Can Motivational Interviewing in Emergency Care Reduce Alcohol Consumption in Young People? A Systematic Review and Meta-analysis

Abstract

Aims: We investigate the effect of motivational interviewing (MI), delivered in a brief intervention during an emergency care contact, on the alcohol consumption of young people who screen positively for present or previous risky alcohol consumption. Methods: MEDLINE, CINAHL, EMBASE, PsycARTICLES, PsycINFO, PSYNDEX and Scopus were searched for randomized controlled trials with adolescents or young adults that compared MI in an emergency care setting to control conditions and measured drinking outcomes. Results: Six trials with 1433 participants, aged 13–25 years, were included in the systematic review and meta-analysis. MI was never less efficacious than a control intervention. Two trials found significantly more reduction in one or more measures of alcohol consumption in the MI intervention group. One trial indicated that MI may be used most effectively in young people with high-volume alcohol consumption. Separate random effects meta-analyses were performed based on the highest impact that MI added on reducing the drinking frequency and the drinking quantity at any point in time during the different study periods. Their results were expressed as standardized mean differences (SMDs). The frequency of drinking alcohol decreased significantly more after MI than after control interventions (SMD ≤ −0.17, P ≤ 0.03). In addition, MI reduced the drinking quantity further than control interventions in a meta-analysis of the subset of trials that were implemented in the USA (SMD = −0.12, P = 0.04). Meta-analyses of the smallest mean differences between MI and control groups detected no differences in alcohol use (SMD ≤ 0.02, P ≥ 0.38). Conclusion: MI appears at least as effective and may possibly be more effective than other brief interventions in emergency care to reduce alcohol consumption in young people.

INTRODUCTION

Binge drinking or heavy episodic drinking is a global health problem among young people, which varies widely between countries. The WHO recently published global data for the year 2010. The worldwide highest binge drinking rates among 15- to 19-year-olds occurred in European countries and Canada, where ≥30% of this age group drink 60 or more grams of pure alcohol on at least one single occasion at least monthly (WHO, 2014). In the USA and Brazil, the binge drinking prevalence in this age group was 10–20% and 20–30%, respectively. The younger people start consuming alcohol, the higher are their risks to drink regularly, to develop an addiction, and to suffer or die from alcohol-related illnesses (Grant and Dawson, 1998). In Germany, for instance, one of the high alcohol consumption countries, 15,680 young people aged 25–35 years and 35,386 young people aged 15–25 years were hospitalized due to alcohol intoxication in 2012. This was ∼1 in every 250 young adults in the younger age group, and twice as many alcohol-related hospitalizations among 15- to 35-year-olds than a decade ago (Federal Statistical Office of Germany, 2014a,b).

The admission to hospital after an excessive spree has been considered a window of opportunity to offer secondary prevention methods (Spirito et al., 2004). Being admitted to the hospital, amongst other factors, was associated with a greater intention to change drinking at the time of the event (Barnett et al., 2002). Some interventions to reduce alcohol-related harms therefore apply motivational interviewing (MI) techniques to offer secondary prevention to hospitalized young people who screen positive for risky alcohol consumption in an emergency care setting (Hofmann and Kohler, 2013). MI is based on the assumption that facilitating and engaging intrinsic motivation is essential for behavioural change. Its central purpose is the examination and resolution of ambivalence. The counsellor is intentionally directive in pursuing this goal. MI is more focused and goal-directed than nondirective counselling (Miller and Rose, 2009; Rollnick and Miller, 2009).

Several systematic reviews have recently discussed the efficacy of MI in medical settings, some of which with a focus on young people, and concluded differently: Most youth reduced alcohol use, regardless of the form of care received during an intervention. Consequently, clear benefits of using MI in emergency care to reduce alcohol or other drug use, and associated injuries or high-risk behaviours remain inconclusive (Yuma-Guerrero et al., 2012; Newton et al., 2013). MI in various inpatient and outpatient settings appears to be modestly advantageous over comparison interventions with particular promise in areas like alcohol and tobacco use among (Lundahl et al., 2013). MI in emergency departments that targeted alcohol problems can reduce alcohol consumption or alcohol-related injuries of adults (D'Onofrio and Degutis, 2002; Havard et al., 2008; Nilsen et al., 2008). Brief interventions with adolescents, in which MI was the predominant approach, were effective in reducing alcohol consumption and related harm (Patton et al., 2014) or alcohol consumption and other drug use (Tait and Hulse, 2003).

We conducted a systematic review and meta-analyses to examine changes in alcohol consumption after brief motivational interventions for young people with existing alcohol use problems, who were admitted to an emergency care unit alcohol positive, with an alcohol-related trauma, or with a history of elevated alcohol consumption.

METHODS

We followed the PRISMA guidelines in conducting this study.

Search strategy

The following databases were searched from inception to 24 September 2013: MEDLINE through PubMed, EMBASE through OvidSP, CINAHL, PsycARTICLES, PsycINFO, PSYNDEX through EBSCOhost, and Scopus. Search terms were (‘motivational interview*’ OR ‘intervention*’) AND (‘alcohol*’ OR ‘drink*’) AND (‘trauma*’ OR ‘injury*’ OR ‘emergency*’). The applied filters were randomized controlled trial (RCT) or clinical trial and adolescent or young adult. The EMBASE search included the filter adult instead of young adult. We also performed a manual search of the references of all retrieved articles (Fig. 1).

Eligibility criteria

Studies were included in the systematic review and meta-analyses if they: evaluated MI or motivational enhancement therapy (MI plus feedback) that targeted alcohol consumption against either another or no intervention in an emergency care setting; included, at least some, young people of age 18 years or younger; employed a RCT design. Studies were excluded if they: allowed participants older than 25 years; did not report drinking outcomes; were not published in English or German in a peer-reviewed source. Studies, study arms or drinking outcome data were also excluded from our review and meta-analyses if: MI was delivered only through a computer-based program; MI was also delivered in standard care; disaggregate outcome data were not available.

Data extraction

We extracted data on the study design and population, details of the intervention, length of follow-up, attrition, drinking outcomes collected, results, and the methods employed for data collection and analysis. Extracted results included the mean and standard deviation of drinking outcomes for the MI and control groups as well as the significance of their difference at baseline and all follow-up times. The significance levels extracted for the treatment, time and interaction effects on the drinking outcomes were adjusted for covariates by most studies and, hence, do typically not correspond to the summarized means, mean differences and standard deviations of the summarized raw data used in our meta-analyses.

Study quality

We assessed the quality of the studies using the Critical Appraisal Skills Programme (CASP) RCT checklist.

Statistical analysis

Alcohol consumption data from the study arms of the trials that fulfilled our study eligibility criteria were pooled. Drinking outcomes were assessed in separate meta-analyses that assessed either drinking frequency, or drinking quantity measures. Due to different measurement scales for these outcomes across trials, effects were expressed as standardized mean differences (SMD) in the meta-analyses. The meta-analyses were performed using a random effects model to account for the heterogeneity between the trials included. The random effects model allowed for a distribution of true effect size (see, e.g. Borenstein et al., 2010). Our goal was to estimate the mean of this distribution. Heterogeneity of effect sizes was assessed by using the I-squared value. Analyses were performed in Stata using the metan, metabias and metafunnel commands.

Either one drinking frequency outcome, or one drinking quantity outcome per trial was used within each analysis to meet the assumption of independence (see, e.g. Murray et al., 2012). We applied the following method to select independent data: We included the data from follow-up during which MI had the strongest effect (SMD) in comparison to the control intervention if studies reported an outcome at more than one follow-up. Meta-analyses of the weakest effects of MI in comparison to the control intervention were performed in addition to establish a lower confidence bound of the SMD.

If a study reported more than one drinking quantity variable, the measure most comparable across studies was included in the meta-analysis (Monti et al., 1999; Bernstein et al., 2010). For one trial, the drinking quantity and its standard error were calculated based on the published frequency of light, moderate and heavy alcohol use (Segatto et al., 2011). For two trials in which pertinent outcomes were not reported, the corresponding authors provided the data upon request (Monti et al., 1999; Cunningham et al., 2009).

Publication bias of the enclosed literature was assessed based on funnel plots and Egger's test for small-study effects.

RESULTS

Study characteristics

Our selection criteria yielded six study arms from six RCTs (Monti et al., 1999, 2007; Spirito et al., 2004; Cunningham et al., 2009; Bernstein et al., 2010; Segatto et al., 2011). For one trial, three separate studies describe the results from the follow-ups after 3, 6 and 12 months (Cunningham et al., 2009, 2012; Walton et al., 2010). We refer to this trial using its first publication by Cunningham et al. (2009). All trials implemented MI in the setting of an emergency department; five in the USA and one in Brazil. There were 1433 participants across the six study arms reviewed and included in the meta-analysis. Sample size varied from 94 to 567 participants. The age of the participants ranged from 13 to 25 years. The share of females was between 9.7 and 67.8%. Participants were followed-up after a minimum of 3 or 6 months, and a maximum of 6 or 12 months. The follow-up rates at completion varied from 69.4 to 93.4%. Alcohol consumption patterns were either collected over 30 days, 3 months, or an unspecific time period (Table 1).

Two study arms from two trials were excluded from the review and meta-analyses. One of them assessed a computerized brief intervention in a three arm trial (Cunningham et al., 2009). The other excluded study arm was a minimally assessed control group, for which drinking outcome data were not reported (Bernstein et al., 2010).

Intervention characteristics

The studied brief interventions aimed at reducing alcohol consumption and associated risks or problems, like drunk driving, violence or alcohol-related injuries. Interventions lasted from 5 to 45 min and were usually longer than 20 min, or had a median length of 37 min. The target group was young people in emergency care who screened positively for past or present risky alcohol consumption. To determine eligibility for a brief intervention, studies screened for hazardous drinking in form of self-reported alcohol consumption, blood alcohol concentration, alcohol in saliva or breath, high-risk behaviour in conjunction with alcohol use, or a combination of these.

Control interventions, or standard care, included written information (e.g. alcohol-use risk handout, educational brochure), a contact list (e.g. community resources, adolescent treatment facilities), a phone follow-up, or personal feedback. MI interventions included a treatment more or less similar to the control intervention plus MI. One trial combined MI with normative resetting and skills training (Cunningham et al., 2009). In two trials, the MI groups also received additional ‘booster’ phone calls after 10 days, or after 1 and 3 months that were not part of the control interventions (Monti et al., 2007; Bernstein et al., 2010) (Table 2). Interventions details are described in the Supplementary Table S1.

Drinking outcomes

A variety of different self-reported outcomes were used to measure alcohol consumption (Table 3). Studies originally reported between two and four drinking outcomes that can be categorized as either quantity or frequency measures, or a combination thereof. Four studies reported the quantity of alcohol as actual or average drinks/units consumed within a given time frame (Monti et al., 1999, 2007; Spirito et al., 2004; Bernstein et al., 2010). Also four studies reported the frequency of drinking days or episodes within a given time frame (Monti et al., 1999; Spirito et al., 2004; Bernstein et al., 2010; Segatto et al., 2011). The three related studies of the same trial reported an Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) cut-off score, which is a combined measure of the drinking frequency and quantity (Cunningham et al., 2009). Four trials included a separate measures for the frequency of heavy drinking (Spirito et al., 2004; Monti et al., 2007; Cunningham et al., 2009; Segatto et al., 2011). One study asked about moderate and low drinking days in addition (Segatto et al., 2011).

Baseline differences

Several sociodemographic baseline characteristics and baseline alcohol use measures were balanced in five of the six study arms included in the review and meta-analyses. In two trials, also measures of alcohol-related risks were balanced at the baseline (P > 0.05, Table 4). In the Bernstein et al. (2010) trial, age, sex, race, primary language, baseline AUDIT score by age group, and consumption variables were similar, but the control group had a higher rate of driving after drinking (P < 0.05). In the Cunningham et al. (2009) trial, age, gender, race, school dropout, reason for the emergency department (ED) visit, past-year substance abuse as well as violence, family public assistance, failing grades, live with parents and gang involvement were similar between the randomized groups at the start of the trial. In the Monti et al. (1999) trial, age, gender, ethnicity, number of years in school, school status, reason for the ED visit, BAC, drinking behaviour and the severity of alcohol involvement, measured by the Adolescent Drinking Inventory, were similar. In the Monti et al. (2007) trial, age, gender, ethnicity, BAC and reasons for treatment were similar, but the control group had more years of school (P < 0.01). In the Segatto et al. (2011) trial, age, gender, ethnicity (P = 0.09), marital status, school status, reason for the ED visit, various measures of alcohol abuse and consumption patterns were similar, but the MI group had more risks associated with alcohol abuse according to an index calculated from an Alcohol Consumption Risk Questionnaire (P = 0.05). In the Spirito et al. (2004) trial, there were no baseline differences between the MI and control groups' sex, race, school status, reason for the ED visit, blood alcohol concentration as well as self-reported alcohol consumption, drinking and driving, and the severity of alcohol involvement. An imbalance in age across conditions was noted by Spirito et al. (2004) in the last year of recruitment, and it was corrected prior to the end of the trial through stratifying participants by age in the trial's final year.

These reported baseline assessments notwithstanding, the drinking frequency (SMD = 0.01, P > 0.81) and the drinking quantity (SMD = 0.09, P > 0.31) of the MI and control groups were statistically indistinguishable in meta-analyses of the drinking outcomes at baseline.

Quality of included studies

The quality of included studies was evaluated using the CASP checklist for RCTs (Table 4). The quality of the six RCTs included was poor to good. All studies addressed a clearly focused question and described their context and the eligibility criteria for the participants. Randomization of the participants into two or three study arms was carried out in all of the trials. The randomization method used was not explained in one study (Monti et al., 1999). Intervening with MI rendered comprehensive blinding impossible, but four trials reported research assistants collecting data were blind (Monti et al., 2007; Cunningham et al., 2009; Bernstein et al., 2010; Spirito et al., 2011). One study states that patients were blinded (Segatto et al., 2011), but, as in the other trials, patients can infer their treatment status from the distinguishable interventions. One trials showed sociodemographic differences between the groups at the beginning (Monti et al., 2007). In two trials, the young patients were treated not only with MI but also with ‘booster’ phone calls that the control group did not receive (Monti et al., 2007; Bernstein et al., 2010). A further trial combined MI with normative resetting and skills training, not received by the control group (Cunningham et al., 2009). Patient follow-up and losses were explained in detail for two trials (Cunningham et al., 2009; Segatto et al., 2011). Dropouts were characterized in three studies (Monti et al., 1999, 2007; Spirito et al., 2004), and included in the analysis of two trials (Cunningham et al., 2009; Bernstein et al., 2010). Statistical power was discussed in the studies of two trials (Cunningham et al., 2009; Bernstein et al., 2010). An intention-to-treat analysis was reported or apparent from participant numbers described for the analyses in four studies (Monti et al., 1999, 2007; Cunningham et al., 2009; Bernstein et al., 2010) and confirmed by the authors of a further study (Spirito et al., 2004). Results were not analysed based on the initial treatment in one study (Segatto et al., 2011). The follow-up rates, which ranged from 69.4 to 93.4%, indicate that attrition is a moderate to severe limitation to all studies (Table 1). Some studies adjusted for attrition and baseline differences in their analyses of drinking outcomes (Cunningham et al., 2009) or attempts to quit drinking (Bernstein et al., 2010), but the unadjusted raw data of the trials were used in our meta-analyses.

Systematic review of effects

In one trial, neither the drinking frequency nor the drinking quantity changed significantly for up to 1 year after the brief intervention during an alcohol-related emergency hospitalization, regardless of the type of intervention used (P > 0.05) (Bernstein et al., 2010). Five RCTs found that young people consumed less alcohol after a hospital stay and brief intervention, regardless of whether MI was used in the intervention or not (P ≤ 0.01). Based on the raw data summarized in Table 3, half of the trials observed some of the lowest amount of drinking at the initial follow-ups after 3 or 6 months, and rising consumption levels afterwards (Monti et al., 1999; Spirito et al., 2004; Cunningham et al., 2009). In four trials, some of the least frequent drinking occurred as well before the end of the study (Monti et al., 1999, 2007; Spirito et al., 2004; Cunningham et al., 2009) (Supplementary Fig. S1). However, t-tests of the equality of the tabulated group means only indicate a significant difference in the mean drinking quantity between the follow-ups after 3 and 12 months in the Spirito et al. (2004) trial (P < 0.04).

Studying moderators of the treatment effect, one trial found that MI reduced alcohol consumption more than a short consultation and written information in the subgroup of young patients with a high Adolescent Drinking Index score: Those who reached the clinical cut-off for referral for alcohol problems (≥15 index points) at baseline mentioned significantly fewer drinking days and high-volume drinking days per month at follow-up (P < 0.05) (Spirito et al., 2004). In another trial, the time-treatment interaction was significant. This indicates that the reduction in alcohol consumption over time with respect to the baseline level was stronger in the MI group than in the control group: MI participants drank on fewer days, had fewer heavy drinking days and drank fewer drinks per week than did control patients at the 6 and 12 months follow-up after accounting for possible baseline differences in these variables, which may have existed despite randomization into the intervention groups (P < 0.01) (Monti et al., 2007). Time–treatment interactions in the other trials were insignificant.

Meta-analyses of effects

The average effect of comparable MI interventions was assessed in terms of a SMD between MI and control group outcomes. The SMD was estimated in separate meta-analyses for the drinking frequency and drinking quantity. A negative SMD implies that MI triggered a higher reduction in alcohol consumption than the control intervention. The estimated SMD varied by the type of drinking outcome studied.

MI reduced alcohol consumption at least as much as a control intervention. Analysing largest mean differences in the drinking behaviour, the frequency of drinking was significantly lower in the MI groups (SMD = −0.17, P = 0.03). MI showed no advantage over control interventions in terms of reducing the drinking quantity in the meta-analysis (SMD = −0.09, P = 0.18). Excluding the RCT from Brazil, which caused a substantial fraction of the heterogeneity in the data, and analysing only the data from trials in the USA indicates that MI is more efficacious than other interventions in reducing drinking frequency (SMD = −0.21, P < 0.01) and drinking quantity (SMD = −0.12, P = 0.04) (Fig. 2). Similar meta-analyses of the weakest effects of MI in comparison to the control intervention indicated no difference in drinking behaviour (SMD ≤ 0.02, P ≥ 0.38) (Fig. 3).

Heterogeneity across the studies in the meta-analyses was quantified using the I-squared statistic (Higgins and Thompson, 2002). The I-squared values of ≤42% for the drinking frequency and ≤24% for the drinking quantity describe the percentage of the variation in the SMDs attributable to heterogeneity rather than chance. Excluding the only trial that was conducted outside the USA reduced the heterogeneity to ≤25% and 0%, respectively.

Risk of publication bias

We investigated the risk of publication bias for the meta-analyses that included the strongest MI effects using funnel plots (Fig. 4). Egger's test is used to test for funnel plot asymmetry, which, among other things, can be caused by selective outcome reporting or publication bias. No evidence of funnel plot asymmetry was detected by Egger's test for the mean differences of the drinking frequency (P = 0.35) and quantity (P = 0.79) outcomes of the published USA and Brazil trials. For the US trials only, Egger's test indicates that publication bias may have affected the meta-analysis of the drinking frequency (P = 0.04), but the test detects no asymmetry in the mean differences of the drinking quantity (P = 0.25). Egger's tests for the mean differences of the weakest MI effects did not indicate publication bias (P > 0.37).

DISCUSSION

To our knowledge, this is the first systematic review and meta-analysis on the efficacy of emergency care triggered brief interventions that include MI to reduce risky drinking behaviour in young people. The included trials not only studied alcohol consumption, but also alcohol-related risks and problems. Our focus was on alcohol consumption as the outcome most comparable across trials. While four trials found MI more successful than control interventions in reducing alcohol-related consequences (Monti et al., 1999, 2007; Cunningham et al., 2009; Bernstein et al., 2010), there was less evidence for an additional impact of MI in emergency care brief interventions when it comes to alcohol use.

One study found no change in alcohol use after either intervention (Bernstein et al., 2010). Young people reduced their alcohol consumption in some way after an alcohol-related emergency care contact in five out of six RCTs, regardless of the content of the brief intervention (Monti et al., 1999, 2007; Spirito et al., 2004; Cunningham et al., 2009; Segatto et al., 2011). There is, however, debate that alcohol consumption may not be reduced as an effect of an intervention, but rather due to regression to the mean or behavioural change resulting from screening (see, e.g. Finney, 2008; Jenkins et al., 2009). The impact of such forces other than the intervention may be mitigated to some extent when differences between randomized groups are assessed. MI contributed to the reduction in alcohol consumption significantly in two studies (P < 0.05) (Spirito et al., 2004; Monti et al., 2007); of which one described a significant additional impact of MI only for the subgroup of more heavy drinkers (Spirito et al., 2004).

Investigating the raw data extracted from the original studies, drinking frequency and/or quantity increased again after an initial decrease in one or both intervention groups in some trials, but usually remained below the preintervention level by the end of study period (Monti et al., 1999, 2007; Spirito et al., 2004; Cunningham et al., 2009). In one of the trials, the control group's drinking frequency after 1 year even appears to have surpassed its baseline level (Spirito et al., 2004). No relapse in alcohol use, but the increase in drinking quantity between 3 and 12 months in the Spirito et al. (2004) trial (P < 0.04), was statistically significant although the phenomenon of reraising average consumption levels occurred frequently in the trials reviewed.

The meta-analysis suggests that MI-based interventions are at least as effective as control interventions and have potential to be more effective. We conducted separate meta-analyses for the correlated drinking frequency and drinking quantity outcomes, as well as for one more, and one less heterogeneous set of trials. Based on the largest effect size differences, the frequency of drinking was reduced significantly more by MI than other brief interventions, regardless of whether we studied US trials on their own or together with a trial from Brazil (SMD ≤ −0.17, P ≤ 0.03). Further meta-analyses indicated no significant SMD in the drinking quantity for all trials (SMD = −0.09, P = 0.18). Within the US trials only, a significantly lower amount of alcohol was consumed after a MI intervention (SMD = −0.12, P = 0.04). Meta-analyses based on the smallest differences in the effect size found no differences in the alcohol consumption after intervention between groups (SMD ≤ 0.02, P ≥ 0.38).

The Brazilian trial differed in several aspects from the US trials in the sample beyond its cultural setting: The experimental intervention was implemented solely by one researcher. The trial relied exclusively on self-report in evaluating the intervention effects, and with only one follow-up after 3 months it was the shortest trial in the sample. The sample was composed of ∼40% of alcohol dependents in each group. This portion could be higher than in the other trials studied because brief interventions usually target risk populations at an early stage of alcohol-related problems (Segatto et al., 2011). In addition, risks associated with alcohol abuse, which may correlate with the readiness to change alcohol consumption, were not balanced at baseline and no intention-to-treat analysis was performed. Despite these differences between trials, results from the meta-analyses with and without the Brazilian trial suggested that some additional reduction in alcohol consumption can potentially be achieved by MI.

Our review and meta-analysis are subject to limitations. The methodological quality of the studies included ranges from poor to good. The major concern with respect to the quality of the meta-analyses, which were based on unadjusted raw data, is the high numbers of dropouts of at least 6.6%, and up to 30.6% of participants. These dropout rates are likely to have introduced a substantial but unexamined bias. Publication bias was indicated for the selective study of drinking frequency outcomes from US trials with the strongest MI effects. Confirming an additional impact of MI on the drinking frequency in the meta-analysis of the USA and Brazil trials, for which a possible publication bias was not indicated, mitigates our concern that estimating a positive upper bound for an additional effect of MI in emergency care brief interventions may be a result caused by publication bias. Within the original studies, few included dropouts in their final assessment (Cunningham et al., 2009) or made a sensitive analysis to investigate the worst-case situation of equating loss to follow-up with a negative answer for change attempts, at least for alcohol-related consequences (Bernstein et al., 2010).

No reviewed study states whether MI was adapted to the special needs and demands of young people as discussed, for instance, by Gillian (1991). Combined interventions, like MI plus normative resetting and skills training or MI plus ‘booster’ phone calls, were tested in some studies (Monti et al., 2007; Cunningham et al., 2009; Bernstein et al., 2010), preventing us from deducing the efficacy of MI on its own. The methodological differences in the analyses and interventions limit the comparability of the study findings reviewed. Finally, we did not study alcohol-related problems as an outcome. However, it is often a history of alcohol-related problems that brings a patient to the attention of a medical provider, and the reduction of these problems rather than alcohol consumption itself might be the aim of an opportunistic brief intervention.

CONCLUSION

Only two of six trials in our systematic review found evidence that brief motivational interventions were more efficacious than other brief interventions in emergency care when it comes to the reduction of alcohol consumption, even though motivational interventions were better able to reduce alcohol-related risks or problems in four trials. In the best case scenario, our meta-analysis suggested that MI decreased alcohol consumption more than brief interventions that use, for instance, educational brochures, contact information for community resources or personal feedback. In the conservative scenario, the meta-analysis detected no differences between brief interventions. The variability in the additional benefits of using motivational techniques in brief emergency care interventions limits their effective use. Future research should therefore assess moderators of the effect size, investigate MI in emergency care settings of further countries and evaluate alcohol-related consequences after MI interventions in emergency care in a meta-analysis. Gaining a better understanding of what makes some brief interventions more effective than others is needed to successfully reduce heavy episodic drinking among young people.

SUPPLEMENTARY MATERIAL

Supplementary material is available at Alcohol and Alcoholism online.

AUTHORS' CONTRIBUTIONS

A.H. had the idea for the systematic review and conducted it. S.K. had the idea for the meta-analysis and conducted it. S.K. and A.H. performed the literature search, analysed the data, interpreted the results and wrote the manuscript.

Conflict of interest statement

None declared.

REFERENCES