|Home | About | Journals | Submit | Contact Us | Français|
Aims: This study examines whether the severity of baseline alcohol consumption/consequences moderates the effect of an alcohol brief intervention (BI) in the emergency department (ED). Methods: Injured patients (N = 494) were recruited from an ED, randomly assigned to receive brief advice or not and completed a 12-month follow-up interview. Results: A significant interaction was found between severity of baseline alcohol consumption (i.e. average weekly, binge drinking) and receipt of a BI on alcohol consumption at 12 months. The form of this interaction indicates that the BI group tended to report lower alcohol consumption at follow-up than the untreated group especially in those who had reported high baseline consumption. Severity of alcohol consequences at baseline did not significantly impact the effect of the BI on 12-month outcomes. Conclusion: ED patients with higher alcohol consumption benefit from BI. In some cases, the BI's effects may be enhanced for patients who are heavier drinkers, perhaps due to a greater opportunity to develop a discrepancy between current behavior and future goals.
Based on data from the 2006 National Household Survey, problematic alcohol use is common, with an estimated 7.6% of the population of the United States meeting criteria for a diagnosable alcohol use disorder and 23% of the US residents engaged in recent risky drinking [Substance Abuse and Mental Health Services Administration (SAMHSA), 2006]. However, a small minority of those with risky alcohol use or a diagnosable alcohol use disorder ever attend any formalized alcohol treatment (SAMHSA, 2006). As a way to bridge this gap between treatment need and the provision of care, alcohol brief interventions (BIs) have been designed for delivery outside traditional specialty addictions treatment settings. The Emergency Department (ED) is a logical setting to implement BIs. Alcohol use is common in individuals presenting for ED services with estimates of up to 36% of all injured patients having some alcohol in their system at the time of presentation to the ED (Cherpitel, 1988, 1993, 1997). Additionally, an ED visit provides a unique ‘teachable moment’ because patients receiving treatment for an alcohol-related injury might be motivated to re-think their alcohol use (Blow et al., 2006; Cherpitel et al., 2006).
Consistent evidence indicates that BIs in the ED are effective and cost effective (D’Onofrio and Degutis, 2002; Crawford et al., 2004; Barrett et al., 2006; Havard et al., 2008; Nilsen et al., 2008), with ED research suggesting a savings of $3.81 for every $1 spent (Gentilello et al., 2005). A recent meta-analysis of BIs in the ED estimates that, on average, participants report about half the rate of alcohol-related injuries in the year following the intervention (Havard et al., 2008); however, the results related to alcohol consumption were not conclusive in ED samples. These findings conflict with meta-analyses from primary care settings, where BIs significantly reduce alcohol consumption (Bertholet et al., 2005; Kaner et al., 2007), but are consistent with meta-analyses from inpatient hospital patients, which generally do not support BI's effectiveness (Emmen et al., 2004).
The lack of consistent effects of BI on alcohol consumption in ED samples may be due to several issues, including sample selection issues with regard to alcohol problem severity. Specifically, there is considerable heterogeneity in the level of alcohol use among ED studies (Havard et al., 2008), and this sample heterogeneity in consumption may be an important factor in the large range of effect sizes for BIs (Noonan and Moyers, 1997). ED studies (and inpatient hospital samples, see Saitz et al., 2007) often include patients with greater alcohol problem severity than primary care samples (Cherpitel, 1999). It is possible that including a high proportion of participants with severe alcohol problems would reduce the impact of BIs on alcohol use because severe alcohol problems may require more intensive intervention approaches. Conversely, the exclusion in some ED studies of more severe drinkers (e.g. AUDIT >14, alcohol dependence, alcohol treatment) may result in restriction of range, preventing detection of changes in consumption (Havard et al., 2008). Clearly, there is a need to more directly examine the potential impact of alcohol use severity on response to BIs delivered in the ED.
Studies that have examined the effectiveness of BI in relation to alcohol problem severity specifically have demonstrated inconsistent findings, perhaps due to variation in definitions of ‘more severe’ drinkers (e.g. diagnosis of alcohol dependence versus heavy drinkers) (Bien et al., 1993; Vasilaki et al, 2006), and/or the selection of outcomes (alcohol use versus alcohol-related problems). A recent meta-analyses evaluating BI's effects on dependent drinkers in multiple settings (including the ED) found that BIs were more effective in decreasing alcohol consumption in studies that excluded alcohol-dependent individuals (Moyer et al., 2002; Vasilaki et al., 2006). In contrast, a recent study of a BI in primary care settings found that those with and without alcohol dependence responded similarly to the intervention (Guth et al., 2008). Regarding BI's effects among heavy drinkers, a meta-analysis which included ED studies showed that the effectiveness of BIs on alcohol consumption was augmented among ‘heavy or low dependent’ drinkers (Vasilaki et al., 2006). However, to date, the literature remains unclear as to the extent to which baseline alcohol problem severity (i.e. consumption and consequences) may moderate the effects of BIs specifically in ED settings.
The present study is a secondary data analysis of a BI study in the ED (Blow et al., 2006) explicitly examining whether alcohol problem severity moderates the impact of a BI. Prior analyses found a general benefit of receiving brief advice on 12-month alcohol consumption (Blow et al., 2006) and that this effect was greatest among participants who attributed their injury to their alcohol use (Walton et al., 2008). The present analyses are possible because no participants were excluded from this study due to the severity of baseline alcohol problems; thus, examination of average consumption, binge drinking and alcohol-related consequences, as moderators of the effectiveness of the BI on 12-month alcohol use, was conducted. The findings will provide important information regarding the appropriateness of BI for ‘more severe’ drinkers in the ED setting, help identify subgroups of drinkers who differentially respond to BIs in the ED and thus have implications for future screening and intervention development efforts in the ED.
As depicted in Fig. Fig.1,1, a total of 4476 adults who presented with acute injuries to a university-affiliated level 1 ED between 8 AM and midnight during a 1-year period were recruited to participate in a computer-based survey of health issues (see Blow et al., 2006, for additional information). Patients were excluded from the study if they were severely injured (e.g. unconscious); were in need of immediate life-saving procedures; presented with a self-inflicted injury, sexual assault, overdose, poisoning, near drowning, chronic injury without specific event associated with re-injury; were pregnant, prisoners, or did not speak English. After completing a computerized survey, those meeting at-risk drinking criteria (n = 575) were randomly assigned to one of four intervention conditions: brief advice with a tailored message booklet, brief advice with a generic message booklet, a tailored message booklet only or a generic message booklet only. To prevent the possibility that staff could unwittingly manipulate assignment to intervention condition, participants were randomly assigned to conditions via the computer using an urn randomization technique. Stratified randomization occurred in blocks of 10, for each gender, in order to equalize randomization over time. Participants received the intervention during their ED visit following a saliva alcohol test to verify their blood alcohol concentration. For participants who screened positive for acute alcohol consumption, the intervention proceeded once their blood alcohol concentration reached 100 mg/dL or less. Follow-up interviews were completed over the telephone, in person or by mail at 12 months with 86% of the original sample. Of the 494 participants with complete follow-up data who comprised the final sample, 71% were male, 86% Caucasian, 5.7% African American and 8.3% of other race/ethnicity. The mean age of the sample was 27.8 (SD = 12.2; median = 22; range: 19–76). The majority of the participants (70%) were never married, and more than 80% had some college education. Participants in the intervention groups did not differ in age, gender, marital status, race/ethnicity or years of education or level of baseline alcohol use (see Table Table1).1). The study was approved by the Institutional Review Board at the University of Michigan.
Participants completed a computerized baseline survey with questions about alcohol use embedded in a larger health and lifestyle assessment that included injury-related questions to encourage response accuracy. The baseline and follow-up measures were identical. This study utilized data on demographics, alcohol use and alcohol-related consequences. All measures of alcohol use were continuous.
Demographic data derived from participants’ surveys included age, gender, race/ethnicity, marital status and years of education completed.
At-risk drinking, determined based on patient report of average alcohol consumption or frequency of binge drinking, was assessed with three questions modeled after the short form of the Alcohol Use Disorders Identification Test (AUDIT-C; Saunders et al., 1993; Gordon et al., 2001).
Average alcohol consumption was calculated based on two questions assessing participants’ typical frequency and quantity of alcohol use in the past 3 months. Specifically, participants were asked to report the number of days per week they had drank any alcohol in the last 3 months (frequency) and indicate the number of drinks they usually had on days that they drank alcohol (quantity), also in the past 3 months. Scores on this measure represent the number of days per week drinking multiplied by the typical number of standard drinks per occasion (range 0–100) (a standard drink is that which contains about 13 g ethanol).
Binge drinking was assessed with a single question asking participants to indicate how often they had consumed more than four drinks, if they were a woman, or more than five drinks, if they were a man, per occasion within the past month. The binge drinking measure reflects the total number of days of binge drinking during the past month (range 0–30).
Alcohol-related consequences were measured with the DrInc SIP (Miller et al., 1995) plus two items from the full DrInc reflecting alcohol-related injuries and arrests for drinking and driving were added: ‘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 measure summed items from the physical, social responsibility, intrapersonal, impulse control and interpersonal consequence subscales to create the total score, with sound internal consistency (Cronbach's alpha = 0.89–0.90) and with a range of 0–17.
During the brief advice session, the research social worker reviewed the booklet, either tailored or generic, with the participant prior to discharge from the ED. The research social workers received training in principles of motivational interviewing (e.g. rolling with resistance, developing discrepancy; Miller and Rollnick, 2002). For the no brief advice condition, participants were only told that, based on their health screen responses, they scored as at risk for hazardous drinking and received either a tailored or a generic booklet to review. The difference between the tailored and genetic booklets was that the tailored booklet included text based on individual screening responses rather than standard text. Because prior research found no evidence for the impact of tailoring (Blow et al., 2006), the tailored and non-tailored conditions were combined for analysis. Thus, the present study examines the impact of brief advice versus a control condition of an advice booklet. Follow-up rates for the brief advice and control conditions were 88% and 84%, respectively; these rates were not significantly different from one another.
Consistent with recommendations for examining moderators (Kraemer et al, 2001), the present study examined the extent to which baseline average alcohol consumption, baseline binge drinking and baseline alcohol-related consequences modified the effect of brief advice on 12-month outcomes. Three separate negative binomial regression models were run that included (1) baseline average alcohol consumption, treatment condition (advice versus no advice) and their interaction; (2) baseline binge drinking, treatment condition (advice versus no advice) and their interaction; and (3) baseline alcohol-related consequences, treatment condition (advice versus no advice) and their interaction.
Results from the primary regression analyses are presented in Table Table22.
After accounting for the main effects of baseline average alcohol consumption and treatment condition, the interaction between these two variables was significant. Looking at the form of the interaction effect (see Fig. Fig.2),2), the impact of the advice condition on 12-month average alcohol consumption is more noticeable in those with greater baseline severity.
The results of the analyses of binge drinking are similar (Fig. (Fig.3).3). The effect of the interaction between binge drinking and treatment condition on 12-month binge drinking is significant after accounting for their main effects.
In the analyses related to consequences, only the main effect of baseline alcohol consequences was significant; no significant interaction was found between baseline consequences and treatment condition.
Baseline severity of alcohol consumption moderated the effectiveness of BIs. The form of this interaction indicates that the BI group tended to report lower alcohol consumption at follow-up than the untreated group especially in those who had reported high baseline consumption, i.e. the effect of the BI was greatest among patients with more severe patterns of alcohol consumption prior to their ED visit. Thus, data from this study suggest that ED patients with higher levels of baseline alcohol consumption appear to be appropriate for BIs. These findings are consistent with Guth et al. (2008) who found that BIs delivered in primary care settings were effective for those with alcohol dependence and with a meta-analysis showing that BI's effectiveness was augmented among ‘heavy or low dependent’ drinkers (Vasilaki et al., 2006). To the best of our knowledge, no prior data from ED samples exist on the extent to which baseline severity moderates the effect of BIs.
Originally, the concept of the BI was to intervene with patients in medical settings before problems become more severe (Babor and Grant, 1992; Bien et al., 1993; Barry, 1999). This approach makes sense from a public health perspective but it is important not to underestimate the extent to which BIs may be able to help the heavier drinkers. One of the strengths of BIs for implementation in the ED is that they can be delivered to a large number of patients, with differing levels of problem severity, who may not present to primary care settings. Our findings suggest that more severe patients should not be excluded from BIs, as done in some prior ED studies (Havard et al., 2008). However, it is likely that heavy drinkers with alcohol dependence would benefit from referral to specialty substance abuse treatment; and prior research shows that BIs based on motivational interviewing increase treatment engagement (Dunn et al., 2001). The BI in the present study focused on alcohol use patterns as the primary target behavior rather than targeting treatment engagement; however, it is possible that the impact of the BI for heavier drinkers could have been impacted at least partially by facilitating treatment use. Thus, although our results suggest that heavier drinkers respond positively to BIs delivered in the ED, it is possible that the effectiveness of BIs for dependent drinkers could be enhanced by also including treatment seeking and engagement more explicitly as target behaviors. Additional research is needed to determine the most effective components and areas of focus of BIs or other referral approaches (e.g. case management) to improve outcomes (and/or increase treatment engagement) among dependent drinkers.
These results should be interpreted with caution for several reasons. First, although the present analyses focused on severity of alcohol use at baseline as a potential moderator, severity likely correlates with numerous other patient attributes. Patients with high levels of psychiatric severity and cognitive impairment, and those with acute intoxication preventing provision of informed consent, were excluded from the present sample, so the results may not apply to individuals with the highest levels of severity. The study was designed to accurately reflect ED patients, but the extent to which the results apply to other settings is unknown.
The present pattern of findings adds to a growing body of literature that indicates that ED patients with more severe alcohol consumption are responsive to BIs. In particular, it may be that these heavier drinkers are most likely to improve following the delivery of BIs in an ED. Additional studies are needed to determine the effectiveness of ED-based BIs on patients with alcohol dependence. Further research is needed also to better identify the essential components of BIs for patients with differing levels of alcohol problem severity.
This project was supported by a grant (#AA11629) from the National Institute on Alcoholism and Alcohol Abuse (NIAAA). We would like to thank Lynn Massey, Scott Kelly, Robert Vaidya, Harpret Otal, Margaret White and Robin Williams for their work on this project.