- Email: [email protected]
Prior suicide attempts are less common in suicide decedents who died by firearms relative to those who died by other means
Journal of Affective Disorders 189 (2016) 106–109
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Short communication
Prior suicide attempts are less common in suicide decedents who died by firearms relative to those who died by other means Michael D. Anestis n University of Southern Mississippi, Hattiesburg, MS 39401, United States
art ic l e i nf o
a b s t r a c t
Article history: Received 25 June 2015 Received in revised form 24 August 2015 Accepted 5 September 2015 Available online 26 September 2015
Background: Suicide prevention efforts often center on the identification of risk factors (e.g., prior suicide attempts); however, lists of risk factors without consideration of context may prove incapable of impacting suicide rates. One contextual variable worth considering is attempt method. Methods: Utilizing data from the National Violent Death Reporting System (2005–2012), I examined suicide deaths (n ¼ 71,775) by firearms and other means to determine whether prior suicide attempts were more common in one group versus the other. Results: Significantly fewer suicide decedents who died by firearms reported a prior history of suicide attempts (12.10%) than did decedents who died by other means (28.66%). This result was further replicated within each state that contributed data to the NVDRS. Limitations: Only 17 states have contributed to the NVDRS thus far and, within those states, not all suicide deaths were reported. Due to the nature of the data, I was unable to test proposed mediators within our model. Conclusions: Suicide decedents who die by firearms may die on their first attempt more often than other decedents due to a capability and willingness to utilize a highly lethal means. Current risk assessment protocols may be ill equipped to identify such individuals prospectively on their own. Broader methods of implementing means restriction (e.g., legislation) may thus be pivotal in suicide prevention efforts. & 2015 Elsevier B.V. All rights reserved.
Keywords: Suicide Firearms Means restriction
1. . Introduction Historically, suicide prevention efforts have centered on the identification of risk factors. Many variables have been identified as risk factors for eventual death by suicide with prior suicide attempts repeatedly proving to be one of the most robust predictors (De Leo et al., 2013; Hawton, 2010) and several risk assessment frameworks and theories emphasizing attempt status as a variable substantially weighted when determining risk levels (Joiner et al., 1999; Klonsky and May, 2015; Posner et al., 2008). Recent research, however, has revealed that our ability to prospectively predict death by suicide is no better now than it was in the mid-20th century (Franklin et al., in preparation). Furthermore, suicide rates have climbed in recent years, indicating that our efforts to identify risk factors may be a suboptimal strategy, at least in isolation (Centers for Disease Control and Prevention, 2015). Consideration of context may thus prove pivotal in understanding the importance of specific variables (Klonsky and May, 2014). n
Fax: þ 1 601 266 5580. E-mail address: [email protected]
http://dx.doi.org/10.1016/j.jad.2015.09.007 0165-0327/& 2015 Elsevier B.V. All rights reserved.
One contextual factor worth considering is suicide method. Each year, more than half of suicide deaths within the US result from self-inflicted gunshot wounds despite the fact that this method is utilized in less than 5% of attempts (Centers for Disease Control and Prevention, 2015). Such findings highlight the high lethality of firearms and prior research has demonstrated that, in contrast to most other methods, firearms almost invariably result in death when utilized in a suicide attempt (Chapdelaine et al., 1991; Spicer and Miller, 2000). Research has repeatedly demonstrated that limiting access to highly lethal means results in decreased overall suicide rates (Carrington, 1999; Kreitman, 1976; Loftin et al., 1991; Nordentoft et al., 2010; Oliver and Hetzel, 1972; Sarchiapone et al., 2011) and that the notion of method substitution – seeking out another suicide method when access to one method is thwarted – is not supported by data (Daigle, 2005; Law et al., 2014; Lester and Abe, 1998). Furthermore, recent research has demonstrated that several state laws regulating access and exposure to handguns are associated with lower overall suicide rates (Anestis et al., 2015a, 2015b; Anestis and Anestis, 2015). In that work, the authors demonstrated that, in states with handgun regulation in place, the impact of the laws on overall suicide rates was largely explained by the proportion of suicide deaths resulting from firearms. Furthermore, the authors in both studies
M.D. Anestis / Journal of Affective Disorders 189 (2016) 106–109
demonstrated that implementation of such laws is prospectively associated with decreases in overall suicide rates whereas repeal is prospectively associated with increased overall suicide rates. Given the frequency with which firearms are utilized in lethal suicidal behavior within the US (Centers for Disease Control and Prevention, 2015), it is vital to more clearly understand to what extent robust predictors of future suicidal behavior overall apply to this particular method. Specifically, if fewer suicide decedents who utilized firearms in their lethal attempt had a history of prior suicide attempts than did suicide decedents who utilized other methods, this would indicate that our risk factor models are less useful in understanding who is most vulnerable to dying by suicide using firearms – the most common method of suicide death in the US (Centers for Disease Control and Prevention, 2015). To address this question, I consulted the National Violent Death Reporting System (NVDRS), which collects information on violent deaths in a selection of states. Thus far only 16 states have participated for the majority of its existence with one additional state (Ohio) joining in 2011. The NVDRS is a voluntary system, meaning that not all suicide deaths are reported; however, reports include information on the circumstances of each death (e.g., prior history of suicide attempt, presence of a suicide note). Initial data collection began in the late 1990s; however, the number of states participating spiked in 2005 and, as such, I opted to examine all reported suicide deaths with known circumstances of death from 2005 through 2012 (the most recent year available) to examine whether a lower proportion of suicide decedents who died from self-inflicted gunshot wounds reported prior suicide attempts than did suicide decedents who died by other methods. Such findings would highlight the importance of emphasizing means restriction alongside assessing general risk factors. Indeed, if prior suicidal behavior is less common among those decedents who utilized firearms in their lethal attempt, this would indicate that many risk assessment approaches are ill equipped on their own to identify the majority of individuals who will die by suicide in the US.
2. Method To compute the proportion of firearm and non-firearm suicide deaths in which the decedent had a prior history of one or more suicide attempts, I consulted the NVDRS website. For each of 16 states that had contributed data since 2005, I entered eight years of data (2005–2012), including total number of suicide deaths, total number of suicide deaths resulting from firearms (with and
107
without prior suicide attempts) and the total number of suicide deaths resulting from methods other than firearms (with and without prior suicide attempts). I also included data from Ohio, which began contributing information on violent deaths in 2011. 2.1. Data analytic plan To test whether a lower percentage of suicide decedents who died by self-inflicted gunshot wound had a prior history of one or more suicide attempts than did suicide decedents who died by other methods, I ran a series of two-sample z-tests. I first ran analyses for each individual state, summed across years. I then ran a single analysis that summed across all states and all years. Cohen's d was computed as an index of effect size.
3. Results Information on the number and nature of suicide deaths by state and overall can be found in Table 1. The extent to which suicides were recorded into the system varied by state and, as such, states with a higher actual incidence of suicides did not necessarily have the highest number of suicides recorded into the NVDRS system. Across all states and years, a total of 71,775 suicide deaths of known circumstances were reported (36,896 by firearms; 34,879 by other means). The initial results indicated that, in each state, the percentage of suicide decedents who died by firearms and who had a history of one or more suicide attempts (range¼6.61–16.23%) was significantly lower than the percentage of suicide decedents who died by other means and who had a history of one or more suicide attempts (range¼20.62–35.56%; z’s 4 4.17; p’so.001; Cohen's d’s 4.25). Furthermore, when summed across all years and all contributing states, results again supported that the percentage of suicide decedents who died by firearms and had a history of one or more suicide attempts (12.10%) was significantly lower than the percentage of suicide decedents who died by other means and had a history of one or more suicide attempts (28.66%; z ¼55.29; po .001; Cohen's d ¼ .41). These results are presented in Table 1.
4. Discussion Suicide is notoriously difficult to predict. Evidence-based risk assessment methods exist (e.g., Joiner et al. (1999) and Posner
Table 1 Descriptive data on suicide deaths (2005–2012) and the percentage of decedents by method with and without a prior history of suicide attempts. State
Suicides
Firearm
Non-firearm
% Firearm prior attempt
% Non-firearm prior attempt
z
p
Cohen's d
Alaska Colorado Georgia Kentucky Maryland Massachusetts New Jersey New Mexico North Carolina Ohio Oklahoma Oregon Rhode Island South Carolina Utah Virginia Wisconsin Total
7304 3337 3397 774 4654 4474 8715 4461 3833 1126 3739 3700 3289 6267 6205 1143 5357 71,775
4081 1742 2093 196 2493 2732 4826 2391 1199 278 771 1748 2159 3927 3044 727 2489 36,896
3223 1595 1304 578 2161 1742 3889 2070 2634 848 2968 1952 1130 2340 3161 416 2868 34,879
11.59% 15.10% 11.18% 10.71% 14.28% 11.31% 10.63% 12.09% 11.84% 7.55% 6.61% 11.33% 7.41% 9.98% 18.59% 16.23% 14.38% 12.10%
32.61% 34.73% 28.37% 29.24% 31.24% 27.73% 26.77% 27.44% 25.06% 23.94% 24.46% 28.28% 20.62% 22.65% 35.56% 26.44% 33.02% 28.66%
21.95 13.18 12.75 5.21 13.89 14.03 19.57 12.98 9.34 5.94 10.88 12.81 11.09 13.69 15.01 4.17 15.85 55.29
o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001
.53 .47 .45 .38 .42 .43 .43 .40 .56 .36 .36 .43 .39 .35 .39 .25 .44 .41
108
M.D. Anestis / Journal of Affective Disorders 189 (2016) 106–109
et al. (2008)); however, even when utilizing such approaches, it remains difficult to determine which individuals will make an attempt and when. Generally speaking, efforts to optimize our predictive abilities have been driven by the assessment of the presence and absence of specific risk and protective factors. A history of prior suicide attempts has repeatedly been established as one of the most robust predictors of future death by suicide (e.g., Hawton (2010)) and is thus rightfully considered a vital variable to assess with any client. The primary aim of our study was to determine whether this predictor is as relevant in suicide resulting from firearms as it is in suicide by other methods. As hypothesized, a smaller percentage of suicide decedents who died from self-inflicted gunshot wounds (12.10%) had a history of one or more suicide attempts than did suicide decedents who died by other methods (28.66%) and this result was replicated within each individual state that contributed data to the NVDRS. This series of results indicates that, for the vast majority of individuals who die by suicide using the most common method in the US, the decedent had never previously attempted suicide. In this sense, many risk assessment protocols are ill equipped on their own to identify such individuals prospectively. Efforts to utilize risk assessment protocols to diminish an individual's desire to die by suicide are pivotal. That being said, if clinicians are ill equipped to identify individuals who will ultimately utilize a gun to die by suicide, other approaches aimed at broadly making access to guns more difficult may be necessary. Given research indicating that state legislation regulating access and exposure to handguns is associated with decreases in overall suicide rates (Anestis et al., 2015a; Anestis and Anestis, 2015), one potential avenue for addressing this point may be legislative efforts. When considering such responses, it is vital to understand the theoretical and empirical rationale underlying the proposition. Prominent theories of suicide (Joiner, 2005; Klonsky and May, 2015) propose that, in addition to the desire for suicide, an individual must be capable of lethal self-harm for serious or lethal suicidal behavior to occur. The principle tenets of such theories have now been supported repeatedly across several samples (e.g., Anestis et al. (2015), Joiner et al. (2009) and Van Orden et al. (2010)). Klonsky and May (2015) further argue that access to and familiarity with lethal means (e.g., handguns) is a component of the capacity for suicidal behavior. In this sense, individuals who have more exposure and access to guns may be more likely to translate suicidal ideation into death by suicide, whereas most suicidal individuals will either not attempt at all or utilize less lethal means. Interventions aimed only at suicidal desire and which do not impact the capability for suicide thus leave clinicians only able to address one component of risk and are largely reliant upon an individual asking for help. If greater access to and familiarity with firearms makes an individual more capable of acting upon suicidal desire, this would explain why such a smaller proportion of suicide decedents who died from self-inflicted gunshot wounds had prior suicide attempts. Simply put, they were capable of dying during their first attempt because they were willing and able (due to capacity and access) to utilize a highly lethal means, whereas most individuals with elevated suicide desire lack this capacity and many lack access to firearms. The fact that our full sample results were replicated within each individual state that participated in the NVDRS supports the notion that this finding is representative of a consistent difference between those who die from gunshot wounds and those who die by other means. An important point to emphasize within the context of these results is that, although only a small minority of suicide decedents engage in non-lethal attempts prior to their lethal attempt, prior attempts are nonetheless one of the most frequently present risk
factors in suicide decedents. As such, these results do not indicate that prior attempts are not an important consideration in determining risk or that this variable should be given less weight than it currently is in accepted assessment approaches. Indeed, past attempts are undoubtedly useful in predicting other clinically relevant outcomes (e.g., future non-lethal attempts), so any move to de-emphasize prior attempts would likely cause substantially more harm than good. Instead, these results highlight the notion that simply identifying and weighting risk factors in the absence of a clear understanding of context represents a suboptimal risk assessment approach. These results indicate that access to lethal means may be useful to assess in all clients regardless of their reason for presenting for mental health care, and that restricting access to such means should be considered at what might be considered a low risk profile in some assessment approaches (e.g., individuals with active suicidal ideation, but no prior attempts and a denial – although not necessarily absence – of a suicide plan). It is also worth noting that the proportion of non-firearm suicide decedents with prior attempts was lower than I anticipated. In this sense, these results highlight the point that contextual factors are not only an important consideration with respect to potential death by suicide resulting from self-inflicted gunshot wounds. Indeed, this unexpected finding may provide further insight into the recent findings of Franklin et al. (in preparation), who noted our general lack of improvement in the prospective prediction of suicide over the past half century. Given the low percentage of decedents across methods (firearms and non-firearms) endorsing prior attempts and prior research indicating that approximately half of suicide decedents have engaged in a nonlethal attempt prior to their death by suicide (Isometsa and Lonnqvist, 1998), it is also entirely possible that the NVDRS is not fully capturing this variable. It is highly unlikely that the NVDRS is capturing the variable in a way that systematically impacts firearm deaths more than non-firearm deaths; however, the percentages reported in this paper should also be considered within the context of those reported in the broader literature. When considering these results, it is important to note limitations. Only a minority of states has participated in the NVDRS thus far and, within those that have participated, not every suicide has been reported. As such, it is unclear that our findings would generalize beyond this sample. Secondly, due to the nature of the variables considered in our model, I was unable to experimentally manipulate them. As such, I could not directly test the effect of individual difference variables on method choice in suicide. Lastly, variables highly relevant to the proposed model (e.g., acquired capability for suicide) were not assessed in the NVDRS and, as such, their roles as moderators or mediators could not be tested. All of this considered, however, I would also note several strengths of the study. By using eight years of data across 17 states, I was able to examine more than 70,000 suicide deaths. Furthermore, whereas much of our understanding of lethal suicidal behavior has been derived from studies that considered only survivors of non-lethal attempts, this sample was directly relevant to the specific research questions addressed in our analyses. Additionally, by using publicly available data, the transparency of the analyses is optimized, allowing for others to easily replicate and expand upon the findings. Lastly, although the NVDRS did not sample the entire country, it did include states that differ widely on their geographic location, racial/ethnic makeup, climate, economic health, and culture, thereby increasing the generalizability of the findings. Ultimately, I believe these findings highlight an important limitation in current risk assessment models and represent further evidence of the importance of supplementing interventions by clinicians with interventions designed to address access and exposure to firearms on a broader level (e.g., legislation).
M.D. Anestis / Journal of Affective Disorders 189 (2016) 106–109
References Anestis, M.D., Anestis, J.C., 2015. Suicide rates and state laws regulating access and exposure to handguns. Am. J. Public Health 105, 2049–2058 10.2105/AJPH.2015. 302753. Anestis, M.D., Khazem, L.R., Law, K.C., Houtsma, C., LeTard, R., Moberg, F., Martin, R., 2015. The association between state laws regulating handgun ownership and statewide suicide rates. Am. J. Public Health 105, 2059–2067. http://dx.doi.org/ 10.2105/AJPH.2014.302465. Anestis, M.D., Khazem, L.R., Mohn, R.S., Green, B.A., 2015. Testing the main hypotheses of the interpersonal-psychological theory of suicidal behavior in a large diverse sample of United States military personnel. Compr. Psychiatry 60, 78–95. Carrington, P.J., 1999. Gender, gun control, suicide and homicide in Canada. Arch. Suicide Res. 5, 71–75. http://dx.doi.org/10.1080/13811119908258316. Centers for Disease Control and Prevention, 2015. Web-Based Injury Statistics Query and Reporting System (WISQARS). Available at 〈http://www.cdc.gov/in jury/wisqars/index.html〉. (accessed 23.06.15). Chapdelaine, A., Samson, E., Kimberly, M.D., Viau, L., 1991. Firearm-related injuries in Canada: issues for prevention. CMAJ: Can. Med. Assoc. J. 145, 1218–1223. Daigle, M.S., 2005. Suicide prevention through means restriction: assessing the risk of substitution. Accid. Anal. Prev. 37, 625–632. http://dx.doi.org/10.1016/j. aap.2005.03.00. De Leo, D., Draper, B.M., Snowdon, J., Kolves, K., 2013. Suicide in older adults: a case-control psychological autopsy study in Australia. J. Psychiatr. Res. 47, 980–988. Franklin, J.C., Riberio, J.D., Fox, K.R., Bentley, K.H., Kleiman, E.M., Jaroszewski, A.C., Chang, B., Nock, M.K., 2015. Risk Factors for Suicidal Thoughts and Behaviors: A Meta-Analysis of 50 Years of Research, in preparation. Hawton, K., 2010. Completed suicide after attempted suicide. Br. Med. J. 341, c3064. Isometsa, E.R., Lonnqvist, J.K., 1998. Suicide attempts preceding completed suicide. Br. J. Psychiatry 173, 531–535. Joiner, T.E., 2005. Why People Die by Suicide. Harvard University Press, Cambridge, MA. Joiner, T.E., Van Orden, K.A., Witte, T.K., Selby, E.A., Riberio, J.D., Lewis, R., Rudd, M. D., 2009. Main predictions of the interpersonal-psychological theory of suicidal behavior: empirical tests in two samples of young adults. J. Abnorm. Psychol.
109
118, 634–646. Joiner, T.E., Walker, R.L., Rudd, M.D., Jobes, D.A., 1999. Scientizing and routinizing the assessment of suicidality in outpatient practice. Prof. Psychol.: Res. Pract. 30, 447–453. http://dx.doi.org/10.1037//0735-7028.30.5.447. Klonsky, E.D., May, A.M., 2015. The three-step theory (3ST): a new theory of suicide rooted in the "ideation to action" framework. Int.J.Cogn.Ther 8, 114–129 10.1521/ijct.2015.8.2.114. Klonsky, E.D., May, A.M., 2014. Differentiating suicide attempters from suicide ideators: a critical frontier for suicidology research. Suicide Life-Threat. Behav. 44, 1–5. Kreitman, N., 1976. The coal gas story. United Kingdom suicide rates, 1960–71. Br. J. Prevent. Soc. Med. 30, 86–93. http://dx.doi.org/10.1136/jech.30.2.86. Law, C., Sveticic, J., de Leo, D., 2014. Restricting access to a suicide hotspot does not shift the problem to another location. An experiment of two river bridges in Brisbane, Australia. Aust. N. Z. J. Public Health 38, 134–138. Lester, D., Abe, K., 1998. The suicide rate by each method in Japan: a test of Durkheim’s theory of suicide. Arch. Suicide Res. 4 (3), 281–285, doi: 0.1080/ 13811119808258302. Loftin, C., McDowall, D., Wiersema, B., Cottey, T.J., 1991. Effects of restrictive licensing of handguns on homicide and suicide in the District of Columbia. N. Engl. J. Med. 323, 1615–1620. http://dx.doi.org/10.1056/ NEJM199112053252305. Nordentoft, M., Qin, P., Helweg-Larsen, K., Juel, K., 2010. Restrictions in means for suicide: an effective tool in preventing suicide: the Danish experience. Suicide Life-Threat. Behav. 37, 688–697. Oliver, R.G., Hetzel, B.S., 1972. Rise and fall of suicide rates in Australia: relation to sedative availability. Med. J. Aust. 2, 919–923. Posner, K., Brent, D., Lucas, C., Gould, M., Stanley, B., Brown, G., Fisher, P., Zelazny, J., Burke, A., Oquendo, M., Mann, J., 2008. Columbia Suicide Severity Rating Scale (C-SSRS). The Research Foundation for Mental Hygiene, Inc, New York, NY. Sarchiapone, M., Mandelli, L., Iosue, M., Andrisano, C., Roy, A., 2011. Controlling access to suicide means. Int. J. Environ. Res. Public Health 8, 4550–4562. Spicer, R.S., Miller, T.R., 2000. Suicide acts in 8 states: incidence and case fatality rates by demographics and method. Am. J. Public Health 90, 1885–1891. http://dx.doi.org/10.2105/AJPH.90.12.1885. Van Orden, K.A., Witte, T.K., Cukrowicz, K.C., Braithwaite, S.R., Selby, E.A., Joiner, T. E., 2010. The interpersonal theory of suicide. Psychol. Rev. 117, 575–600.
Contents lists available at ScienceDirect
Journal of Affective Disorders journal homepage: www.elsevier.com/locate/jad
Short communication
Prior suicide attempts are less common in suicide decedents who died by firearms relative to those who died by other means Michael D. Anestis n University of Southern Mississippi, Hattiesburg, MS 39401, United States
art ic l e i nf o
a b s t r a c t
Article history: Received 25 June 2015 Received in revised form 24 August 2015 Accepted 5 September 2015 Available online 26 September 2015
Background: Suicide prevention efforts often center on the identification of risk factors (e.g., prior suicide attempts); however, lists of risk factors without consideration of context may prove incapable of impacting suicide rates. One contextual variable worth considering is attempt method. Methods: Utilizing data from the National Violent Death Reporting System (2005–2012), I examined suicide deaths (n ¼ 71,775) by firearms and other means to determine whether prior suicide attempts were more common in one group versus the other. Results: Significantly fewer suicide decedents who died by firearms reported a prior history of suicide attempts (12.10%) than did decedents who died by other means (28.66%). This result was further replicated within each state that contributed data to the NVDRS. Limitations: Only 17 states have contributed to the NVDRS thus far and, within those states, not all suicide deaths were reported. Due to the nature of the data, I was unable to test proposed mediators within our model. Conclusions: Suicide decedents who die by firearms may die on their first attempt more often than other decedents due to a capability and willingness to utilize a highly lethal means. Current risk assessment protocols may be ill equipped to identify such individuals prospectively on their own. Broader methods of implementing means restriction (e.g., legislation) may thus be pivotal in suicide prevention efforts. & 2015 Elsevier B.V. All rights reserved.
Keywords: Suicide Firearms Means restriction
1. . Introduction Historically, suicide prevention efforts have centered on the identification of risk factors. Many variables have been identified as risk factors for eventual death by suicide with prior suicide attempts repeatedly proving to be one of the most robust predictors (De Leo et al., 2013; Hawton, 2010) and several risk assessment frameworks and theories emphasizing attempt status as a variable substantially weighted when determining risk levels (Joiner et al., 1999; Klonsky and May, 2015; Posner et al., 2008). Recent research, however, has revealed that our ability to prospectively predict death by suicide is no better now than it was in the mid-20th century (Franklin et al., in preparation). Furthermore, suicide rates have climbed in recent years, indicating that our efforts to identify risk factors may be a suboptimal strategy, at least in isolation (Centers for Disease Control and Prevention, 2015). Consideration of context may thus prove pivotal in understanding the importance of specific variables (Klonsky and May, 2014). n
Fax: þ 1 601 266 5580. E-mail address: [email protected]
http://dx.doi.org/10.1016/j.jad.2015.09.007 0165-0327/& 2015 Elsevier B.V. All rights reserved.
One contextual factor worth considering is suicide method. Each year, more than half of suicide deaths within the US result from self-inflicted gunshot wounds despite the fact that this method is utilized in less than 5% of attempts (Centers for Disease Control and Prevention, 2015). Such findings highlight the high lethality of firearms and prior research has demonstrated that, in contrast to most other methods, firearms almost invariably result in death when utilized in a suicide attempt (Chapdelaine et al., 1991; Spicer and Miller, 2000). Research has repeatedly demonstrated that limiting access to highly lethal means results in decreased overall suicide rates (Carrington, 1999; Kreitman, 1976; Loftin et al., 1991; Nordentoft et al., 2010; Oliver and Hetzel, 1972; Sarchiapone et al., 2011) and that the notion of method substitution – seeking out another suicide method when access to one method is thwarted – is not supported by data (Daigle, 2005; Law et al., 2014; Lester and Abe, 1998). Furthermore, recent research has demonstrated that several state laws regulating access and exposure to handguns are associated with lower overall suicide rates (Anestis et al., 2015a, 2015b; Anestis and Anestis, 2015). In that work, the authors demonstrated that, in states with handgun regulation in place, the impact of the laws on overall suicide rates was largely explained by the proportion of suicide deaths resulting from firearms. Furthermore, the authors in both studies
M.D. Anestis / Journal of Affective Disorders 189 (2016) 106–109
demonstrated that implementation of such laws is prospectively associated with decreases in overall suicide rates whereas repeal is prospectively associated with increased overall suicide rates. Given the frequency with which firearms are utilized in lethal suicidal behavior within the US (Centers for Disease Control and Prevention, 2015), it is vital to more clearly understand to what extent robust predictors of future suicidal behavior overall apply to this particular method. Specifically, if fewer suicide decedents who utilized firearms in their lethal attempt had a history of prior suicide attempts than did suicide decedents who utilized other methods, this would indicate that our risk factor models are less useful in understanding who is most vulnerable to dying by suicide using firearms – the most common method of suicide death in the US (Centers for Disease Control and Prevention, 2015). To address this question, I consulted the National Violent Death Reporting System (NVDRS), which collects information on violent deaths in a selection of states. Thus far only 16 states have participated for the majority of its existence with one additional state (Ohio) joining in 2011. The NVDRS is a voluntary system, meaning that not all suicide deaths are reported; however, reports include information on the circumstances of each death (e.g., prior history of suicide attempt, presence of a suicide note). Initial data collection began in the late 1990s; however, the number of states participating spiked in 2005 and, as such, I opted to examine all reported suicide deaths with known circumstances of death from 2005 through 2012 (the most recent year available) to examine whether a lower proportion of suicide decedents who died from self-inflicted gunshot wounds reported prior suicide attempts than did suicide decedents who died by other methods. Such findings would highlight the importance of emphasizing means restriction alongside assessing general risk factors. Indeed, if prior suicidal behavior is less common among those decedents who utilized firearms in their lethal attempt, this would indicate that many risk assessment approaches are ill equipped on their own to identify the majority of individuals who will die by suicide in the US.
2. Method To compute the proportion of firearm and non-firearm suicide deaths in which the decedent had a prior history of one or more suicide attempts, I consulted the NVDRS website. For each of 16 states that had contributed data since 2005, I entered eight years of data (2005–2012), including total number of suicide deaths, total number of suicide deaths resulting from firearms (with and
107
without prior suicide attempts) and the total number of suicide deaths resulting from methods other than firearms (with and without prior suicide attempts). I also included data from Ohio, which began contributing information on violent deaths in 2011. 2.1. Data analytic plan To test whether a lower percentage of suicide decedents who died by self-inflicted gunshot wound had a prior history of one or more suicide attempts than did suicide decedents who died by other methods, I ran a series of two-sample z-tests. I first ran analyses for each individual state, summed across years. I then ran a single analysis that summed across all states and all years. Cohen's d was computed as an index of effect size.
3. Results Information on the number and nature of suicide deaths by state and overall can be found in Table 1. The extent to which suicides were recorded into the system varied by state and, as such, states with a higher actual incidence of suicides did not necessarily have the highest number of suicides recorded into the NVDRS system. Across all states and years, a total of 71,775 suicide deaths of known circumstances were reported (36,896 by firearms; 34,879 by other means). The initial results indicated that, in each state, the percentage of suicide decedents who died by firearms and who had a history of one or more suicide attempts (range¼6.61–16.23%) was significantly lower than the percentage of suicide decedents who died by other means and who had a history of one or more suicide attempts (range¼20.62–35.56%; z’s 4 4.17; p’so.001; Cohen's d’s 4.25). Furthermore, when summed across all years and all contributing states, results again supported that the percentage of suicide decedents who died by firearms and had a history of one or more suicide attempts (12.10%) was significantly lower than the percentage of suicide decedents who died by other means and had a history of one or more suicide attempts (28.66%; z ¼55.29; po .001; Cohen's d ¼ .41). These results are presented in Table 1.
4. Discussion Suicide is notoriously difficult to predict. Evidence-based risk assessment methods exist (e.g., Joiner et al. (1999) and Posner
Table 1 Descriptive data on suicide deaths (2005–2012) and the percentage of decedents by method with and without a prior history of suicide attempts. State
Suicides
Firearm
Non-firearm
% Firearm prior attempt
% Non-firearm prior attempt
z
p
Cohen's d
Alaska Colorado Georgia Kentucky Maryland Massachusetts New Jersey New Mexico North Carolina Ohio Oklahoma Oregon Rhode Island South Carolina Utah Virginia Wisconsin Total
7304 3337 3397 774 4654 4474 8715 4461 3833 1126 3739 3700 3289 6267 6205 1143 5357 71,775
4081 1742 2093 196 2493 2732 4826 2391 1199 278 771 1748 2159 3927 3044 727 2489 36,896
3223 1595 1304 578 2161 1742 3889 2070 2634 848 2968 1952 1130 2340 3161 416 2868 34,879
11.59% 15.10% 11.18% 10.71% 14.28% 11.31% 10.63% 12.09% 11.84% 7.55% 6.61% 11.33% 7.41% 9.98% 18.59% 16.23% 14.38% 12.10%
32.61% 34.73% 28.37% 29.24% 31.24% 27.73% 26.77% 27.44% 25.06% 23.94% 24.46% 28.28% 20.62% 22.65% 35.56% 26.44% 33.02% 28.66%
21.95 13.18 12.75 5.21 13.89 14.03 19.57 12.98 9.34 5.94 10.88 12.81 11.09 13.69 15.01 4.17 15.85 55.29
o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001 o.001
.53 .47 .45 .38 .42 .43 .43 .40 .56 .36 .36 .43 .39 .35 .39 .25 .44 .41
108
M.D. Anestis / Journal of Affective Disorders 189 (2016) 106–109
et al. (2008)); however, even when utilizing such approaches, it remains difficult to determine which individuals will make an attempt and when. Generally speaking, efforts to optimize our predictive abilities have been driven by the assessment of the presence and absence of specific risk and protective factors. A history of prior suicide attempts has repeatedly been established as one of the most robust predictors of future death by suicide (e.g., Hawton (2010)) and is thus rightfully considered a vital variable to assess with any client. The primary aim of our study was to determine whether this predictor is as relevant in suicide resulting from firearms as it is in suicide by other methods. As hypothesized, a smaller percentage of suicide decedents who died from self-inflicted gunshot wounds (12.10%) had a history of one or more suicide attempts than did suicide decedents who died by other methods (28.66%) and this result was replicated within each individual state that contributed data to the NVDRS. This series of results indicates that, for the vast majority of individuals who die by suicide using the most common method in the US, the decedent had never previously attempted suicide. In this sense, many risk assessment protocols are ill equipped on their own to identify such individuals prospectively. Efforts to utilize risk assessment protocols to diminish an individual's desire to die by suicide are pivotal. That being said, if clinicians are ill equipped to identify individuals who will ultimately utilize a gun to die by suicide, other approaches aimed at broadly making access to guns more difficult may be necessary. Given research indicating that state legislation regulating access and exposure to handguns is associated with decreases in overall suicide rates (Anestis et al., 2015a; Anestis and Anestis, 2015), one potential avenue for addressing this point may be legislative efforts. When considering such responses, it is vital to understand the theoretical and empirical rationale underlying the proposition. Prominent theories of suicide (Joiner, 2005; Klonsky and May, 2015) propose that, in addition to the desire for suicide, an individual must be capable of lethal self-harm for serious or lethal suicidal behavior to occur. The principle tenets of such theories have now been supported repeatedly across several samples (e.g., Anestis et al. (2015), Joiner et al. (2009) and Van Orden et al. (2010)). Klonsky and May (2015) further argue that access to and familiarity with lethal means (e.g., handguns) is a component of the capacity for suicidal behavior. In this sense, individuals who have more exposure and access to guns may be more likely to translate suicidal ideation into death by suicide, whereas most suicidal individuals will either not attempt at all or utilize less lethal means. Interventions aimed only at suicidal desire and which do not impact the capability for suicide thus leave clinicians only able to address one component of risk and are largely reliant upon an individual asking for help. If greater access to and familiarity with firearms makes an individual more capable of acting upon suicidal desire, this would explain why such a smaller proportion of suicide decedents who died from self-inflicted gunshot wounds had prior suicide attempts. Simply put, they were capable of dying during their first attempt because they were willing and able (due to capacity and access) to utilize a highly lethal means, whereas most individuals with elevated suicide desire lack this capacity and many lack access to firearms. The fact that our full sample results were replicated within each individual state that participated in the NVDRS supports the notion that this finding is representative of a consistent difference between those who die from gunshot wounds and those who die by other means. An important point to emphasize within the context of these results is that, although only a small minority of suicide decedents engage in non-lethal attempts prior to their lethal attempt, prior attempts are nonetheless one of the most frequently present risk
factors in suicide decedents. As such, these results do not indicate that prior attempts are not an important consideration in determining risk or that this variable should be given less weight than it currently is in accepted assessment approaches. Indeed, past attempts are undoubtedly useful in predicting other clinically relevant outcomes (e.g., future non-lethal attempts), so any move to de-emphasize prior attempts would likely cause substantially more harm than good. Instead, these results highlight the notion that simply identifying and weighting risk factors in the absence of a clear understanding of context represents a suboptimal risk assessment approach. These results indicate that access to lethal means may be useful to assess in all clients regardless of their reason for presenting for mental health care, and that restricting access to such means should be considered at what might be considered a low risk profile in some assessment approaches (e.g., individuals with active suicidal ideation, but no prior attempts and a denial – although not necessarily absence – of a suicide plan). It is also worth noting that the proportion of non-firearm suicide decedents with prior attempts was lower than I anticipated. In this sense, these results highlight the point that contextual factors are not only an important consideration with respect to potential death by suicide resulting from self-inflicted gunshot wounds. Indeed, this unexpected finding may provide further insight into the recent findings of Franklin et al. (in preparation), who noted our general lack of improvement in the prospective prediction of suicide over the past half century. Given the low percentage of decedents across methods (firearms and non-firearms) endorsing prior attempts and prior research indicating that approximately half of suicide decedents have engaged in a nonlethal attempt prior to their death by suicide (Isometsa and Lonnqvist, 1998), it is also entirely possible that the NVDRS is not fully capturing this variable. It is highly unlikely that the NVDRS is capturing the variable in a way that systematically impacts firearm deaths more than non-firearm deaths; however, the percentages reported in this paper should also be considered within the context of those reported in the broader literature. When considering these results, it is important to note limitations. Only a minority of states has participated in the NVDRS thus far and, within those that have participated, not every suicide has been reported. As such, it is unclear that our findings would generalize beyond this sample. Secondly, due to the nature of the variables considered in our model, I was unable to experimentally manipulate them. As such, I could not directly test the effect of individual difference variables on method choice in suicide. Lastly, variables highly relevant to the proposed model (e.g., acquired capability for suicide) were not assessed in the NVDRS and, as such, their roles as moderators or mediators could not be tested. All of this considered, however, I would also note several strengths of the study. By using eight years of data across 17 states, I was able to examine more than 70,000 suicide deaths. Furthermore, whereas much of our understanding of lethal suicidal behavior has been derived from studies that considered only survivors of non-lethal attempts, this sample was directly relevant to the specific research questions addressed in our analyses. Additionally, by using publicly available data, the transparency of the analyses is optimized, allowing for others to easily replicate and expand upon the findings. Lastly, although the NVDRS did not sample the entire country, it did include states that differ widely on their geographic location, racial/ethnic makeup, climate, economic health, and culture, thereby increasing the generalizability of the findings. Ultimately, I believe these findings highlight an important limitation in current risk assessment models and represent further evidence of the importance of supplementing interventions by clinicians with interventions designed to address access and exposure to firearms on a broader level (e.g., legislation).
M.D. Anestis / Journal of Affective Disorders 189 (2016) 106–109
References Anestis, M.D., Anestis, J.C., 2015. Suicide rates and state laws regulating access and exposure to handguns. Am. J. Public Health 105, 2049–2058 10.2105/AJPH.2015. 302753. Anestis, M.D., Khazem, L.R., Law, K.C., Houtsma, C., LeTard, R., Moberg, F., Martin, R., 2015. The association between state laws regulating handgun ownership and statewide suicide rates. Am. J. Public Health 105, 2059–2067. http://dx.doi.org/ 10.2105/AJPH.2014.302465. Anestis, M.D., Khazem, L.R., Mohn, R.S., Green, B.A., 2015. Testing the main hypotheses of the interpersonal-psychological theory of suicidal behavior in a large diverse sample of United States military personnel. Compr. Psychiatry 60, 78–95. Carrington, P.J., 1999. Gender, gun control, suicide and homicide in Canada. Arch. Suicide Res. 5, 71–75. http://dx.doi.org/10.1080/13811119908258316. Centers for Disease Control and Prevention, 2015. Web-Based Injury Statistics Query and Reporting System (WISQARS). Available at 〈http://www.cdc.gov/in jury/wisqars/index.html〉. (accessed 23.06.15). Chapdelaine, A., Samson, E., Kimberly, M.D., Viau, L., 1991. Firearm-related injuries in Canada: issues for prevention. CMAJ: Can. Med. Assoc. J. 145, 1218–1223. Daigle, M.S., 2005. Suicide prevention through means restriction: assessing the risk of substitution. Accid. Anal. Prev. 37, 625–632. http://dx.doi.org/10.1016/j. aap.2005.03.00. De Leo, D., Draper, B.M., Snowdon, J., Kolves, K., 2013. Suicide in older adults: a case-control psychological autopsy study in Australia. J. Psychiatr. Res. 47, 980–988. Franklin, J.C., Riberio, J.D., Fox, K.R., Bentley, K.H., Kleiman, E.M., Jaroszewski, A.C., Chang, B., Nock, M.K., 2015. Risk Factors for Suicidal Thoughts and Behaviors: A Meta-Analysis of 50 Years of Research, in preparation. Hawton, K., 2010. Completed suicide after attempted suicide. Br. Med. J. 341, c3064. Isometsa, E.R., Lonnqvist, J.K., 1998. Suicide attempts preceding completed suicide. Br. J. Psychiatry 173, 531–535. Joiner, T.E., 2005. Why People Die by Suicide. Harvard University Press, Cambridge, MA. Joiner, T.E., Van Orden, K.A., Witte, T.K., Selby, E.A., Riberio, J.D., Lewis, R., Rudd, M. D., 2009. Main predictions of the interpersonal-psychological theory of suicidal behavior: empirical tests in two samples of young adults. J. Abnorm. Psychol.
109
118, 634–646. Joiner, T.E., Walker, R.L., Rudd, M.D., Jobes, D.A., 1999. Scientizing and routinizing the assessment of suicidality in outpatient practice. Prof. Psychol.: Res. Pract. 30, 447–453. http://dx.doi.org/10.1037//0735-7028.30.5.447. Klonsky, E.D., May, A.M., 2015. The three-step theory (3ST): a new theory of suicide rooted in the "ideation to action" framework. Int.J.Cogn.Ther 8, 114–129 10.1521/ijct.2015.8.2.114. Klonsky, E.D., May, A.M., 2014. Differentiating suicide attempters from suicide ideators: a critical frontier for suicidology research. Suicide Life-Threat. Behav. 44, 1–5. Kreitman, N., 1976. The coal gas story. United Kingdom suicide rates, 1960–71. Br. J. Prevent. Soc. Med. 30, 86–93. http://dx.doi.org/10.1136/jech.30.2.86. Law, C., Sveticic, J., de Leo, D., 2014. Restricting access to a suicide hotspot does not shift the problem to another location. An experiment of two river bridges in Brisbane, Australia. Aust. N. Z. J. Public Health 38, 134–138. Lester, D., Abe, K., 1998. The suicide rate by each method in Japan: a test of Durkheim’s theory of suicide. Arch. Suicide Res. 4 (3), 281–285, doi: 0.1080/ 13811119808258302. Loftin, C., McDowall, D., Wiersema, B., Cottey, T.J., 1991. Effects of restrictive licensing of handguns on homicide and suicide in the District of Columbia. N. Engl. J. Med. 323, 1615–1620. http://dx.doi.org/10.1056/ NEJM199112053252305. Nordentoft, M., Qin, P., Helweg-Larsen, K., Juel, K., 2010. Restrictions in means for suicide: an effective tool in preventing suicide: the Danish experience. Suicide Life-Threat. Behav. 37, 688–697. Oliver, R.G., Hetzel, B.S., 1972. Rise and fall of suicide rates in Australia: relation to sedative availability. Med. J. Aust. 2, 919–923. Posner, K., Brent, D., Lucas, C., Gould, M., Stanley, B., Brown, G., Fisher, P., Zelazny, J., Burke, A., Oquendo, M., Mann, J., 2008. Columbia Suicide Severity Rating Scale (C-SSRS). The Research Foundation for Mental Hygiene, Inc, New York, NY. Sarchiapone, M., Mandelli, L., Iosue, M., Andrisano, C., Roy, A., 2011. Controlling access to suicide means. Int. J. Environ. Res. Public Health 8, 4550–4562. Spicer, R.S., Miller, T.R., 2000. Suicide acts in 8 states: incidence and case fatality rates by demographics and method. Am. J. Public Health 90, 1885–1891. http://dx.doi.org/10.2105/AJPH.90.12.1885. Van Orden, K.A., Witte, T.K., Cukrowicz, K.C., Braithwaite, S.R., Selby, E.A., Joiner, T. E., 2010. The interpersonal theory of suicide. Psychol. Rev. 117, 575–600.