by analytical weights for each survey year. Random-effects meta-analysis was used to estimate pooled risk estimates and 95% confidence intervals. In order to determine trends in associations, we used generalized least squares for trend estimation of summarized dose-response data (15, 16). In-hospital fatality rates for a diagnosis of gunshot injury were also calculated, followed by stratified analyses in subgroups of gender, age category, race, marital status, and region. Formal tests for interaction were performed using meta-regression to compare stratumspecific associations of in-hospital mortality and FRH with each of these characteristics. We used Stata 12.1 (StataCorp LP, College Station, Texas; 2009) to manage the data and conduct the analyses in order to account for the complex sample survey design and to calculate standard errors, variances, and 95% confidence intervals for weighted estimates. All P values were 2-sided.
Firearm-related Hospitalizations in the United States
RESULTS
100,000 hospitalizations among women. By race, the lowest overall rate was seen among whites (33.48 per 100,000 hospitalizations), while among blacks the rate was 208.9 per 100,000. A strong evidence for moderate-to-high heterogeneity in the subgroups was noted. Table 1 presents the associations of FRH with patient sociodemographic and hospital characteristics. We observed that hospitalizations due to firearm injury were much more likely to involve men than women (odds ratio (OR) = 13.74, 95% CI: 9.99, 18.90), to involve the age group 15–44 years (OR = 19.67, 95% CI: 14.26, 27.12) as compared with ≤15 years, and to involve blacks (OR = 6.27, 95% CI: 5.17, 7.60) as compared with whites. There was a significant trend for the association between overall FRH and categories of age (P-trend < 0.0001). A similar and significant trend was observed across all 4 type categories (accidents, attempted suicides, assaults, and undetermined intent). A reversal of risk of overall FRH, indicating a relative risk reduction of 75%, was observed in the oldest age group (OR = 0.25, 95% CI: 0.16, 0.40). This relative risk reduction for persons aged 65 years or more persisted across the cause categories. Overall, persons with FRHs were 6.27 times more likely to be black than to be white, with the risk ranging from 4.6 to 9.3 times in each of the cause categories except self-inflicted gunshots, where a reduction of 11% was observed. Accidental FRHs (OR = 0.54, 95% CI: 0.35, 0.82) were distinctly less likely among persons in the “other” race category as compared with whites, and self-inflicted gunshot injuries showed a trend towards lower likelihood (OR = 0.76, 95% CI: 0.52, 1.12) among persons of “other” race, while gunshot injuries due to assault had a 97% greater likelihood of involving “other” racial groups. While unmarried status was a risk factor for FRH overall (OR = 2.94, 95% CI: 2.42, 3.57), the strength of association was 5.6 times greater for hospitalizations due to assault-related firearm injuries than for other etiologies.
From 3,257,720 hospitalizations recorded in the NHDS during 2000–2010, we identified 1,545 FRHs; after weighting, these represented an estimated rate of 57.28 (95% confidence interval (CI): 52.93, 61.64) FRHs per 100,000 hospitalizations (Figure 1). Rates of FRH during 2000– 2010 ranged from a low of 38.57 for 2009 to a high of 70.32 for 2003. Heterogeneity between survey years was moderately high (I 2 = 63.5%, P = 0.002), but an overall significant trend for reduction in rates from 2000 to 2010 was observed (P-trend = 0.0016). An annual reduction of 1.38 (standard deviation (SD), 1.39) in the rate of FRH was seen, with an effect size of 0.63. Figures 2–4 present the rates of FRH by cause, gender, and race. The pooled rate across the study period was highest when the intent was known to be assault (30.67 FRHs per 100,000 hospitalizations; 95% CI: 28.02, 33.31), with the nexthighest rate being FRH due to accident (17.94, 95% CI: 15.47, 20.40), while the lowest rate was associated with attempted suicide (2.93, 95% CI: 1.67, 4.20). Annual reductions in rates of FRH due to assault (−0.42 (SD, 1.7)) and attempted suicide (−0.45 (SD, 0.8)) were similar, but the magnitudes of reduction as measured by effect size were 0.25 and 0.58, respectively. The magnitude of reduction in the hospitalization rate from 2000 to 2010 was significant for undetermined intent (annual change: −0.40 (SD, 0.7); effect size = 0.59, P-trend = 0.003) and for self-inflicted injuries (−0.45 (SD, 0.8); effect size = 0.58, P-trend = 0.002). A significant declining trend over the years was observed among men (annual change: −3.21 (SD, 3.6); effect size = 0.90, P-trend = 0.004) and among whites (annual change: −1.31 (SD, 2.7); effect size = 0.47, P-trend = 0.031) and blacks (annual change: −7.03 (SD, 12.7); effect size = 0.55, P-trend = 0.049). The hospitalization rate among men was 124.57 FRHs per 100,000 hospitalizations as compared with 9.78 per
Survey Year 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Overall (I 2 = 63.5, P = 0.002) 30
40
50
3
60
70
Rate
95% CI
62.47 57.47 60.97 70.32 55.95 57.74 55.66 53.76 64.44 38.57 53.58 57.28
53.72, 71.22 49.30, 65.65 52.51, 69.43 61.13, 79.51 48.34, 63.56 50.05, 65.42 48.13, 63.20 46.25, 61.27 52.22, 76.67 29.01, 48.12 41.93, 65.23 52.93, 61.64
80
Rate Figure 1. Rate of all firearm-related hospitalizations per 100,000 hospitalizations in each survey year, National Hospital Discharge Survey, 2000– 2010. The pooled rate of firearm-related hospitalization was calculated using random-effects meta-analysis. The average annual change in rates was derived from meta-regression. P values for trend across years were calculated using frequencies per year, weighted using survey analytical weights. Annual change = −1.38; P-trend = 0.0016. Bars, 95% confidence interval (CI).
4 Kalesan et al.
Cause and Survey Year Assault 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 41.2%, P = 0.083) Accident 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 64.8%, P = 0.002) Undetermined intent 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 66.6%, P = 0.001) Suicide 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 83.2%, P < 0.0001) 0
10
20 30 Rate
Rate
95% CI
31.51 28.72 39.52 31.84 25.49 35.65 27.99 30.13 27.09 29.73 30.67
25.45, 37.56 22.92, 34.53 32.63, 46.42 26.10, 37.58 20.38, 30.59 29.62, 41.68 22.57, 33.42 21.77, 38.49 19.08, 35.10 21.05, 38.41 28.02, 33.31
18.93 16.05 18.72 20.10 17.77 21.12 15.94 22.75 22.71 8.21 17.27 17.94
14.11, 23.75 11.73, 20.37 14.04, 23.41 15.18, 25.01 13.48, 22.06 16.48, 25.77 11.90, 19.97 17.86, 27.64 15.46, 29.97 3.80, 12.62 10.65, 23.88 15.47, 20.40
5.33 6.45 7.81 4.42 4.78 2.65 1.58 4.06 2.98 5.58 4.31
2.84, 7.82 3.70, 9.20 4.75, 10.88 2.28, 6.55 2.57, 7.00 1.01, 4.30 0.29, 2.87 0.99, 7.13 0.32, 5.64 1.82, 9.33 3.05, 5.58
4.58 7.08 2.88 1.93 6.34 1.43 1.43 7.53 0.29 1.01 2.93
2.27, 6.89 4.19, 9.96 1.02, 4.75 0.52, 3.35 3.79, 8.89 0.22, 2.63 0.20, 2.65 3.35, 11.71 –0.54, 1.11 –0.59, 2.61 1.67, 4.20
40
Figure 2. Rates of firearm-related hospitalization per 100,000 hospitalizations, by cause, National Hospital Discharge Survey, 2000–2010. The pooled rate of firearm-related hospitalization in each category was calculated using random-effects meta-analysis. The average annual change in rates was derived from meta-regression. P values for trend across years in each category were calculated using frequencies per year, weighted using survey analytical weights. Annual change: assault, −0.42; accident, −0.29; undetermined intent, −0.40; suicide, −0.45. P-trend: assault, 0.19; accident, 0.32; undetermined intent, 0.0029; suicide, 0.002. Bars, 95% confidence interval (CI).
Among the hospital characteristics, government hospitals as compared with proprietary hospitals (OR = 11.91, 95% CI: 7.67, 18.49) and large hospitals with 500 or more beds as compared with hospitals with fewer than 100 beds (OR = 14.42, 95% CI: 8.97, 23.19) had a greater likelihood of admitting people with firearm injuries. The durations (days)
of hospital stays due to any type of firearm injury (mean = 4.81 (SD, 6.3)) did not differ significantly (P = 0.72) from lengths of stay for persons admitted for other reasons (mean = 4.75 (SD, 7.1)). Leading diagnoses for FRHs were internal injury to the thorax, abdomen, or pelvis (20.1%), fracture of the lower limb (16.3%), open wound of the head, neck,
Firearm-related Hospitalizations in the United States
Gender and Survey Year Men 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 45.9%, P = 0.047) Women 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 81.0%, P < 0.0001)
0
50
100 Rate
150
Rate
95% CI
140.77 125.48 132.31 144.10 131.70 119.03 121.72 120.55 122.63 87.59 120.22 124.57
120.18, 161.35 106.58, 144.37 112.83, 151.79 123.58, 164.61 113.47, 149.92 101.84, 136.22 104.34, 139.09 103.02, 138.07 96.40, 148.87 65.33, 109.85 93.29, 147.16 116.35, 132.79
8.75 11.12 11.68 18.72 3.50 14.72 9.31 6.41 23.28 3.38 5.92 9.78
4.50, 13.00 6.45, 15.80 6.87, 16.50 12.54, 24.90 1.02, 5.98 9.66, 19.78 5.2, 13.34 3.02, 9.79 13.69, 32.87 –0.33, 7.08 0.83, 11.00 6.78, 12.78
5
200
Figure 3. Rates of firearm-related hospitalization per 100,000 hospitalizations, by gender, National Hospital Discharge Survey, 2000–2010. The pooled rate of firearm-related hospitalization for each gender was calculated using random-effects meta-analysis. The average annual change in rates was derived from meta-regression. P values for trend across years in each category were calculated using frequencies per year, weighted using survey analytical weights. Annual change: men, −3.21; women, −0.40. P-trend: men, 0.0038; women, 0.21. Bars, 95% confidence interval (CI).
or trunk (12.1%), open wound of the lower limb (11.7%), and fracture of the upper limb (11.0%). The most common primary surgical procedures performed on FRH victims were operations on the musculoskeletal system (24%), diagnostic and therapeutic procedures (16%), operations on the digestive system (14%), and operations on the integumentary system (10%); 17% had no major procedure, and 3% had operations on the nervous system. From 2000 to 2010, overall in-hospital deaths (firearmrelated and non–firearm-related) accounted for 2.04% of all hospitalizations (2,053.3 deaths per 100,000 hospitalizations; 95% CI: 2,037.9, 2,068.7). Among the FRHs only, in-hospital mortality was 8.1%, while 91.8% of the patients were discharged alive. In comparison, the in-hospital mortality rate among non–firearm-related hospitalizations was 2.04%. By type of firearm injury, the mortality rate was 39.3% for attempted suicides, 5.6% for accidents, 4.6% for assaults, and 17.2% for undetermined intent. The rates of nonfatal and fatal FRHs were 52.75 (95% CI: 48.51, 56.92) and 4.36 (95% CI: 2.78, 5.95) per 100,000 hospitalizations, respectively. Table 2 shows results from stratified analysis for the risk of in-hospital mortality and FRH. Among hospitalizations related and not related to firearm injuries, the in-hospital mortality rates were 7,453.2 (95% CI: 6,081.8, 8,824.6) and 2,050.1 (95% CI: 2,034.7, 2,065.4) per 100,000 hospitalizations, respectively. The overall risk of in-hospital death was 3.6 times greater among FRHs than among other hospitaliza-
tions (OR = 3.62, 95% CI: 2.72, 4.81). Risk of mortality remained high in all subgroups based on gender, age, race, marital status, and region and did not show significant interactions across strata. DISCUSSION
Using national hospitalization data from 2000–2010, we found that the overall rate of hospitalization due to firearm injuries during this period was 57 per 100,000 hospitalizations. Of those FRHs, 53% were due to assault, 33% were due to accidental injury, and 9% were of undetermined intent, while 5% were related to attempted suicide. A declining trend of moderate magnitude in the rate of overall FRH across the 11-year period was observed and was driven by reduction in rates of FRH associated with attempted suicide and undetermined intent. There was no significant change in accidental or assault-related FRHs. Our main finding, the pattern of a moderate declining trend in overall FRHs, is in line with a trend analysis that used US hospital emergency data from the National Electronic Injury Surveillance System for 1985–1995 (1). This suggests that the decline we observed was probably a continuation of the momentum set prior to the 21st century. A subanalysis of data from the same study, for the period 1992–1995, demonstrated that almost three-fourths of FRHs were due to assault (1). This estimate was higher than what we observed; in our
6 Kalesan et al.
Race and Survey Year White 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 78.3%, P < 0.0001) Black 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 68.2%, P = 0.001) Other 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Subtotal (I 2 = 79.7%, P < 0.0001) 0
100
200 Rate
300
Rate
95% CI
38.40 36.07 39.49 36.79 34.86 30.59 32.16 32.85 51.26 11.83 32.21 33.48
28.90, 47.89 27.11, 45.02 29.99, 48.99 27.43, 46.16 26.61, 43.10 22.90, 38.28 24.31, 40.02 24.73, 40.98 35.65, 66.87 5.02, 18.64 20.82, 43.59 27.69, 39.27
213.05 220.40 194.00 317.69 206.09 237.47 228.09 199.28 165.12 146.76 164.43 208.90
169.65, 256.45 178.19, 262.61 154.08, 233.92 265.59, 369.79 166.59, 245.59 195.37, 279.56 186.76, 269.41 160.69, 237.87 112.11, 218.13 96.22, 197.31 109.58, 219.28 185.09, 232.72
53.61 33.83 51.57 70.32 38.41 40.77 31.76 33.55 45.46 45.70 46.27 44.49
39.70, 67.53 22.96, 44.69 38.31, 64.83 61.13, 79.51 27.47, 49.35 29.61, 51.94 21.87, 41.64 23.33, 43.77 28.73, 62.20 25.29, 66.11 23.79, 68.76 36.14, 52.85
400
Figure 4. Rates of firearm-related hospitalization per 100,000 hospitalizations, by race, National Hospital Discharge Survey, 2000–2010. The pooled rate of firearm-related hospitalization in each racial group was calculated using random-effects analysis. The average annual change in rates was derived from meta-regression. P values for trend across years in each category were calculated using frequencies per year, weighted using survey analytical weights. Annual change: white, −1.31; black, −7.03; other, −1.03. P-trend: white, 0.031; black, 0.049; other, 0.19. Bars, 95% confidence interval (CI).
study, only half of all FRHs were assault-related. The drop from 75% to 53% could be indicative of a rapid reduction in assault-related FRHs during the early 1990s that began in 1993 and continued until 1999, thereafter plateauing in the 21st century. Our finding that assault was the leading cause of FRH between 2000 and 2010, followed by unintentional injuries, is concordant with results from a cross-sectional analysis of the National Inpatient Sample from 1997 (17). Additionally, only 5% of the FRHs in our study were related to attempted suicide; this is consistent with estimates from the National Electronic Injury Surveillance System (1). Echoing the gender disparity in national firearm fatality rates (18), we found substantial gender heterogeneity, where rates of FRH for men were almost 14 times those of
women throughout the study time period. The rates in men showed a significant fall across the 11-year period, with an effect size of 0.9, and rates among women showed high heterogeneity with no change, suggesting that the overall decline in FRHs could be attributed to the drop among men. The rate of FRH was disproportionately higher among blacks in our study, with rates that were 6-fold greater than those of whites throughout the study period, and this is in line with the racial differences documented in national firearm-related homicide estimates (19). We also demonstrated a significant declining trend in FRHs among both whites and blacks. We found that all causes of FRH were more prevalent among men and among persons aged 15–45 years. These associations concord with results from trauma centers (20), state-specific reports
Firearm-related Hospitalizations in the United States
7
Table 1. Associations of Different Types of Firearm-related Hospitalizations With Sociodemographic and Hospitalization Characteristics, National Hospital Discharge Survey, 2000–2010 Cause of Firearm-related Hospitalization
All Causes Accident ORa
Male gender (referent: female)
95% CI
Attempted Suicide
Assault
Undetermined Intent
ORa
95% CI
ORa
95% CI
ORa
95% CI
ORa
95% CI
13.74
9.99, 18.90
11.42
6.75, 19.32
10.63
3.52, 32.06
18.18
13.76, 24.04
15.48
4.86, 49.31
15–44
19.67
14.26, 27.12
13.51
6.91, 26.41
3.25
0.16, 64.30
21.75
14.69, 32.21
8.31
3.62, 19.08
45–64
3.20
2.12, 4.84
1.90
1.08, 3.36
3.30
0.72, 15.06
2.91
1.79, 4.70
0.77
0.20, 3.02
0.12, 0.85
—
Age group, years (referent: 0–14)
65–99
0.25
0.16, 0.40
0.45
0.24, 0.85
0.35
0.09, 1.37
0.31
b
—
Race (referent: white) Black
6.27
5.17, 7.60
4.65
3.03, 7.14
0.89
0.51, 1.53
9.27
7.69, 11.17
6.51
3.17, 13.37
Other
1.26
1.03, 1.55
0.54
0.35, 0.82
0.76
0.52, 1.12
1.97
1.39, 2.79
1.58
0.69, 3.61
2.94
2.42, 3.57
1.82
1.22, 2.71
1.18
0.67, 2.08
5.62
3.96, 7.97
3.18
0.90, 11.27
1.66
1.37, 2.02
1.56
1.02, 2.41
1.12
0.38, 3.36
1.74
1.29, 2.34
1.77
0.71, 4.45
South
2.33
1.76, 3.08
3.16
1.79, 5.57
1.66
0.53, 5.19
1.97
1.49, 2.61
1.97
0.79, 4.88
West
2.73
2.14, 3.48
2.05
1.20, 3.47
4.52
2.77, 7.38
3.10
2.67, 4.24
3.04
1.15, 8.02
Unmarried marital status (referent: married) Region (referent: Northeast) Midwest
Type of hospital ownership (referent: proprietary) Government
11.91
7.67, 18.49
5.56
3.54, 12.17
1.15
0.59, 2.24
12.57
6.94, 22.79
3.80
1.58, 9.16
4.56
3.03, 6.84
2.93
1.83, 4.67
0.51
0.34, 0.76
4.51
2.40, 8.48
0.77
0.17, 3.51
100–199
2.55
1.52, 4.29
2.78
1.48, 5.21
0.80
0.10, 6.22
1.65
0.42, 6.55
4.05
3.56, 4.60
200–299
4.41
2.76, 7.05
3.73
2.27, 6.13
0.75
0.30, 1.87
5.73
2.45, 13.42
3.45
1.05, 11.31
9.01
5.14, 15.75
4.68
2.77, 7.92
0.90
0.39, 2.12
12.59
3.34, 47.44
1.34
1.04, 1.73
14.42
8.97, 23.19
7.62
4.58, 12.67
3.94
1.19, 13.03
23.86
11.25, 50.61
11.87
4.39, 32.08
Nonprofit Hospital size, no. of beds (referent: 6–99)
300–499 ≥500
Abbreviations: CI, confidence interval; OR, odds ratio. a Unadjusted odds ratios and 95% confidence intervals were calculated by means of logistic regression using survey analytical weights for each year and then pooled using random-effects meta-analysis. b For undetermined intent, the odds ratio and 95% confidence interval in the age category 65–99 years were not estimable because of an inadequate sample size.
(21, 22), and studies using other nationally representative data (1, 17, 23). Trauma centers and national data from emergency departments have established that the risk of in-hospital death is disproportionately higher in firearm-related self-inflicted injuries than in other types of self-inflicted injuries (24–26). Therefore, it was not surprising that in our study, risk of inhospital mortality among FRHs was greatest when the injury was self-inflicted. A closer look at the risk of in-hospital mortality in our study found a differential in risk of in-hospital mortality by gender; the risk was doubled in women compared with men. This concords with similar relationships described in studies using data from the National Inpatient Sample and the National Electronic Injury Surveillance System (1, 17), and it suggests greater severity of firearm-related injuries in hospitalized women as compared with men,
eventually leading to death. On the other hand, the relatively lower risk of in-hospital mortality among blacks in comparison with whites, when taken together with the known increased fatality due to firearm injuries among blacks (18), suggests that a higher proportion of blacks compared with whites die immediately upon injury and that the few who survive and reach the hospital are less severely injured. Rates of nonfatal firearm crimes in the United States declined sharply (from 7.3 per 1,000 persons to 2.9 per 1,000 persons) along with overall crime rates in the 1990s (2). The decline in overall crime rates (3) is generally attributed to an increase in the sheer numbers of police, increased incarceration for major crimes (incapacitation and deterrence), and reduced use of crack cocaine. Gun control laws are generally not thought to have contributed to this trend because of the existence of active black markets in guns and the lack of
8 Kalesan et al.
Table 2. Risk of In-hospital Mortality According to Firearm-related Hospitalization Status, National Hospital Discharge Survey, 2000–2010 In-hospital Mortality Rate per 100,000 Hospitalizations Hospitalization Related to Firearm
Total
Hospitalization Not Related to Firearm
Odds Ratioa
95% Confidence Interval
P Value
7,453.2
2,050.1
3.62
2.72, 4.81
