Attribution of Alcohol to Violence-Related Injury: Self and Other’s Drinking in the Event

Abstract

Objective:

Alcohol can result in harm (including injury) not only to the drinker but also to others; however, little research exists on the additional proportion of violence-related injuries that can be attributed to the perpetrator.

Method:

Data are reported from emergency department studies in 14 countries on the prevalence of patients’ self-report of drinking within the 6 hours before the violence-related injury event, patients’ belief that the event would not have happened if they had not been drinking at the time, and patients’ perception that the perpetrator had been drinking. Alcohol-attributable fraction was calculated based on the patients’ perception that their own drinking was causally related to the event and on their perception that the perpetrator had been drinking.

Results:

Across all countries, 62.9% of the violence-related injuries involved alcohol use on the part of the victim, the perpetrator, or both. Rates of others definitely drinking, as perceived by the victim, ranged from 14% to 73% across countries and was positively associated with patients’ own drinking in the event and with attributing a causal association between their drinking and the event. Estimates of alcohol-attributable fraction were 38.8% when the victim and perpetrator were considered together compared with 23.9% when only the patient was considered and varied by country-level drinking pattern.

Conclusions:

These findings suggest adjustments that could be made to global burden of disease estimates because of violence-related injury morbidity to better reflect alcohol-attributable fraction when drinking by others and country-level drinking patterns are taken into account.

Alcohol use can result in harm (including injury) not only to the drinker but also to others (often called externalities), and this is an area of increasing research interest from a societal perspective as well as from an individual-level perspective (Greenfield et al., 2009; Room et al., 2010). From the perspective of societal costs of injuries, alcohol-related harm can result in direct costs such as those incurred by governments, health insurance companies, or hospital emergency departments (EDs) as well as in indirect costs because of disability and death. From the individual-level perspective, drinking by others may affect a person through risk taking, clumsiness, or inattention, which may result in an accident, or through intentional harm from another, which may result in a violence-related injury. This last possibility is the subject of this study.

Prior research suggests that perpetrators of violence-related events are likely to have been drinking at the time (Brecklin and Ullman, 2002; Laslett et al., 2010); however, epidemiological studies in EDs have generally obtained data only on self-reports of one’s own drinking in the injury event, and no prior ED studies have obtained reports from the victims regarding their perception of drinking by the perpetrator. The aim of this study was to help fill this gap in the literature.

Estimates of the fraction of violence attributable to alcohol use have been based on both individual- and population-level data, both of which have been found to vary across countries and cultures (Room and Rossow, 2001). At the aggregate level, an increase in a population’s drinking level tends to be followed by an increase in rates of violence. There are also cultural differences in alcohol’s role in violence, so that the same increase in consumption seems to produce different levels of change in rates of violence in different societies. For example, an increase in consumption has been found to have a greater effect on rates of violence in some northern and eastern European countries, whose drinking patterns are characterized by acute intoxication, compared with southern European countries, which exhibit a Mediterranean drinking pattern of frequent but lighter drinking (Bye, 2008; Lenke, 1990; Rossow, 2001).

The literature is relatively small on harm in the form of violence-related injury from others’ drinking. An early U.S. study of domestic physical assaults found that 40% of the victims reported that the perpetrator had been drinking, whereas the police judged that the perpetrator had been drinking in only 21% of the cases (Bard and Zacker, 1974). Few general population studies have obtained reports on harm, including that related to physical assault, from others’ drinking. A 1989 Canadian national survey reported a 20% prevalence of assaults because of another’s drinking on a lifetime basis and a 7% prevalence for the last 12 months (Eliany et al., 1992), whereas a Canadian community-based study found that 51% of perpetrators and 30% of victims reported having been drinking at the time of the event (Pernanen, 1991). A California survey found a prevalence for physical abuse because of another’s drinking of 30% on a lifetime basis and 6% for the last 12 months (Jones and Greenfield, 1991), whereas U.S. data from the 2005 National Alcohol Survey found a prevalence for assault because of others’ drinking of 28% on a lifetime basis and 2.4% for the last 12 months (Greenfield et al., 2009). Generally, studies have found that heavy drinkers are more likely than others to suffer violence. This was true for British arrestees (Norton and Morgan, 1989), in the 1989 Canadian population survey (Eliany et al., 1992), in the 2005 U.S. survey (Greenfield et al., 2009), and in the 1994 and 1995 Australian national surveys (Laslett et al., 2010; Makkai, 1997).

Victimization surveys have provided some data about the victim’s perception that the perpetrator had been drinking. In the 1995 Australian population-based Personal Safety Survey, more than two thirds of the assailants were reported by the victim to have been drinking (Laslett et al., 2010), whereas in the U.S. National Violence Against Women Survey, 63.5% of the victims perceived that their assailants had been using alcohol and/or other drugs (Brecklin and Ullman, 2002).

Studies in emergency departments

Studies in hospital EDs have been an important source of data documenting the association of drinking at the time of the event with being a victim of violence, but these studies have provided little data on alcohol involvement of the perpetrator. A review of probability samples of ED patients reporting alcohol and injuries resulting from violence found significantly higher rates for a positive blood alcohol concentration (BAC) as well as higher rates for reporting drinking prior to the event for those sustaining violence-related injuries than for patients sustaining injuries from other causes, with rates ranging from 17% to 70% for BAC and 27% to 55% for reporting drinking prior to the event (Cherpitel, 1994). A subsequent review of ED studies (since 1995; Cherpitel, 2007) found rates for violence-related injury ranging from 22% to 70% for a positive BAC and 36% to 84% for self-reported drinking before the event. A study of violence-related injuries in ED studies across 15 countries found rates ranging from 17% to 78% for a positive BAC and from 25% to 78% for self-reported drinking within 6 hours before the event (Wells et al., 2007), with those who reported drinking being 22 times more likely across all countries to be admitted for a violence-related injury compared with those not drinking (Cherpitel and Ye, 2010). In a meta-analysis of ED studies across six countries, the estimated attributable fraction of injury from drinking (or the proportion of injury that would be eliminated in the absence of exposure to alcohol) was found to be greater for violence-related injuries than for injuries from other causes, with 43% of violence-related injuries attributed to drinking in the 6 hours preceding the event and 27% to a usual pattern of heavy drinking (compared with 6% of injuries from other causes for both risk factors) (Cherpitel et al., 2005). These ED studies have included all patients admitted after having suffered violence; undoubtedly, some of these patients will have perpetrated the violent event even though they sustained an injury themselves.

Data are reported here from ED studies in 14 countries on the prevalence of patients’ self-report of drinking within the 6 hours before the violence-related injury event, the patients’ belief that the event would not have happened if they had not been drinking at the time (causal attribution), and the patients’ perception that the perpetrator had been drinking. Alcohol-attributable fractions (AAFs) are calculated based both on the patients’ perception that their own drinking was causally related to the event and on their perception that the perpetrator had been drinking. Because we were also interested in how cross-country variation in AAFs might vary by country-level drinking pattern, AAFs are also calculated for countries grouped according to aggregate-level drinking pattern.

Findings here will contribute to our knowledge regarding the magnitude of the additional burden that drinking by someone other than the victim of violence places on the ED. Findings also point to modifications of the AAF for violence-related injury when the perpetrator is considered alongside the victim and will inform burden-of-disease estimates of alcohol as a risk factor in violence-related injury, potentially suggesting adjustments that could be made to these estimates to better reflect AAF when drinking by others is taken into account and when country-level drinking patterns are also taken into account.

Method

Samples

Data analyzed here include 28 ED studies from 14 countries comprising the Emergency Room Collaborative Alcohol Analysis Project (ERCAAP), the World Health Organization (WHO) Collaborative Study on Alcohol and Injuries, and National Institute on Alcohol Abuse and Alcoholism (NIAAA) Collaborative Study on Alcohol and Injury (Table 1). All studies used a similar methodology (Cherpitel et al., 2003). Although most studies used only one ED site, multiple ED sites in a country or region were selected based on the diversity and size of the population served in their respective locales and to be representative of the different kinds of health care delivery systems or health plans available in that locale.

Table 1.

Drinking of the other person as reported by patients with an injury resulting from violence

Country	n	Yes/ definitely, %	Suspected, %	Don’t know/ unsure, %	No, %
Total	1,798	44.5	8.0	23.5	24.0
Argentina	55	25.7	16.2	20.7	37.3
Belarus	45	73.3	8.9	11.1	6.7
Brazil	45	31.1	15.6	26.7	26.7
China	488	25.2	9.8	27.7	37.3
Czech Republic	18	38.9	0.0	22.2	38.9
India	188	54.3	7.4	19.7	18.6
Mexico	72	37.5	15.3	26.4	20.8
Mozambique	108	26.9	5.6	40.7	26.9
New Zealand	26	50.0	0.0	20.9	29.1
South Africa	286	55.3	2.9	31.1	10.7
Sweden	35	14.3	0.0	51.4	34.3
Ireland	250	68.4	4.4	11.6	15.6
Korea	157	59.2	12.1	7.6	21.0
Switzerland	25	44.0	24.0	8.0	24.0

In all studies, probability samples of injured patients 18 years and older were obtained by approaching consecutive arrivals to each ED. In most studies, patients were sampled with equal representation of each shift for each day of the week. In three studies (Argentina, New Zealand, and South Africa), patients from some shifts and/or days of the week were oversampled, and the relevant analysis was adjusted by weighting as described below. Only patients arriving within 6 hours of the event were included in the sample. Patients were approached with an informed consent to participate in the study as soon as possible following arrival in the ED, and those who were severely injured were followed into the hospital and interviewed once their condition had stabilized. Completion rates averaged 82% across all studies. Reasons for noninterviews included refusing, being incapacitated, leaving before completing the interview, being in police custody, and having language barriers.

Measures

After obtaining informed consent, a cadre of interviewers trained at each site administered a 25-minute questionnaire that included items having to do with the type and cause of injury, whether the injury involved violence, drinking before the event, and whether patients attributed a causal relationship of their drinking with the injury. All patients were asked, “Did you get into a fight? Were you beaten, attacked, or raped?” If they answered that the injury was intentional and inflicted by someone else, they were asked, “In your opinion, had the person(s) who harmed you or that you fought with been drinking alcohol?” and the following options were coded as follows: yes/definitely, suspected, don’t know/unsure, and no. Patients were also asked, “In the 6 hours before and up to you having your injury, did you have any alcohol to drink—even one drink?” and coded as yes or no. Finally, patients who reported drinking before injury were asked, “Do you think that your injury would have happened even if you had not been drinking?” (causal attribution) and coded as yes, no, or not sure.

At the aggregate level, countries were grouped according to the level of detrimental drinking pattern (DDP), which is a measure developed by WHO from aggregate survey data and/or key informant surveys for a number of countries around the world (Rehm et al., 2001). This measure, ranging in scores from 1 to 4, is an indicator of the “detrimental impact” on health and other drinking-related harms at a given level of alcohol consumption; the higher the score, the higher the postulated detrimental effect of the same per capita consumption of alcohol (Gmel et al., 2007; Rehm et al., 2003a, 2003b).

Data analysis

Current analysis focused on violence-related injuries (i.e., patients whose injuries were the result of someone else’s intentional aggression), with 1,798 patients from 14 countries. Some counties are disproportionally represented in the total sample (e.g., 16% of the sample is from South Africa, whereas only 1% is from the Czech Republic). As described above, three studies that were sampled disproportionately were weighted so that the weighted prevalence rates from these studies represent the adjusted estimates from representative sampling. Tables 1 and 2 report prevalence rates of others’ drinking and cross-classifications of others’ drinking with self-drinking before injury.

Table 2.

Classification of patients injured by violence by the patient’s own drinking and the other person’s drinking

Variable	n	Both positive, %	Self, but not other, %	Other, but not self, %	Neither self nor other, %
Total	1,798	30.6	18.4	13.9	37.1
Argentina	55	18.6	32.8	7.2	41.5
Belarus	45	60.0	17.8	13.3	8.9
Brazil	45	13.3	22.2	17.8	46.7
China	488	15.4	19.3	9.8	55.5
Czech Republic	18	16.7	11.1	22.2	50.0
India	188	23.4	10.6	30.9	35.1
Mexico	72	26.4	23.6	11.1	38.9
Mozambique	108	11.1	17.6	15.7	55.6
New Zealand	26	44.9	39.4	5.1	10.6
South Africa	286	41.8	16.4	13.5	28.3
Sweden	35	11.4	42.9	2.9	42.9
Ireland	250	52.8	14.4	15.6	17.2
Korea	157	49.7	18.5	9.6	22.3
Switzerland	25	36.0	24.0	8.0	32.0

Alcohol-attributable fraction.

Because theoretically, a proportion of the violence-related injuries can be attributed to others’ drinking, estimates of the AAF are based on drinking by both the patient and the perpetrator (Table 3). Given the small sample size for some countries, analysis of AAF combined the samples across all 14 countries. AAF was calculated separately for self-drinking and others’ drinking and then combined for a single estimate of the AAF, including both self-drinking and others’ drinking. First, alcohol involvement was conservatively derived by cross-classifying patients’ self-drinking and perception of others’ drinking into three groups: (a) self-drinking and others definitely drinking; (b) self-drinking and others suspected, not sure/don’t know, or not drinking; and (c) no self-drinking and others definitely drinking (first column of figures).

Table 3.

Calculation of fraction of injury from violence causally attributed to the patient’s own drinking and to others’ drinking

Drinking by self and others (using conservative defi nition for others’ drinking)	(1)	(2)	(3)	(4)	(5)
	%	% of those who had been drinking who see it as causing the injuries	% of total injured where own drinking is seen as cause	Where own drinking is not causal and others had been drinking: Estimated %where others’ drinking would be seen as causal	% of total injuries where drinking by someone seen as causal
Self and others both positive	30.6	52.6a	16.1b	7.6c	23.7d
Self positive, others not	18.4	42.1	7.8	0	7.8
Self not, others positive	13.9	0	0	7.3e	7.3
Total	62.9		23.9	14.9	38.8

^aSelf-attributable question was not asked in India study; 52.6% was estimated using all studies combined without India (n = 1,610) and applied to India data.

^b16.1% = 30.6% × 52.6%; this is the proportion of all injuries in which own drinking is seen as causal.

^c7.6% = (30.6% − 16.1%) × 52.6%; i.e., for those who did not attribute their injury causally to their own drinking (30.6% − 16.1%). It is assumed 52.6% of those reporting the other was drinking would see the other’s drinking as causal. By using 52.6% in this step of calculation, we assume the same proportion in attributing injuries to self-drinking (Step ^b above) and to other’s drinking.

^d23.7% = 16.1% + 7.6%.

^e7.3% = 13.9% × 52.6%. Here, as in Step ^c above, we use 52.6% as the estimate of the fraction in which other’s drinking is seen as causal.

To estimate AAF because of patients’ self-report of drinking (for the two groups in which the patient has been drinking), the proportion of those patients who said the injury would not have happened if they had not been drinking (causal attribution) (Table 3, column 2) was multiplied by the proportion of violence-related injuries in the respective group (Table 3, column 1) to obtain the proportion of violence-related injury because of the patient’s own drinking (Table 3, column 3). We assume here that when a patient defined the injury as caused by their own drinking, this means it was not caused by the perpetrator’s drinking.

Next, to estimate AAF because of others’ drinking, excluding injuries the patient attributed to their own drinking, the proportion attributed to the patient’s own drinking (Table 3, column 3) was subtracted from the total proportion in the group in which both the patient and others had been drinking (Table 3, column 1). Assuming that the proportion of others’ drinking that would be causally attributed is the same as the proportion causally attributed to one’s own drinking, the remainder proportion in which the perpetrator has been drinking was multiplied by the causal attribution fraction (column 2) to yield column 4, as a conservative estimate, the proportion of injuries causally attributable to another’s drinking.

For each of the three groups, AAF for the patients’ drinking is added to that for others’ drinking to provide a total AAF for that group (Table 3, column 5), and AAFs for all three groups are then added together to provide an overall AAF for violence-related injury, using the conservative definition.

A similar procedure was conducted for three more liberal estimates of AAF, which additionally included (a) half of the patients who were unsure whether the event would have happened if they had not been drinking, (b) all of those who the patient suspected had been drinking, and (c) both half of the patients who were unsure of causal attribution and all of those who the patient suspected had been drinking (the most liberal definition). Finally, AAF estimates based on the most conservative estimate were calculated for countries grouped according to aggregate-level detrimental drinking pattern.

Results

Overall, about half of those with a violence-related injury reported drinking within the 6 hours before the event and, of these, 49% reported a definite causal association of their drinking with the event. Table 1 shows the prevalence rates of perceived others’ drinking (definitely, suspected, unsure, and no) in the violence-related injury event by country. Rates ranged from 14% to 73% (average rate = 44.5) for the conservative definition of others’ drinking, which included only those who the victim thought had definitely been drinking, and from 14% to 82% (average rate = 52.2%) when the liberal definition was used, which also included all of those suspected of drinking (but did not include those whom the victim reported they didn’t know or were unsure they had been drinking).

Rates for others’ drinking were found to vary by whether patients reported drinking in the event and by whether patients attributed a causal association of their drinking with the event (not shown). Among those who reported drinking, 62.4% also believed the perpetrator had definitely been drinking, compared with 27.3% of those who reported not drinking in the event (p < .01). Among those drinking who attributed a causal association of their drinking with the event, 67% perceived that the perpetrator had been drinking, compared with 57.1% of those drinking but who did not attribute a causal association (p < .05).

Table 2 shows the proportion of those in the four categories of (a) both drinking, (b) only the victim drinking, (c) only the perpetrator drinking, and (d) neither drinking, across countries, defining perpetrator drinking as only instances where the victim perceived the perpetrator as definitely drinking. Again, a great deal of variability is seen across countries for both self-drinking and others’ drinking. For the total combined data, 30.6% reported drinking by both perpetrator and victim, 18.4% reported drinking by the victim only, and 13.9% reported drinking by the perpetrator only.

Table 3 shows the AAF based on the conservative definition of the patients’ definite causal attribution of their own drinking to the event and of the patients’ belief that the perpetrator had definitely been drinking. Across all countries, 62.9% of the violence-related injuries involved alcohol on the part of the victim, the perpetrator, or both (Table 3, column 1). Following the procedure discussed above, among those patients who reported drinking before the event and definitely perceived that the perpetrator had been drinking, 23.7% of events were attributable to alcohol.

Likewise, among those who reported drinking before the event but were not definitely sure that the perpetrator had been drinking, 7.8% of events were attributable to alcohol, whereas among those who had not been drinking but definitely believed the perpetrator had been drinking, 7.3% of events were attributable to alcohol. Altogether, given this conservative definition of alcohol involvement, 38.8% of the injuries can be attributed to alcohol when causal attributions for the victim and the perpetrator are considered together, compared with 23.9% if only the patient’s causal attribution is taken into account. That is, the AAF for violence-related injury increases by 62% when others’ drinking is also taken into consideration (14.9% / 23.9%).

When the more liberal definitions of alcohol involvement were used, higher rates for AAF were found (not shown): 45.4% when included in the estimate were half of the patients who were unsure whether the event would have happened if they had not been drinking, 41.8% when included were all of the others who the patient suspected of drinking, and 49% when included were half of the patients who were unsure of causal attribution as well as all of the others who the patient suspected of drinking (the most liberal definition).

AAF was calculated, based on the most conservative estimate, for the countries grouped according to DDP level (not shown). Those countries with a DDP of 3 (Brazil, India, Ireland, Korea, Mozambique, South Africa, Sweden) showed the highest rate of AAF (48.7 based on own and others’ drinking), which was higher than those countries considered to have the highest level of DDP (Belarus and Mexico; AAF = 39.3).

Discussion

Although epidemiological studies in the general population as well as in the ED have typically obtained self-reports of one’s own drinking and associated problems, there has been relatively little study of respondents’ reports of drinking and harm by others (Room et al., 2010). The present study reports on others’ drinking as a factor in injuries treated in EDs, in the context of data also on the patient’s drinking in the event.

When drinking by the victim and perpetrator are considered together, AAF ranged from the most conservative estimate of 38.8% to the most liberal of 49%, with the increase in AAF ranging from 14.9% for the most conservative estimate to 19.6% when also included were half of the patients who were unsure of causal attribution and all of the others who the patient suspected of drinking.

In an earlier meta-analysis of ED studies across six countries (Cherpitel et al., 2005), AAF was estimated at 43% for violence-related injury, which only accounted for patients’ self-reported drinking. This is similar to the AAF estimates derived in the current analysis, which included both the patients’ and the perpetrators’ drinking (39%–49%), and much higher than the AAF estimate for the patients’ self-reported drinking reported here (24%–29%). In both approaches, alcohol involvement was first estimated by prevalence rates of patients’ self-report of alcohol use before injury, and a proportion of this alcohol involvement prevalence was assigned as the AAF.

The differences in AAF estimates between the earlier study and the present analysis arise from how this “assigned proportion” was determined. Typically, the estimated AAF is a function of the proportion of the population reporting drinking before injury and of the ratio of the injury rate among those who reported drinking to the injury rate among those who did not report drinking (i.e., the relative risk). In the prior study, the AAF was based on an estimate of the relative risk derived from a case-control analysis, whereas in the present study the AAF was estimated based on the patients’ individual assessment of whether, among those reporting drinking before their injury, the injury would have occurred had they not been drinking.

As shown in Table 3, 52.6% of those reporting drinking by both patients and perpetrators attributed their injuries to their own drinking, and this same proportion was then applied to the proportion caused by others’ drinking, which is equivalent to a relative risk estimate slightly greater than 2. However, relative risk estimates for violence-related injury associated with patients’ self-reported drinking were much larger than 2 in the earlier study using the case-control method (ranging from 3.2 to 35 in the 13 ED studies).

Very high relative risk estimates have also been derived using the case-crossover method, with one pooled analysis generating a relative risk of 14 (Borges et al., 2006), which translates to about 93% for the “assigned proportion.” In both this case-crossover study and the earlier case-control study, relative risk estimates failed to control for important confounding variables such as environmental factors. As shown by Macdonald et al. (2005) in EDs from six countries, risk of violence-related injury associated with a BAC of greater than .08% was reduced and became nonsignificant in two countries when place of drinking was controlled in the model, which suggests context as an important variable that should be taken into account in future studies of AAF.

Subjective causal attribution, although having its own limitations, is a more conservative approach and might thus provide more valid estimates of AAF for violence-related injury than using relative risk estimates derived from either case-control or case-crossover analysis described above, both of which may produce an overestimation of AAF. Prior analysis of the ERCAAP data comparing AAF based on causal attribution of drinking to violence-related injury with that based on any self-reported drinking before the injury found that AAF based on causal attribution was smaller than that based on self-report in all 12 studies and only half of the AAF based on self-report in 5 of the studies (Bond and Macdonald, 2009).

Considerable variability was found across countries in the likelihood of patients drinking before the violence-related injury, attributing a causal association of their drinking with the event, and perceiving that the perpetrator had been drinking, the latter of which was positively associated with both patients’ own drinking and their attributing a causal association between their drinking with the event. Cross-country variation would be expected, given the varying drinking patterns across cultures, and this is supported by findings here where AAF varied across DDP level. Prior research has reported higher rates of violence attributable to alcohol in countries characterized by episodic heavy drinking patterns compared with those characterized by a more integrated pattern of drinking associated with a lower DDP level (Gmel et al., 2007; Rehm et al., 2001, 2003a, 2003b). The highest rates of AAF for violence-related injury here were found for those countries exhibiting frequent heavy drinking patterns (DDP level of 3) and not those characterized by more infrequent but heavy drinking patterns, considered the most detrimental pattern level (DDP of 4). This may be explained by the fact that, although the relative risk of injury may be higher for those countries with a DDP score greater than 3 (characterized by infrequent but heavy drinking patterns), those countries with more frequent heavy drinking patterns (DDP of 3) would actually have a higher prevalence of alcohol-related injury and, therefore, a higher AAF.

Some limitations apply to this study. Although a patient’s causal attribution of his or her own drinking to the event appears to be a valid manner of assessing causality, the victim’s perception of others’ drinking may not be valid or, if valid, may have little to do with causality. As noted earlier, a study of domestic violence found that 40% of the victims reported the perpetrator had been drinking, but only 21% of the police reports found this to be the case (Bard and Zacker, 1974); however, it remains unknown as to which reports may be more accurate. For the present analyses, the same rate reported for causal attribution of the patients’ injury to their own drinking was also applied to others’ drinking. Unfortunately, no information was available for the latter estimate, and estimates of such rates have not previously been reported in the literature. The rate of 52.5% for self-attribution is near 50%, or a 50/50 chance of being attributed, which may underestimate or overestimate the true rate for attribution of the patient’s injury to others’ drinking, which likely varies according to a number of factors including context in which the injury occurred, time of day, and location of the injury event. For example, someone injured in a bar may be more likely to assume that the perpetrator had been drinking (and to attribute a causal association of this to the event) compared with someone injured at a sporting event.

Also, as noted earlier, there was large variability in sample size across countries, and the combined country analysis of AAF does not take into account cross-country differences. In addition, these EDs are not necessarily representative of their respective countries.

Although AAF implies causality, and prior epidemiological analysis points to a causal role of alcohol in violence-related injuries (Macdonald et al., 2005), only experimental studies are able to exert the proper controls to determine a true causal association. Alcohol has been shown to inhibit the brain’s normal ability to perceive and respond to situation cues, such as discriminating between threatening and nonthreatening situations (Pihl and Peterson, 1993); acts directly as a chemical disinhibitor of aggressive behavior impulses; and often serves as an excuse for inappropriate behavior (Collins and Schlenger, 1988). Experimental studies have found that aggressive behavior varies with the amount of alcohol ingested (Taylor and Gammon, 1975) and increases with provocation in intoxicated individuals (Bailey and Taylor, 1991). Episodes of heavy drinking can lead to an alcohol-induced psychotic disorder or alcoholic hallucinosis that includes violent behavior, and individuals with a prior personality disorder, when drinking heavily, may also be more prone to express violence (Taylor and Gammon, 1975) given the disinhibition properties of alcohol mentioned above. During alcohol withdrawal, agitation and distress also can lead to violent behavior, which are relationships that have been known for some time (Schuckit, 1995).

In addition to mechanisms that may operate at the individual level, at the societal level, alcohol-related violence is associated with social expectancies and societal contexts where drinking takes place, especially in societies where public drinking is common (Collins and Schlenger, 1988). The cross-country variation in AAF estimates found in this study calls attention to the importance of taking social context into account in the study of alcohol-related violence and the potential underlying mechanisms of this relationship.

Drinking can result in injury not only to the drinker but also to others, depending on the cause of injury (for example, those related to motor vehicle crashes or other types of negligence, or to violence), but drinking’s effect on others has not usually been taken into account in estimates of the attributable fraction of alcohol to injury morbidity. Alcohol is a leading risk factor in the global burden of disease, and 12% (by current estimates) of the alcohol-related disease burden is caused by intentional injuries (Rehm et al., 2003a, 2003b), many of which come to the attention of ED personnel.

Despite study limitations, findings here across 14 countries contribute to our knowledge regarding the magnitude of the additional burden that drinking by the perpetrator of a nondrinking victim of violence places on the ED. Findings are also important for refining the AAF for violence-related injury to take into account drinking by the perpetrator as well as drinking by the victim. Results like those in the present study will inform revisions of global burden of disease estimates of alcohol’s contribution to injury from violence, suggesting adjustments that could be made to these estimates to better reflect AAF when drinking by others is also taken into account. Additionally—and perhaps most important—findings suggest cultural variation in AAF based on country-level drinking patterns, with those countries exhibiting frequent heavy drinking patterns demonstrating the highest AAFs. Such country-level variation is important to consider in assigning AAF estimates in global burden of disease estimates.