You Are Here:
Summary of the Evidence
The authors of this article are responsible for its contents, including any clinical or treatment recommendations. No statement in this article should be construed as an official position of the Agency for Healthcare Research and Quality or the U.S. Department of Health and Human Services.
Address correspondence to: Evelyn P. Whitlock, MD, MPH, Kaiser Permanente Center for Health Research, 3800 North Interstate Avenue, Portland, OR 97227-1110.
Select for copyright information. The USPSTF recommendations based on this review are online.
The summaries of the evidence briefly present evidence of effectiveness for preventive health services used in primary care clinical settings, including screening tests, counseling, and chemoprevention. They summarize the more detailed Systematic Evidence Reviews, which are used by the U.S. Preventive Services Task Force (USPSTF) to make recommendations.
Background: Primary health care visits offer opportunities to identify and intervene with risky or harmful drinkers to reduce alcohol consumption.
Purpose: To systematically review evidence for the efficacy of brief behavioral counseling interventions in primary care settings to reduce risky and harmful alcohol consumption.
Data Sources: Cochrane Database of Systematic Reviews, Database of Research Effectiveness (DARE), MEDLINE®, Cochrane Controlled Clinical Trails, PsycInfo, HealthSTAR, CINAHL databases, bibliographies of reviews and included trials from 1994 through April 2002; update search through February 2003.
Study Selection: An inclusive search strategy (alcohol* or drink*) identified English-language systematic reviews or trials of primary care interventions to reduce risky/harmful alcohol use. Twelve controlled trials with general adult patients met our quality and relevance inclusion criteria.
Data Extraction: Investigators abstracted study design and setting, participant characteristics, screening and assessment procedures, intervention components, alcohol consumption and other outcomes, and quality-related study details.
Data Synthesis: Six to 12 months after good-quality, brief, multi-contact behavioral counseling interventions (those with up to 15 minutes of initial contact and at least 1 followup), participants reduced the average number of drinks per week by 13% to 34% more than controls did, and the proportion of participants drinking at moderate or safe levels was 10% to 19% greater compared with controls. One study reported maintenance of improved drinking patterns for 48 months.
Conclusions: Behavioral counseling interventions for risky/harmful alcohol use among adult primary care patients could provide an effective component of a public health approach to reducing risky/harmful alcohol use. Future research should focus on implementation strategies to facilitate adoption of these practices into routine health care.
Alcohol misuse, including risky and harmful drinking, alcohol abuse, and dependence, is associated with numerous health and social problems and more than 100 000 deaths per year.1 Risky drinkers consume alcohol above recommended daily, weekly, or per-occasion amounts. Harmful drinkers experience harm associated with their alcohol use but do not meet criteria for alcohol abuse or dependence.2 Persons who misuse alcohol have elevated risks for a host of health problems,3-6 including violence-related trauma and injury.4 Most individuals who consume alcohol do so in moderation and without adverse consequences, however, and observational research suggests light or moderate use may be beneficial for some people.7-20
The assumption underlying brief behavioral counseling interventions in primary care is that, for identified risky or harmful drinkers, reducing overall alcohol consumption or adopting safer drinking patterns (that is, fewer drinks per occasion and not drinking before driving) will reduce the risk for medical, social, and psychological problems.21 Little experimental evidence supports this assumption, and most epidemiologic evidence relates health outcomes to existing drinking behaviors rather than to changes in drinking behaviors. Cross-sectional and cohort studies have consistently related high average alcohol consumption to short- or long-term health consequences.4,22 A meta-analysis of studies examining the association between all-cause mortality and average alcohol consumption found that men averaging at least 4 drinks per day and women averaging 2 or more drinks per day experienced significantly increased mortality relative to nondrinkers.23 Studies also relate heavy per-occasion alcohol use ("binge drinking") to acute injury risks and alcohol-related life problems.4,22 Injury rates are higher for binge drinkers who consume 5 or more drinks on one occasion as infrequently as 3 to 6 times per year, even when average intake is not excessive.24
In the United States, the National Institute on Alcohol Abuse and Alcoholism (NIAAA) has proposed epidemiologically based alcohol use guidelines to limit risks for short- and long-term drinking-related consequences by establishing age- and sex-specific recommended consumption thresholds.25 Maximum recommended consumption is 1 or less standard drink per day for adult women and for anyone older than 65 years of age, and 2 or fewer standard drinks per day for adult men. These guidelines do not apply to persons (such as adolescents, pregnant women, and persons with alcohol dependence or medical conditions or medication use) for whom alcohol intake is contraindicated, or to circumstances (driving) in which no consumption is considered safe.
Primary care clinicians commonly see patients with a range of alcohol-related risks and problems. In Wisconsin, about 20% of primary care patients were found to exceed NIAAA guidelines and to qualify as risky drinkers.26 Across multiple primary care populations, 4% to 29% are risky drinkers, 0.3% to 10% are harmful drinkers, and 2% to 9% exhibit alcohol dependence.27 Prevalence of these forms of alcohol misuse generally is higher in males and younger persons of all races and ethnicities.28
The NIAAA and others encourage physicians to identify patients with alcohol-related risks or problems and to provide office-based brief interventions or referrals as needed.25,29,30 In everyday practice, screening and screening-related assessment procedures are necessary to identify the range of alcohol users in order to offer appropriate treatment.31,32 Even so, few primary care clinicians use recommended screening protocols or offer treatment.33
To assist the U.S. Preventive Services Task Force (USPSTF) in updating its 1996 recommendation,34 the Oregon Evidence-based Practice Center systematically reviewed the evidence on primary care-based behavioral counseling interventions for risky/harmful alcohol use; systematic evidence reviews and meta-analyses since the last USPSTF report35-39 did not adequately address the key questions posed by the USPSTF. This review was exempted by the Institutional Review Board at Kaiser Permanente Northwest (FWA 00002344-IRB 00000405). Our review addressed the following questions:
- Do behavioral counseling interventions in primary care reduce risky or harmful alcohol use? What are elements of effective interventions? Do such interventions improve health outcomes?
- What methods were used to identify risky/harmful drinkers for behavioral counseling interventions in primary care?
- What adverse effects are associated with interventions addressing risky/harmful drinkers in primary care?
- What health care system influences are present in effective interventions for risky and harmful drinkers in primary care?
We concentrated our review on the program elements of brief primary care interventions for risky and harmful drinkers and their effects on alcohol use, health outcomes, and intermediate alcohol-related outcomes. Appendix Figure 1 shows the analytic framework and key questions guiding the entire systematic evidence review. Methods not described in this section appear in the Appendix, Appendix Figures 2 and 3, and Table 1.
No consistent definitions for the drinking patterns that should be the focus of primary care interventions are available from existing guidelines or research; however, it is commonly held that less severe alcohol problems are appropriate for brief interventions in primary care, whereas more severe problems need specialty addiction treatment.40 We adapted the following definitions from a recent systematic review of primary care screening for alcohol problems.2
Risky or hazardous drinkers are at risk from consumption that exceeds daily, weekly, or per-occasion thresholds (other terms further distinguish risky/harmful users who exceed longer-term thresholds—"high-average" or "heavy users"—from "heavy occasional" or "binge" drinkers, who exceed per-occasion thresholds).
Harmful drinkers experience physical, social, or psychological harm from their above-threshold alcohol use without meeting criteria for dependence.
Alcohol-abusing/-dependent drinkers continue to use alcohol despite significant negative physical, psychological, and social consequences,41 generally meet criteria for abuse or dependence as outlined in the Diagnostic and Statistical Manual of Mental Disorders, fourth edition,42 and are candidates for specialty addiction treatment.
Our review focuses on studies oriented toward the risky/hazardous/harmful category, which we refer to as "risky/harmful" drinkers. Fiellin and colleagues2 similarly divide the literature on screening instruments for alcohol problems into studies that focus primarily on risky, heavy, or harmful drinking and studies that focus on detecting alcohol abuse or dependence.
Among the brief intervention studies targeting risky/harmful drinkers selected for this review, we classified intervention groups into 1 of 3 levels of intensity:
- "Very brief interventions" had 1 session, up to 5 minutes long.
- "Brief interventions" had 1 session, up to 15 minutes long.
- "Brief multi-contact interventions" had an initial session up to 15 minutes long, plus followup contacts.
We used the definition of primary care recommended by the Institute of Medicine43 (see Inclusion and Exclusion Criteria in the Appendix) to identify relevant medical settings for our review.
Inclusion and Exclusion Criteria
We included English-language reports of randomized or nonrandomized controlled clinical trials of nondependent drinkers 12 years of age or older who received a primary care behavioral counseling intervention primarily to reduce alcohol intake. We excluded studies based in hospitals or emergency departments, specialty addiction treatment settings, behavioral health departments, and schools or community agencies without health clinics. We also excluded studies among comorbid patient populations because of limited generalizability to primary care. We excluded studies rated as having poor quality, as described below.
We identified 5 recent systematic reviews addressing primary care brief interventions to reduce risky/harmful alcohol use35-39 and 3 addressing screening2,44,45 from the Cochrane Database of Systematic Reviews and Database of Research Effectiveness (DARE). Relevant trials were identified from searches of MEDLINE®, Cochrane Controlled Clinical Trials, PsycINFO, HealthSTAR, and CINAHL databases (1994 to April 2002), reference lists of systematic reviews, the USPSTF 1996 recommendation,34 and experts. We conducted separate searches in MEDLINE® and PsycINFO from 1994 through April 2002 to identify any literature on harms related to alcohol screening, screening-related assessment, or intervention. None was found. The Appendix contains further search strategy details, along with information on our abstract and article review processes. We used USPSTF internal validity criteria46 (Table 1), supplemented by specific quality criteria addressing study randomization, attrition, and intention-to-treat analyses from the Cochrane Drug and Alcohol Group (CDAG)47 (Appendix Figure 3), to grade the quality of trials that met inclusion and exclusion criteria. We assigned each study's final quality rating according to investigator team consensus. Minimal to no attrition, non-differential attrition, and replacement of missing values in the outcome analyses were key features of trials rated good quality. Studies receiving a consensus rating of poor quality (n = 27) were excluded from the review (Appendix Table 2). Major quality problems included non-random assignment, non-comparable baseline conditions, attrition rates greater than 30%, and inadequate or unavailable alcohol consumption outcomes. Seventeen studies met final setting and quality criteria (although 1 did not have study results available in time for our review).48 Twelve of the 16 reviewed studies addressed non-pregnant adults and are the basis of this report. The others addressed pregnant women (n = 3) and adolescents (n = 1) and are reviewed elsewhere.40 A database search update through February 2003 revealed no new trials.
For all 12 included studies, 1 author abstracted relevant information using data-abstraction forms. The Appendix describes the data abstraction. A second author checked all data in the final evidence tables.
We examined intervention groups (n = 15) from included studies (n = 12) by levels of intensity and use of 5 key intervention components (feedback, advice, goal-setting, further assistance, and followup) identified from previous research.25,31,34,49 We recorded 3 commonly reported alcohol use outcomes that measured different but comparably important improvements in alcohol use at the end point nearest to 12 months followup:
- Mean drinks per week or the reduction in mean drinks per week (followup minus baseline).
- Percentage of participants without binge drinking (usually defined as ≥ 5 drinks per occasion).
- Percentage of participants achieving recommended drinking levels or patterns (as defined by the study).
Where possible, we converted alcohol outcomes into consistent measures across studies and conveyed final results as "net" (that is, intervention minus control); the Appendix further describes our calculations. We did not undertake a quantitative synthesis of alcohol outcomes because of the lack of a clearly superior measure among the 3 alcohol use outcomes available and because of our judgment, supported by that of the USPSTF, that a qualitative synthesis that includes all outcomes would be most informative. Graphs displaying trial results by alcohol use outcome, with sex subgroups (where available), can be accessed elsewhere.40
Role of the Funding Source
This research was funded by the Agency for Healthcare Research and Quality (AHRQ) under a contract to support the work of the USPSTF. The USPSTF members participated in the initial design and reviewed interim results and the final evidence review. AHRQ had no role in study design, data collection, or synthesis, although AHRQ staff reviewed interim and final evidence reports and distributed the initial evidence report for external content review by 11 outside experts, including representatives of professional societies and federal agencies. The subsequently revised systematic evidence review on which this manuscript is based is available at http://www.ncbi.nlm.nih.gov/books/bv.fcgi?rid=hstat3.chapter.45217.40
Characteristics of Behavioral Counseling Intervention Trials Reviewed
Table 2 and Appendix Table 3 detail the 12 trials of primary care interventions for risky/harmful alcohol use. Seven trials50-56 were judged good quality, and the rest were fair.57-61 All were randomized, controlled trials conducted in multiple primary care practices (ranging from 3 to 47 practices per study), except for 1 controlled clinical trial.61 All but 3 trials51,54,59 involved more than 300 participants. The studies examined drinking outcomes after at least 12 months of followup, except for 1 with 6-month results52 and 1 with at least 9 months of followup.58
About one third of study participants were women; the exceptions were some older international studies that did not target women.54,57,58 Adults 65 years of age or older were included in 9 trials50,52,54-56,58-61 and were specifically targeted in another.51 Rates of participation of nonwhite persons were not reported in many older international studies and were low (4% to 27%) where reported in recent U.S. studies.50,52,53,56
The trials generally targeted risky or harmful drinkers or both and excluded known or suspected dependent drinkers, using variable criteria. However, more recent studies (those published after 1996) were more likely to include binge drinkers in addition to persons with high average consumption. These studies tended to define lower thresholds for risky weekly or average use and often excluded heavier drinkers who were at a lower threshold of average use or had any evidence of dependence or abuse. Generally, thresholds for risky alcohol consumption were lower for women than men. More details on inclusion and exclusion criteria applied within each trial are available in Appendix Table 3 and in the Systematic Evidence Review.40
On the basis of our definitions, 2 studies evaluated very brief interventions,58,61 6 evaluated brief interventions,54,56-60 and 7 evaluated brief multi-contact interventions.50-53,55,57,60 Twelve of the 15 interventions were delivered all or in part by the patient's usual primary care physician. Four of these used physicians to deliver initial and repeated intervention contacts,52,55,59,61 whereas others used health educators and counselors50,56 or clinic nurses51,53 for some contacts.
Effectiveness of Behavioral Counseling Interventions on Risky/Harmful Alcohol Use
All 7 trials testing brief multi-contact behavioral counseling interventions50-53,55,57,60 reported mean drinks per week or average daily consumption outcomes. Five studies50-53,55 reported the proportion of participants with safe or moderate alcohol use. Four studies reported the proportion of participants not bingeing.50-53 Six of these trials50-53,55,57 reported a significant effect on at least 1 drinking outcome (Table 2). The seventh fair-quality study, delivered entirely by research personnel outside the clinical setting, found no significant effect on mean drinks per week, the only outcome measure it reported.60 Four good-quality trials51-53,55 reported that weekly drinking was reduced 13% to 34% more in intervention groups than in controls (that is, 13% to 34% net reduction), resulting in 2.9 to 8.7 fewer mean drinks per week at followup in intervention compared with control participants (data shown elsewhere).40 One fair-quality brief multi-contact intervention significantly reduced mean daily alcohol consumption,57 while 1 good-quality trial did not significantly change average use.50 All 5 good-quality trials50-53,55 found significant effects on recommended or safe alcohol use, resulting in 10% to 19% more intervention participants than controls reporting recommended or safe drinking patterns (data shown elsewhere).40 Two of 4 good-quality trials reported significantly reduced binge drinking.51,53 In trials with at least 49% binge users in the study sample at baseline,51-53 binge drinking remained fairly common (31% to 69%) among intervention participants after intervention.
Of the 8 trials testing very brief interventions58,61 or brief interventions,54,56-60 all reported mean drinks per week or average daily consumption outcomes. Six intervention groups from 5 studies54,56,58,59,61 reported the proportion of participants with safe or moderate alcohol use; 3 reported the proportion not bingeing.54,56,59 Statistically significant results were limited to 3 studies,54,57,58 although results tended to favor intervention groups over control groups. One fair-quality very brief intervention58 improved daily alcohol intake and the proportion of participants drinking moderately among males only. This result may have been due to limited power given the relatively small number of females in the study, or the very brief intervention could have been contaminated—interventionists also delivered a brief intervention protocol (which similarly improved outcomes in males) as part of the same study. A trial testing both brief and brief, multi-contact interventions found an average intake effect for both, although potential for contamination was not clear.57 A good-quality brief intervention targeting males significantly improved the proportion with safe or moderate use and the proportion not bingeing.54
All interventions that showed statistically significant improvements in alcohol outcomes of any intensity included at least 2 of 3 key elements—feedback, advice, and goal-setting. Since most effective interventions were multi-contact ones, they also provided further assistance and followup. A few also reported tailoring intervention elements to each participant.50,52
We found no consistent differences between women and men in the effectiveness of interventions, particularly brief multi-contact interventions (data displayed and discussed in detail elsewhere).40 One intervention that targeted older adults51 appeared as effective as or more effective than a similar intervention in younger adults.53
Effectiveness of Behavioral Counseling Interventions on Health and Related Outcomes
About half of intervention studies reported morbidity-related outcomes, such as problem scores,54,58,59,61 psychological scores,54,59 and lifestyle improvements or reduced accidents and injuries.51,53,54 In 2 of the 4 studies examining problem scores, those in all groups generally improved, with no differences between intervention and control groups at followup.54,61 The other 2 studies showed no changes from baseline to followup within or between groups.58,59 With other outcomes, studies generally found no improvement or similar improvements in interventions and controls over the duration of the trials.51,53,54,59 Of the 5 trials that examined health care utilization,53,54,56,59 only 1 found reduced self-reported hospital days at 12 months.53 In a study evaluating brief interventions and brief, multi-contact interventions,60 quality-of-life measures, including those related to alcohol-related problems, improved among the subset of intervention and control participants who reduced drinking by at least 20%.62
We identified 4 reports of long-term health outcomes following 3 intervention trials.63-66 In 1 good-quality brief multi-contact intervention trial,53 fewer hospital days were self-reported by the intervention group than controls after 48 months (429 vs. 664 days; P < 0.05), and there was a trend toward reduced all-cause mortality in intervention participants compared with controls (3 vs. 7 deaths; P > 0.10).64 However, other morbidity-related outcomes did not significantly differ between groups. Significantly greater reductions in alcohol use among intervention participants than controls were maintained at 48 months.
In a second study, a brief single-contact intervention had no long-term effects on morbidity, mortality, or alcohol consumption at 10-year followup.66
The third study,65 an intensive population-based intervention that alternately enrolled annual cohorts in screening and non-screening study groups over many years, reported health outcomes but not alcohol consumption outcomes (the Malmö Screening and Intervention Study). Men age 32 to 37 years who were invited to participate had significantly lower total mortality rates (24/100,000 person-years) than non-invited controls (30/100,000 person-years) (P < 0.02), and had significantly reduced alcohol-related mortality after 3 to 21 years.65 In a nested, randomized, controlled trial within the Malmö Study, men age 45 to 49 years with elevated serum gamma-glutamyltransferase levels who were randomly assigned to control groups had more alcohol-related deaths after a median of 13 years (relative risk, 2.0 [95% CI, 1.1 to 3.7]; P = 0.026) than those assigned to intensive intervention.63 Since this trial did not report alcohol use outcomes and it selected drinkers on the basis of confirmed elevations in serum gamma-glutamyltransferase levels, participants may have been more severely affected than in other studies we reviewed.
Methods Used To Identify Risky and Harmful Alcohol Users
In the 12 trials reviewed, methods to identify alcohol users appropriate for brief interventions in primary care (Table 2, Appendix Table 3, and Table 11 from the Systematic Evidence Review40) typically included screening (identifying patients with probable risky/harmful alcohol use) and screening-related assessment (confirming screening results and distinguishing patients suitable for brief interventions from those needing specialty care referral). Screening typically involved self-administered questionnaires or brief interviews to assess average quantity or frequency and binge use.
In recent U.S. studies,50-53,56,60 about 8% to 18% of patients screened "positive," and about half of these remained eligible for primary care intervention after assessment (data shown elsewhere40). Processes to identify patients were generally embedded, at least initially, within assessment of other behavioral health risks. Screening and assessment steps included an added respondent burden for research; however, this burden applied equally to intervention and control participants in all but 1 study.57 Many of the trials we reviewed used validated screening instruments (CAGE [4-item screening questionnaire to detect alcoholism], AUDIT [alcohol use disorders identification test—10-item instrument for risky/harmful use]) that have been shown to have reasonable-to-good test performance among primary care populations.2,44,45 Test performance is summarized elsewhere.40 Validated instruments were used alone (for example, AUDIT) or in combination (CAGE plus standardized questions on quantity and frequency) to detect patients with at-risk or harmful drinking, or alcohol abuse or dependence. Research personnel generally provided all or most of the screening and assessment for participants. Screening and assessment steps for each study, and their yields, are examined in greater detail elsewhere.40
Adverse Effects of Screening and Intervention
We found no research that addressed adverse effects associated with alcohol use screening or assessment, or with behavioral counseling interventions for alcohol use. Three good-quality intervention trials reported greater dropout rates among participants receiving alcohol interventions than among controls,50,55,56 while 1 good-quality trial reported higher dropout among controls.54 Differential dropout rates did not affect outcomes since they were addressed analytically; however, dropout may indicate discomfort or dissatisfaction with the intervention, among other plausible explanations. These findings occurred in a minority of trials and cannot be explained with the available data.
Health Care System Supports and Influences
In all 12 trials, additional staff or systems support were required to provide screening and assessment services and, in some cases, intervention support. To identify potential study participants for screening and assessment, 2 studies used systems that highlight upcoming appointments,50,52 while others used practice registries.54,55,59 In nearly every study, research staff conducted the screening and assessment outside the routine care encounter. Most of these processes took more than 30 minutes, although time estimates also include research-related procedures.
Provider training sessions, reported in many studies,50,52-54,58,59,61 ranged from 15 minutes to 2.5 hours. Several recent studies reported both initial and ongoing training.52,53 Only 3 studies reported using incentives for participating providers or patients.51,53,60 Besides usual care physicians, studies also used research staff50,56,58,60 or non-physician health care staff51,53 to deliver some or all of the intervention. Research staff often performed important support functions, such as prompting the provider and supplying intervention materials to the chart.50,52,56,60 None reported using electronic medical record support.