Methods

Data

This study used restricted data for decedents aged 18 and older from the 2003–6 NVDRS. The NVDRS is a state-based surveillance system that provides detailed accounts of violent deaths that occur in the USA. Currently, 17 states (Alaska, California, Colorado, Georgia, Kentucky, Maryland, Massachusetts, New Jersey, New Mexico, North Carolina, Oklahoma, Oregon, South Carolina, Rhode Island, Utah, Virginia and Wisconsin) contribute data to the NVDRS. It should be noted that, as only four California counties (Los Angeles, Riverside, San Francisco and Santa Clara), which account for 39% of the California population, provided data, California was excluded from the analysis.

The NVDRS includes all suicides, homicides, legal intervention deaths, unintentional firearm deaths and undetermined deaths. The data were collected from coroner/medical examiner records, police reports, death certificates, toxicology laboratories, crime laboratories and Alcohol, Tobacco, Firearms and Explosives (ATF) firearm trace reports. Some of the decedents’ information was obtained from family members and friends (proxies). Decedent variables included: demographic information; the date, time and place of injury and death; toxicology results; data on wounds; circumstances preceding the death; information on mental health and substance abuse diagnoses and treatment; and details on the suicide method.

Measures

The main outcome measure was suicide by firearm versus suicide by all other methods (including death from a sharp or blunt instrument, poisoning, hanging, falling, drowning, fire or burns, a motor vehicle, or other). The coroner or medical examiner and the death certificate determined the manner of death. The cause of death identified by the coroner or medical examiner did not always match that of the death certificate. Of all the cases that were defined as suicides, fewer than 1% (n = 200) were recorded as non-suicides on their death certificates.16 These cases were excluded from the analysis. The suicide method was unknown for fewer than 1% of the cases.

The independent variables included age, marital status, race, veteran status and region of residence at the time of death (New England, Middle Atlantic, East and West North Central, South Atlantic, East South Central, West South Central, Mountain and Pacific). In addition, mental health status, blood alcohol concentration (BAC), alcohol problems, suicidal behaviours and life events were examined for their association with firearm use among suicide decedents.

Mental health status

Proxies reported if the decedent or others perceived that he or she was depressed, sad, down, blue or unhappy shortly before the suicide was completed. In addition, proxies also provided information on whether the decedents had an American Psychiatric Association Diagnostic and Statistical Manual (DSM-IV) diagnosis at the time of death. Mental health treatment was assessed from evidence gathered at the scene or from proxies. Mental health diagnosis and treatment were recoded into three categories: received a DSM-IV diagnosis and treatment; had a DSM-IV diagnosis with no treatment; or had neither treatment nor a diagnosis.

Suicidal behaviour

Family members or friends reported whether the suicide decedent disclosed his or her intention to complete a suicide or had a history of suicide attempts.

Alcohol use

BAC, based on toxicological data, was dichotomised into <0.08% versus ⩾0.08%. The BAC was not available for 32% of the male and 30% of the female decedents. An alcohol problem, assessed from a family member or a friend, denotes whether the decedent had an alcohol problem/dependence.

Life events

Family members or friends reported whether the decedent experienced a crisis within 2 weeks of the suicide or if a crisis was imminent. In addition, they were asked if the decedent experienced any of the following events/problems: the recent death of a family member or friend, financial problems, physical health problems, job problems, relationship problems (intimate and/or not intimate), legal problems (criminal and/or non-criminal) and the suicide of a friend or relative in the past 5 years.

Analysis

Pearson χ² tests were used to compare the characteristics of those who completed suicide with a firearm with those who used other methods. We used logistic regressions to estimate the odds of using a firearm as a function of selected risk factors and precipitating circumstances. We then ran separate regressions by gender because men and women have different rates of firearm use. For both regression models, the dependent variable was set to 1 or 0, depending on whether the decedent used a firearm or not. Dummy variables were the key independent variables. Owing to the listwise deletion procedure, inclusion of the BAC in the female model led to a small cell size in some of the independent variables (ie, veteran status, region); therefore, BAC was excluded to maximise statistical power. The analysis for men was conducted using decedents with BAC data (excluding 32.3% of the cases). Analyses were performed using SPSS V14.0.

Results

Of the 25 491 suicides in 2003–6, 52.2% were completed with firearms. Firearms accounted for 58.1% and 31.2% of the suicides among men and women, respectively. The percentage of suicides involving firearms ranged from 22.8% in Massachusetts to 65.5% in South Carolina. There were significant associations with the demographic, health and mental health variables.

As table 1 shows, among the male suicide decedents, firearms were the most common method used among older people, married people, white people, veterans and those who lived in the South Atlantic and Pacific divisions. Furthermore, compared with non-firearm decedents, firearm users were also more likely to have: had a BAC ⩾0.08% at the time of their death, an acute crisis during the previous week, physical health problems, and experienced the death of a relative or friend, but were less likely to be Hispanic; received a mental health diagnosis and treatment; had a history of suicide attempts, an alcohol problem, a legal problem, or a job problem.

Among women, there were also a number of significant differences between firearm and non-firearm suicide decedents (table 1). Female firearm decedents were more likely to be married, white and veterans and to live in the South Atlantic and Pacific regions. Higher percentages of female firearm decedents were reported to have had an acute crisis during the previous week, relationship problems and financial problems, and to have experienced the death (suicide or non-suicide) of a relative or friend. However, lower percentages of female firearm decedents were reported to have received a mental health diagnosis and treatment and to have had a history of suicide attempts and physical health problems.

Table 2 presents the adjusted odds ratios with 95% CIs for the use of firearms versus other means of suicide. All the independent variables were entered simultaneously into gender-stratified multivariate logistic models. Several variables were significantly different between the firearm and non-firearm users. In particular, among the male decedents, being older and a veteran, residing in regions with higher rates of firearm availability (outside the Northeast region of the USA), and having a BAC of ⩾0.08% at the time of death, an acute crisis during the previous weeks, physical health problems and relationship problems were all associated with higher odds of using firearms. Conversely, men who were American Indian or Hispanic or who had a DSM-IV diagnosis with treatment, a history of suicide attempts, a legal problem or alcohol problems had lower odds of using firearms to complete suicide.

Among the female decedents, the factors that significantly affected the use of firearms included: being older, married and a veteran; living outside the Northeast; having an acute crisis during the previous week; having experienced the death of a relative or friend; being depressed; and having relationship problems. Women who had received a DSM-IV diagnosis and treatment, had previously attempted suicide, or had physical health problems were less likely to use firearms.

Discussion

The choice of a firearm as a suicide method appears to be precipitated by stressful life events. Recent crises and relationship problems were also associated with firearm suicide for men. Older people were particularly disposed to use firearms to complete suicide. A surprising and noteworthy finding is that firearm suicide decedents did not appear to have longstanding mental health or substance abuse problems for which they sought help. This finding is consistent with a study involving the National Mortality Followback Survey.23 Joe et al23 found an inverse relationship between treatment for a mental health problem, including antidepressant drug use, and suicide by firearms. This inverse relationship may be due to the fact that people with a history of mental illness have less access to firearms. Several studies have shown that people with a history of mental illness were less likely to possess firearms.242526

According to our findings, male decedents who chose firearms to complete suicide consumed a large amount of alcohol to the point of impairment shortly before their death, but were not likely to have a major or ongoing depressive illness and/or experience alcohol dependence. This finding concurs with those of earlier studies on firearm suicide conducted in Canada27 and the UK.28 Alcohol intoxication may have acted as an agent of emotional disinhibition by fostering impulsive behaviour that facilitated the use of firearms among suicidal persons.2930 There was no apparent connection between alcohol use and the choice of suicide method among women.

Suicidal individuals who are contemplating using a firearm appear unlikely to have had contact with mental healthcare providers. Therefore, interventions that are aimed at preventing or deterring firearm suicide need to be established outside the purview of providers of mental healthcare, including emergency medicine, internal medicine and paediatric providers who are caring for these patients.

The findings on the odds of firearm use appear to follow the pattern of household possession of firearms. The likelihood of using firearms as a suicide method varies substantially by geographic region in the USA and parallels patterns of household possession of firearms as measured by the Behavioural Risk Factor Surveillance System (BRFSS).31 The correlational analysis between BRFSS firearm prevalence data and NVDRS data on the proportion of suicides involving firearms yielded significant coefficients for men (r = 0.94, p<0.001) and women (r = 0.85, p<0.01). These findings lend support to the hypothesis that geographic location seems to be an important factor in the choice of particular suicide methods.6283233

Major strengths of this study were the use of the NVDRS and the ability to explore at the individual level an array of potential suicide-related risk factors and precipitating circumstances obtained from multiple sources in a large, population-based sample.3435 The NVDRS also has important limitations. Firstly, data on the circumstances preceding death were not routinely collected by the participating states. Secondly, the accuracy and completeness of proxy-derived information may lead to misclassifications or under-reporting. However, previous studies have shown that information obtained from informants may be sufficiently valid and reliable for analytical purposes.3637 Thirdly, ATF trace data were available for only a small number of suicide decedents (5%). Finally, there was an apparent lack of uniformity in forensic investigations conducted across participating jurisdictions.1519

Conclusions

It is extremely difficult to identify potential firearm suicide attempters. The precursors of firearm use are complex and defy conventional views about suicidal behaviour. Our analysis of the NVDRS revealed that firearm suicide could result from many factors other than mental illness. Rather, suicide with firearms appears far more unpredictable and impulsive.

Suicide by firearms in general appears to be chosen more often by older people, veterans, those who have experienced recent life events, and men with physical health problems. Another important finding is the effects of geographic location on the likelihood of using firearms for suicide among men and women after controlling for potential confounding factors. As the results show, living in the East South Central and West South Central regions seems to exert a stronger influence on the choice of firearms as a suicide method. It is interesting that the proportion of suicides completed with a firearm is a surrogate for the availability of firearms. Regions with the highest proportion of firearm-related suicides also had the highest rates of access to firearms, as measured by the proportion of households owning firearms.3813

Reducing access to firearms through policy initiatives or by removing guns from high-risk individuals could reduce suicide rates or force attempters to switch to means that are typically less lethal.13 However, our results underscore the difficulty of screening for potential suicidal behaviour involving firearms. As the data demonstrate, many firearm suicide decedents did not exhibit classic psychiatric markers of suicidality (ie, more acute psychopathology). Equally important, the use of highly lethal means provides a limited window for rescue and psychiatric treatment.39 Furthermore, once patients have been identified as suicidal, Kaplan et al40 found that in only half the instances would physicians inquire whether or not the patient had access to a firearm. Universal or more targeted restriction of highly lethal suicide means must be front and centre in any effort to prevent or mitigate this human tragedy. As recently noted, “restricting access to lethal methods decreases suicide by those methods” and “where the method is common, restriction of means has led to lower overall suicide rates”.21