Evidence reviews for the US Preventive Services Task Force (USPSTF) use an analytic framework to visually display the key questions that the review will address to allow the USPSTF to evaluate the effectiveness and safety of a preventive service. The questions are depicted by linkages that relate interventions and outcomes. A dashed line depicts a health outcome that follows an intermediate outcome. Refer to the USPSTF Procedure Manual for further details.7
Articles could appear in more than 1 key question (KQ). Reasons for exclusion: Aim: Study aim was not relevant. Setting: Study was not conducted in a country relevant to United States practice or not conducted in, recruited from, or feasible for primary care or a health system. Outcomes: Study did not have relevant outcomes or had incomplete outcomes. Population: Study was not conducted in an included population. Intervention: Intervention was out of scope. Comparator: Study did not have a comparison group. Screener: Study did not use an included screener. Design: Study did not use an included design. Quality: Study was poor quality. KQ indicates key question; USPSTF, United States Preventive Services Task Force.
NR indicates not reported.
a4+ drinks includes modified 3-Item Alcohol Use Disorders Identification Test (AUDIT-3; lower threshold for females and older adults) and the Substance Use Brief Screen (SUBS). 6+ drinks includes AUDIT-3. Quant × freq includes the first 2 items from the AUDIT (range, 0-8). Maximum drinks asks “During the last 12 months, what was the LARGEST number of drinks that you drank in a single day?” 6+, 5/4+, and 4+ drinks are variations of a screening test that quantifies the number of occasions per year on which a certain amount of drinks (4-6, depending on the test) were consumed in 1 day.
bScreened group: all participants.
cStudy enrolled male participants only.
fScreened group: participants aged 65 and older.
Weights are from random-effects analysis.
eMethods. Literature Search Strategies for Primary Literature
eTable 1. Inclusion and Exclusion Criteria
eTable 2. Quality Assessment Criteria
eTable 3. Summary Study Population Characteristics for Key Questions 4 and 5
eTable 4. Positive and Negative Predictive Values for a Range of Sensitivity and Specificity Based on US Prevalence of Unhealthy Alcohol Use
eTable 5. Study and Population Characteristics for Key Question 2, By Population
eFigure 1. Test Accuracy of the AUDIT-C at the Optimal Cutoff to Detect Unhealthy Alcohol Use
eFigure 2. Test Accuracy of the AUDIT-C at Cutoff of ≥3 to Detect Unhealthy Alcohol Use Among Females
eFigure 3. Test Accuracy of the AUDIT-C at Cutoff of ≥4 to Detect Unhealthy Alcohol Use Among Males
eFigure 4. Unhealthy Use, AUDIT, Cutoff ≥8, Adolescents, Young Adults, Adults, Older Adults
eFigure 5. Test Accuracy of the AUDIT to Detect the Full Spectrum of Unhealthy Alcohol Use or Alcohol Use Disorder, at Cutoffs of ≥3, 4, or 5, in US Primary Care
eFigure 6. Funnel Plot of Between-Group Difference in Change From Baseline in Drinks per Week by its Standard Error (Key Question 4a)
eFigure 7. Forest Plot of Odds Ratios for Exceeding Recommended Limits (Key Question 4a), Comparing Alcohol Counseling Interventions and Control Groups, by Population
eFigure 8. Forest Plot of Odds Ratios for Reporting a Heavy Use Episode (Key Question 4a), Comparing Alcohol Counseling Interventions and Control Groups, by Population
eFigure 9. Forest Plot of Odds Ratios for Reporting Abstinence During Pregnancy (Key Question 4a), Comparing Alcohol Counseling Interventions and Control Groups, Among Trials in Pregnant Women
eFigure 10. Forest Plot of Odds Ratios for Mortality (Key Question 4b), Comparing Alcohol Counseling Interventions and Control Groups, by Population
Customize your JAMA Network experience by selecting one or more topics from the list below.
O’Connor EA, Perdue LA, Senger CA, et al. Screening and Behavioral Counseling Interventions to Reduce Unhealthy Alcohol Use in Adolescents and Adults: Updated Evidence Report and Systematic Review for the US Preventive Services Task Force. JAMA. 2018;320(18):1910–1928. doi:10.1001/jama.2018.12086
Unhealthy alcohol use is common, increasing, and a leading cause of premature mortality.
To review literature on the effectiveness and harms of screening and counseling for unhealthy alcohol use to inform the US Preventive Services Task Force.
MEDLINE, PubMed, PsycINFO, and the Cochrane Central Register of Controlled Trials through October 12, 2017; literature surveillance through August 1, 2018.
Test accuracy studies and randomized clinical trials of screening and counseling to reduce unhealthy alcohol use.
Data Extraction and Synthesis
Independent critical appraisal and data abstraction by 2 reviewers. Counseling trials were pooled using random-effects meta-analyses.
Main Outcomes and Measures
Sensitivity, specificity, drinks per week, exceeding recommended limits, heavy use episodes, abstinence (for pregnant women), and other health, family, social, and legal outcomes.
One hundred thirteen studies (N = 314 466) were included. No studies examined benefits or harms of screening programs to reduce unhealthy alcohol use. For adolescents (10 studies [n = 171 363]), 1 study (n = 225) reported a sensitivity of 0.73 (95% CI, 0.60 to 0.83) and specificity of 0.81 (95% CI, 0.74 to 0.86) using the AUDIT-C (Alcohol Use Disorders Identification Test–Consumption) to detect the full spectrum of unhealthy alcohol use. For adults (35 studies [n = 114 182]), brief screening instruments commonly reported sensitivity and specificity between 0.70 and 0.85. Two trials of the effects of interventions to reduce unhealthy alcohol use in adolescents (n = 588) found mixed results: one reported a benefit in high-risk but not moderate-risk drinkers, and the other reported a statistically significant reduction in drinking frequency for boys but not girls; neither reported health or related outcomes. Across all populations (68 studies [n = 36 528]), counseling interventions were associated with a decrease in drinks per week (weighted mean difference, −1.6 [95% CI, −2.2 to −1.0]; 32 studies [37 effects; n = 15 974]), the proportion exceeding recommended drinking limits (odds ratio [OR], 0.60 [95% CI, 0.53 to 0.67]; 15 studies [16 effects; n = 9760]), and the proportion reporting a heavy use episode (OR, 0.67 [95% CI, 0.58 to 0.77]; 12 studies [14 effects; n = 8108]), and an increase in the proportion of pregnant women reporting abstinence (OR, 2.26 [95% CI, 1.43 to 3.56]; 5 studies [n = 796]) after 6 to 12 months. Health outcomes were sparsely reported and generally did not demonstrate group differences in effect. There was no evidence that these interventions could be harmful.
Conclusions and Relevance
Among adults, screening instruments feasible for use in primary care are available that can effectively identify people with unhealthy alcohol use, and counseling interventions in those who screen positive are associated with reductions in unhealthy alcohol use. There was no evidence that these interventions have unintended harmful effects.
Unhealthy alcohol use (including use that exceeds recommended limits, use that is having negative effects on health, or alcohol use disorder) was estimated to be the third leading preventable cause of mortality in the United States in 2000,1 with 9.8% of deaths attributable to alcohol consumption from 2006 to 2010.2 Unhealthy alcohol use is relatively common; in 2016 in the United States, 26% of adults and 4.9% of adolescents reported heavy use episodes (≥5 drinks on the same occasion on ≥1 day in the previous month, also referred to as binge episodes) and 6.6% of adults reported engaging in heavy drinking (≥5 drinks on the same occasion on ≥5 days) in the previous month.3 Alcohol use can exacerbate or cause a wide range of medical conditions commonly encountered in the primary care setting, including gastrointestinal, cardiopulmonary, dermatologic, reproductive, and neurologic conditions.4 Alcohol also interacts dangerously with many commonly used prescription and over-the-counter medications.5 Screening and counseling to reduce unhealthy alcohol use may prevent deleterious health effects and help prevent progression to more severe forms of unhealthy use.
In 2013, the US Preventive Services Task Force (USPSTF) recommended that clinicians screen adults 18 years or older for alcohol misuse and provide brief behavioral counseling interventions to those engaged in risky or hazardous drinking behaviors (B recommendation).6 The USPSTF concluded, however, that the evidence in adolescents was insufficient to evaluate the balance of benefits and harms of screening and behavioral counseling interventions to reduce alcohol misuse (I statement). This review was prepared to inform an updated recommendation by the USPSTF on the evidence related to screening test accuracy and benefits and harms of screening and counseling for unhealthy alcohol use in populations and settings relevant to US primary care.
An analytic framework was developed with 5 key questions (KQs) (Figure 1) that examined the benefits (KQ1) and harms (KQ3) of screening for unhealthy alcohol use, screening test accuracy (KQ2), and benefits (KQ4) and harms (KQ5) of counseling interventions for unhealthy alcohol use. A draft of the analytic framework, review questions, and inclusion and exclusion criteria was posted on the USPSTF website from August 25, 2016, to September 21, 2016, for the purpose of gathering public input. Detailed methods (eg, more detailed information about inclusion and quality rating criteria, methods for grading the strength of evidence for key questions, expert review, and public comment process) are available in the full evidence report at http://www.uspreventiveservicestaskforce.org/Page/Document/UpdateSummaryFinal/unhealthy-alcohol-use-in-adolescents-and-adults-screening-and-behavioral-counseling-interventions.
MEDLINE, PubMED (for publisher-supplied records only), PsycINFO, and the Cochrane Central Register of Controlled Trials were searched from January 1, 2011, to October 12, 2017, and supplemented by checking reference lists from the prior 2013 review and other relevant reviews, covering literature published since January 1, 1985. ClinicalTrials.gov was searched for ongoing trials. From October 12, 2017, through August 1, 2018, surveillance was conducted through article alerts and targeted searches of journals with a high impact factor and journals relevant to the topic to identify major studies that might affect the conclusions or understanding of the evidence and therefore the related USPSTF recommendation. The last surveillance, conducted on August 1, 2018, identified no new studies. However, 1 recently published diagnostic accuracy study was subsequently identified that met the inclusion criteria; that study did not change the conclusions and therefore is cited in the Discussion section only.
Two reviewers, applying a priori inclusion criteria, independently reviewed 17 149 unique citations and 570 full-text articles (Figure 2; eTable 1 in the Supplement). The review included English-language fair- and good-quality studies conducted among adolescents (12 years or older) or adults in countries categorized as “very high” on the United Nations Human Development Index.8 For benefits and harms of screening (KQ1 and KQ3) and interventions (KQ4 and KQ5), randomized clinical trials were included, as were nonrandomized controlled intervention studies with an eligible control group (eg, usual care, minimal intervention, attention control) that reported an alcohol use outcome. A minimum of 6 months of follow-up was required to assess intervention benefits (KQ1 and KQ4), but there was no minimum requirement for harms (KQ3 and KQ5). For screening test accuracy (KQ2), studies of test accuracy reporting sensitivity and specificity compared with a structured or semistructured clinical interview, or computer-based versions of structured assessments, were included.
For studies of benefits or harms of unhealthy alcohol screening (KQ1 and KQ3) and screening test accuracy (KQ2), studies that were restricted to participants with unhealthy alcohol use were excluded. For benefits or harms of unhealthy alcohol screening (KQ1 and KQ3), trials were sought that tested the effects of a screening program compared with usual care or a similar unscreened control group. Screening test accuracy (KQ2) evidence was limited to screening instruments named in national-level recommendations related to screening for unhealthy alcohol use or that had evidence to support their use based on the previous review (AUDIT [Alcohol Use Disorders Identification Test], AUDIT-C [AUDIT Consumption], SASQ [Single Alcohol Screening Question], and variations of these). Additionally, instruments were selected that target important subpopulations, ie, adolescents (National Institute on Alcohol Abuse and Alcoholism [NIAAA] 2-item screener, BSTAD [Brief Screener for Tobacco, Alcohol, and Other Drugs], and variations of these), pregnant women (TWEAK, T-ACE), or older adults (CARET [Comorbidity Alcohol Risk Evaluation Too]), or that cover both drug and alcohol use (ASSIST [Alcohol, Smoking, and Substance Involvement Screening Test]). For benefits or harms of unhealthy alcohol screening (KQ1 and KQ3) and of counseling interventions to reduce unhealthy alcohol use (KQ4 and KQ5), studies using any screening instrument were eligible.
For evaluating counseling interventions to reduce unhealthy alcohol use (KQ4 and KQ5), trials of behavioral counseling—with or without referral—were included if they were conducted in or recruited from primary care or a health care system or could feasibly be implemented in or referred from primary care. Since pharmacotherapy is primarily relevant to patients with moderate or severe alcohol use disorder (AUD), studies of pharmacotherapy treatment were excluded.
Trials were required to enroll participants through screening for unhealthy alcohol use for at least half of their sample. Screening had to take place in settings comparable or applicable to primary care with a defined population (eg, primary care clinic, Special Supplemental Nutrition Program for Women, Infants, and Children, college freshmen orientation). Trials that identified patients though behavioral or mental health clinics, substance abuse treatment centers, emergency department and trauma centers, work sites (including occupational screening), inpatient or residential facilities, or other institutions (eg, correctional facilities) were excluded. Studies of participants with alcohol dependence or severe AUD (or >50% of the enrolled sample having alcohol dependence or severe AUD) were excluded. Also excluded were studies limited to treatment-seeking individuals, those with concomitant psychotic disorders, those presenting in an emergency setting, and others not generalizable to primary care (eg, inpatients, those court-mandated to treatment, those who were incarcerated).
Included trials were critically appraised by 2 independent reviewers using criteria defined by the USPSTF and for test accuracy studies, supplemented with criteria from the Quality Assessment of Diagnostic Accuracy Studies 2 (QUADAS-2) (eTable 2 in the Supplement).7,9 Disagreements were resolved by a third reviewer.
Studies were rated as poor quality and excluded if there was an important limitation such as, among treatment trials, very high attrition (generally >40%); differential attrition between intervention groups (generally >20%); substantial lack of baseline comparability between groups without adjustment; or major concerns about the trial conduct, analysis, or reporting of results. For diagnostic accuracy studies, examples of important limitations warranting a “poor” quality rating included use of a reference standard that was not likely to categorize participants accurately, having the participant complete the screener after participating in an in-depth interview on his or her alcohol use, and/or lack of assurance that the study sample was representative of a relevant population. One reviewer abstracted descriptive and outcome data from fair- and good-quality studies into standardized evidence tables; a second checked for accuracy and completeness.
Summary tables of study, population, and intervention characteristics were created, along with forest plots of outcomes, to examine the size, consistency, and precision of effects. Studies were grouped according to population: adolescents (≈12 to 18 years), young adults (≈18 to 25 years), general adult populations (≈18 years or older), older adults (≈65 years or older), and pregnant and postpartum (up to 1 year after childbirth) women.
For the analysis of screening test accuracy, data were not pooled because of variability in cutoffs, populations, and screening tests. Contingency tables were used to calculate confidence intervals for sensitivity and specificity. If contingency tables were not reported, they were estimated using the reported sensitivity, specificity, and prevalence. Positive and negative predictive values were estimated based on the population prevalence of unhealthy alcohol use3 and 3 combinations of sensitivity and specificity. This article reports the test accuracy to screen for the full spectrum of unhealthy alcohol use (inclusive of exceeding limits and AUD). Test accuracy for other conditions (alcohol dependence, AUD, and exceeding limits) can be found in the full report.
For intervention effectiveness, meta-analysis was conducted for 4 alcohol use outcomes: drinks per week, drinking that exceeded recommended limits, heavy use episodes, and abstinence (for pregnant women). All related outcomes were converted to drinks per week, such as when provided with other time frames (eg, drinks per month) or with grams of ethanol rather than drinks. The conversion factor of 14 g of ethanol was used for 1 standard drink, since this is the definition of a standard drink in the United States. To determine whether meta-analyses were appropriate, clinical and methodological heterogeneity were assessed. In general, when at least 5 similar studies were available or when there were fewer studies but statistical heterogeneity was very low, quantitative synthesis was conducted and reported. Few health outcomes were reported in enough trials to consider pooling; however, a meta-analysis of mortality and alcohol problems or consequences was conducted.
Random-effects models were performed using the DerSimonian and Laird method to estimate pooled effects.10 For analyses that showed statistically significant pooled effects but that had fewer than 10 trials and I2 values larger than 50%, a sensitivity analysis was performed that used a more conservative pooling method to determine whether statistical significance was sustained (profile likelihood model or, if the profile likelihood model did not converge, a restricted maximum likelihood analysis with the Knapp-Hartung correction for small samples). For outcomes with 10 or more trials in the meta-analysis (drinks per week, exceeding recommended limits, and heavy use episodes), funnel plots were generated and the Egger test was used to examine funnel plot asymmetry to explore small-study effects, which can be related to publication bias.11 Additionally, for the outcome drinks per week, which was the most commonly reported outcome, meta-regression and subgroup analyses were conducted to explore factors associated with effect size.
Stata version 13.1 (StataCorp) was used for all analyses. All significance testing was 2-tailed, and results were considered statistically significant if the P value was .05 or less.
Two reviewers independently assessed 17 149 unique abstracts and 570 full-text articles for inclusion (Figure 2). One hundred thirteen studies (N = 314 466) were included. Overall, 0 studies were included for KQ1, 45 studies (56 articles) for KQ2, 0 studies for KQ3, 68 studies (103 articles) for KQ4, and 11 studies (12 articles) for KQ5.
Key Question 1a. Does primary care screening for unhealthy alcohol use in adolescents and adults, including pregnant women, reduce alcohol use or improve other risky behaviors?
No eligible studies were identified.
Key Question 1b. Does primary care screening for unhealthy alcohol use in adolescents and adults, including pregnant women, reduce morbidity or mortality or improve other health, social, or legal outcomes?
Key Question 2. What is the accuracy of commonly used instruments to screen for unhealthy alcohol use?
Forty-five studies12-56 were included (reported in 56 publications12-67) (Table 1) that addressed the accuracy of screening instruments: 10 in adolescents,20,21,25,34,36,38,39,43,50,51 5 in young adults,14,23,29,40,47 27 in general adult populations,12,15-18,22,24,26-28,30-33,35,37,41,42,44-46,48,49,52-54,56 1 in older adults,13 and 2 in pregnant19 or postpartum55 women. One study in a general adult population provided subgroup analyses of pregnant women and older adults,26,59 and 1 study of participants aged 12 to 20 years provided subgroup analyses of young adults (18 to 20 years).21 The majority of studies were conducted in the United States (28/45 [62%]) and recruited patients from primary care (23/45 [51%]) (Table 1). The number of study participants ranged from 95 to 166 165. A variety of 1- and 2-item screening tests were used in the included studies, as well as the AUDIT, AUDIT-C, and ASSIST. Reference standards used in the included studies were most commonly structured diagnostic interviews, and the interview sometimes was used in combination with other instruments (eg, Timeline Followback). Most studies were fair quality (28/45 [62%]).
For adolescents, just 1 study (n = 225) in a German high school reported on test accuracy for detecting the full spectrum of unhealthy alcohol use (eFigure 1 in the Supplement), finding a sensitivity of 0.73 (95% CI, 0.60 to 0.83) and specificity of 0.81 (95% CI, 0.74 to 0.86) for the optimal cutoff of 5 or higher on the AUDIT-C (male and female participants combined). The majority of the test accuracy evidence for adolescents was to detect AUD and is available in the full evidence report.
For adults, studies of the NIAAA-recommended single-item question (How many times in the past year have you had 5 or 4 [males or females, respectively] or more drinks in a day?) reported sensitivity ranging from 0.73 to 0.88 (95% CI range, 0.65 to 0.89) and specificity ranging from 0.74 to 1.0 (95% CI range, 0.69 to 1.0) for detecting the full spectrum of unhealthy alcohol use (4 studies [n = 44 461]) (Figure 3, labeled “5/4+ drinks”). All of these studies were conducted in the United States, primarily in primary care settings. Other 1- and 2-item screening tests (8 studies [n = 48 211]) generally showed sensitivities of 0.70 or higher, although the standard of 6 or more drinks per occasion tended to have lower sensitivity than the 5/4 or more drinks standard, often with nonoverlapping confidence intervals. Other adult populations (young adults, older adults, pregnant women) had results in similar ranges.
For the AUDIT-C, sensitivity for detecting the full spectrum of unhealthy alcohol use in adults was similar to the 1- and 2-item screeners, excluding 1 Veterans Affairs–based study in HIV-positive patients and matched controls45 that had substantially lower sensitivity. In most studies, the range of sensitivities was 0.73 to 0.97 for female participants (5 studies [n = 2714]; 95% CI range, 0.62 to 0.99) (eFigure 2 in the Supplement) and 0.82 to 1.0 for male participants (4 studies [n = 1038]; 95% CI range, 0.75 to 1.0) (eFigure 3 in the Supplement) at the standard score cutoffs of 3 or higher for female participants and 4 or higher for male participants, but the range of reported specificity was much wider (0.28 to 0.91 [95% CI range, 0.21 to 0.93] for female participants and 0.34 to 0.89 [95% CI range, 0.25 to 0.92] for male participants). Several studies reported sensitivities of 0.80 or higher at optimal cutoffs on the AUDIT-C, with associated specificities generally in the range of mid-0.70s to mid-0.80s (eFigure 1 in the Supplement). Evidence on the use of the AUDIT-C was very sparse in the adult subpopulations of younger adults, older adults, and pregnant women.
For the AUDIT, when using the recommended score cutoff of 8 or higher, studies (7 studies [n = 8852]) reported a wide range of sensitivity for detecting the full spectrum of unhealthy alcohol use in general adult populations (range, 0.38 to 0.73 [95% CI range, 0.33 to 0.84]) but high specificity (range, 0.89 to 0.97 [95% CI range, 0.84 to 0.98]) (eFigure 4 in Supplement). Sensitivity was relatively high (0.82) in young adults at the standard score cutoff of 8 or higher, but data were sparse in this population (2 studies [n = 660]). In many studies, sensitivity improved at lower cutoffs. Studies conducted in US primary care settings generally showed a more optimal balance of sensitivity and specificity at cutoffs of 3, 4, or 5 (3 trials [n = 2782]; sensitivity range, 0.64 to 0.86 [95% CI range, 0.57 to 0.91]; specificity range, 0.74 to 0.94 [95% CI range, 0.68 to 0.95]) (eFigure 5 in the Supplement).
At a sensitivity of 0.80 and a specificity of 0.90, the positive predictive value was estimated at 74% and the negative predictive value was estimated at 93% among adults with heavy use episodes in the past month (eTable 4 in the Supplement), using prevalence estimates for the US general population. Among population groups with lower prevalence of unhealthy alcohol use—older adults, pregnant women, and adolescents—the estimated positive predictive value was much lower, ranging from 26% to 46%.
Key Question 3. What are the harms of screening for unhealthy alcohol use in adolescents and adults, including pregnant women?
Key Question 4a. Do counseling interventions to reduce unhealthy alcohol use, with or without referral, reduce unhealthy alcohol use or improve other risky behaviors in screen-detected persons?
Sixty-eight trials70-137 (n = 36528) were included (reported in 100 publications70-169) that addressed the effect of a counseling intervention on alcohol use. Two of the trials targeted adolescents,92,109 22 targeted college-aged or young adults,71,75,79,83,87,96,98-101,103-105,107,108,111-113,125,129,133 29 addressed general adult populations,70,72-74,76,80-82,84,85,88,91,93-95,97,106,114,120,122,124,126-128,131,134,137 4 focused on older adults,86,90,110,136 and 11 targeted pregnant77,78,115,116,118,119,123,130,132 or postpartum89,117 women. Details of the included trials are reported in eTable 3 in the Supplement. Most trials were conducted in the United States (41/68 [60%]) and in primary care settings (42/68 [62%]). Trials were typically limited to participants who reported a prespecified level of alcohol use (most commonly either more than 7 [female participants] or 14 [male participants] drinks per week on average, or 4 [female participants] or 5 [male participants] or more drinks on a single occasion) or scored above a predetermined cutoff on a screening instrument such as the AUDIT. Ten trials were rated as good quality71,86,88,91,92,100,123,129,133,136 and the remaining were fair quality. Nineteen trials (28%) were included in the previous review.
Most interventions involved 1 to 2 sessions (90% involved 4 or fewer sessions), with a median of 30 minutes of contact time (88% involved 2 hours of contact or less) (Table 2). Almost all interventions involved at least basic education; general feedback, such as how the participant’s drinking compared with recommended limits; and suggestions about how to reduce alcohol use. Many, particularly those in primary care settings, used a SBIRT (screening, brief intervention, and referral to treatment) approach, consistent with approaches recommended by several health organizations. The most commonly reported intervention element was the use of personalized normative feedback sessions, in which participants were shown how their alcohol use compared with that of others; this technique was used in 62% of the included interventions and 89% of the interventions in younger adults. Motivational techniques were also common, particularly in combination with personalized normative feedback.
Most trials in adolescents and young adults involved 1 or 2 in-person or web-based personalized normative feedback sessions in school or university settings. Counseling interventions targeting adults other than college students (including pregnant and postpartum women) were more likely to take place in primary care settings, have multiple sessions, and involve the primary care team in some way; 33% of the interventions were delivered by the primary care clinician in trials of general and older adult populations.
Six trials (in 7 intervention groups) incorporated feedback on how an individual’s alcohol consumption was affecting his or her health, such as elevated liver enzyme levels, symptoms or medical conditions that could be exacerbated by alcohol use, and potentially dangerous alcohol use with prescribed medications.73,86,93,110,132,134
The most commonly reported alcohol use outcome was drinks per week, reported in 45 trials. On average, individuals in intervention groups reduced their drinking by 1.6 drinks per week more than those in control groups after 6 to 12 months (32 trials and 37 analysis groups [n = 15 974]; weighted mean difference [WMD] between groups in change from baseline, −1.6 [95% CI, −2.2 to −1.0]; I2 = 63%) (Figure 4, Table 3). This analysis included only 1 trial in adolescents, with separate entries for moderate- and high-risk users, and so is primarily reflective of adult unhealthy alcohol users. Baseline use levels were highly variable, with trial baseline means ranging from 3.8 to 59.3 drinks per week across all populations, and larger effects were typically seen with larger baseline use levels. The intervention group means changed from 20.5 drinks per week at baseline to 15.6 drinks per week at follow-up; control group means changed from 20.1 at baseline to 17.4 at follow-up. Excluding trials in adolescents and young adults, whose drinking patterns were generally typified by heavy use episodes rather than daily heavy drinking, the mean drinks per week in adult populations changed from 26.0 at baseline to 19.1 at follow-up in the intervention groups and from 25.6 at baseline to 21.6 in the control groups.
Trials that could not be included in the meta-analysis generally showed effects of a similar or slightly smaller size, favoring the intervention group (eg, between-group differences in change ranging from 0.9 to 1.8 drinks/wk, or posttest differences of 2.3 drinks/wk, or 10% to 20% relative reductions in use). The associations remained statistically significant when limited to trials conducted in primary care settings (21 trials [n = 7803]; WMD, −2.4 [95% CI, −3.4 to −1.3]; I2 = 70%), in the United States (18 trials [n = 8766]; WMD, −1.3 [95% CI, −1.9 to −0.6]; I2 = 64%), and in US-based primary care settings (9 trials [n=4989]; WMD, −1.8 [95% CI, −2.9 to −0.6]; I2 = 77%) (Figure 5). For trials with multiple follow-up assessments, effects were typically maintained between 6 and 12 months of follow-up; however, in several trials of young adults, group differences at 6 months’ follow-up were no longer statistically significant at 12 months’ follow-up.75,87,99,125 Seven trials70,88,90,95,102,107,114 reported follow-up at 24 months or beyond, and group differences were maintained in 4 of these through 24 months90,107,112 to 48 months.88
A small-studies effect was identified for drinks per week (Egger test bias coefficient, −1.04; P = .03) (eFigure 6 in the Supplement), and earlier publication date, younger population age (young adults vs other adults), and higher baseline drinking levels were also associated with larger effect sizes (Figure 5). These factors were not independent of each other, however, and it could not be determined which had a causal association with effect size. Smaller trials were more likely to have been published before 2007 and to have been conducted among heavier drinkers. Older trials were also primarily conducted among general adult populations in primary care settings, whereas many of the newer trials were conducted among young adults in college settings, with baseline use levels that were considerably lower than those in trials targeting general adult populations. Associations between effect size and intervention elements or other populations or study characteristics were generally not found.
The intervention was associated with a reduction in the odds of exceeding recommended drinking limits at 6 to 12 months of follow-up (15 trials [16 effects; n = 9760]; odds ratio [OR], 0.60 [95% CI, 0.53 to 0.67]; I2 = 24%) (Table 3; eFigure 7 in the Supplement), although this outcome was reported in only 24% (16/68) of the included studies. Between 15% and 76% of participants exceeded recommended drinking limits at follow-up in the intervention groups, compared with 29% to 82% in the control groups. Similarly, there was a reduction in the pooled odds of reporting an episode of heavy use (12 trials [14 effects; n = 8108]; OR, 0.67 [95% CI, 0.58 to 0.77]; I2 = 24% (Table 3; eFigure 8 in the Supplement), which was also relatively sparsely reported. Between 10% and 76% of intervention participants reported heavy use episodes at follow-up, compared with 13% to 92% in control groups. Small-studies effects were not detected for either of these outcomes. The 9 trials in pregnant women were most likely to report the odds of abstinence rather than the aforementioned outcomes; abstinence was higher in the intervention groups compared with the control groups (5 trials [n = 796]; pooled OR, 2.26 [95% CI, 1.43 to 3.56]; I2 = 0%) ( Table 3; eFigure 9 in the Supplement). Among trials reporting abstinence before delivery, abstinence ranged from 72% to 90% among intervention participants and from 55% to 74% among control participants. Other alcohol use outcomes were very sparsely reported and generally showed no statistically significant differences between groups.
Few changes in other behavioral outcomes such as drug use, sex after alcohol use, and seeking help for unhealthy alcohol use were noted, and those outcomes were only rarely reported. One trial82 in a general adult population found a reduction in self-reported drinking and driving, but 2 trials, in younger125 and older86 adults, did not. The latter trial in older adults also reported that participants reduced the likelihood of using alcohol in the face of symptoms or comorbidities that could be exacerbated by alcohol and with medication that could interact negatively with alcohol.86
Key Question 4b. Do counseling interventions to reduce unhealthy alcohol use, with or without referral, reduce morbidity or mortality or improve other health, social, or legal outcomes in screen-detected persons?
The most commonly reported health outcome was alcohol-related problems or consequences, measured using a variety of instruments. A pooled analysis showed a statistically significant, but very small, standardized mean difference in change between groups of −0.04 (18 trials [n = 9894]; 95% CI, −0.09 to −0.01; I2 = 3%). This effect size (Hedges g) can be interpreted as a Cohen d, where a small effect is typically considered to be 0.20 to 0.50.170 Mortality was reported in 8 trials, primarily as part of the description of the participant retention. The pooled association was not statistically significant (9 trials [n = 4533]; OR, 0.64 [95% CI, 0.34 to 1.19]; I2 = 0%) (eFigure 10 in the Supplement) and also may represent an overestimate of the true effect, since some trials that did not report deaths likely had no deaths, particularly trials among young adults. Trials were not powered for this outcome and many had very few events, resulting in imprecise results.
One trial, the Trial for Early Alcohol Treatment (TrEAT), described ascertainment methods.88 The effect on mortality at 4 years, 0.8% (3/392) of intervention participants dying compared with 1.8% (7/382) of control participants, was not statistically significant. Differences in mortality between groups were statistically significant at 3 years of follow-up, when there had been only 1 death among intervention participants but 7 among controls. This trial also reported statistically significant reductions in days of hospitalization (420 in the intervention group vs 664 in the control group) and controlled substance or liquor violations (2 in the intervention group vs 11 in the control group) at 4 years of follow-up. Other trials reported a wide variety of health outcomes, generally at 6 to 12 months of follow-up, with few findings of benefit for intervention over control groups.
Key Question 5. What are the harms of interventions to reduce unhealthy alcohol use in screen-detected persons?
Only 6 of the included trials (n = 3650) of counseling interventions to reduce unhealthy alcohol use reported on harms.72,103,105,113,116,136 In all cases, authors reported no harms in both groups. Further, no pattern of unexpected paradoxical increases in alcohol use was noted with these interventions.
The evidence in this review is summarized in Table 4. No evidence was found for screening programs to reduce unhealthy alcohol use or improve health, compared with usual care without screening. Multiple screening instruments are available that can detect unhealthy alcohol use with reasonable accuracy and that require 1 or 2 minutes to administer. For example, studies of adults found that the NIAAA-recommended single question had sensitivity ranging from 0.73 to 0.88 and specificity from 0.74 to 1.0 for detecting unhealthy alcohol use. For the AUDIT-C, sensitivity was similar, but the range of reported specificity was wider. For the full AUDIT, range of sensitivity was wide (0.38-0.73) using the recommended score cutoff of 8 or higher, but range of specificity was high (0.89-0.97). This pattern supports the use of a brief screener to identify excess use, followed by assessment with a more detailed instrument with greater specificity (eg, the AUDIT), as is currently done in some health care systems.171-173 If used as an initial screening test, data for the AUDIT from US primary care settings suggests that lower cutoffs (eg, 3, 4, or 5) may be preferable to provide a more optimal balance of sensitivity and specificity for detecting the full spectrum of unhealthy alcohol use. Given the relatively brief time required for follow-up questions after a positive screen to confirm the presence of unhealthy alcohol use and determine its extent (if present), clinicians may prioritize sensitivity over specificity for the initial screening and may consider calibrating the optimal cutoff for their setting.
One limitation of the evidence on the accuracy of screening instruments is that studies sometimes used variations of the standard instruments and cutpoints, and the gold standard was also heterogeneous across studies (eg, the definition of “exceeding recommended limits” varied across countries). This likely increased the variability in results but also supports the robustness of these tools, even with modifications. Use of the USAUDIT and USAUDIT-C, designed to use the United States’ standard drink size and to return results consistent with NIAAA recommendations, is likely to improve on the performance of the standard AUDIT and AUDIT-C.174 No studies on the USAUDIT or USAUDIT-C were published during the search window; however, a newly published study conducted among college students confirms that the performance characteristics of these instruments are improved over those of the standard AUDIT and AUDIT-C for determining whether someone exceeds the NIAAA-recommended drinking limits.175
Among adults identified through screening, counseling interventions to reduce unhealthy alcohol use were associated with reductions in alcohol use (by a mean of 1.6 drinks/wk) and in the odds of exceeding recommended drinking limits (by 40%) and heavy use episodes (by 33%) at 6 to 12 months of follow-up. Based on these findings, among adult unhealthy alcohol users, and assuming 33% of control group participants were drinking within recommended limits at follow-up (the median of the included trials), such interventions would result in an absolute increase of 14 percentage points in the likelihood of drinking within recommended limits, meaning 7 adults would need to be treated to achieve 1 drinking within recommended limits (number needed to treat [NNT], 7.2 [95% CI, 6.2 to 11.5]). Among pregnant women, counseling interventions were associated with an odds ratio of 2.26 for remaining abstinent from alcohol during pregnancy, for an NNT of 6.0 (95% CI, 4.3 to 12.5), assuming a baseline rate of 62% of women being abstinent from alcohol. Very limited data suggested that benefits from alcohol use counseling interventions can be maintained over 2 to 4 years.
Although many trials reported health, social, legal, and related outcomes, no specific outcomes were widely reported. Very limited information on harms of the included intervention was found, but the fact that most results favored the intervention groups across a wide range of outcomes, even though differences were not always statistically significant, suggests very low risk of harm. Several studies reported on the acceptability of their interventions to participants and generally reported positive to very positive ratings.79,97,116,117,130
Findings in the current review were generally consistent with the findings of the previous USPSTF review.176 For test accuracy, the previous reviewers concluded that a single-question screener, the AUDIT-C, and the AUDIT appeared to be the best overall instruments for screening adults for the full spectrum of unhealthy alcohol use in primary care, with ranges of sensitivities and specificities solidly in the range of the sensitivities and specificities seen in this review among studies of adults. In the current review, original studies were examined rather than existing systematic reviews, and at least 60% of the studies included in this review were newly published since the previous review. Among the newly included evidence are 10 studies in adolescents, who were not previously represented.
For counseling interventions, overall, the pooled effect size for drinks per week was larger in the previous review,176 although results were quite similar for general and older adult populations and for other drinking outcomes. One of the main differences between the 2 reviews is the inclusion of studies conducted outside of primary care settings in the current review, which resulted in the inclusion of a substantial number of studies in college settings. Consistent with the previous review was the finding of a fairly large but statistically nonsignificant association between interventions and reduced all-cause mortality (OR, 0.64 [95% CI, 0.34 to 1.19] in the current review; relative risk, 0.52 [95% CI, 0.22 to 1.22] in the previous review).
Areas for future research include direct comparisons of screening programs with usual care (without universal screening); further evaluations of the versions of the AUDIT and AUDIT-C recently developed for the United States (USAUDIT and USAUDIT-C); interventions to reduce unhealthy alcohol use in populations of adolescents, young adults, and older adults in health care settings; and exploration of more intensive intervention approaches with young adults. One important limitation of evidence on the benefits and harms of alcohol counseling interventions is the lack of a consistently reported group of outcomes. It would be beneficial for trials to routinely report outcomes with the greatest clinical meaning, such as the proportion of participants drinking within recommended limits, and to report health (including alcohol-related medical conditions), social, and legal outcomes. It would also be useful for trials to commit a priori to reporting subgroup effects in important subpopulations, such as by age group, sex, race/ethnicity, and baseline severity.
This evidence review has several limitations. First, comparative effectiveness trials—which have the potential to identify important features or mechanisms of change—were not included; however, other reviews that have included comparative effectiveness reviews have had very limited success in identifying mechanisms of change. Second, evidence regarding use of medication in treatment of AUD was not included. While this is primarily relevant to individuals being treated for more severe disorders rather than to most people with unhealthy alcohol use in screen-detected samples, medication would likely be appropriate for some patients identified through screening. A previous review found that multiple medications were associated with reductions in drinking and maintenance of abstinence for people with moderate to severe AUD, with NNTs from 12 to 20.177 Third, among adolescents, trials addressing prevention of unhealthy alcohol use were not included. This was outside the scope of the review but may be an important body of literature to consider when developing recommendations for adolescents.
Among adults, screening instruments feasible for use in primary care are available that can effectively identify people with unhealthy alcohol use, and counseling interventions in those who screen positive are associated with reductions in unhealthy alcohol use. There was no evidence that these interventions have unintended harmful effects.
Corresponding Author: Elizabeth A. O’Connor, PhD, Kaiser Permanente Research Affiliates Evidence-based Practice Center, Center for Health Research, Kaiser Permanente Northwest, 3800 N Interstate Ave, Portland, OR 97227 (firstname.lastname@example.org).
Accepted for Publication: July 27, 2018.
Author Contribution: Dr O’Connor had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: O’Connor, Perdue, Patnode, Jonas.
Acquisition, analysis, or interpretation of data: O’Connor, Perdue, Senger, Rushkin, Bean, Jonas.
Drafting of the manuscript: O’Connor, Rushkin, Bean, Jonas.
Critical revision of the manuscript for important intellectual content: O’Connor, Perdue, Senger, Patnode, Jonas.
Statistical analysis: O’Connor, Perdue, Jonas.
Administrative, technical, or material support: Perdue, Senger, Rushkin, Patnode, Bean, Jonas.
Conflict of Interest Disclosures: All authors have completed and submitted the ICMJE Form for Disclosure of Potential Conflicts of Interest and none were reported.
Funding/Support: This research was funded under contract HHSA-290-2015-000017-I-EPC5, Task Order No. 2 from the Agency for Healthcare Research and Quality (AHRQ), US Department of Health and Human Services, under a contract to support the USPSTF.
Role of the Funder/Sponsor: Investigators worked with USPSTF members and AHRQ staff to develop the scope, analytic framework, and key questions for this review. AHRQ had no role in study selection, quality assessment, or synthesis. AHRQ staff provided project oversight, reviewed the report to ensure that the analysis met methodological standards, and distributed the draft for peer review. Otherwise, AHRQ had no role in the conduct of the study; collection, management, analysis, and interpretation of the data; and preparation, review, or approval of the manuscript findings. The opinions expressed in this document are those of the authors and do not reflect the official position of AHRQ or the US Department of Health and Human Services.
Additional Contributions: We gratefully acknowledge the following individuals for their contributions to this project: Ernest Sullivent, MD, MPH, and Amanda Borsky, DrPH, MPP (AHRQ); current and former members of the US Preventive Services Task Force who contributed to topic deliberations; Richard Saitz, MD, MPH (Boston University Schools of Medicine and Public Health; Associate Editor, JAMA), for his content expertise and review of the draft report; Meghan Soulsby Weyrich, MPH (University of California Davis, Center for Healthcare Policy and Research), for her assistance in data abstraction; and Smyth Lai, MLS, and Katherine Essick, BS (Kaiser Permanente Center for Health Research), for technical and editorial assistance. USPSTF members did not receive financial compensation for their contributions.
Additional Information: A draft version of the full evidence report underwent external peer review from 5 content experts (Michael Fleming, MD, MPH, Northwestern University; John Higgins-Biddle, PhD, University of Connecticut; Alison Moore, MD, MPH, University of California San Diego; Emily E. Tanner-Smith, PhD, University of Oregon; Michael Weaver, MD, University of Texas) and 9 federal partner reviewers from the National Institute of Alcohol Abuse and Alcoholism, National Institute of Mental Health, Office of Research on Women’s Health, National Center of Birth Defects and Developmental Disabilities, and the National Institute of Dental and Craniofacial Research. Comments from reviewers were presented to the USPSTF during its deliberation of the evidence and were considered in preparing the final evidence review. Peer reviewers and those commenting on behalf of partner organizations did not receive financial compensation for their contributions.
Editorial Disclaimer: This evidence report is presented as a document in support of the accompanying USPSTF Recommendation Statement. It did not undergo additional peer review after submission to JAMA.