The Accessibility of Firearms and Risk for Suicide and Homicide Victimization Among Household Members

Firearms cause an estimated 31000 deaths annually in the United States (1). Data from the 16-state National Violent Death Reporting System indicate that 51.8% of deaths from suicide in 2009 (n= 9949) were firearm-related; among homicide victims (n= 4057), 66.5% were firearm-related. Most suicides (76.4%) occurred in the victims homes. Homicides also frequently occurred in the home, with 45.5% of male victims and 74.0% of female victims killed at home (2). Firearm ownership is more prevalent in the United States than in any other country; approximately 35% to 39% of households have firearms (3, 4), and 22% of persons report owning firearms. The annual rate of suicide by firearms (6.3 suicides per 100000 residents) is higher in the United States than in any other country with reported data, and the annual rate of firearm-related homicide in the United States (7.1 homicides per 100000 residents) is the highest among high-income countries (4). Results from ecological studies suggest that state restrictions on firearm ownership are associated with decreases in firearm-related suicides and homicides (5). Specific characteristics about storage and types of firearms seem to increase suicide risk. Firearms that are stored loaded or unlocked are more likely to be used than those that are unloaded or locked (6, 7), and adolescent suicide victims often use an unlocked firearm in the home (8). The apparent increased risk for suicide associated with firearms in the home is not unique to persons with a history of mental illness (7) and may be more of an indicator of the ease of impulsive suicide. Impulsiveness may be a catalyst in using a firearm to commit suicide and may also play a role in firearm-related homicide. Researchers have estimated higher odds of homicide victimization among women than men (9, 10). Because most homicide victims know their perpetrators (9), this finding may indicate an impulsive reaction to domestic disputes. To our knowledge, this is the first systematic review and meta-analysis to estimate the association between firearm accessibility and suicide or homicide victimization. Supplement. Original Version (PDF) Methods We used Cochrane Collaboration methods (11) throughout the review process. Data Sources and Searches We searched PubMed, EMBASE, the Cochrane Central Register of Controlled Trials, and Web of Science without date, geographic, or language limitations. We also examined bibliographies of included articles to identify additional references. In addition, we searched the gray literature for papers related to firearms and suicide or homicide. The Appendix and Appendix Table 1 (both available at www.annals.org) present details of our search strategy and screening process. Appendix Table 1. Search Strategy Study Selection Study Design Study designs eligible for inclusion in our review were randomized, controlled trials; nonrandomized, controlled trials; pre- or postintervention evaluations; and observational studies (for example, cohort or casecontrol studies) if a comparator was available. Because we were concerned with the individual effects of firearm accessibility, we included only studies with individual-level data and excluded those with population-level data (for example, ecological studies). Types of Participants Participants were not restricted by age, sex, or country of residence. Types of Exposures Studies needed to assess whether firearms were available for all participants. In addition, included studies needed to assess outcomes between participants with and without access to firearms. Specifically, studies needed to compare firearm ownership or availability (that is, accessibility) with no firearm ownership or availability (that is, no accessibility) or provide adequate data to estimate the effect that firearms had on selected harms outcomes. Firearm accessibility could be defined as self- or proxy-reported or assumed from other types of exposure data (for example, firearm purchase records). Types of Outcome Measures The primary outcomes of interest were suicide or homicide victimization (that is, being a victim of homicide rather than a perpetrator). Data Extraction and Quality Assessment Two authors independently extracted relevant data into a standardized, prepiloted data extraction form. Assessment of Risk of Bias Two authors independently assessed the risk of bias for each study by using the NewcastleOttawa Scale (12, 13). We resolved disagreements by discussion or by involving the third author to adjudicate (Table 1 and Appendix Table 2). Table 1. Summary of Critical Appraisal of Included Studies Using the NewcastleOttawa Scale for Assessing the Quality of Observational Studies* Appendix Table 2. Detailed Risk of Bias Results Using the NewcastleOttawa Scale for Assessing Quality for Observational Studies Data Synthesis and Analysis When necessary, we calculated the odds ratio (OR) and 95% CI for dichotomous outcomes, although published adjusted estimates were preferentially used if provided in the report. We pooled data across studies and estimated summary effect sizes by using fixed- and random-effects models. The choice of model was determined by the significance of the maximum likelihood estimate of the heterogeneity parameter ( 2) (14). If the estimate of 2 did not significantly differ from 0, the fixed-effects model was used (14). We present 2 estimates of heterogeneitythe I 2 statistic and the coefficient. Estimates of the former are interpreted as the percentage of variability in effect estimates due to heterogeneity rather than chance, whereas the latter can be interpreted as the clinical heterogeneity as determined by the estimated SD of underlying effects across studies. Unlike the I 2 statistic, the coefficient does not change with the number of patients included in the studies in a meta-analysis (15). We used R, version 3.0.0 (R Foundation for Statistical Computing, Vienna, Austria), for statistical analyses. The coefficient was measured on the log OR scale. This review is registered in PROSPERO (CRD42013004469). Results Search Results The database searches yielded 6902 references (Figure 1). We removed 2929 duplicates and an additional 2881 clearly irrelevant references. We then identified 2382 records through gray-literature searches. We closely reviewed 3474 titles and abstracts. After this screening, we selected 70 articles for full-text review. We identified an additional 4 studies by cross-referencing bibliographies (1619). Overall, 16 observational studies met our inclusion criteria. The Appendix shows the disposition of studies excluded after full-text review. Figure 1. Summary of evidence search and selection. Fourteen of the included studies estimated the odds of suicide in the context of firearm accessibility (68, 10, 1625), and 6 studies estimated the odds of homicide victimization in this context (9, 10, 2224, 48). Four studies reported both outcomes (10, 2224). Study Characteristics Demographic Characteristics Persons who completed suicide (mean, 75% [range, 70% to 85%]) (68, 10, 1621, 23) and homicide victims (mean, 79% [range, 63% to 92%]) (9, 10, 23, 48) were more commonly men. Most persons who completed suicide were white (range, 78% to 98%) (6, 8, 10, 1619, 21, 23, 26), whereas most homicide victims were non-Hispanic black or another race (range, 47% to 88%) (9, 10, 23, 48). Four (28.6%) of the 14 suicide studies were among adolescents only (6, 8, 16, 17), and 10 (71.4%) were among adults only (7, 10, 1825). All studies of outcomes of homicide victimization were among adults only (9, 10, 2224, 48). Firearm Access Among 11 U.S. casecontrol studies using survey data, proportions of firearm access ranged from 62.7% to 75.4% among case patients and from 26.4% to 50.8% among controls participants. One non-U.S. study (20) used survey data to estimate the proportion of case patients (23.9%) and control participants (18.5%) with firearm access, and another non-U.S. study (25) assumed firearm access from military duty and estimated the proportion of case patients (41%) and control participants (17%) with access. Among U.S.-based studies with reported data, the proportion of completed suicides using a firearm ranged from 47% to 73% (6, 7, 10, 16, 17, 2124); 3 studies did not report adequate data (8, 18, 19). One nonU.S.-based study of civilians reported that 13% of suicides were completed using a firearm (20), whereas another non-U.S. study of military personnel reported that 52% of suicides were completed using a firearm (25). The proportion of homicides using a firearm ranged from 50% to 76% (13, 15, 2729). Studies of Suicide Eleven of 14 studies (78.6%) interviewed proxies to determine firearm accessibility among decedents or control participants (68, 10, 1621, 23), whereas 3 studies (21.4%) used firearm purchase records or military duty to determine accessibility among decedents or control participants (22, 24, 25) (Table 2). Twelve studies (85.7%) defined suicide as self-inflicted, intentional death by any means (6, 7, 10, 1623, 25), whereas 2 studies (14.3%) defined suicide as injury related only to firearms or firearm- or violence-related injury (8, 24). All suicides were reported consecutively or identified using death certificates. In casecontrol studies, various types of control participants were identified, such as inpatients who attempted suicide (14.3%) (16, 17), community or school control participants (42.9%) (68, 18, 20, 21), decedents from causes other than suicide (28.6%) (18, 19, 24, 25), participants in a national health survey (7.1%) (10), or living HMO-based control participants (7.1%) (22). Table 2. Characteristics of Included Studies of Suicide and Homicide Victimization Studies of Homicide Victimization Three of 6 studies (50.0%) interviewed proxies to determine firearm accessibility in the home of decedents or control participants (Table 1) (9, 10, 23). Two studies (33.0%) used firearm purchase records to determine firearm acc