Alcohol and Risk of Violent Death to Persons Living in the Home

Unhooked Science Readings

The unhooked.com science section contains selected educational readings from the scientific and popular literature about alcoholism, addiction, and recovery. The views expressed in the articles are those of their authors and not necessarily those of the science pagemaster or the webmaster or of the person who suggested the article to the list. This material is made available solely for the nonprofit educational use of unhooked.com readers as an aid in their personal recovery, and no other use is authorized or intended. Click here for the current Science Section reading list.

Alcohol and Illicit Drug Abuse and the Risk of Violent Death in the Home

Frederick P. Rivara, MD, MPH; Beth A. Mueller, DrPH; Grant Somes, PhD; Carmen T. Mendoza, MSPH; Norman B. Rushforth, PhD; Arthur L. Kellermann, MD, MPH

Context.--While acute alcohol and illicit drug use are common in homicide and suicide victims, the role of chronic substance use in violent death is unclear.

Objective.--To measure the magnitude of risk of violent death in the home associated with alcohol use or chronic abuse and use of illicit drugs.

Data Sources.--Data obtained from a case-control study of risk factors for homicide and suicide in 3 large metropolitan areas of the United States.

Design.--Matched case-control study including 388 homicide cases, 438 suicide cases, and equal numbers of controls matched for age, sex, race, neighborhood, and county. Data were analyzed by means of conditional logistic regressions in which other potential risk factors for violent death were also considered.

Outcome and Exposure Measures.--Homicide and suicide victims were identified from medical examiner reports in Shelby County, Tennessee; King County, Washington; and Cuyahoga County, Ohio. Structured interviews were conducted with proxy respondents close to the decedents to obtain information about alcohol or illicit drug use, and history of alcohol-related hospitalization or trouble at work because of drinking by the subject. Data about alcohol use by others living in the same house as the subject were also obtained.

Results.--The risks of homicide and suicide associated with alcohol or illicit drug use were elevated, as were the risks of violent death associated with several indicators of chronic alcohol abuse. In addition, nondrinkers living in a home with alcohol users were at increased risk of homicide (odds ratio, 1.7; 95% confidence interval, 0.98-3.0), and non-drug-using individuals residing in homes with illicit drug users were at greatly increased risk of homicide (odds ratio, 11.3; 95% confidence interval, 4.4-28.8).

Conclusions.--Alcohol and illicit drug use appear to be associated with an increased risk of violent death. The risk of homicide was increased for non-substance-abusing individuals living in households in which other members abused alcohol or drugs. The concept of the individual at risk of homicide should be broadened to include not only the abuser but also those who may be at risk because of their exposure to others.

ALCOHOL is a factor in a large proportion of violent deaths, including both homicides and suicides. The classic study by Wolfgang1 in 1958 reported evidence of alcohol consumption in 64% of homicide cases in Philadelphia, Pa. Since then, homicide studies in the United States and other countries have reported high rates of alcohol use by the victim, perpetrator, or both. Similar rates of alcohol and other drug use have also been detected in suicide victims.2

Although many reports have demonstrated a high prevalence of substance use in victims of violent death, few have collected comparable data from a control group to assess the independent effects of alcohol and other drug use while controlling for the effects of other previously identified personal and environmental risk factors. Moreover, nearly all studies have focused exclusively on intoxication or drug use in the victim at the time of death and have not examined the role of chronic substance use problems by the victim or other members of the household in the occurrence of violent death in the home.

To clarify the relationship between substance abuse and risk of violent death, we used data from 2 large, population based, case-control studies of homicides and suicides that occurred in the home. We were interested in determining the magnitude of the risk of violent death in the home associated with alcohol and drug abuse, relative to that for individuals without these exposures, after adjusting for effects of other known and potential risk factors for violent death.

SUBJECTS AND METHODS

Identification of Cases

The data used for this study consisted of medical examiner reports and interview information on 438 suicides and 388 homicides occurring in the victims' residences. The study was limited to violent death in the home because we were interested specifically in factors in the home environment that were associated with risk of violent death. Suicide cases were identified from incidents occurring between August 23, 1987, and April 30, 1990, in Shelby County, Tennessee, and King County, Washington. Homicide cases were identified from incidents occurring in the same 2 counties between August 23, 1987, and August 23, 1992, and from incidents occurring in Cuyahoga County, Ohio, between January 1, 1990, and August 23,1992. The methods used for this study have been described previously3,4 and are briefly reviewed below.

Selection of Cases and Recruitment of Case Proxies

All medical examiner reports of homicide and suicide were screened during the study period in each county to identify events that occurred in the victims' home. A home was defined as any house, apartment, or dwelling occupied by a victim as that person's principal residence. If more than 1 individual died in the same incident, only the oldest victim was included as a case in the study to maintain statistical independence, since some of the analyses were of variables pertaining to the whole household. Also excluded were cases designated by the medical examiner in each area as fatal child abuse because the circumstances related to such events were thought to differ enough to require separate study.

Police or medical examiner investigators at the scene identified persons close to the victim who might provide an interview at a later date. Most often this was a relative, although friends and neighbors were also considered proxy respondents for obtaining information concerning characteristics and behavior of the deceased. Approximately 3 weeks later, each proxy was sent a letter outlining the project. A follow-up telephone call was made a few days later to arrange a time and place for a personal interview, at which time informed consent was obtained.

Selection of Controls

After each interview with a case proxy, we sought a control subject matched to the victim by age (15-24, 25-40, 41-60, or ---61 years), sex, race, neighborhood, and county. To minimize selection bias, the controls were identified by a previously validated procedure for the random selection of a matching household in the neighborhood5,6 After marking off a 1-block avoidance zone around the home of the case subject, interviewers started a neighborhood census at a randomly selected point along a predetermined route. Households in which no one was home were visited twice more, at different times of the day and days of the week. An adult (a person aged 18 years or older) in the first household with a member who met the matching criteria was offered a $10 incentive and asked to provide an interview concerning behaviors and exposures of individuals in the household. Because of the necessity of relying on proxy respondents to obtain data for the cases, whenever possible we attempted to interview a proxy respondent for the actual matching control subject as well. For controls who were living alone, all interviews were done with the controls themselves.

Interviews

Case and control interviews were identical in format, order, and content and used a brief, highly structured questionnaire. The order of the questionnaire was such that more sensitive questions were broached later in the interview. The interviews were designed to ascertain exposure to a variety of individual, household, and environmental factors that may be associated with the occurrence of violent death in the home. Questions about alcohol use were adapted from the Short Michigan Alcohol Screening Test.? Interviewers queried about current drinking by controls, members of the household, or the case at the time of death; whether case, control, or household members had ever been in trouble at work because of drinking, or ever been hospitalized because of drinking; use of illicit drugs, such as marijuana, cocaine, heroin, barbiturates, or amphetamines; and history of ever being arrested.

To confirm the reliability of the interviewers, supervisory personnel contacted a 10% random sample and administered an abbreviated version of the questionnaire. Concordance between these responses and those given to the interviewers was 95%. Although efforts were made to conduct every interview in person, proxy respondents for the cases were more likely than controls or their proxies to request a telephone interview (approximately 40% and 10%, respectively), and slightly greater proportions of individuals responding by telephone were relatives of the deceased. However, the age and sex distributions were similar for those responding in person and via telephone within the case and control groups. Respondents for both homicide and suicide cases were more likely to be male than were respondents for controls, but they had similar age distributions. Approximately 45% of both homicide and suicide control respondents were proxies. For both cases and controls, most proxy respondents were relatives (62% and 93% of homicide case and control proxies, and 78% and 68% of suicide case and control proxies).

Data Analysis

Standard descriptive statistics were used to summarize information related to the scene of each violent death and personal characteristics of the subjects. Analysis of site-specific data to compute risk estimated for the associations of alcohol and other drug use with homicide and suicide within each geographical location determined that the magnitude of these risks did not measurably differ across sites. Multivariate analyses using conditional logistic regressions with data for all sites were conducted to estimate the risks of violent death associated with the presence of various histories of alcohol and illicit drug use (marijuana, cocaine, heroin, barbiturates, or amphetamines) and behaviors reported in the questionnaire. These included patterns and extent of substance use and abuse by case or control, and by other members of their respective households. Because of the manner in which the questions were asked, many items were evaluated separately for subjects who lived alone and those living with others. Factors that were considered as potentially affecting the relationship of substance use and violent death included living alone, history of psychiatric illness or depression in the case or control ("was the individual depressed or having psychiatric problems at the time of death?"), possible criminal history of the case or control (as indicated by whether the subject, by respondent report, had ever been arrested), and presence of a gun in the home. In addition, the potential effects of alcohol use by the case or control were considered in the risk estimates for drug use and vice versa. Only those variables that meaningfully altered the risk estimates were retained in the final estimates.

Human Subjects

The study method was reviewed and approved on an annual basis by the institutional review board of each participating university. A $10 incentive was offered to eligible respondents to encourage participation, and informed con- sent was obtained before each interview. Respondents were assured that their answers would be kept confidential and that they could terminate or revoke the interview at any time. When a telephone interview was requested, written consent was obtained by mail. All of the information collected for this study is protected by a federal certificate of confidentiality.

RESULTS

Study Population

There were 1860 homicides and 803 suicides in the study counties during the study period. Of these, 444 homicides (23.9%) and 565 suicides (70.4%) took place in the home of the victim. After excluding the younger victims in 25 incidents of multiple deaths, 1 delayed death, 5 late changes to the death certificate, 3 cases excluded by the medical examiner staff, and 2 deaths not reported to the study team, there were 420 homicides and 554 suicides available for study. Interviews were obtained with proxies in 389 (92.6%1 of the homicide cases and 442 (79.8°/t) of the suicide cases. The households that agreed to be interviewed did not differ from the households of those who refused with respect to the age, sex, or race of the victim or the method of death.

Interviews with a matching control or proxy were obtained for 99% of the cases, yielding 388 matched pairs for homicide and 438 for suicide. Three hundred fifty-seven homicide pairs and 404 suicide pairs were matched for all .3 variables; 27 and 33, respectively, for 2 variables; and 4 and 1, respectively, for 1 variable (sex). The demographic characteristics of the cases and controls were fairly similar, with the exception that twice as many cases as controls within each group were found to live alone. Most homicide (63.1%) and suicide (71.9%) victims were male (Table 1). Homicide victims were generally slightly younger than suicide victims, with nearly one half (47.7%) being between the ages of 25 and 44 years. Differences in race were also observed between the 2 groups; most (89%) suicide victims were white, compared with 32.9% of homicide victims. The heads of households for suicide victims were somewhat better educated than those of homicide victims and had higher socioeconomic status on the Hollingshead score.

Homicide

Subjects who reportedly drank alcohol, were ever in trouble at work from drinking, or who were ever hospitalized because of a drinking problem were at increased risk of violent death by homicide (Table 2). Reported use of illicit drugs or ever being arrested were also associated with an increased risk of homicide. Subjects who reportedly used both alcohol and drugs were at markedly increased risk of homicide, relative to those who used neither substance (odds ratio [OR], 12.0: 95% confidence interval [CI], 5.7-25.4). Having a "psychiatric problem or depression" at the reference date was associated with a nearly 3-fold increased risk of homicide (OR. 2.7: 95% CI. 1.7-4.3). This risk, however, was confined to those who were not reportedly receiving medication for a psychiatric illness (OR. 4.0: 95% CI, 2.2-7.4). All estimates were adjusted for reported use of illicit drugs and/or alcohol.

The risk of homicide associated with drinking among males was 3.0 (95% CI. 1.9-4.7), whereas a modest elevation not reaching statistical significance was noted among females (Table 3). This pattern was also observed when the risk of homicide associated with illicit drug use was examined separately for males and females. Use of alcohol was associated with approximately 2-fold increased risks of homicide within all age groups examined. However, illicit drug use was observed to be associated with an increased risk of homicide only among those younger than 50 years (OR, 6.2; 95% CI, 3.1-12.6). Risk estimates of the 48 case-control pairs whose race was other than white did not differ from those calculated for the whole study population.

Living alone, rather than with other individuals, affected the alcohol-homicide and illicit drug use-homicide associations somewhat differently. Relative to those who lived with others, subjects who lived alone appeared to have greater risk of homicide regardless of whether they drank alcohol (Table 4). Among those living in households with other members, the risk of homicide associated with the case only or in addition to other household members' use of alcohol was 3.5 (95% CI, 2.3-5.3). The risk for nondrinking subjects living with others who drank was 1.7 (95% CI, 0.98-3.0), relative to those living with others in a nondrinking household. There were no differences in this risk for males compared with females. Relative to those living with others in a nondrinking household, nondrinking subjects who lived alone were at nearly 5-fold increased risk of homicide (OR, 4.8; 95% CI, 2.6-9.0); drinking subjects who lived alone had the greatest increased risk of homicide (OR, 9.8; 95% CI, 5.0-19.4). In apparent contrast, the greatest elevations in risk of homicide associated with illicit drug use were observed among subjects who were living with others. Relative to those living with others in households without illicit drug use, subjects who either used illicit drugs, or who lived in homes where illicit drug use occurred, had similar, and markedly elevated, risks of homicide (Table 4). Reportedly drug-free subjects who lived alone were also at increased risk of homicide, although the magnitude of their risk was lower (OR, 4.8; 95% CI, 2.8-8.2); those living alone with reported illicit drug use had a risk of 7.3 (95% CI, 2.0-26.2).

Suicide

Factors noted to be associated with risk elevations for death by homicide were generally also associated with suicide occurrence (Table 2). Drinking alcohol was associated with a nearly 2-fold increased risk of suicide (OR, 1.8; 95% CI, 1.3-2.5). and problem drinking associated with trouble at work or resulting in hospitalization was associated with 6-and 10-fold increased risks. Reported use of drugs was associated with a 7-fold increased risk of suicide. The highest elevation in risk was observed for those who reportedly used both alcohol and illicit drugs (OR, 16.6: 95% CI, 7.0-39.2). Subjects who reportedly had a psychiatric problem or depression, with or without use of psychiatric medications, were at greatly increased risk of suicide (OR. 106.6: 95% CI, 33.2-342.1), unlike the case for homicide, where the risk was restricted to those who reportedly were not receiving medication for their condition.

Among males, the risks of suicide associated with use of alcohol was 2.3 (95% CI. 1.5-3.4); among females the risk was modest and did not reach statistical significance (Table 3). The risks of suicide associated with use of illicit drugs, however, were markedly elevated among both males and females. A modest, but not statistically significant, increased risk of suicide associated with drinking alcohol was noted for those older than 50 years (OR. 1.3; 95% CI, 0.8-2.0); the risk among those younger than 50 years was 2.8 (95% CI, 1.7-4.6). Use of illicit drugs was associated with a greater than 6-fold increased risk of suicide among those younger than 50 years; the risk among those in the older age group could not be estimated because it was reported too infrequently. Similar risk elevations were noted among the nonwhite subjects; in particular, however, the risk of suicide associated with drinking alcohol was markedly elevated in this group (OR. 12.0: 95% CI. 2.8-50.8).

As with homicide, the risk of suicide was somewhat modified by whether subjects lived alone or with others. Relative to nondrinkers living with others, non-drinking subjects who lived alone were at nearly 3-fold increased risk of suicide (OR, 2.7: 95% CI, 1.5-4.9). Use of alcohol, whether or not the subject lived alone, was associated with 2- to 7-foldincreased risk of suicide (Table 4). Unlike the case with homicide, however, nondrinking individuals who lived with others who drank were not at increased risk of suicide. Drug use, either by the subject or by others in the home, was associated with increased risks of suicide, with the greatest elevation noted for those who used drugs and lived alone (OR. 25.9: 95% CI. 5.7-117.4).

COMMENT

In this study, a history of alcohol and illicit drug use by the subject was strongly associated with increased risks of both homicide and suicide, and increases were generally observed among all sex and racial subgroups examined. Of note, in addition to the effects of drinking or illicit drug use by the subject, these behaviors by others within the same household affected the risk of violent death. Even in the absence of illicit drug use by the subject, increased risks, in particular for homicide, were observed if others in the household reportedly used illicit drugs. Nondrinking subjects living in homes with others who drank were also at increased risk of homicide. These findings have important implications for interventions to reduce the risk of violent death in the home.

Homicide

Most previous studies have been limited to demonstrating that many homicide victims have been drinking at the time of death. According to reviews of the literature,9-11 most studies report that at least 40% and as many as 70% of homicide victims have positive blood alcohol concentrations at autopsy. These studies also found higher rates in males than females and in young adults than in older adults and the elderly. However. few studies have examined the role of chronic alcohol abuse in victims or have used appropriate control groups for comparison. The National Mortality Follow-back Survey found that homicide and suicide victims were more likely to be moderate and heavy drinkers than were individuals who died of other causes between 25 and 64 years of age. In a study of nonfatal injuries, 41% of individuals injured in assaults reported getting drunk once or more per week, compared with 12% and 21% of those involved in motor vehicle crashes and falls, respectively. There is previous evidence that, among women, alcohol appears to be associated with violence as well. Women with alcoholism were much more likely than a random sample to experience spousal violence14 In our study, a more modest association of alcohol use and homicide was observed for women than for men. We also found no differences in the risks for women compared with men living in a household in which another member of the household had trouble with alcohol.

In contrast to the lack of work on alcoholism in victims of violence, many studies have demonstrated a high rate of chronic alcohol abuse among perpetrators15,16 Alcohol use increases the likelihood that the drinker will be involved in violence either as a victim or as a perpetrator,17 and a recent study of youth violence indicates that many victims are in fact also offenders,i8 The present study did not evaluate the potential role of victim status as previous perpetrator; however, it did clearly demonstrate that alcohol use and abuse, known to be strongly associated with offender status, play a similar role for status as victim in acts of violence. The association of alcohol consumption and homicide likely results from many factors. By impairing a person's ability to process and interpret information correctly, alcohol may increase the potential for miscommunication in interpersonal interactions,11 and this could lead to conflict and escalation into violence. Alcohol impairs judgment, possibly causing individuals to place themselves in situations at high risk of violence. Laboratory studies demonstrate that drinking increases aggressive behavior20; it also changes one's expectation of violence, in that drinkers are likely to believe that alcohol use leads to aggressive behavior, creating a self-fulfilling prophecy21 Chronic alcohol abuse may be a marker for antisocial personality, which is associated with increased rates of violence and victimization. Several longitudinal studies noted that, relative to nonaggressive children, aggressive and antisocial children were significantly more likely to be heavy drinkers as adults.22,23

In addition to these potential alcohol-related pathways to violence involving pharmacological effects or antisocial tendencies, the association of illicit drug use and risk of violent death may also (and perhaps largely) result from drug-seeking activities, such as interaction with drug dealers and theft to obtain resources for drug purchases. Illicit drug use by members of the household increases a woman's risk of death at the hands of a spouse, lover, or close relative 28-fold.24 The increase in youth violence during the past decade is concurrent with the crack cocaine epidemic25 A nearly ubiquitous presence of handguns in an environment of robbery and burglary for procuring illicit drugs can only serve to increase the likelihood of violence. That the true risk factor may be the drug culture environment is supported by our finding that even non-drug users who lived in households where illicit drug use occurred were at greatly increased risk of homicide.

Suicide

Numerous studies have documented positive drug and alcohol test results in a substantial proportion of suicide victims2 As with homicide, however, the relationship of suicide to chronic substance abuse is less clear. The 10-fold increased risk of suicide we observed associated with hospitalization for drinking is consistent with a risk of 6.9 reported in a 40-year follow-up study of 40 000 Norwegian men26 A follow-up of 13 673 participants in the Epidemiologic Catchment Area surveys reported an 18-fold increased risk of attempted suicide associated with active alcoholism and a 62-fold increased risk associated with use of cocaine27 It has also been noted that depressed patients who have alcoholism are more likely to be suicidal (OR, 1.59) than depressed patients who are not alcohol abusers.

Alcohol problems appear to be more common in male than in female suicide attempters. Merrill et a129 found that alcohol-related problems such as alcohol dependence, social problems, or physical illness were present in 34% of men and 15% of women who attempted suicide. In our data, the association of alcohol use and suicide was stronger in males than females. The same was true for illicit drug use, although the association of drug use with suicide was strong in both sexes. Alcohol may play a role in suicides involving impulsive acts, particularly among the young, as found in our study. In 1 study of teenage suicide, victims who used firearms were 5 times more likely to have been drinking than teenagers who used other methods of suicide.30

Having 1 addictive disorder, such as alcoholism, is associated with a 7-fold increased risk of also having a second addictive disorder, and 37% of those who abuse alcohol and one half of those who abuse other drugs also have a mental disorder21 We found that the use of both alcohol and illicit drugs was associated with a 16-fold increased risk of suicide, markedly higher than that observed for individual use of either substance. As a greatly increased risk of homicide was also observed for subjects with multisubstance use, identification of such individuals coupled with effective interventions may serve to prevent an even wider spectrum of violence. Furthermore, given our findings related to increased risk among nonsubstance users in households where substance use occurred, one could argue that these are also clues that might be useful as indicators of individuals who may be at high risk of violent death. Primary prevention of alcohol and other drug abuse is important; however, further thought concerning prevention of their potentially violent sequelae is called for. Exploration of potential interventions similar to some of the alcoholism intervention programs conducted in emergency departments32 and identification of points of contact in which they might be employed (schools. community health centers, etc) are warranted.

Limitations

A number of important limitations must be considered in interpreting our findings. First, given the violent nature of the case subjects' death, it is possible that case proxy respondents were more likely than control respondents to report that the case had ever been arrested, or had psychiatric problems, and that this bias may have been responsible for at least part of the elevated risks we observed. It is also possible that subjects with an arrest history were more likely to have continued contact with individuals involved in criminal activities, placing themselves in relatively high-risk environments for homicide, or that antisocial behavior associated with a criminal history is also associated with suicide. In an attempt to control for this. to at least some extent, in the reporting of a psychiatric history, we asked further questions concerning whether the subject was using medication for the psychiatric problem, with the argument that those reportedly using medication would have the greatest likelihood of truly being under care for a diagnosed psychiatric problem. The psychiatric problem-homicide association was observed only for those who reportedly were not receiving medication; for suicide, marked associations were observed whether or not medication use was reported. Whether reporting bias may have accounted for some level of these associations is unclear; however, given their magnitudes, it is plausible that they represent true associations, indicating yet another group of individuals at high risk of violent death.

Second, this study was limited to homicides and suicides that occurred in the home. The dynamics of violent death outside of the home may be quite different. Third, the study relied on proxy respondents for all the cases and for as many of the controls as was possible. The proxy's knowledge of alcohol and other drug use by the victim may have been inaccurate or possibly biased, as described above with respect to our evaluation of previous psychiatric or arrest history. Mis-classification may have been more of a problem for cases than controls, since some of the control interviews were with the study subjects themselves. However, restriction of the analyses to case-control pairs in which both respondents were proxies (187 homicide pairs and 221 suicide pairs) resulted in similarly elevated risk estimates. Fourth, the interviews were conducted usually within 3 to 4 weeks of the death, and some of the perpetrators of the homicides were still unknown. It is possible, although we believe unlikely, that some of the case respondents were in fact perpetrators. Fifth, our research was conducted in 3 urban counties and may not be generalizable to more rural areas, where demographic and social characteristics may be quite different. Sixth, the mechanisms behind the associations of drinking and drug use with violent death are not clear. Some subjects may have been drinkers or used other drugs as a result of living or having lived previously in violent households: conversely, drinking or other drug use might have contributed directly as a causal factor to the risk of violence. It is likely that both pathways are important, further emphasizing the need for both primary prevention of substance abuse and secondary intervention strategies aimed at reducing its impact.

Finally, alcohol and other drug abuse may play a role solely as markers for other risk factors, such as antisocial personality, or for environmental conditions that place individuals at risk. While it is possible that our measurement of these aspects was flawed because of the limitations of obtaining this information from proxy respondents, given the magnitude of our risk estimates, it is likely that our findings represent true associations. We also attempted to reduce any such confounding by examining the potential effects of other factors known or suspected to influence the risk of homicide and suicide, such as arrest history, presence of a gun in the home,3,4 and mental illness or depression.29 Although some of these have been previously demonstrated to have strong, independent associations with our outcomes, none was observed to meaningfully change the magnitude of the associations for alcohol or other drug use with the occurrence of homicide or suicide.

Implications

This study supports the need to address alcohol and other drug abuse and its relationship to risk of violent death. The problem of substance abuse should be addressed on multiple levels, including primary and secondary prevention. and should include screening for alcohol and other substance abuse problems in the emergency department, particularly in trauma patients32 or at other points of contact, as well as in other settings for persons with depression, mental illness, or suicide attempts. Brief interventions by primary care physicians can have a significant effect on weekly excessive alcohol use and binge drinking.33 In addition, people with alcoholism should be assessed for concomitant depression and treated accordingly34 Our concept of the individual at risk for violent death should be broadened to include not only the substance abuser, but also those who may be at risk because of the presence of others within the household who are substance abusers. Community interventions can play an important role as well, particularly in decreasing alcohol abuse. Increasing the age for legal purchase of alcohol,35 increasing the excise tax on alcohol,36 countering alcohol advertising and promotion, and server management and responsibility training37 can decrease the rate of alcohol abuse and violence associated with alcohol use. Community policing and other interventions appear to have attenuated the rise in violence associated with drug abuse25 The effect of such measures on suicide is unknown.

The problem of violence in our society is a complex one and will require multifaceted solutions. This study indicates that addressing the problem of substance abuse on many levels may have a substantial impact on rates of violence and should be pursued.

From the Harborview Injury Prevention and Research Center, Seattle, Wash (Drs Rivara and Mueller); Department of Epidemiology, University of Washington, Seattle (Drs Rivara and Mueller and Ms Mendoza); Department of Preventive Medicine, University of Tennessee, Memphis (Dr Somes); Department of Biology, Department of Epidemiology and Biostatistics, and Center for Adolescent Health, Case Western Reserve University, Cleveland, Ohio (Dr Rushforth); and Emory Center for Injury Control, Emory University, Atlanta, Ga (Dr Kellermann).

Corresponding author: Frederick P. Rivara, MD, MPH, Harborview Injury Prevention and Research Center, 325 Ninth Ave, PO Box 359960, Seattle, WA 98104-2499.