IN THE UNITED STATES, there are more than 100 million emergency-department (ED) visits each year (Burt et al., 2007), with injuries making up 37% of ED visits (McCaig and Burt, 2003). Numerous studies have found a relationship between alcohol use and injury (Cherpitel, 1988a, b, 1989, 1993b, 1999;

Hingson and Howland, 1987; Hingson et al., 1985; Kerr et al., 1987). Alcohol is involved in up to 50% of inpatient cases admitted because of trauma (Rivara et al., 1993; Soderstrom et al., 1988), and in nearly 50% of all motor-vehicle crash deaths, suicides, and homicides (Perrine et al., 1989). Although estimates vary, on average one quarter of injured adults presenting to the ED have positive blood alcohol concentrations (BACs) (range: 4%-46%; Cherpitel et al., 2005), and nearly one in five injured adult ED patients has a diagnosis of alcohol use disorder (Maio et al., 1997; Soderstrom et al., 1997).

The health care context of the ED visit provides an opportunity to screen for alcohol use disorders and identify individuals at risk for problem drinking, even if the visit is not explicitly alcohol related (Fleming et al., 2007). The ED visit may also represent a unique "teachable moment," wherein patients may be more receptive to interventions for alcohol misuse (Cherpitel, 1993a, 1997; Conigrave et al., 1991; Dowey, 1993; Maio, 1995; Monti et al., 2001b). Because most injured ED patients (97%) are treated and released (McCaig and Burt, 2001), it is desirable to intervene during the ED visit, where lengthy wait times lend themselves to the delivery of brief interventions (BIs) adapted from motivational interviewing (MI; Burke et al., 2003; Miller and Rollnick, 2002). Overall, reviews have concluded that BIs targeting alcohol use delivered in primary care settings reduce drinking by 20%-30%; that these effects are maintained over time; and that BIs are efficacious as both single-session, stand-alone treatments and as a prelude to other treatments. Furthermore, in primary care settings, BI effects are generally stronger for women than men (Babor and Grant, 1992; Barry, 1999; Bien et al., 1993; Burke et al., 2003; Dunn et al., 2001; Fleming et al., 1997, 1999; Poikolainen, 1999).

Interventions delivered in the ED have the potential to reduce costs associated with injury and other alcohol-related health consequences (Fleming et al., 2002; Gentilello, 2005; Kunz et al., 20041. Previous researchers have examined the effectiveness of BIs in the ED, using both adolescent samples (Gregor et al., 2003; Monti et al., 1999, 2001a) and adult samples (Bazargan-Hejazi et al., 2005; Blow et al., 2006; Gentilello et al., 1999; Johnston et al., 2002; Leontieva et al., 2005; Longabaugh et al., 2001; Mello et al., 2005; Neumann et al., 2006; Welte et al., 1998). To date, the findings have been mixed, primarily demonstrating modest reductions in either alcohol consequences (Bazargan-Hejazi et al., 2005; Blow et al., 2006; Longabaugh et al., 2001; Mello et al., 2005; Monti et al., 1999) or alcohol consumption (Blow et al., 2006; Gentilello et al., 1999; Neumann et al., 2006). A prior article from this study examined the effectiveness of a tailored BI on reducing alcohol misuse among injured ED patients (Blow et al., 2006). Overall, drinking was reduced by 48.5% in the advice with tailored booklet condition, and young women in the advice conditions significantly decreased their frequency of heavy drinking, compared with those women in the no advice conditions.

Additional information concerning moderators of treatment efficacy might assist in streamlining or amplifying the effectiveness of BIs in ED settings. In particular, identifying pre-intervention characteristics that may enhance or attenuate intervention outcomes among subgroups would enable further focusing of BIs to participant characteristics. Two participant characteristics that may have a significant impact for MI-based interventions are readiness to change drinking and self-efficacy. In addition, in the ED setting, participants may be more receptive to a BI if the reason for their ED visit (i.e., injury) occurs recently after drinking alcohol or if they perceive the injury as alcohol related.

Although readiness to change has been conceptualized as an important factor in predicting response to a BI, surprisingly little research has been conducted in this area. According to Prochaska and DiClemente's (1983) transtheoretical model of change, individuals with greater readiness to change should be more receptive to interventions and should show greater treatment engagement and outcome. A recent ED study found that readiness to change predicted decreases in alcohol problem scores at 3 months after intervention, but moderators could not be examined because all participants received a BI (Leontieva et al., 2005). Another BI study of at-risk drinkers in the ED found that stage of change was not a moderator of treatment outcome; however, contemplation stage of change predicted worse outcomes at a 6-month follow-up, compared with precontemplation and action (Neumann et al., 2006). Two non-ED-based studies did not find a moderating effect for readiness to change on intervention response (Heather et al., 1996; Maisto et al., 2001). Instead, one study found that, among the subset of participants noted as not ready to change, those who received the BI showed greater reductions in drinking (Heather et al., 1996). The other study found that, contrary to hypotheses, individuals who received the BI and had greater readiness to change at baseline reported the highest levels of alcohol use at 12 months (Maisto et al., 2001). Thus, to date, research findings supporting stage of change as a moderator of treatment outcome are inconsistent.

Although enhancing self-efficacy is a standard component of BIs (Barry, 1999; Miller and Rollnick, 2002), there is also a paucity of research on self-efficacy as a moderator of BIs. Among substance use treatment samples, perceived confidence in making changes is an important predictor of reductions in drinking over time (Blume et al., 2003). Data from Project MATCH (Matching Alcoholism Treatments to Client Heterogeneity) indicated that higher baseline self-efficacy was associated with better drinking outcomes at 1 year after treatment (Ilgen et al., 2006). Similarly, in a study of behavioral couples who were treated for alcohol use disorders, lower self-efficacy at the end of treatment was associated with poorer drinking outcomes during a 1-year follow-up (McKay et al., 1993). When examined as a moderator of outcome in ED-based studies, however, findings have generally not demonstrated a moderating effect for self-efficacy on the effectiveness of BIs (Bazargan-Hejazi et al., 2005; Monti et al., 2001a).

Another potentially important moderator for ED-based BIs is the extent to which alcohol prompted the ED visit. Theoretically, whether alcohol was consumed before the injury and whether the injury is attributed to alcohol consumption should affect treatment response because of increased salience of alcohol's role in the injury, greater motivation to change, and the opportunity for intervening immediately following the alcohol event, thereby highlighting the use of a "teachable moment" (Barnett et al., 2006; Cherpitel et al., 2006b). Previous studies have found no link between drinking before injury and intervention effectiveness (Longabaugh et al., 2001). To our knowledge, previous studies have not examined patients' attributions concerning whether their injury was related to their alcohol use.

This article fills an important gap in the literature by identifying moderators of treatment outcome among injured ED patients who participated in a randomized controlled trial of a BI (Blow et al., 2006). By examining moderators of treatment effectiveness, BIs may be adapted to address the needs of subgroups that are less responsive. It was conceptualized that characteristics of the participants drinking in relation to their injury would increase the saliency of the "teachable moment" during the ED visit and thus would increase the potency of the advice session. The moderators selected for inclusion were based on theory and prior findings. Specifically, the primary hypotheses were that the effectiveness of the advice session on drinking outcome would be augmented among participants with greater self-efficacy and readiness to change, recent/acute alcohol consumption, and attribution of their injury to alcohol use.

Method

Design

Injured patients (age 19 and older) presenting to a Midwest Level 1 ED (between 8 AM and midnight) completed a computer-based survey of health issues (see Blow et al., 2006, for additional information) and were entered in a monthly drawing for $100. Eligible participants were randomly assigned to one of four intervention conditions: (1) advice + tailored booklet, (2) advice + generic booklet, (3) no advice + tailored booklet, or (4) no advice + generic booklet. Because extensive research demonstrates that brief alcohol interventions are effective, a traditional "no treatment/control" condition was not included but was rather approximated by the no advice + generic booklet condition. Following a saliva alcohol test to verify competency (e.g., BAC = 100 mg/dl or less), participants received the appropriate intervention before discharge and completed follow-up interviews at 3 and 12 months. The study protocol was approved by the university institutional review board for the use of human subjects.

Participants

Participants eligible for the study included ED patients age 19 years or older who had been injured in the last 24 hours. Trained research staff determined injury classification (based on definitions of E-codes 800-968; International Classification of Diseases, 1995). Patients were excluded if they were severely injured (e.g., unconscious, intubated, or in need of immediate life-saving medical procedures); presented with a self-inflicted injury, sexual assault, overdose, poisoning, near drowning, chronic injury without specific event associated with re-injury; or were pregnant, prisoners, or did not speak English. During the recruitment period, 6,047 potentially eligible patients presented to the ED; 507 (8.4%) patients were missed, mostly because of research staff being occupied with other patients or unable to locate the patient or because of medical staff presence with the patient. Of the 5,540 patients approached to participate, 4,476 (80.8%) consented and 1,064 (19.2%) refused. (Assuming all the missed patients were eligible for screening, a more conservative estimate of the response rate is 74%; 4,476 participants/6,047 patients presenting to the ED during the sampling frame). The majority of refusals reflect patients who reported feeling too sick, in too much pain, or too emotionally distressed to participate. Our institutional review board did not allow collecting additional information on missed or refused patients without consent.

Participants were eligible for the randomized controlled trial if they met at-risk drinking criteria based on either average consumption in the past 3 months (15 or more drinks/week for men younger than 65, 12 or more drinks/ week for women younger than age 65 and men age 65 and older, and 9 or more drinks/week for women age 65 and older) or heavy drinking (5 or more drinks/occasion on at least four occasions in the last month [e.g., weekly heavy drinking] for men younger than age 65 and 4 or more drinks/ occasion on at least four occasions in the past month for women younger than age 65 and for men and women age 65 and older).

Among the 14.5% of patients (n = 649) who screened positive for at-risk drinking, 89% (n = 575) received a BI during their ED visit. The 74 patients who left the ED before receiving the BI are not described in this article. Three- and 12-month follow-up interview rates exceeded 85%; all subsequent data are for participants with complete follow-up data (n = 494). Note that 90%-95% (depending on a 3-month or 12-month interview) of follow-up data were obtained by phone. The remainder were obtained in person or by mail. Follow-up interviewers were blind to intervention condition assignment. BI participants were mostly male (71%), white (86%; 5.7% African-American, 8.3% other), never married (70%; 17% married), and college educated (80% at least some), with an average (SD) age of 27.8 (11.2; median = 22; range: 19-76). As indicated in Blow et al. (2006), no significant differences were found across the intervention groups in age (F = 0.47, 3/ 490 df, p = .7066), gender ([chi square] = 4.62, 3 df, p = .2018), marital status ([chi square] = 1.97, 6 df, p = .92), race/ethnicity ([chi square] = 1.06, 6 df, p = .98), or years of education ([chi square] = 3.02, 9 df, p = .96).

Measures

ED baseline screening was computerized with alcohol questions embedded in a health and lifestyle assessment that included questions concerning type and mechanism of current injury. The follow-up interview included identical questions to that of the baseline assessment.

Alcohol consumption questions included typical quantity (drinks per occasion) and frequency (days per week) of alcohol use in the past 3 months. A composite quantity/ frequency score was calculated to reflect average weekly consumption. Heavy drinking was assessed by asking participants to indicate how frequently they consumed 4+/5+ drinks per occasion in the past month for men/women. The Drinker Inventory of Consequences (DrInC) Short Inventory of Problems was used to assess alcohol-related consequences (Miller et al., 1995). We added two items to reflect alcohol-related injuries and arrests for drinking and driving: "While drinking or intoxicated, I have been physically hurt, injured, or burned" and "I have been arrested for driving under the influence of alcohol." This revised Short Inventory of Problems retained the physical, social responsibility, intrapersonal, impulse control, and interpersonal consequence subscale scores, as well as the total score, with adequate internal consistency (Cronbach's [alpha] = .89- .90).

Participants were asked whether they drank alcohol within 6 hours before their injury (yes/no) and whether they attributed their injury to alcohol use (yes, maybe, no). Regarding self-efficacy, participants rated how confident they were that they could control or abstain from drinking if they wanted to change their drinking habits on a 5-point Likert scale (ranging from not at all to extremely). Readiness to change was assessed using two questions: (1) "Do you plan to cut down on your drinking in the next 30 days?" (yes/no) and (2) "Do you plan to cut down on your drinking in the next 6 months?" (yes/no). Based on these questions and previous algorithms used to categorize stage of change for substance use (e.g., Carey et al., 2002), participants were categorized into one of three stages: (1) precontemplation (no plans to cut down or stop drinking in next 6 months), (2) contemplation (plans to cut down or stop drinking in the next 6 months but not in the next month), and (3) action (plans to cut down or stop drinking in the next month).

Intervention

During the advice session, the tailored or generic booklet was reviewed with the participant before discharge from the ED. Fidelity to intervention sessions was maintained via several strategies. A masters-level clinical supervisor "shadowed" therapists on a monthly basis to observe advice sessions and provide retraining as necessary. A psychologist trained in MI conducted half-day training seminars twice yearly to review principles of MI (e.g., rolling with resistance, developing discrepancy), including those encompassed in the FRAMES acronym (i.e., Feedback, Responsibility, Advice, Menu, Empathy, Self-Efficacy; Miller and Rollnick, 1991). All therapists had master's degrees in social work or related fields. For the no advice conditions, participants were told that, based on their responses to the health screen, they were at risk for problem drinking and were given either a tailored or generic booklet to review.

Analyses

Analyses were conducted using SAS 9.1 statistical software (SAS Institute, Inc., Cary, NC). Descriptive analyses were conducted to examine variable distributions and determine the appropriateness of the analytic approach (e.g., negative binomial distribution and a log-link function for all outcome measures). The primary analysis strategy evaluated the effects of the intervention (advice, no advice) at baseline and over the 3-month and 12-month follow-up periods using generalized estimating equations because of the correlated structure of our repeated measures (Liang and Zeger, 1986). Dependent variables included the following: average weekly alcohol consumption, frequency of heavy drinking in the past month, and consequences as measured by the DrInC (summary of all item responses).

Our prior article (Blow et al., 2006) showed limited influence of tailoring, compared with generic booklets. As a result, for the purposes of examining moderators of treatment outcome, we limited our analyses to treatment effects comparing the advice and no advice conditions. This approach also provides a more powerful test of the main effects of the advice condition. Independent variables included advice (yes/no) and time (baseline, 3 months, 12 months). In addition, gender and age ([less than or equal to] 22 and >22) were included as covariates. Gender was included because the literature and our prior work (Blow et al., 2006) shows that BIs may be more effective for women than men. Age (median age = 22) was included in the analyses because of the large proportion of college students in the sample and prior research from this study (Blow et al., 2006) showing differential effects of BIs based on age. Moderators of treatment outcome were examined via interactions for Treatment Condition x Hypothesized Moderator (drinking within 6 hours before injury, stage of change, alcohol-related injury attributions, and self-efficacy) x Time. Models were constructed separately for the three drinking outcomes, with main effects (independent variables and moderators) and covariates entered on the first step, Treatment x Moderator interactions entered on the second step, Time x Moderator interactions entered on the third step, and three-way interactions (Time x Treatment x Moderator) entered on the fourth step. Note that the size of the smallest subgroup for the categorical predictors used in the three-way interaction analyses (n = 36; for the no treatment by positive attribution of injury to alcohol) suggests the sample distribution was adequate for conducting these analyses. Thus, multivariate analyses examined evidence for significant changes over time in the outcome variables based on moderators of effectiveness for the advice/no advice conditions.

Results

Drinking characteristics of the sample and attrition analyses

On average, participants reported consuming 20.5 (16.8) drinks per week, had engaged in 7.1 (6.0) heavy drinking episodes in the past month, and endorsed 4.7 (5.5) drinking consequences at baseline. Participants lost to follow-up did not differ from complete cases on the following baseline measures: frequency of heavy drinking, alcohol-related negative consequences, stage of change, alcohol-related injury attributions, alcohol use within 6 hours before the injury, and gender. Those participants lost to follow-up had higher average consumption (26.3 vs 20.5 drinks per week, p < .01), were significantly older (79.0% of cases with missing data were >22 vs 49.5% for complete cases, p < .001), and had lower self-efficacy for controlling or abstaining from drinking (4.2 vs 4.5 on a 5-point scale, p < .01).

Descriptive information for moderators at baseline

The majority of participants (53.4%) were classified in the precontemplation stage of change, whereas 8.7% were in contemplation and 37.9% were in action. Regarding self-efficacy, 89.9% indicated they were very or extremely confident they could cut down or stop drinking, whereas 8.9% indicated they were somewhat confident and 1.2% indicated they were a little or not at all confident. Approximately one quarter (27.1%) of participants indicated they drank within 6 hours before their injury, and 15.6% perceived their injury to be related to alcohol use. Table 1 shows descriptive data for change in the alcohol-related dependent variables (i.e., quantity/frequency, heavy drinking, and DrInC consequences) over time, based on the moderator variables, and associated individual significance levels. Note that, for most levels of the moderator variables, quantity/frequency, heavy drinking, and DrInC consequences significantly decreased over the follow-up period.

Moderators of intervention outcomes

Average consumption. First, we examined the main effects of the moderator variables on average weekly consumption over time. Overall, average weekly consumption decreased over time ([chi square] = 97.2, 2 df, p < .001) and was lower among women than among men ([chi square] = 51.0, 1 df, p < .001). In addition, there were significant main effects for self-efficacy ([chi square] = 10.3, 1 df, p < .01) and readiness to change ([chi square] = 24.7, 1 df, p < .001). Specifically, participants who reported higher levels of self-efficacy had lower weekly consumption, whereas those with higher readiness to change had greater weekly consumption. Baseline age older than age 22 ([chi square] = 0.1, 1 df, p = .718), alcohol use within 6 hours before injury ([chi square] = 0.9, 1 df, p = .331), and alcohol-related injury attributions ([chi square] = 3.2, 1 df, p = .074) were not significantly associated with alcohol consumption. Second, we examined two-way interactions between treatment moderators and time, and between treatment moderators and intervention type; these findings were not significant and are not presented. Third, we examined all three-way interactions (Treatment x Time x Moderator [self-efficacy, readiness to change, alcohol-related injury attributions, and alcohol use within 6 hours before injury]), and only one significant three-way interaction was identified: Alcohol Attribution x Treatment x Time ([chi square] = 7.9, df = 2, p < .05). Individuals who attributed their injury to alcohol and received advice had significantly lower levels of average weekly alcohol consumption from baseline to 12-month follow-up, compared with those who attributed their injury to alcohol but did not receive advice (Figure 1). Furthermore, as shown in Table 2, participants in the advice condition who attributed their injury to alcohol showed a 44.6% change in mean average consumption from baseline to 12 months, compared with a 16.7% change in consumption in the no advice by alcohol attribution group.

[FIGURE 1 OMITTED]

Heavy drinking. First, we examined main effects over time and found significant effects for the following variables: age, gender, self-efficacy, and stage of change (see Table 1 for descriptive data). Overall, heavy drinking decreased over time ([chi square] = 69.3, 2 df, p < .001). Women reported fewer episodes of heavy drinking than did men ([chi square] = 6.0, 1 df, p < .05). Participants who were older than age 22 ([chi square] = 19.2, 1 df, p < .001) and those with greater self-efficacy ([chi square] = 10.9, 1 df, p < .001) had fewer heavy drinking episodes. In addition, participants who reported greater readiness to change reported more frequent heavy drinking in the past month ([chi square] = 38.6, 1 df, p < .001). There were no significant main effects for attributions for the ED-related incident ([chi square] = 3.3, 1 df, p = .071) or alcohol use within 6 hours before injury ([chi square] = 0.7, 1 dr, p = .401). Second, there were no significant two-way interactions between the moderator variables and treatment condition or time. Third, regarding all possible three-way interactions tested, there was a significant three-way interaction for Alcohol Attribution x Treatment x Time ([chi square] = 6.6, 2 df, p < .05). Specifically, individuals who attributed their injury to alcohol and received advice had significantly fewer heavy drinking episodes than participants who attributed their injury to alcohol but did not receive advice (Figure 2). Furthermore, as shown in Table 3, participants in the advice condition who attributed their injury to alcohol showed a 50.5% change in mean heavy drinking episodes from baseline to 12 months, compared with a 13.2% change in the no advice by alcohol attribution group.

[FIGURE 2 OMITTED]

Alcohol consequences: DrInC. First, we examined main effects over time and found that negative consequences decreased over time ([chi square] = 39.7, 2 df, p < .001), and women reported experiencing fewer consequences than men ([chi square] = 13.3, 1 df, p < .001). In addition, those with greater self-efficacy reported fewer consequences ([chi square] = 19.4, 1 df, p < .001). Conversely, participants who attributed their injury to alcohol reported a greater number of consequences, compared with those who did not make alcohol-related attributions ([chi square] = 16.6, 1 df, p < .001), and those who reported greater readiness to change also had a greater number of consequences ([chi square] = 87.3, 1 df, p < .001). There were no main effects of age ([chi square] = 2.9, 1 df, p = .087) or alcohol use within 6 hours before injury on drinking consequences ([chi square] = 1.2, 1 df, p = .282). Second, none of the possible two-way or three-way interactions tested reached significance, indicating that there were no significant moderators of treatment or time on negative consequences of drinking.

Discussion

This article provides important information regarding moderators of the effectiveness of brief alcohol interventions in an ED setting. To our knowledge, there have been very few prior articles to examine this issue in any setting despite the call from researchers to identify the essential components of brief alcohol interventions (Burke et al., 2003). Identification of moderators provides preliminary data regarding the mechanisms of change and has important implications for focusing future brief alcohol interventions in the ED to augment the effectiveness of these interventions. Overall, our findings suggest that BIs may be particularly effective for individuals who attribute their injury to alcohol.

Furthermore, attributing the current injury to alcohol seems to maintain the efficacy of the BI, with these effects being most pronounced 1 year after the ED visit. Prior ED research of BIs has typically evaluated outcomes at a single follow-up period. For example, some studies have published only short-term (e.g., 3 months or 6 months) outcomes (Bazargan-Hejazi et al., 2005; Monti et al., 1999), whereas other studies have published only longer term (e.g., 12 months) outcomes (Gentilello et al., 1999; Longabaugh et al., 2001; Mello et al., 2005). Among ED studies with multiple follow-up periods, one study (Gentilello et al., 1999) found BI effects were more pronounced at 1 year than at 6 months due to increases in drinking and injury in the control group. However, another study found BI effects were more robust at 6 months than at 12 months (Neumann et al., 2006). These discrepancies may result from variations in the impact and extent of the control condition. For example, it may be that an ED visit for injury attenuates drinking among control participants; however, these effects may not be maintained over the long term with drinking (or injury) returning to baseline levels by 12 months. In contrast, reductions in drinking among intervention subgroups (such as those making a connection between their drinking and their injury) may be maintained over the long term, with one study showing lasting effects on injury at 3 years (Gentilello et al., 1999). Future research with multiple, long-term follow-ups are required to test these suppositions.

Researchers have theorized that BIs are effective when delivered during a health care visit because of the notion of a "teachable moment," during which patients may be more receptive to feedback regarding their drinking, particularly when delivered in the context of an empowering, nonconfrontational motivational interview. Findings from this study suggest that participants who attributed their injury to drinking initially had significant reductions in alcohol use, regardless of intervention type, but that long-term reductions for these participants were best achieved via therapist-delivered BIs. In addition, actual alcohol use within 6 hours before the injury was not related to intervention effectiveness, whereas the individual's perception that alcohol caused the injury was related to intervention effectiveness. Indeed, a recent study found that alcohol attribution was not simply a function of the amount of alcohol consumed and a feeling of drunkenness but was influenced by usual drinking pattern and cultural context (Cherpitel et al., 2006a). In addition, our findings are consistent with Barnett et al. (2006), who found that drinking history was associated with reduced motivation to change among individuals who were mandated to a BI following an alcohol-related incident, whereas aversiveness of the incident was associated with greater motivation. Sanchez-Craig (1990) has suggested that the effectiveness of BIs result from increasing responsibility for drinking. The present findings are consistent with this notion and suggest that attribution of injury to alcohol is a piece of information that therapists may use to develop a discrepancy between current behaviors and health and social goals, and these findings might tip the decisional balance in favor of consideration of reducing or stopping alcohol misuse. When an individual does not perceive his/her injury as alcohol related, it might be an important target of the initial MI intervention. Heightening awareness regarding the relationship between alcohol and injury may increase the effectiveness of the intervention. If the connection between the current injury and alcohol use is not relevant, the relationship between past or future injuries and alcohol use could be discussed.

In addition, although pre-intervention self-efficacy and readiness to change were unrelated to intervention outcomes over time, further research is needed to determine whether BIs impact stage of change. Assessing stage of change immediately following the intervention would provide a more precise assessment of changes in stage of change as moderators of intervention outcome. We found that, before the intervention, higher self-efficacy was associated with lower levels of alcohol use and fewer negative consequences, whereas higher stage of change was associated with higher levels of alcohol use and more negative consequences. Regarding self-efficacy, it seems that those with higher levels of self-efficacy are indeed successful at maintaining lower levels of alcohol use and consequences, regardless of specific interventions. Conversely, those who are consuming alcohol at higher levels may be more likely to endorse readiness to change, but this has little impact on actual intervention response. It seems that stage of change, as measured in this study, reflects baseline problem severity--with precontemplators showing lower levels of drinking and consequences, and less interest in changing drinking than other groups. These findings are consistent with others who have found a positive relationship between stage of change and alcohol use (Leontieva et al., 2005; Shealy et al., 2007).

There are several limitations of this study that must be noted. The site of the study was a suburban university-affiliated ED, which serves a highly educated, predominantly white population. Replication is required in ED settings serving more diverse patient populations and should include admitted patients, pregnant women, suicidal patients, and patients presenting for sexual assault. In addition, as participant recruitment of injured patients occurred between the hours of 8:00 AM and midnight, results may not generalize to injured patients presenting on midnight shifts or to patients presenting with medical complaints. Although the omission of a true control condition was based on extensive research demonstrating the effectiveness of BIs for alcohol, the enhanced usual care condition (i.e., no advice + generic booklet) precludes direct comparison of findings to a no-intervention condition. Given the time constraints in a busy ED setting, it was not feasible to include several multi-item scales in the assessment. In addition, self-efficacy and stage of change were not measured immediately after the intervention; implementation of posttests should be conducted to further examine these variables as potential moderators. Although the follow-up rate was excellent (85%), it is possible that findings were biased because of attrition. Future studies are needed to replicate our findings with established measures (e.g., rulers for stage of change, self-efficacy) and to expand the type of moderators examined (e.g., therapeutic alliance). Finally, future studies should assess the timing of the BI relative to length of time in the ED to determine whether this impacts intervention effectiveness.

Despite these limitations, this article provides novel information regarding potential moderators of the effectiveness of brief alcohol interventions in the ED. Findings suggest that BIs may reduce drinking outcomes via identification of the link between alcohol and injury and suggest that interventions focusing on increasing change talk in this regard may improve outcomes and reduce future injuries. In addition, interventions aimed at supporting self-efficacy may be particularly important for reducing alcohol use. These suggestions are consistent with a MI philosophy and framework, but staff delivering these interventions require training and supervision to master these techniques. Given time and caseload pressures facing physicians in the fast-paced ED setting, it may be most beneficial and cost effective to train other health care providers (e.g., social workers as used in the present study, nurses, mental health workers) to deliver BIs. Findings from this study shed some light on the "black box" of brief alcohol interventions that may assist clinicians and researchers in streamlining and increasing the potency of these approaches.

Acknowledgments

We thank Lynn Massey, Scott Kelly, Robert Vaidya, Harpret Otal, Margaret White, and Robin Williams for their work on this project.

Received: September 28, 2007. Revision: March 28, 2008.

References

BABOR, T.F. AND GRANT, M. (Eds.) Project on Identification and Management of Alcohol-Related Problems. Report on Phase II: A Randomized Clinical Trial of Brief Interventions in Primary Health Care, Geneva, Switzerland: World Health Organization, 1992.

BARNETT, N.P., GOLDSTEIN, A.L., MURPHY, J.G., COLBY, S.M., AND MONTI, P.M. "I'll never drink like that again": Characteristics of alcohol-related incidents and predictors of motivation to change in college students. J. Stud. Alcohol 67: 754-763, 2006.

BARRY, K.L. Brief Interventions and Brief Therapies for Substance Abuse. Treatment Improvement Protocol (TIP) Series, No. 34, DHHS Publication No. (SMA) 99-3353, Rockville, MD: Center for Substance Abuse Treatment, Substance Abuse and Mental Health Services Administration, 1999.

BAZARGAN-HEJAZI, S., BING, E., BAZARGAN, M., DER-MARTIROSIAN, C., HARDIN, E., BERNSTEIN, J., AND BERNSTEIN, E. Evaluation of a brief intervention in an inner-city emergency department. Ann. Liner. Med. 46: 67-76, 2005.

BIEN, T.H., MILLER, W.R., AND TONIGAN, J.S. Brief interventions for alcohol problems: A review. Addiction 88:315-335, 1993.

BLOW, F.C., BARRY, K.L., WALTON, M.A., MAID, R.F., CHERMACK, S.T., GINGHAM, C.R., IGNACIO, R.V., AND STRECHER, V.J. The efficacy of two brief intervention strategies among injured, at-risk drinkers in the emergency department: Impact of tailored messaging and brief advice. J. Stud. Alcohol 67: 568-578, 2006.

BLUME, A.W., SCHMALING, K.B., AND MARLATT, A.G. Predictors of change in binge drinking over a 3-month period. Addict. Behav. 28: 1007-1012, 2003.

BURKE, B.L., ARKOWITZ, H., AND MENCHOLA, M. The efficacy of motivational interviewing: a meta-analysis of controlled clinical trials. J. Cons. Clin. Psychol. 71: 843-861, 2003.

BURT, C.W., McCAIG, L.F., AND RECHTSTEINER, E.A. Ambulatory Medical Care Utilization Estimates for 2005. Advance Data from Vital and Health Statistics, No. 388, Hyattsville, MD: National Center for Health Statistics, 2007.

CAREY, K.B., PURNINE, D.M., MAISTO, S.A., AND CAREY, M.P. Correlates of stages of change for substance abuse among psychiatric outpatients. Psychol. Addict. Behav. 16: 283-289, 2002.

CHERPITEL, C.J. Alcohol consumption and casualties: A comparison of two emergency room populations. Brit. J. Addict. 83:1299-1307, 1988a.

CHERPITEL, C.J. Drinking patterns and problems associated with injury status in emergency room admissions. Alcsm Clin. Exp. Res. 12: 105-110, 1988b.

CHERPITEL, C.J. Breath analysis and self-reports as measures of alcohol-related emergency room admissions. J. Stud. Alcohol 50: 155-161, 1989.

CHERPITEL, C.J. Alcohol and violence-related injuries: An emergency room study. Addiction 88: 79-88, 1993a.

CHERPITEL, C.J. Alcohol, injury, and risk-taking behavior: Data from a national sample. Alcsm Clin. Exp. Res. 17: 762-766, 1993b.

CHERPITEL, C.J. Alcohol and injuries resulting from violence: A comparison of emergency room samples from two regions of the U.S.J. Addict. Dis. 16: 25-40, 1997.

CHERPITEL, C.J. Drinking patterns and problems: A comparison of primary care with the emergency room. Subst. Abuse 20: 85-95, 1999.

CHERPITEL, C.J., BOND, J., YE, Y., GORGES, G., ROOM, R., POZNYAK, V., AND HAD, W. Multi-level analysis of causal attribution of injury to alcohol and modifying effects: Data from two international emergency room projects. Drug Alcohol Depend. 82: 258-268, 2006a.

CHERPITEL, C.J., YE, Y., BOND, J., REHM, J., CREMONTE, M., NEVES, O., MOSKALEWICZ, J., SWIATKIEWICZ, G., AND GIESBRECHT, N. The effect of alcohol consumption on emergency department services use among injured patients: A cross-national emergency room study. J. Stud. Alcohol 67: 890-897, 2006b.

CHERPITEL, C.J., YE, V., BOND, J., REHM, J., POZNYAK, V., MACDONALD, S., STAFSTROM, M., AND HAD, W. Emergency Room Collaborative Alcohol Analysis Project (ERCAAP) and the WHO Collaborative Study on Alcohol and Injuries. Multi-level analysis of alcohol-related injury among emergency department patients: A cross-national study. Addiction 100: 1840-1850, 2005.

CONIGRAVE, K.M., BURNS, F.H., REZNIK, R.B., AND SAUNDERS, J.B. Problem drinking in emergency department patients: The scope for early intervention. Med. J. Aust. 154: 801-805, 1991.

Dowey, K.E. Alcohol-related attendances at an accident and emergency department. Ulster Med. J. 62: 58-62, 1993.

DUNN, C., DEROO, L., AND RIVARA, F.P. The use of brief interventions adapted from motivational interviewing across behavioral domains: A systematic review. Addiction 96:1725-1742, 2001.

FLEMING, M.F., BALOUSEK, S.L., KLESSIG, C.L., MUNDT, M.P., AND BROWN, D.D. Substance use disorders in a primary care sample receiving daily opioid therapy. J. Pain g: 573-582, 2007.

FLEMING, M.F., BARRY, K.L., MANWELL, L.B., JOHNSON, K., AND LONDON, R. Brief physician advice for problem alcohol drinkers: A randomized controlled trial in community-based primary care practices. JAMA 277: 1039-1045, 1997.

FLEMING, M.F., MANWELL, L.B., BARRY, K.L., ADAMS, W., AND STAUFFACHER, E.A. Brief physician advice for alcohol problems in older adults: A randomized community-based trial. J. Fam. Pract. 48: 378-384, 1999.

FLEMING, M.F., MUNDT, M.P., FRENCH, M.T., MANWELL, L.B., STAUFFACHER, E.A., AND BARRY, K.L. Brief physician advice for problem drinkers: Long-term efficacy and benefit-cost analysis. Alcsm Clin. Exp. Res. 26: 36-43, 2002.

GENTILELLO, L.M. Alcohol interventions in trauma centers: The opportunity and the challenge. J. Trauma 59 (3 Suppl.): S18-S20, 2005.

GENTILELLO, L.M., RIVARA, F.P., DONOVAN, D.M., JURKOVICH, G.J., DARANC1ANG, E., DUNN, C.W., VILLAVECES, A., COPASS, M., AND RIES, R.R. Alcohol interventions in a trauma center as a means of reducing the risk of injury recurrence. Ann. Surg. 230: 473-480; discussion 480-483, 1999.

GREGOR, M.A., SHOPE, J.T., BLOW, F.C., MAID, R.F., WEBER, J.E., AND NYPAVER, MM. Feasibility of using an interactive laptop program in the emergency department to prevent alcohol misuse among adolescents. Ann. Emer. Med. 42: 276-284, 2003.

HEATHER, N., ROLLNICK, S., BELL, A., AND RICHMOND, R. Effects of brief counselling among male heavy drinkers identified on general hospital wards. Drug Alcohol Rev. 15: 29-38, 1996.

HINGSON, R. AND HOWLAND, J. Alcohol as a risk factor for injury or death resulting from accidental falls: A review of the literature. J. Stud. Alcohol 48: 212-219, 1987.

HINGSON, R.W., LEDERMAN, R.I., AND WALSH, D.C. Employee drinking patterns and accidental injury: A study of four New England states. J. Stud. Alcohol 46: 298-303, 1985.

ILGEN, M., TILT, Q., FINNEY, J., AND MOOS, R.H. Self-efficacy, therapeutic alliance, and alcohol-use disorder treatment outcomes. J. Stud. Alcohol 67: 465-472, 2006.

INTERNATIONAL CLASSIFICATION OF DISEASES. Practice Management Information Corporation, 1995.

JOHNSTON, B.D., RIVARA, F.P., DROESCH, R.M., DUNN, C., AND COPASS, M.K. Behavior change counseling in the emergency department to reduce injury risk: A randomized, controlled trial. Pediatrics 110: 267-274, 2002.

KERR, M.H., CAMPBELL, F.C., AND RUTHERFORD, W.H. Unemployment, alcohol and injury in west Belfast. Injury 18: 313-314, 1987.

KUNZ, F.M., JR., FRENCH, M.T., AND BAZARGAN-HEJAZI, S. Cost-effectiveness analysis of a brief intervention delivered to problem drinkers presenting at an inner-city hospital emergency department. J. Stud. Alcohol 65: 363-370, 2004.

LEONTIEVA, L., HORN, K., HAQUE, A., HELMKAMP, J., EHRLICH, P., AND WILLIAMS, J. Readiness to change problematic drinking assessed in the emergency department as a predictor of change. J. Crit. Care 20:251-256, 2005.

LIANU, K.Y. AND ZEOER, S.L. Longitudinal data analysis using generalized linear models. Biometrika 73: 13-22, 1986.

LONGABAUGH, R., WOOLARD, R.E., NIRENBERG, T.D., MINUGH, A.P., BECKER, B., CLIFFORD, P.R., CARTY, K., LICSW, SPARADEO, F., AND GOGINENI, A. Evaluating the effects of a brief motivational intervention for injured drinkers in the emergency department. J. Stud. Alcohol 62: 806-816, 2001.

MCCAIG, L.F. AND BURT, C.W. National Hospital Ambulatory Medical Care Survey: 1999 Emergency Department Summary. Advance Data from Vital and Health Statistics, No. 320, DHHS Publication No. (PHS) 2001-1250, Hyattsville, MD: National Center for Health Statistics, 2001.

MCCAIG, L.F. AND BURT, C.W. National Hospital Ambulatory Medical Care Survey: 2001 Emergency Department Summary. Advance Data from Vital and Health Statistics, No. 335, DHHS Publication No. (PHS) 2003-1250, Hyattsville, MD: National Center for Health Statistics, 2003.

MCKAY, J.R., MAISTO, S.A., AND O'FARRELL, T.J. End-of-treatment self-efficacy, aftercare, and drinking outcomes of alcoholic men. Alcsm Clin. Exp. Res. 17: 1078-1083, 1993.

MAIO, R.F. Alcohol and injury in the emergency department: Opportunities for intervention. Ann. Emer. Med. 26: 221-223, 1995.

MAIO, R.F., WALLER, P.F., BLOW, F.C., HILL, E.M., AND SINGER, K.M. Alcohol abuse/dependence in motor vehicle crash victims presenting to the emergency department. Acad. Emer. Med. 4: 256-262, 1997.

MAISTO, S.A., CONIGLIARO, J., MCNEIL, M., KRAEMER, K., CONIGLIARO, R.L., AND KELLEY, M.E. Effects of two types of brief intervention and readiness to change on alcohol use in hazardous drinkers. J. Stud. Alcohol 62: 605-614, 2001.

MELLO, M.J., NIRENBERG, T.D., LONGABAUGH, R., WOOLARD, R., MINUGH, A., BECKER, B., BAIRD, J., AND STEIN, L. Emergency department brief motivational interventions for alcohol with motor vehicle crash patients. Ann. Emer. Med. 45: 620-625, 2005.

MILLER, W.R. AND ROLLNICK, S. Motivational Interviewing: Preparing People to Change Addictive Behavior, New York: Guilford Press, 1991.

MILLER, W.R. AND ROLLNICK, S. Motivational Interviewing: Preparing People for Change, 2nd Edition, New York: Guilford Press, 2002.

MILLER, W.R., TONIOAN, J.S., AND LONGBAUGH, R. The Drinker Inventory of Consequences (DrInC): An Instrument for Assessing Adverse Consequences of Alcohol Abuse (Test Manual). NIAAA Project MATCH Monograph Series, Vol. 4, NIH Publication No. 95-3911, Bethesda, MD: National Institute on Alcohol Abuse and Alcoholism, 1995.

MONTI, P.M., BARNETT, N.P., O'LEARY, T.A., AND COLBY, S.M. Motivational enhancement of alcohol-involved adolescents. In: MONTI, P.M., COLBY, S.M., AND O'LEARY, T.A. (Eds.) Adolescents, Alcohol and Substance Abuse: Reaching Teens through Brief Intervention, New York: Guilford Press, 2001a, pp. 145-182.

MONTI, P.M., COLBY, S.M., BARNETT, N.P., SPIRITO, A., ROHSENOW, D.J., MYERS, M., WOOLARD, R., AND LEWANDER, W. Brief intervention for harm reduction with alcohol-positive older adolescents in a hospital emergency department. J. Cons. Clin. Psychol. 67: 989-994, 1999.

MONTI, P.M., ROHSENOW, D.J., SWIFT, R.M., GULLIVER, S.B., COLBY, S.M., MUELLER, T.I., BROWN, R.A., GORDON, A., ABRAMS, D.B., NIAURA, R.S., AND ASHER, M.K. Naltrexone and cue exposure with coping and communication skills training for alcoholics: Treatment process and l-year outcomes. Alcsm Clin. Exp. Res. 25: 1634-1647, 2001b.

NEUMANN, T., NEUNER, B., WEISS-GERLACH, E., TONNESEN, H., GENTILELLO, L.M., WERNECKE, K.D., SCHMIDT, K., SCHRODER, T., WAUER, H., HEINZ, A., MANN, K., MULLER, J.M., HAAS, N., Kox, W.J., AND SPIES, C.D. The effect of computerized tailored brief advice on at-risk drinking in subcritically injured trauma patients. J. Trauma 61: 805-814, 2006.

PERRINE, M.W., PECK, R.C., AND FELL, J.C. EpidemioIogic perspectives on drunk driving. In: SURGEON GENERAL'S WORKSHOP ON DRUNK DRIVING: BACKGROUND PAPERS, Rockville MD: Office of the Surgeon General, 1989, pp. 35-76.

POIKOLAINEN, K. Effectiveness of brief interventions to reduce alcohol intake in primary health care populations: A meta-analysis. Prey. Med. 28: 503-509, 1999.

PROCHASKA, J.O. AND DICLEMENTE, C.C. Stages and processes of self-change of smoking: Toward an integrative model of change. J. Cons. Clin. Psychol. 51: 390-395, 1983.

RIVARA, F.P., JURKOVICH, G.J., GURNEY, J.G., SEGUIN, D., FLIGNER, C.L., RIES, R., RAISYS, V.A., AND COPASS, M. The magnitude of acute and chronic alcohol abuse in trauma patients. Arch. Surg. 128: 907-912; discussion 912-913, 1993.

SANCHEZ-CRAIG, M. Brief didactic treatment for alcohol and drug-related problems: An approach based on client choice. Brit. J. Addict. 85: 169-177, 1990.

SHEALY, A.E., MURPHY, J.G., BORSAR1, B., AND CORREIA, C.J. Predictors of motivation to change alcohol use among referred college students. Addict. Behav. 32: 2358-2364, 2007.

SODERSTROM, C.A., DISCHINGER, P.C., SMITH, G.S., HEGEL, J.R., MCDUFF, D.R., GORELICK, D.A., KERNS, T.J., Ho, S.M., AND READ, K.M. Alcoholism at the time of injury among trauma center patients: Vehicular crash victims compared with other patients. Accid. Anal Prey. 29: 715-721, 1997.

SODERSTROM, C.A., TRIFILLIS, A.L., SHANKAR, B.S., CLARK, W.E., AND COWLEY, R.A. Marijuana and alcohol use among 1023 trauma patients: A prospective study. Arch. Surg. 123: 733-737, 1988.

WELTE, J.W., PERRY, P., LONGABAUGH, R., AND CLIFFORD, P.R. An outcome evaluation of a hospital-based early intervention program. Addiction 93: 573-581, 1998.

* This research was supported by National Institute on Alcohol Abuse and Alcoholism grant AA 11629.

MAUREEN A. WALTON, M.P.H., PH.D., ([dagger]) ABBY L. GOLDSTEIN, PH.D., ([dagger]) STEPHEN T. CHERMACK, PH.D., ([dagger]) RYAN J. McCAMMON, M.S., REBECCA M. CUNNINGHAM, M.D., ([dagger]) KRISTEN L. BARRY, PH.D., ([dagger]) AND FREDERIC C. BLOW, PH.D. ([dagger])

Department of Psychiatry, Rachel Upjohn Building, University of Michigan, 4250 Plymouth Road, Ann Arbor, Michigan 48109

([dagger]) Correspondence may be sent to Maureen A. Walton at the above address or via email at: waltonma@umich.edu. Abby L. Goldstein is with the Department of Psychology, York University, Toronto, Ontario, Canada. Stephen T. Chermack, Kristen L. Barry, and Frederic C. Blow are also with Health Services Research and Development, Department of Veterans Affairs, Ann Arbor, MI. Rebecca M. Cunningham is with the Department of Emergency Medicine, University of Michigan, Ann Arbor, MI.

TABLE 1. Mean (SD) quantity/frequency, heavy drinking, and DrInC
consequences for moderator variables, by follow-up (N = 494)

Moderator                               Baseline       3 months
variable                               Mean (SD)      Mean (SD)

Quantity/frequency
  Stage of change
    Precontemplation                  18.5 (16.1)    11.4 (10.5)
    Contemplation                     21.8 (15.5)    14.8 (13.7)
    Preparation/action                23.0 (17.8)    18.0 (15.6)
  Drink w/in 6 hours before injury
    Yes                               23.2 (19.8)    16.9 (16.4)
    No                                19.5 (15.5)    13.0 (11.5)
  Self-efficacy
    Low (not at all-moderate)         32.9 (24.0)    21.6 (22.4)
    High (very-extremely)             19.1 (15.2)    13.2 (11.4)
  Attribute injury to alcohol
    Yes                               25.3 (21.1)    19.7 (18.8)
    No                                19.6 (15.8)    13.0 (11.5)
Heavy drinking
  Stage of change
    Precontemplation                   6.0 (5.3)      3.9 (4.6)
    Contemplation                      8.6 (7.4)      5.9 (6.6)
    Preparation/action                 8.3 (6.4)      6.2 (6.1)
  Drink w/in 6 hours before injury
    Yes                                8.7 (6.7)      6.3 (6.2)
    No                                 6.5 (5.6)      4.4 (5.1)
  Self-efficacy
    Low (not at all moderate)         10.1 (7.9)      6.9 (7.9)
    High (very-extremely)              6.7 (5.7)      4.7 (5.2)
  Attribute injury to alcohol
    Yes                               10.0 (7.0)      7.5 (6.8)
    No                                 6.6 (5.7)      4.4 (5.1)
DrInC consequences
  Stage of change
    Precontemplation                   3.4 (4.0)      2.3 (3.7)
    Contemplation                      7.5 (5.9)      3.9 (3.5)
    Preparation/action                 7.9 (7.7)      7.1 (8.1)
  Drink w/in 6 hours before injury
    Yes                                8.7 (7.9)      5.5 (7.3)
    No                                 4.2 (5.0)      3.5 (5.3)
  Self-efficacy
    Low (not at all-moderate)         13.6 (10.1)     9.2 (11.0)
    High (very-extremely)              4.5 (5.0)      3.5 (4.8)
  Attribute injury to alcohol
    Yes                               12.0 (9.1)      7.9 (10.3)
    No                                 4.2 (4.7)      3.4 (4.4)

Moderator                              12 months          p
variable                               Mean (SD)

Quantity/frequency
  Stage of change
    Precontemplation                  10.6 (11.0)       <.0001
    Contemplation                     17.3 (11.5)         NS
    Preparation/action                15.6 (12.8)       <.0001
  Drink w/in 6 hours before injury
    Yes                               15.6 (15.2)       <.0001
    No                                12.4 (10.4)       <.0001
  Self-efficacy
    Low (not at all-moderate)         16.8 (11.9)         NS
    High (very-extremely)             12.7 (11.9)       <.0001
  Attribute injury to alcohol
    Yes                               17.8 (18.5)        <.05
    No                                12.4 (10.2)      -0.0001
Heavy drinking
  Stage of change
    Precontemplation                   3.7 (4.6)        <.0001
    Contemplation                      8.3 (7.1)          NS
    Preparation/action                 5.6 (5.7)        <.0001
  Drink w/in 6 hours before injury
    Yes                                6.1 (6.2)        <.0001
    No                                 4.6 (5.3)        <.0001
  Self-efficacy
    Low (not at all moderate)          6.6 (7.0)          NS
    High (very-extremely)              4.8 (5.4)        <.0001
  Attribute injury to alcohol
    Yes                                6.8 (5.9)        <.001
    No                                 4.7 (5.5)        <.0001
DrInC consequences
  Stage of change
    Precontemplation                   1.9 (2.8)        <.0001
    Contemplation                      3.9 (3.6)          NS
    Preparation/action                 6.9 (8.1)        <.001
  Drink w/in 6 hours before injury
    Yes                                5.4 (7.8)        <.0001
    No                                 3.3 (4.5)         <.01
  Self-efficacy
    Low (not at all-moderate)          7.5 (8.2)         <.05
    High (very-extremely)              3.3 (5.1)        <.0001
  Attribute injury to alcohol
    Yes                                7.1 (9.6)        <.0001
    No                                 3.3 (4.5)        <.0001

Notes: DrInC = Drinker Inventory of Consequences; NS =
not significant; w/in = within.

TABLE 2. Average weekly consumption: Means by follow-up and alcohol/
injury attribution, by advice interaction

                            Alcohol/injury attribution
                                    Mean (SD)

                              Yes              No
                          attribution,    attribution,
                             advice          advice
Follow-up                   (n = 41)        (n = 212)

Baseline                   23.3 (19.3)     20.6 (17.4)
3 months                   22.3 (23.6)     13.5 (11.6)
12 months                  12.9 (11.9)     12.6 (9.6)

Baseline to 12 months
  Difference in mean          -10.4           -8.0
  change in mean              -44.6           -38.8
  p                           .024            <.001

                            Alcohol/injury attribution
                                    Mean (SD)

                               Yes             No
                          attribution,    attribution,
                            no advice       no advice
Follow-up                   (n = 36)        (n = 205)

Baseline                   27.5 (23.1)     18.6 (13.9)
3 months                   16.7 (10.5)     12.4 (11.3)
12 months                  22.9 (22.7)     12.1 (10.8)

Baseline to 12 months
  Difference in mean          -4.6            -6.5
  change in mean              -16.7           -34.9
  p                           .585            <.001

Table 3. Heavy drinking: Means by follow-up and alcohol/injury
attribution, by advice interaction

                            Alcohol/injury attribution
                                   Mean (SD)

                              Yes              No
                          attribution,    attribution,
Follow-up                    advice          advice
                            (n = 41)        (n = 212)

Baseline                    9.3 (6.5)       6.7 (5.5)
3 months                    8.0 (8.0)       4.6 (4.8)
12 months                   4.6 (4.5)       5.1 (5.4)

Baseline to 12 months
  Difference in mean          -4.7            -1.6
  change in mean              -50.5           -23.9
  p                           .007            <.001

                            Alcohol/injury attribution
                                   Mean (SD)

                               Yes             No
                          attribution,    attribution,
Follow-up                   no advice       no advice
                            (n = 36)        (n = 205)

Baseline                   10.6 (7.5)       6.4 (5.9)
3 months                    7.0 (5.0)       4.1 (5.4)
12 months                   9.2 (6.4)       4.2 (5.5)

Baseline to 12 months
  Difference in mean          -1.4            -2.2
  change in mean              -13.2           -34.4
  p                           .136            <.001