David Hemenway, PhD
Potential Conflicts of Interest: None disclosed. Forms can be viewed at www.acponline.org/authors/icmje/ConflictOfInterestForms.do?msNum=M13-2657.
Requests for Single Reprints: David Hemenway, PhD, Harvard School of Public Health, 677 Huntington Avenue, Boston, MA 02115; e-mail, firstname.lastname@example.org.
Hemenway D.; Guns, Suicide, and Homicide: Individual-Level Versus Population-Level Studies. Ann Intern Med. 2014;160:134-135. doi: 10.7326/M13-2657
Download citation file:
Published: Ann Intern Med. 2014;160(2):134-135.
In a systematic review published in this issue, Anglemyer and colleagues (1) found that access to firearms is associated with the risk for completed suicide (odds ratio, 3.2) and being the victim of homicide (odds ratio, 1.9). These results are consistent with those of previous literature reviews (2, 3). The lack of recent research on these issues is telling: The current review found no relevant study published after 2005. However, I believe that this review provides too conservative an estimate of the relationship among guns and suicide and homicide at a societal level.
The evidence that a gun in the home increases the risk for suicide is overwhelming, even stronger than Anglemyer and colleagues’ robust findings. They examined only individual-level studies, stating, “[W]e decided not to include population-level data because we were concerned about ecological bias; for example, gun ownership data on a population level may not reflect the persons who actually commit suicide, so no true link between gun ownership and harms outcomes can be made.” Different disciplines use different research methods. The National Research Council (4) team that critiqued the scientific knowledge about firearms gave little weight to evidence from case–control studies, possibly because the team included few public health experts. Anglemyer and colleagues display an opposite and potentially equally misleading bias by excluding population-level evidence.
Everyone who studies epidemiology learns about the ecological fallacy, but too few seem to learn when it is and is not a serious problem. I believe that the ecological problem is minimal when guns in the home and suicide are examined. Why? Because we know that most guns used in suicides are guns obtained from the home that are used by someone living there. Suicide perpetrators almost never use a gun belonging to another family. We also know that the gunshot wound caused the death in these cases.
The evidence from many population-based studies of guns and suicide is as strong as that from individual-level studies. Across U.S. cities, states, and regions, higher levels of household gun ownership are associated with higher rates of firearm-related and overall suicide. There is no association between gun ownership levels and suicide by means other than guns. These studies have controlled for such factors as rates of urbanization, poverty, education, alcohol use, unemployment, divorce, depression, suicidal ideation, and suicide attempts. The association holds for men, women, adolescents, and all adult age groups (5). The effect of a gun in the home is not only significant but important. Indeed, differences in suicide rates among the states are better explained by household gun ownership levels than by levels of mental health problems, suicidal ideation, or suicide attempts (6). Studies also show that adults in households with firearms are no more depressed or suicidal than those in households without firearms (7), yet they are far more likely to die of suicide.
The few U.S. ecological studies that did not find a significant relationship between guns and suicide used poor proxies for gun ownership. The only individual-level study (8) in Anglemyer and colleagues’ review that did not show a statistically significant association was from New Zealand, where gun safety laws are stronger than those in the United States (for example, guns in the home must be locked up), and was underpowered—only 20 firearm-related suicides were included among the cases. The risk for suicide was 40% greater in homes where persons owned a gun, a result that is significant at the 10% level but not at the 5% level.
The evidence that Anglemyer and colleagues summarized about the relationship between guns and homicide is not nearly as strong as that between guns and suicide, perhaps largely because the authors focused exclusively on individual-level studies of victims. Most victims of firearm-related suicide shoot themselves with a gun from their own home. In contrast, victims of firearm-related homicide are, by definition, shot by someone else and most perpetrators do not use a gun from the victim's home. Only in domestic homicides does the perpetrator often kill the victim with a household firearm, and in these instances a man typically shoots his female partner; relatively few male homicide victims are shot by their partners. Not surprisingly, the evidence on the relationship between homicide and household firearms is strongest for female victims. Studies of battered women—not included in Anglemyer and colleagues’ review—confirm the strength of this relationship (9).
Why having a gun in the home would substantially increase the risk for being murdered by a person who is not a family member is uncertain, and Anglemyer and colleagues do not provide an explanation. Although there is little evidence that having a gun reduces the risk for homicide victimization (10), there is not yet compelling evidence that having a gun substantially increases the risk for homicide victimization for most men. What does put men at substantially increased risk for homicide victimization is other men having access to guns.
I suggest that future analysis of the association between guns and homicide should not be through individual-level studies of victims but through individual-level studies of perpetrators or ecological, population-based studies. Such studies that have been done provide evidence that more guns in the community substantially increase the risk for homicide victimization (3).
Anglemyer and colleagues aptly summarize the evidence that, for most families, bringing a gun into the home substantially increases the risk for suicide for all family members and the risk for women being murdered in the home (11). Evidence not included in their review also indicates that a gun in the home increases the risk for homicide victimization for others in society. This increased risk may be due to someone in the family shooting others (for example, the Sandy Hook Elementary School shooting) or the gun being stolen and used by criminals. Obtaining a firearm not only endangers those living in the home but also imposes substantial costs on the community.
The In the Clinic® slide sets are owned and copyrighted by the American College of Physicians (ACP). All text, graphics, trademarks, and other intellectual property incorporated into the slide sets remain the sole and exclusive property of the ACP. The slide sets may be used only by the person who downloads or purchases them and only for the purpose of presenting them during not-for-profit educational activities. Users may incorporate the entire slide set or selected individual slides into their own teaching presentations but may not alter the content of the slides in any way or remove the ACP copyright notice. Users may make print copies for use as hand-outs for the audience the user is personally addressing but may not otherwise reproduce or distribute the slides by any means or media, including but not limited to sending them as e-mail attachments, posting them on Internet or Intranet sites, publishing them in meeting proceedings, or making them available for sale or distribution in any unauthorized form, without the express written permission of the ACP. Unauthorized use of the In the Clinic slide sets will constitute copyright infringement.
Andrew Anglemyer, PhD, MPH
University of California San Francisco
January 28, 2014
In Dr Hemenway’s editorial, he correctly suggests our systematic review and meta-analysis (1) “provides too conservative an estimate of the relationship among guns and suicide and homicide at a societal level” (2). While we agree that our methods have likely underestimated the true relationship between guns and suicide or homicide victimization, we feel that our inclusion of only studies where the individual’s personal firearm experience was assessed does provide an estimate of the minimum likely risk.
The narrative around studies of firearms ownership and harms is usually around discrediting the causality assumptions, and causal inference is particularly problematic in ecologic studies. Critics are quick to point out conflicting “evidence” from various sources of rates and their interpretations. In fact, a commonly cited review of ecologic data by Kates and Mauser only adds fodder to the on-going debate (3). We purposefully excluded ecologic data, in part, to obviate this common critique regarding the interpretability of population-level data.
Dr Hemenway correctly points out that most individuals commit suicide with a firearm within their own home, thus potentially justifying using aggregate population-level data for suicide outcomes. Unfortunately, this would likely not be enough to convince doubters of the use of rate data to assert correlations between firearm access and suicide. More importantly, in our review we had two primary outcomes of interest-- suicide and homicide victimization. While we can make a reasonable assumption that the firearms used in the suicides were obtained from the home (albeit, still an assumption that we wanted to avoid), we cannot reasonably make that same assumption about the firearms used in homicides. In fact, even Dr Hemenway writes “most perpetrators [of homicides] do not use a gun from the victim’s home” (2). We feel that it would not have been prudent to use different inclusion criteria for different outcomes (i.e., include ecologic data for suicide outcomes and not for homicide outcomes) as the focus of our review was the access to firearms among cases, not among the perpetrators, and the result focuses on the home. Moreover, our results may help households understand what the risks are of keeping a firearm in their home, particularly if there is a household member who is depressed or if there is a violent relationship; results from ecologic studies may be better for evaluating policies.
Notwithstanding our review, ecologic studies are a good source of data for establishing trends and positing relationships. We agree that one potential future methodological review could compare the summary estimates we obtained from individual-level data with the summary estimates one would have obtained had ecologic data been included. Operationally, of course, there are a number of issues that a reviewer would need to consider: which rates of ownership to use (e.g., regional rates, state rates, city rates), which timeframe (i.e., most recent data, most reliable) and which types of suicide (i.e., include only firearm-related suicides or any suicides)? For these reasons, among others, ecologic data are seldom seen in meta-analyses of individual-level data. Assuming that the reviewer were able to overcome the operational obstacles of disparate rates over disparate regions describing dissimilar outcomes, he or she would have to ensure that the interpretation of the results from such a review are truly reflective of aggregated population-level data, and not reflective of individual-level data. Of course, there are conceptual steps to calculate ecologic relative risk (RR) estimates so that conclusions about individual-level behavior can be inferred, though there are inherent assumptions (4). The reviewer could pool these new “individual-level” RR estimates to get a pooled RR from ecologic studies.
To make comparisons between results from the meta-analysis of ecologic studies with the meta-analysis of the individual-level data we summarize, there are a number of approaches a reviewer could use. Namely, he or she could pool both results together to get a summary estimate from both ecologic and individual-level data, though combining estimates that were derived from different sources with different assumptions can be problematic. He or she could also perform a sensitivity analysis comparing the pooled estimates with and without ecologic data. Or he or she could perform a subgroup analysis comparing the pooled estimates from ecologic data and from individual-level data separately. We feel that there are indeed benefits from reviewing ecologic data, and Dr Hemenway’s suggestion for future research will compliment our current review nicely. The two reviews together could provide a more nuanced understanding of the true relationship between access to firearms and harms outcomes.
1. Anglemyer A, Horvath T, and Rutherford G. The Accessibility of Firearms and Risk for Suicide and Homicide Victimization Among Household Members. A Systematic Review and Meta-Analysis. Ann Intern Med 2014; 160: 101-10.
2. Hemenway D. Guns, Suicide, and Homicide: Individual-Level Versus Population-Level Studies. Ann Int Med 2014; 160: 134-135.
3. Kates D and Mauser G. Would Banning Firearms Reduce Murder and Suicide? A Review of International and Some Domestic Evidence. Harv J Law Public Policy 2007; 30 (2): 649-694.
4. Schuessler A. Ecological Inference. Proc Natl Acad Sci USA 1999; 96: 10578-10581.
Robert B. Sklaroff, M.D.
Nazareth Hospital, Philadelphia PA
February 4, 2014
Potential Politicization of the Gun Control Controversy
TO THE EDITOR:
Because Public Health research predictably guides generation of Public Policy, it is necessary to scrutinize the political science underlying the paired gun-control review (1) and editorial (2); challenges are detected to fundamental standards that may compromise an otherwise sound meta-analysis of available literature. The last sentences of each are revelatory, for the former finds “restricting [access to a firearm in the home] may effectively prevent injury” and the latter concludes “obtaining a firearm not only endangers those living in the home, but also imposes substantial costs on the community.” Notwithstanding unaddressed Second Amendment constraints, the authors of both unabashedly campaign to restrict the right to bear arms, thereby ignoring—for example—the human compulsion to manifest reasonable self-defense.
The intuitive deduction, that availability of a firearm will increase the risk that momentary depression will yield suicide, is consistent with modern lay culture—recalling the 1945 movie “Spellbound”—and medical scholarship—recalling an essay published last year in this journal (3). Yet, it is undermined by the editorialist, who has argued that the widespread ownership of firearms in private hands in the U.S. promotes the spread of the "disease" of gun violence (4). He invoked a generalized reference to his book when claiming “There is no association between gun ownership levels and suicide by means other than guns. These studies have controlled for…depression [and] suicidal ideation.” If true, this assertion would undermine efforts to include scrutiny of mental health data during any mandated background-checks; alas, it is untrue, for a profile has been generated of psychiatric patients at high risk for suicide (5).
This latter citation was among the articles cited in the review (#26), prompting confusion when noting it was among three articles cited in the online Appendix—which purports to show “the disposition of studies excluded after full-text review”—along with two others (#32 and #60) “because the study populations were contained in previously published data included in this review.” Noting there are 59 published references and 97 online references, merely counting the number of citations associated with a particular reason for exclusion yields the observation that there is an admixture of articles that were included and articles that were excluded (i.e., some were among the first #1-59 and at least one was among the latter #60-97). The authors should have provided a cross-walk “pairing” of how one set of data was subsuming another set of myriad peer-reviewed studies, precluding concern that any undue selectivity existed.
Therefore, author-bias—seeking the ability to generate the above preordained outcome—could have clouded how subsidiary observations were drawn regarding, for example, the allegation of enhanced risk of being killed by a household member. And, overall, adopting a purely academic approach could have yielded insights, for example, as to the type of mental health diagnoses that might predispose to criminal gun-use; indeed, this entire body of work could then have been compared/contrasted with lethal violence committed via non-household, unregistered firearms, yielding far more useful insights as to what societal interventions might be optimal.
1. Andrew Anglemyer, Tara Horvath, George Rutherford; The Accessibility of Firearms and Risk for Suicide and Homicide Victimization Among Household Members - A Systematic Review and Meta-analysis. Ann. Intern. Med. 2014 Jan;160(2):101-110.
2. David Hemenway; Guns, Suicide, and Homicide: Individual-Level Versus Population-Level Studies. Ann. Intern. Med. 2014 Jan;160(2):134-135.
3. Carl E. Fisher, Jeffrey A. Lieberman; Getting the Facts Straight About Gun Violence and Mental Illness: Putting Compassion Before Fear. Ann. Intern. Med. 2013 Sep;159(6):423-424.
4. Wheeler, Timothy J. (September 2005). "Private Guns, Public Health". The Freeman. In http://en.wikipedia.org/wiki/Private_Guns,_Public_Health#cite_ref-free_1-0.
5. Brent DA, Perper JA, Moritz G, Baugher M, Schweers J, Roth C. Firearms and adolescent suicide. A community case-control study. Am J Dis Child. 1993; 147:1066-71.
Results provided by:
Copyright © 2016 American College of Physicians. All Rights Reserved.
Print ISSN: 0003-4819 | Online ISSN: 1539-3704
Conditions of Use
This PDF is available to Subscribers Only