Ready access to firearms at home is an international phenomenon. Many nations, both developing and developed, have high levels of per capita firearm ownership.^1,2 In the US, where 41% to 49% of households own a firearm, about 75% of handgun owners and 38% of all firearm owners claim protection of themselves, their families, and their property as the main reason for storing firearms at home.^3–5 However, far from being a source of protection, access to firearms at home has been repeatedly linked with household members’ risk of violent death.^6–11 This link is particularly well reported for suicide. Ecological studies (table 1) have consistently linked state, regional, and national levels of firearm availability with firearm related suicide rates among various demographic groups^2,12–28 as well as overall suicide rates,^{2,12,18,20,21,23,24} particularly among people under age 25.^22,23,29

More convincingly, individual level studies also evince a link between firearm ownership and risk of suicide (table 2). Among adults, purchase of a handgun is linked with an increased risk of suicide, an association that can persist for five years after the purchase.^30–32 On the balance, individual level studies also suggest a positive independent association between access to firearms at home and risk of suicide among adults. While two of these studies^33,34 found firearm ownership to be independent of risk of suicide, three regional and two national studies have linked household access to firearms with the risk of suicide.^7,8,35–37 These findings among adults are further corroborated by studies of adolescents that have uniformly found a positive association between access to firearms at home and risk of suicide.^38–44

Even more convincingly, individual level studies that have specifically examined firearm related suicides uniformly report a positive association between household access to firearms and risk of suicide by firearms. Regional studies have found that adults’ risk of suicide by firearm is 3 to 12 times higher among handgun purchasers as compared with controls.^30–32 Similarly, adolescents with access to firearms at home are four times as likely to commit suicide by firearm than other adolescents.³⁹ Most persuasively, in a nationally representative sample, household access to firearms was associated with a 17-fold increase in the risk of suicide by firearm relative to those without such access.⁸

Despite the ubiquity of firearms in several nations across the globe and the abundant evidence linking access to firearms at home with the risk of violent death—particularly suicide by firearm—information on factors that may protect against household members’ risk of suicide by firearm (for example, placement of lock triggers) is conspicuously sparse, and the sum of the available information is inconclusive. Seven studies have examined the link between risk of suicide and the manner in which firearm(s) are stored (table 3). In two studies of adult suicide victims and community controls, unsafe firearm storage independently predicted suicide.^36,37 In a study comparing seven adolescent suicide victims with no apparent psychopathology to 60 suicide victims with a psychiatric disorder and 38 community controls with no disorder, the presence of a loaded firearm in the home was significantly associated with suicide completion.⁴³ In four other studies, risk of suicide among adolescents was independent of the manner in which firearms were stored in their residences.^38,39,41,42 While these studies provide useful preliminary information, they are limited by their focus on small geographical areas, small sample sizes, and in some instances, under-representation of racial and ethnic minorities (see Miller and Hemenway⁴⁵ and Brent⁴⁶ for reviews). These studies are further compromised by not considering study participants’ intention to die.

Here we present the first study to address these shortcomings by using a nationally representative sample to estimate the protective effect of storing firearms locked and/or unloaded. We greatly, but not completely, reduce the bias introduced when intention to die is not considered by including in the analyses proxies for intention to die. Furthermore, our focus is on suicides that are likely to have been impulsive. There is compelling evidence that a considerable proportion of suicides are impulsive acts. Internationally, 24% to 53% of suicides are contemplated for as little as five minutes.^47–51 Clearly, any measure that can prolong the period between the initial decision to commit suicide and the suicidal act can potentially decrease the fleeting intention to die and possibly prevent the suicidal behaviour. Reducing ease of access to firearms through safer storage practices is potentially one such method. Our aim is to estimate the protective effect of safer firearm storage practices (that is, storing firearms locked and/or unloaded) on the risk of suicide by firearm among people with relatively low intention to die.

METHODS

Sample

We use data from the 1993 National Mortality Followback Survey (NMFS⁵²). This is a survey of social, medical, and lifestyle practices of the decedent as reported by the next of kin (hereafter, informant). The 1993 NMFS is based on a sample of 22 957 deaths representing 2 215 000 people. A 10% random sample of death certificates from all states in the union (except South Dakota), and independent vital registration areas of the District of Columbia and New York City were drawn from the 1993 Current Mortality Sample. Informants were identified either on the death certificate as having provided information or were identified by the funeral director. When an informant was not identified in this manner or could not be located, another person familiar with the decedent’s life history was recruited by personalised letter. Eighty three per cent of informants participated in the survey.

To reduce the likelihood of misinformation provided by respondents not well acquainted with the decedent, we included in the analyses only information reported by the spouse (38%), parent (32%), sibling (14%), or offspring (14%). As an added precaution, we further focused the study sample to decedents who had resided with their informant for at least a year. To focus on impulsive suicides, we further focused our study sample to decedents who, according to the informant, had not expressed a wish to die during the last month of life. Finally, the study sample included only records with complete information on all variables of interest.

Recognising that there may be residual confounding by intention to die, we identified two groups in addition to the study sample just described. The first group is comprised of decedents who did not have any of the three strongest correlates of suicide (other than demographics) in the study sample:(1) consumed alcohol in the last year of life, (2) experienced an employment change in the last year of life, and (3) visited a mental health professional in the last year of life. The second group is comprised of only suicide victims. We reasoned that, on average, compared with the study sample, intention to die is relatively lower among decedents who did not have any of the three strongest correlates of suicide, and relatively higher among those who died of suicide. Thus, for this analysis, we make the simplifying assumption that if the risk associated with unsafe firearm storage is highest among those with the highest intention to die, and lowest among those with the lowest intention to die, then we have reasonably succeeded to group decedents by, and partially control for, their intention to die within limitations imposed by the data.

Measurement

The dependent variable in this study is an indicator of whether the decedent committed suicide with a firearm according to the cause of death stated on the death certificate. Use of firearms in the suicide was determined based on the following ICD-9 codes for external cause of injury (E-codes): E955.0 (suicide by handgun) and E955.1–E955.4 (suicide by all other and unspecified firearms)*. The primary exposures of interest are whether the decedent kept any firearm(s) in or around home during the last year of life, and whether the firearms were stored unloaded and/or locked in the decedent’s home. We consider firearms stored in a locked drawer, cabinet, or closet and those stored with a trigger lock or other locking mechanism as locked, and disassembled firearms to be unloaded.

The regression models include the following variables that refer to the last year of life, except where, as will be noted, the question refers to the last month of life. (1) Sociodemographic characteristics—gender, age, education level, and race/ethnicity^53–57; US region of residence,¹³(2) Correlates of mental wellbeing and stress—three categorical variables reflecting the frequency of religious activities,^58–61 moderate and/or vigorous physical activities (for example, heavy housework, jogging,),^62,63 and interaction with family or friends.^61,64 Also included were indicators of whether or not the decedent consumed alcohol^35,57,65; experienced a demotion, job loss, quit a job, or retired^66–68; avoided or refused needed health care^69,70; or saw a mental health professional.^35,57,71 Categorical variables reflecting the number of depressive symptoms during the last month of life^35,71–74 were also included. Up to eight depressive symptoms were endorsed, allowing a possible range of 0–8; the α reliability for this scale is 0.77.

Analysis

Using logistic regression, we first estimated the risk of suicide by firearm associated with having access to firearm(s) at home, and then estimated the protective effect of safer storage practices among decedents with relatively low, medium, and high intention to die. All covariates deemed relevant based on our review of the literature were simultaneously included in the models. Observations are weighted to account for the study design and for non-response of proxy respondents^75,76 using sudaan (version 8.0.1, Research Triangle Institute, Research Triangle Park, NC) and the weights provided with the NMFS.⁵²

RESULTS

The study sample consists of 4996 decedents representing 721 436 people aged 15 and older. Compared with the original NMFS sample, decedents in our study sample are slightly younger (71 years v 72 years) and somewhat over-representative of males (57%v 51%), decedents with a college education (26%v 23%), non-Hispanic whites (87%v 85%), and residents of the North east (17%v 21%). Midwestern residents are under-represented (30%v 24%).

Table 4 shows that relative to those who perished because of other causes, decedents who committed suicide by firearm were significantly more likely to be younger (50 years v 71 years), male (87%v 57%), and Hispanic (6%v 2%). They were more likely to have resided in the south (49%v 34%) and less likely to have resided in the north east (10%v 17%). Decedents who committed suicide by firearm were more likely to have: never engaged in religious activities (57%v 37%); exercised frequently (55%v 35%); consumed alcohol (59%v 24%); experienced an employment change (17%v 2%); avoided or refused needed health care (23%v 14%); exhibited fewer depressive symptoms (32%v 23%, no symptoms), and visited a mental health professional in the past year (14%v 5%). They were also more likely than others to have stored a firearm in/around their residence (91%v 35%) and were less likely to keep it unloaded (62%v 81%).

Multivariate analysis of the study sample shows that (table 5(A)), risk of committing suicide varies by age and gender, but is independent of level of education and race/ethnicity. Other predictors of suicide by firearm include decedents’ alcohol consumption, employment change, and visit(s) to a mental health professional during the last year of life. Among decedents with access to firearms (table 5(B)), similar associations were found between the risk of suicide by firearm and alcohol consumption, employment change, and visit(s) to mental health professionals. Analysis of the original NMFS sample (analyses not shown) yield similar effect sizes.

Table 6 shows the risk of suicide by firearm relative to death from other causes among decedents in the study sample, as well as among the two groups with relatively lower and higher intention to die. The risk of suicide by firearms is highest among those with the highest intention to die (that is, subsample consisting entirely of suicide victims; OR = 36.7, 95% CI = 14.2 to 95.0), medium among those with relatively moderate intention to die (that is, study sample, OR = 17.7, 95% CI = 10.2 to 30.9), and lowest among those with the lowest intention to die (that is, subsample excluding those with any of the three strongest correlates of suicide; OR = 14.0, 95% CI = 5.7 to 34.3).

The protective effect of safer firearm storage is strongest among those with lowest relative intention to die (locked: OR = 0.18, 95% CI = 0.06 to 0.49; unloaded: OR = 0.22, 95% CI = 0.08 to 0.57) and medium among those with relatively moderate intention to die (locked: OR = 0.39, 95% CI = 0.24 to 0.66; unloaded: OR = 0.30, 95% CI = 0.18 to 0.49). Among those with the strongest intention to die, there is no association between safer firearm storage and risk of suicide by firearm.

DISCUSSION

In line with previous research (for example, Krug et al,⁵³ Blakely et al,⁷⁷ Shenassa et al⁷⁸), we found an increased risk of suicide among males, more frequent drinkers, and persons with recent employment changes. Also in accord with previous works,^{7,8,30–32,36–44} we found that household access to firearms is associated with an increased risk of suicide by firearm. Decedents with access to firearms at home were 18 times as likely to commit suicide by a firearm than to die from other causes. This points to the importance of factors that may protect against the risk of suicide among populations with access to firearms at home. In this first study to analyse the protective effect of safer firearm storage practices in a nationally representative sample, we found that firearm owners who keep their firearms locked or unloaded were at least 60% less likely to die from firearm related suicide than those who store their firearms unlocked and/or loaded.

These findings must be considered in light of the study’s limitations and strengths. One shortcoming of this study regards the assumption that household firearm ownership is a reliable proxy for firearm access. Evidence suggests that considering firearm ownership alone is likely to miss some decedents with ready access to firearms, but not necessarily at home.⁷⁹ Thus, these analyses may have underestimated the proportion of decedents with ready access to a firearm. This underestimation was probably more prevalent among decedents who did not die of a firearm related injury. This would result in a degree of overestimation of our odds ratios, as it is likely that some members of the comparison groups have had easy access to firearms. However, it is highly unlikely that the large effect size can be explained away by this potential source of misclassification.

We partially addressed one important limitation of existing literature on suicide and firearms by using proxies for decedents’ intention to die. We examined three groups that, on average, had relatively high, medium, and low intention to die. As expected, access to firearms at home was most lethal among decedents with highest intention to die, for whom safer firearm storage practices were not protective. For the group with the moderate intention to die, access to firearms at home was moderately lethal, as was the protective effect of keeping firearms locked or unloaded. The most striking results regard the group with the lowest relative intention to die. For this group, access to firearms at home is associated with relatively low risk of suicide by firearms, while the protective effect of storing firearms locked and/or unloaded was greatest. Considering that all those with a stated intention to die were excluded from the study, and considering that this group excludes those with any of the three strongest correlates of suicide, it appears that the decedents in this group had a fairly low intention to die and may have engaged in an impulsive act. This suggests that safer firearm storage may be most effective in reducing impulsive suicidal acts.

Our findings are most generalisable to populations with low to moderate intention to die. However, even among those who are highly motivated to die,^80,81 lack of access to firearms can lower suicide mortality. Even if all of those who committed suicide by firearm simply switched to the next most lethal method, still a significant reduction in suicide mortality would occur.⁷⁸

Our findings further support the utility of devices intended to reduce unauthorised or impulsive use of firearms, including various types of locks and personalisation technology,^82–87 and the need for a consumer product regulation approach to reducing firearm related injuries and fatalities.^{82–86,88–92} The importance of this approach is even more evident when considering the difficulty of changing individuals’ firearm storage practices,^82–87,93 and the failure of otherwise safety conscious people to store their firearms safely.^94–97 Although firearm training has frequently been proposed as a means of promoting safer firearm storage, information on the effectiveness of these programmes is sparse. At least two community based programmes appear to have improved firearm storage practices by distributing gun locks or safes to firearm owners.^98,99 In contradiction, a nationally representative survey of firearm owners suggests that receiving firearm training may actually predict unsafe storage practices,⁵ and at least one intervention in a paediatric primary care setting did not promote safer firearm storage.¹⁰⁰ Thus, as it is currently provided, the efficacy of firearm training in regard to firearm storage practices is questionable.

Internationally, the World Health Organisation has called for greater emphasis on passive approaches to suicide prevention.¹⁰¹ Although our findings and recommendations are most directly applicable in nations with higher levels of per capita firearm ownership, such as Australia, Finland, Germany, Guinea, New Zealand, Sweden, the United States, and Russia,^1,101,102 they also underscore the importance of policies that make unauthorised and impulsive use of firearms less likely in countries with strict gun control policies. In these countries, efforts to increase the safety of firearm storage among those who keep firearms at home, such as members of military, is still warranted.¹⁰³

In the US, legislation mandating minimum safety features on new firearms can be an essential part of the solution, one that is supported by the vast majority of American adults.¹⁰⁴ Currently, minimum safety criteria exist for handguns imported into the US.¹⁰⁵ However, the net effect of these criteria is modest given that they do not apply to firearms manufactured within the US, and they are, alone, not enough to ensure that safety features are consistently applied to new handguns.¹⁰⁵ Clearly, implementing federally mandated safety standards may be necessary to reduce the risk of violent death associated with firearm ownership. And to the extent that changes in the legal and social milieu are lasting, they are likely to reduce the suicide risk of future generations.^106,107