Suicide in the Home in Relation to Gun Ownership

Case Identification

Table 1. Table 1. Characteristics of Shelby County, Tennessee, and King County, Washington, as of 1980.*

Shelby County, Tennessee, and King County, Washington, are the most populous counties in their states, containing the cities of Memphis and Seattle, respectively. Although both counties are urban in character, King County is predominantly white and enjoys a relatively high standard of living. In contrast, Shelby County has a large black minority, and a substantial proportion of its citizens live below the poverty level (Table 1).^{22 , 23}

All suicides involving a resident of either county that occurred between August 23, 1987, and April 30, 1990, were studied to identify those that took place in the home of the victim. Any death ruled a suicide was included, regardless of the method used. Self-inflicted injuries that were not immediately fatal were included if death followed within three months. Cases in which there was potential litigation over the official cause of death were excluded at the request of the medical examiners.

Case Selection and Recruitment of Case Proxies

A home was defined as any house, apartment, or dwelling occupied by a victim (a case subject) as that person's principal residence. Suicides occurring in adjacent structures (e.g., a garage) or the surrounding yard were also included. In the event of a murder—suicide, the suicide victim was included only if older than the murder victim; in double suicides, only the older victim was included.

Reports made at the scene were collected to ensure that the study criteria were met. In King County, the medical examiner's staff conducted all investigations of the scene. In Shelby County, police officers conducted the investigations. In addition to data obtained at the scene, investigators obtained the names of persons close to the victim who might provide an interview at a later date. These lists were supplemented with names obtained from newspaper accounts, obituaries, and calls to funeral homes. In this way, proxies for the victims were identified.

Approximately three weeks after each death, those chosen to be interviewed were sent a signed letter that outlined the nature of the project. A $10 incentive was offered, and a follow-up telephone call was made a few days later to arrange a time and place for the interview. At the time of this meeting, informed consent was obtained. The proxies were told that they could decline to answer any question. They were also free to terminate the interview at any time.

Selection and Recruitment of Controls

After each interview with a case proxy, we sought a control subject matched with the case subject according to sex, race, age range (0 to 14, 15 to 24, 25 to 40, 41 to 60, and 61 years or older), and neighborhood of residence. To minimize selection bias, the controls were identified by a previously validated procedure for the random selection of matching households in nearby locations.^{24 , 25} After marking off a one-block avoidance zone around the home of the case subject, the interviewers started a neighborhood census at a randomly selected point along a predetermined route radiating out from the case subject's residence. Households in which no one was home were visited twice more, at different times of day and on different days of the week. After the census was completed, an adult (18 years old or older) in the first household in which a member met the matching criteria was offered a $10 incentive and asked to consent to an interview. Whenever possible, attempts were made to interview someone other than the actual matching control subject (a control proxy). When no interview was granted, the next matching household on the route was approached.

Interviews

The interviews with proxies for case subjects and their controls were virtually identical in format, order, and content. Each was brief, highly structured, and arranged so that the more sensitive questions were not broached until later in the interview. Items drawn from the Short Michigan Alcoholism Screening Test,²⁶ the Hollingshead two-factor index of social position,²⁷ and a 1978 poll by Decision Making Information²⁸ were included. Particularly sensitive questions were preceded by "permissive" statements, such as the following: "Half of all homes in America contain one or more firearms. Are guns of any kind kept in your home?" To confirm the reliability of the interviewers, supervisory personnel contacted a random sample of 10 percent of those interviewed and administered an abbreviated form of the questionnaire. The rate of concordance between these responses and those given to the interviewers equaled or exceeded 96 percent.

Statistical Analysis

Reports from the scene of the suicide were analyzed for descriptive purposes only. Interview data were used in the risk assessment because these were collected in an analogous manner from the case proxies and the controls. Since any member of a control household might decide to remove firearms from the home after a death in the neighborhood, the answers were adjusted to reflect the state of affairs on the date of each suicide. MantelHaenszel chi-square analysis was used to calculate the crude odds ratio associated with each variable. Subsequent multivariate analyses used conditional logistic regression, the appropriate technique for a matched-pairs design.²⁹

Potentially confounding variables were identified and controlled for through a three-step process. First, models containing closely related variables (such as those describing the use of alcohol in the home) were constructed to identify the variable or variables in each set that were most predictive of being a case subject or a control. Whenever both measures of exposure pertaining to the individual and measures pertaining to the household were found to be significant, only the variable describing exposure for the individual was retained. Next, a model that incorporated the variables selected in this initial step was constructed, to select the variables that remained significant after we controlled for the effects of the remaining variables in the model. Finally, a third model was constructed in which gun ownership was added to the list of covariates to assess the relation between firearms and suicide in the home after adjustment for the effects of the remaining variables. At this stage, we also looked for two-way interaction terms and checked to see whether there was an interaction between gun ownership and any of the matching variables.

After calculating an adjusted odds ratio for suicide based on all the pairs for which data were complete (360 pairs), we analyzed a series of subgroups representing various strata in the full study sample. To exclude the possibility of confounding due to preexisting mental illness or depression, we calculated the relative odds of suicide among case subjects who had no history of psychiatric problems. To limit bias resulting from potentially faulty reporting about case subjects who lived alone, we conducted an analysis restricted to pairs in which interviews were obtained with proxies who lived in the same home as the victim. To determine whether gun ownership was associated with an increased risk of suicide by means other than firearms, an analysis stratified according to the method of suicide was also conducted.

After these calculations, we performed three final analyses to see whether different patterns of gun ownership or storage were associated with a greater or lesser risk of suicide. In each comparison we used the final model as previously defined, but we subdivided gun ownership into three categories instead of comparing owners with nonowners. One analysis compared handgun ownership, ownership only of long guns, and ownership of no guns. A second compared ownership of loaded guns, ownership of unloaded guns, and ownership of no guns. The third compared guns kept in a locked place, guns kept in an unlocked place, and no guns.

Study Population

Table 2. Table 2. Suicides Committed in the Home of the Victim from August 23, 1987, through April 30, 1990.

Eight hundred three suicides took place during the 32 months of our study. In King County, 555 suicides were identified, 385 of which (69 percent) occurred in the home of the victim. In Shelby County, 248 suicides occurred, 180 of which (73 percent) took place in the victim's home. After we excluded the younger victims in five cases of double death, one case involving delayed death, three cases already excluded by the medical examiner's staff, and two cases determined long after the time of death to have involved suicide, 554 cases of suicide were available for study (Table 2).

Most of the suicides we studied occurred at the victim's residence. One fourth were committed in an adjoining yard, an accessory building, or a place of work attached to the home. Notes were left in 36 percent of cases. A gun was the means of death most commonly used; slightly more than half of all suicides in the home in King County involved a firearm, as did 73 percent of those in Shelby County. Handguns were used in 72 percent of the suicides involving a firearm. In four of every five such suicides, the investigators specifically noted that the gun had been kept in the victim's home. In the remaining 20 percent of cases, the origin of the firearm was not noted.

Comparability of Case Subjects and Controls

Eighty-three percent of proxies for the case subjects in King County and 73 percent of those in Shelby County agreed to be interviewed. The households of those who agreed to an interview and the households of those who did not agree did not differ with respect to the age, sex, or method of death (firearm vs. other means) of the victim. Those who declined to be interviewed, however, were somewhat more likely to be black (13.4 percent, vs. 8.1 percent among those who were interviewed).

Table 3. Table 3. Demographic Characteristics of 438 Pairs of Case Subjects and Controls.

Interviews with matching controls or their proxies were obtained for 99 percent of the case subjects, yielding 438 matched pairs. Four hundred four pairs were matched for all three variables, 33 for two variables, and 1 for a single variable only (sex). The demographic characteristics of the households of the case subjects and controls were highly similar, except that 36 percent of the case subjects lived alone, as compared with only 18 percent of the matching controls (Table 3). Although every effort was made to conduct the interviews in person, the proxies for case subjects were more likely than controls or their proxies to request a telephone interview (39 percent vs. 10 percent). Despite efforts to interview proxies for each matching control, only 50 percent of the control interviews were conducted in this manner. However, similar rates of gun ownership were reported in telephone interviews and face-to-face interviews, and the proxies for the controls reported rates of gun ownership similar to those reported by the controls themselves.

Univariate Analysis

Univariate comparisons revealed that alcohol was more commonly consumed in the households of case subjects than in those of controls. Alcohol was also more commonly consumed by case subjects than by matching controls. Behavioral correlates of alcoholism (such as trouble at work, problems at home, or hospitalization due to drinking) were also reported by substantially higher percentages of the proxies of case subjects than of controls. Illicit-drug use was reported by 19.2 percent of case proxies but by only 3.1 percent of matched controls.

To the question "Many people occasionally have quarrels or fights — has anyone in this household ever been hit or hurt in a fight in the home?" 13.6 percent of the case proxies answered in the affirmative, but only 3.9 percent of controls. The combination of alcohol and domestic violence was particularly striking. Almost 15 percent of case proxies reported physical fights while drinking, as compared with only 1.2 percent of matched controls.

The case subjects were far more likely than the controls to have been arrested (27.8 percent vs. 8.5 percent). A history of depression or mental illness was reported for 83.5 percent of the case subjects but for only 6.4 percent of the controls. Thirty-six percent of the case subjects took prescribed psychotropic medication, as compared with 3.5 percent of matched controls.

Sixty-five percent of the case subjects had had one or more firearms in their home, as compared with 41 percent of the matched controls (crude odds ratio, 3.2; 95 percent confidence interval, 2.4 to 4.4). Handguns were kept in 49.5 percent of case households but only 23.4 percent of control households (crude odds ratio, 3.7; 95 percent confidence interval, 2.7 to 5.1). In homes with firearms, a gun was the method chosen for suicide in 86 percent of cases. In homes where firearms were not usually kept, only 6 percent of case subjects killed themselves with a gun. The last 172 case proxies to be interviewed were asked how long a gun had been kept in the victim's home. Only 5 of the 162 who answered (3 percent) reported that the gun had been obtained within two weeks of the suicide.

Multlvariate Analysis

Table 4. Table 4. Variables Included in the Final Conditional Logistic-Regression Model Derived from Data on 360 Matched Pairs of Case Subjects and Controls.*

Table 5. Table 5. Suicide in the Home in Relation to Gun Ownership, According to Subgroup.

Six potentially confounding variables were retained in our final conditional logistic-regression model: failure to graduate from high school, living alone, consumption of alcoholic beverages, previous hospitalization due to drinking, current use of prescription medication for depression or mental illness, and use of illicit drugs. Each of these variables was strongly and independently associated with an increased risk of suicide in the home. Our final model did not include any interaction terms between gun ownership and these covariates or between gun ownership and the matching variables (Table 4). Controlling for the effects of these six covariates revealed that keeping one or more firearms was strongly associated with an increased risk of suicide in the home (adjusted odds ratio, 4.8; 95 percent confidence interval, 2.7 to 8.5). Stratified analyses using the same model demonstrated that the presence of guns in the home was associated with an increased risk of suicide among women as well as men, across all age strata, and among whites (Table 5). No subgroup analysis limited to nonwhites could be performed. Restricting the analysis to matched pairs with data obtained from case proxies who lived in the home of the victim verified the association. Another analysis restricted to case subjects with no history of depression or mental illness revealed that guns were even more strongly associated with suicide in this group than in the study population overall. An analysis stratified according to the method of suicide revealed that the link between gun ownership and suicide was entirely due to much higher odds of suicide with a firearm. Suicide by any other means was not significantly associated with the presence of a gun in the home (Table 5).

Table 6. Table 6. Risk of Suicide in the Home in Relation to Various Patterns of Gun Ownership.

Subsequent analyses revealed that case subjects in households with loaded firearms (odds ratio, 9.2) were at higher risk of suicide than those in homes with unloaded firearms (odds ratio, 3.3), as compared with those in homes with no guns (Table 6). Households with guns kept in an unlocked place were associated with a higher risk of suicide than households in which guns were kept in a locked place, and homes with one or more handguns were associated with a risk of suicide almost twice as high as that in homes containing only long guns. However, homes with guns of any sort were associated with a significantly higher risk of suicide than homes without guns, regardless of the type of gun or the method of storage.

Our study was restricted to suicides occurring in the victim's home because a previous study has indicated that most suicides committed with guns occur there³⁰ and because almost half the homes in America contain one or more firearms.²⁸ If readily available firearms increase the risk of suicide, this effect should be most noticeable in an environment where guns are commonly kept. Our results offer strong evidence that the ready availability of guns increases the risk of suicide in the home. Homes with handguns and homes where firearms were not locked up or were kept loaded were even more likely to be the scene of a suicide than homes where firearms were kept securely stored. Few victims acquired their guns within hours or days of their death; the vast majority had guns in the home for months or years.

Case–control studies offer many advantages over geographic comparisons and time-series designs, but they are prone to potential sources of bias.³¹ We minimized selection bias by including all victims of suicide in the home as eligible case subjects and by using an explicit protocol for the random selection of matched controls. High response rates for both case proxies and controls (80 percent) limited any potential nonresponse bias. To minimize differential recall between case proxies and controls, we delayed our interview to allow for the initial grief process and selected relatively objective variables for our final regression analysis.

The possible misreporting of sensitive information by controls was our greatest concern, since we had no way to verify their accounts independently. The underreporting of risk factors such as a history of family violence, alcohol abuse, or illicit-drug use would amplify the apparent effects of these factors among the case subjects and bias our adjusted odds ratio for gun ownership toward the null hypothesis. In contrast, underreporting of gun ownership among controls would tend to bias the odds ratio upward. For two reasons, we think misreporting of gun ownership was not a problem. First, in a pilot study of owners of registered handguns, we demonstrated that responses to our questions about gun ownership were generally valid.³² Second, the rates reported by our control subjects in both cities for gun ownership in general and handgun ownership in particular are comparable to those noted in earlier social surveys³³ and are higher than those predicted by Cook's 1979 gun-prevalence index.³⁴

Three additional limitations warrant comment. This study addresses only suicides in the home and does not examine the relation between the availability of firearms and suicides out of the home. However, during the study period the proportion of out-of-home suicides involving firearms in each county exceeded the prevalence of gun ownership in the respective control group. Second, our research was conducted in two metropolitan counties with relatively small rural populations. Therefore, our conclusions may not be generalizable to rural communities. Finally, we cannot exclude the possibility that gun owners (and people who live in homes with guns) may be psychologically predisposed to commit suicide. Although this idea seems unlikely, one cannot readily control for "psychological confounding" of this sort in a case–control study.

The odds ratio associated with the presence of one or more firearms in the home increased from 3.2 to 4.8 after adjustment for confounding variables. An explanation of this effect can be found by examining the relation between gun ownership and two of the risk factors included in our final regression model. Use of psychotropic medication and the presence of guns in the home were both risk factors for suicide, but persons taking psychiatric medication were actually less likely to live in a home with a firearm than those not taking such medication. This was especially true among the case subjects. Fewer case subjects and controls who lived alone owned guns than did persons living with others. Since both living alone and taking psychotropic medication were strongly associated with suicide but negatively correlated with gun ownership, controlling for their effects increased the apparent strength of the association between suicide and the presence of a gun in the home.

Two case–control studies have previously explored the relation between firearms in the home and suicide. In 1988, Brent and colleagues reported that guns were more likely to be present in the homes of adolescent suicide victims than in the homes of demographically similar suicidal inpatients.³⁵ They subsequently verified this finding in a larger series of adolescent suicide victims, a group who had attempted suicide, and a third group of nonsuicidal controls with other psychiatric illnesses.³⁶

In our previous study, we compared suicide rates in King County, Washington, and Vancouver, British Columbia, and found little evidence of a link between firearm regulations and rates of suicide in the community.¹³ However, the ecologic analysis employed in that study is best used to generate hypotheses.²¹ Validation requires a more rigorous study design, such as the one used here.

The ready availability of firearms appears to be associated with an increased risk of suicide in the home. Although this risk is present in households where there is mental illness, it is also evident in households in which no one was previously known to be mentally ill or depressed. People who own firearms should carefully weigh their reasons for keeping a gun in the home against the possibility that it may someday be used in a suicide.