JAMAControversies - August 5, 1998
Does Owning a Firearm Increase or Decrease the Risk of Death?
Peter Cummings, MD, MPH; Thomas D. Koepsell, MD, MPH
During a visit for preventive care, a patient asks, "If I buy a gun, will it increase or decrease my risk of dying prematurely?" What evidence should the physician seek in formulating a response?
BALANCING RISKS AND BENEFITS
A firearm can potentially offer both benefits and risks to the health of its owner. Balancing risks is something that clinicians commonly do in medicine. For example, when considering a prescription for hormone replacement therapy, clinicians must weigh a somewhat higher risk for endometrial cancer against a lower risk for heart disease. Although the pros and cons of gun ownership have been the subject of impassioned debate, using an evidence-based approach to address the patient's question is appealing.
Having a gun in the home might affect the risk of homicide, suicide, or unintentional firearm injury. The risk of homicide might decrease if the gun is used successfully to ward off an assailant, but the risk might increase if the gun is used against the owner by an intruder or another household member. The risk of suicide might increase if the owner chooses to use a gun for a suicide attempt, rather than a less lethal method. The risk of unintentional gunshot injury might be greater if a gun is in the home.
To address the question about the net effects of having a gun on the risk of dying prematurely, investigators must study all completed homicides and suicides as outcomes, not just those involving guns. Imagine, for example, that persons with guns had a suicide rate similar to that of other persons. If this were true, then it would be proper to conclude that gun ownership did not increase the risk of suicide, even if the gun owners were more likely to commit suicide with a gun and were less likely to use other methods.
STUDY DESIGN ISSUES
In evaluating choices regarding health care, evidence from randomized controlled trials is generally preferred because these studies offer advantages with regard to the reduction of potential biases. A randomized trial of gun ownership, however, will probably never be done because few people would volunteer for random assignment of gun ownership and the trial would have to be very large since the outcomes of interest (eg, fatalities) are not that common.
A study design that clinicians usually prefer not to rely on for inferences about causation is the case series, which only records information about persons with the outcome of interest. In this context, a similar study design is the exposure series, which records the outcomes only of exposed persons. Several surveys, analogous to exposure series, have asked some variation of this question: "Have you used a gun to protect yourself or your property?" Published estimates of the number of times that a gun was used in the United States for protection in a single year have ranged from 62,000 to 23 million.[2-6] One study, which asked for details about gun use, estimated that about 400,000 adults felt that they saved a life by using a gun in 1993.
The health effects of tobacco use or the use of bicycle helmets or guns cannot be judged by surveying only those who use them. The testimony of users is sometimes of interest, but this alone cannot validly estimate the effect of an exposure, including guns, for several reasons: (1) Clinicians cannot be confident that persons who had used a gun could accurately judge what would have happened if they had not used a gun. (2) This design only counts episodes of presumed benefit and omits the most unsuccessful gun users, those who died. (3) This design offers no comparison with persons who did not have access to a gun, and therefore provides no estimate of the net benefit associated with access to a gun.
Ecologic studies of homicide or suicide and gun prevalence in populations or of gun ownership regulations are of interest in studying the effects of public policy regarding firearms.[7,8] However, studies of this type, which compare outcome rates of populations, are not well suited for assessing the health effects of gun ownership for individuals. If researchers knew that a population with many guns had a higher (or lower) rate of homicide than another population with fewer guns, they could not be certain if individual gun owners within each population had a higher (or lower) homicide rate compared with nonowners. The patient, however, wants to know about his or her risks and benefits if he or she buys a gun, not about the effects of gun prevalence on the population.
To examine exposures that are difficult to study in a randomized trial, clinicians generally turn to evidence from study designs that observe and compare the experience of individuals. One of these, the cohort study, would assess the gun ownership status of people at some point in time, and determine their health outcomes at a later time. In a cohort study, assignment to a gun-owning household is not random; it would be necessary to measure potential confounding factors, which might be related both to each person's risk for the outcomes and to the likelihood that a person had a gun in the home. For example, persons might be induced to buy a gun if they lived in a neighborhood with a high level of violent crime. Failure to control for neighborhood, therefore, might create an association between gun ownership and death by homicide, even if gun ownership had no causal relationship to homicide. A cohort study of gun access has not been done; like a controlled trial, it would have to be very large.
When an outcome is rare, but the exposure common, researchers usually turn to another observational design, the case-control study, which is more efficient than a cohort study in this situation. As with a cohort study, it is necessary to ascertain possible confounders and to control for these in the analysis. If guns offer net protection against death, then the cases will be less likely to own guns than the controls and the estimated relative risk (RR) for death, comparing gun owners with nonowners, will be less than 1. If, however, guns increase the risk of death for owners, then the cases will be more likely to own guns than the controls and the RR estimates will be higher than 1. The case-control design appears to offer the best hope for making the comparison of interest to the patient.
PUBLISHED CASE-CONTROL STUDIES
Six published case-control studies have examined the association between a gun in the home and suicide.[9-14] All of these studies reported that the RR of suicide was greater among persons in a home with a gun compared with other persons. Only 2 case-control studies analyzed homicide as an outcome.[14,15] Both reported a higher RR of being killed in a homicide among those with a gun compared with those without a gun.[14,15]
The published case-control studies have used different control groups. Two studies of suicide used as the control group patients hospitalized for psychiatric care.[9,10] While many case-control studies have used hospitalized controls, there is always concern that this group may not accurately reflect the exposure experience of the population from which the cases were derived. Two studies went to great effort to find controls who were closely matched to the cases on neighborhood and other characteristics; this involved a systematic canvass of each case's neighborhood.[11,15] Three studies used other means of finding population controls, such as the electoral rolls in New Zealand or the computerized membership files of a health plan to which the cases belonged.[12-14]
One study included only suicides in the home. Persons without a gun might choose a suicide method, such as jumping off a bridge, that required them to leave their home, so the RR for all suicides might be less than the estimate of 4.8 reported by this study. The degree of bias, however, may not be great, as only 30% of suicides in the study's region were outside the home and 46% of these involved a firearm. Another study only included homicides in the home. As with suicide, having a gun in the home might affect the location of some homicides, without affecting the overall risk of homicide. For example, someone's risk of homicide in the home might be smaller, but the risk outside the home greater, if an assailant knew about a gun in the home and decided, therefore, to attack the victim outside the home. Given these possibilities, it is probably best if investigators examine all homicides and suicides, regardless of location or means used, to assess and balance all the risks and benefits.
Interviews were used in most studies to determine gun exposure status; bias could have occurred if there was deception by some respondents and if the degree of deception differed between case and control subjects. One study relied solely on records of past handgun purchases and therefore deception regarding exposure and differential recall for cases and controls were both eliminated as sources of bias. However, since gun exposure status was not measured at the date of death for the cases or on the same date for the matched controls, any random misclassification may have caused the investigators to underestimate the true strength of an association.
Four studies made some attempt to measure and control for suicidal intent or psychiatric diagnoses based on information reported and evaluated after the outcome occurred.[9,10,12,13] Knowledge of the outcome might bias these measurements. It may be possible to reduce potential bias if investigators rely on information collected prior to the outcome and if the evaluators are unaware of each subject's case or control status.
Most studies examined information about gun exposure beyond mere ownership. For example, 3 studies reported evidence that access to a handgun was more strongly associated with suicide or homicide than access to a long gun.[11,12,15] The strength of association between death and a gun in the home was greater if the home had a loaded gun, an unlocked gun, or multiple guns.[11,12,14,15] One study examined the time interval from handgun purchase to death: the RR of homicide did not vary by time since purchase; however, the RR of suicide was greater within the first year after handgun purchase (5.7) and less thereafter (1.7). This suggests that some people may buy a gun expressly to kill themselves, whereas later a gun may enhance the likelihood of suicide because it is readily available at a time of despondency.
Persons who have firearms in their homes may differ from other persons in their risk for homicide or suicide, aside from their gun exposure status. For this reason, most studies make some effort to measure and control for potential confounding factors. These included neighborhood, age, sex, education, socioeconomic status, marital status, psychiatric history, medications, use of alcohol and illicit drugs, criminal history, and previous household violence. Studies have examined whether the association between a gun in the home and death varies by sex, age, race, or neighborhood median income; important variations have not been found.[11,14,15]
Clinicians and researchers never have all the evidence that they desire. In addition to having more studies similar to those that have been done, it would be desirable to have evaluations of other outcomes, such as unintentional gunshot injuries, both fatal and nonfatal, and other nonfatal outcomes, such as rape or assault. More accurate information is needed regarding the effect of firearm storage practices on the balance of risks. Clinicians and researchers also might wish to know if there are settings or subgroups for which benefits do outweigh the risks.
What advice, if any, can clinicians offer to their patients who are considering the purchase of a gun? Based on criteria for judging whether an association is causal, the evidence from comparative observational studies appears consistent with the inference that owning a gun increases the risk of suicide. Most studies show a moderately strong association, the biological mechanism is plausible, the exposure precedes the outcome, the association has been replicated in several populations, and there is evidence of a dose response (greater risk with more or more available guns). Evidence that a gun in the home increases the risk of homicide comes from only 2 studies and seems weaker; however, these studies offer no support for the view that gun ownership confers a net benefit in terms of protection against homicide.
Based on the evidence currently available, it appears that gun ownership is associated with a net increase in the risk of death for a typical individual. Clinicians might advise their patients accordingly.
From the Harborview Injury Prevention and Research Center and the Department of Epidemiology, School of Public Health and Community Medicine, University of Washington, Seattle.
Corresponding author: Peter Cummings, MD, MPH, Harborview Injury Prevention and Research Center, Box 359960, 325 Ninth Ave, Seattle, WA 98104-2499 (e-mail: email@example.com).
1. Cummings P, Weiss NS. Case series and exposure series: the role of studies without controls in providing information about the etiology of injury or disease. Inj Prev. 1998;4:54-57.
2. Kleck G. Crime control through the private use of armed force. Soc Problems. 1988;35:1-21.
3. McDowall D, Wiersema B. The incidence of defensive firearm use by US crime victims, 1987 through 1990. Am J Public Health. 1994;84:1982-1984.
4. Rand MR. Guns and Crime: Handgun Victimization, Firearm Self-defense, and Firearm Theft. Washington, DC: US Dept of Justice, Bureau of Justice Statistics; 1994.
5. Kleck G, Gertz M. Armed resistance to crime: the prevalence and nature of self-defense with a gun. J Crim Law Criminol. 1995;86:150-187.
6. Cook PJ, Ludwig J. Guns in America: National Survey on Private Ownership and Use of Firearms. Washington, DC: National Institute of Justice; 1997.
7. Loftin C, McDowall D, Wiersema B, Cottey TJ. Effects of restrictive licensing of handguns on homicide and suicide in the District of Columbia. N Engl J Med. 1991;325:1615-1620.
8. Sloan JH, Kellermann AL, Reay DT, et al. Handgun regulations, crime, assaults, and homicide: a tale of two cities. N Engl J Med. 1988;319:1256-1262.
9. Brent DA, Perper JA, Goldstein CE, et al. Risk factors for adolescent suicide: a comparison of adolescent suicide victims with suicidal inpatients. Arch Gen Psychiatry. 1988;45:581-588.
10. Brent DA, Perper JA, Allman CJ, Moritz GM, Wartella ME, Zelenak JP. The presence and accessibility of firearms in the homes of adolescent suicides: a case-control study. JAMA. 1991;266:2989-2995.
11. Kellermann AL, Rivara FP, Somes G, et al. Suicide in the home in relation to gun ownership. N Engl J Med. 1992;327:467-472.
12. Brent DA, Perper JA, Moritz G, Baugher M, Schweers J, Roth C. Firearms and adolescent suicide: a community based case-control study. AJDC. 1993;147:1066-1071.
13. Beautrais AL, Joyce PR, Mulder RT. Access to firearms and the risk of suicide: a case-control study. Aust N Z J Psychiatry. 1996;30:741-748.
14. Cummings P, Koepsell TD, Grossman DG, Savarino J, Thompson RS. The association between purchase of a handgun and homicide or suicide. Am J Public Health. 1997;87:974-978.
15. Kellermann AL, Rivara FP, Rushforth NB, et al. Gun ownership as a risk factor for homicide in the home. N Engl J Med. 1993;329:1084-1091.
16. Zimring FE. Policy research on firearms and violence. Health Aff (Millwood). 1993;12:109-122.
17. Hill AB. The environment and disease: association or causation. Proc R Soc Med. 1965;58:295-300.
© 1995-1998 American Medical Association. All rights reserved.