Skip to main content
PMC home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2023 Jul 1.
Published in final edited form as: Eur Child Adolesc Psychiatry. 2021 Mar 16;31(7):1–11. doi: 10.1007/s00787-020-01712-3

A RISK ALGORITHM THAT PREDICTS ALCOHOL USE DISORDERS AMONG COLLEGE STUDENTS

CORINA J BENJET a, PHILIPPE MORTIER b, GLENN KIEKENS c, DAVID D EBERT d, RANDY P AUERBACH e, RONALD C KESSLER f, PIM CUIJPERS g, JENNIFER G GREEN h, MATTHEW K NOCK i, KOEN DEMYTTENAERE j, YESICA ALBOR k, RONNY BRUFFAERTS l
PMCID: PMC9336831  NIHMSID: NIHMS1824117  PMID: 33723648

Abstract

Aim:

To assess the one-year incidence of alcohol use disorders (AUD) among incoming first year students, predictors of AUD-incidence, prediction accuracy and population impact.

Design:

Prospective cohort study.

Participants:

First-year college students (baseline: N=5,843; response rate=51.8%; 1-year follow-up: n=1,959; conditional response rate=41.6%).

Setting:

KU Leuven University, Belgium

Measurements:

AUD were evaluated with the AUDIT and baseline predictors with the Composite International Diagnostic Interview Screening Scales (CIDI-SC), adapted for the World Mental Health International College Student Survey initiative (WMH-ICS).

Findings:

The incidence of AUD at follow-up was 3.9% (SE=0.4). The most important individual-level baseline predictors of AUD incidence were being male (OR=1.53; 95% CI = 1.12 – 2.10), a break-up with a romantic partner in the past year (OR=1.67; 95% CI =) 1.08 – 2.59, hazardous drinking (OR=3.36; 95% CI = 1.31 – 8.63), and alcohol use characteristics at baseline (ORs between 1.29 and 1.38). Multivariate cross-validated prediction (cross-validated AUC=0.887) show that 55.5% of incident AUD cases occurred among the 10% of students at highest predicted risk (20.1% predicted incidence in this highest-risk subgroup). Four out of five students with incident AUD would hypothetically be preventable if baseline hazardous drinking were to be eliminated along with a reduction of one standard deviation in alcohol use characteristics scores, and another 15.0% would potentially be preventable if all 12-month stressful events were eliminated.

CONCLUSIONS:

Screening at college entrance is a promising strategy to identify students at risk of AUD onset, which may improve the development and deployment of targeted preventive interventions.

Keywords: alcohol use disorder, hazardous drinking, university students, risk algorithm, incidence

INTRODUCTION

The college years are a developmentally crucial period when students make the transition from late adolescence to emerging adulthood [1]. Apart from personal, social, and intellectual challenges and achievements, the college years are also a peak period for the prevalence of mental disorders [2–3], with around one third of incoming college students meeting criteria for a 12-month mental disorder [4–5]. The prevalence of AUD among college students is lower than the estimates for some other mental disorders, with 12-month prevalence in students across 21 countries around 5% [6]. However, the college years may carry especially high risk for onset of alcohol use disorders (AUD – either abuse or dependence) as previous research suggests that up to 70% of AUD among college students has its onset during and not prior to college entrance [6]. AUD in college is associated with deleterious psychological, social, and physical health consequences [7], including violence [8], accidents and injuries [9], and risky sexual behaviors [10]. Additionally, over 90% of students with AUD do not perceive their symptoms to be a problem [11]. From a public health perspective, early identification of students that will make the transition from non-problematic alcohol use to a more severe level of alcohol consumption would facilitate effective deployment of targeted preventive interventions during college and thereby reduce the incidence, prevalence, severity, duration, and consequences of future AUD as well as of mental disorders that are influenced by AUD [12]. To guide allocation of resources and clinical decision-making, colleges need tools that accurately identify students at high risk of developing AUD. Although there are many studies that estimate the prevalence of AUD in college, studies on the incidence of AUD, and predictors of incidence, among college students are scant. Using longitudinal data from the National Epidemiologic Survey on Alcohol and Related Conditions (NESARC), the incidence of AUD in the general population is estimated to be 1.45% per year, and approximately 4% among 20-29 year olds, but they did not assess incidence specifically for college students [13].

Given the high availability of internet access and geographic proximity to centralized student services, college campuses may be ideally situated to access large groups of young adults for screening and referral to adequate care [14]. Web-based screening provides a practical alternative for students with drinking problems who may be less likely to seek clinical services [15], and further, it may offer personalized feedback and access to online self-help interventions [16–17]. Despite these potential advantages, it is currently unknown how many first-year college students effectively make the transition from non-problematic use of alcohol to a more problematic use of alcohol or to AUD, and how accurately these screening tools can identify the high risk students that will make the transition, without identifying too many false-positive cases (a concern raised for the screening of suicidal behaviors which limits the feasibility of screening [18]) that would put undue demands on college mental health centers). The development of powerful risk screening algorithms may remediate this.

The present study addresses these shortcomings by examining the first onset of AUD during the college years in a large, longitudinal survey of college students (Leuven College Surveys – see: www.mindmates.be/page.phpid28), a part of the WHO World Mental Health Surveys International College Student initiative (WMH-ICS initiative, see: http://www.hcp.med.harvard.edu/wmh/college_student_survey.php). We build on earlier work on the development of concentration-of-risk models that estimated and accurately predicted incidence or persistence of mental disorders and self-injurious thoughts and behaviors [19–21]. Consistent with recommendations to develop such risk algorithms to target high-risk individuals for preventive interventions [22–23], we examine the strength of multivariate associations in our model of baseline predictors to determine whether a well-defined subset of students at highest risk of incidence of AUD can be detected.

METHOD

Procedures

Full procedures of the LCS have been reported previously [5, 21]. Briefly, the LCS consists of a series of web-based self-report surveys of KU Leuven students. In the academic years 2014-2016, all 13,103 Dutch-speaking incoming freshmen aged 18 years or older were eligible for the baseline survey. A total of 5,844 students completed the baseline survey (51.8% response rate after adjusting for potential non-participation due to college attrition). Students were contacted for the follow-up survey 12 months after the baseline assessment. A total of 1,959 of the original baseline respondents responded to the follow-up survey (corresponding to a 41.6% conditional response rate after adjusting for non-participation due to college attrition). Informed consent was obtained from all students. Participants with 12-month suicidal behaviors or non-suicidal self-injury received links to local mental health resources. The study’s protocol was approved by the University Hospital Leuven Biomedical Ethical Board.

Measures

Socio-demographic variables.

The university’s students’ administration office provided socio-demographic characteristics, including gender, age, nationality, parents’ financial situation, parents’ education, parental familial composition, university group membership, student situation (full-time student versus other) and type of secondary school education.

Alcohol use disorder was assessed using the Alcohol Use Disorders Identification Test (AUDIT) [24]. The AUDIT consists of a total score (range 0-40) and allows for the calculation of three subscales: the consumption subscale (consisting of three items assessing the frequency and quantity of alcohol use), the dependence subscale (consisting of three items assessing perceived control over drinking, failure to comply to normal expectations due to drinking, and withdrawal symptoms), and the alcohol-related problems subscale (consisting of four items measuring guilt or remorse after drinking, memory lapses after drinking, alcohol-related injuries, and concerns of family, friends or professionals regarding one’s drinking). In line with prior recommendations [25], we defined AUD (alcohol abuse or dependence) as either a total AUDIT score of 16+ or a total AUDIT score of 8-15 with a score of 4+ on the AUDIT dependence subscale. Those students who did not meet the criteria for AUD were divided into either a hazardous drinking group when they had a total AUDIT score of 8-15 with a score of 0-3 on the AUDIT dependence subscale, or a no hazardous drinking group as having a total AUDIT score of 0-7. This version of AUDIT scoring has concordance with clinical diagnosis in the range AUC = 0.78–0.91 [26].

Traumatic experiences in childhood-adolescence (i.e. prior to the age of 17) were assessed using 19 items adapted from the Composite International Diagnostic Interview (CIDI 3.0) childhood section [27], the Adverse Childhood Experience Scale [28], and the Bully Survey [29]. Items assessed parental psychopathology (i.e., any serious mental or emotional problems, substance use problems, suicidal behaviors or death by suicide, criminal activities, or interpersonal violence), physical abuse, emotional abuse, sexual abuse, neglect, bully victimization (i.e., either direct verbal or physical bullying, as well as indirect bullying [e.g., spreading rumors], or cyberbullying), and dating violence. Response options consisted of five-point Likert items (“never”, “rarely”, “sometimes”, “often”, and “very often”). To obtain dichotomously coded variables (i.e., potential risk factors), cut-off values consisted of “rarely” for all items, except bully victimization which had a cut-off of “sometimes”, in line with a previous recommendation [30].

Stressful events experienced in the 12-months before the baseline survey were assessed using items from well-validated screeners [31–33], and included relevant stressful experiences among young adults, including life-threatening illness or injury of a family member or close friend [34], accidents or death of a family member or close friend [35], interpersonal events (e.g., break-up with a romantic partner, serious betrayal by someone other than one’s partner) [36], and other stressful experiences (e.g., physical or sexual assault, and legal problems such as time spent in jail [37–39].

Mental disorders were assessed using the Composite International Diagnostic Interview Screening Scales (CIDI-SC) [40] for major depressive episode, mania/hypomania (broad mania), generalized anxiety disorder (GAD), panic attacks, and drug use disorder (abuse or dependence either on cannabis, cocaine, or any other street drug, or on a prescription drug either used without a prescription or used more than prescribed to get high, buzzed, or numbed out). The CIDI-SC scales have concordance with blinded clinical diagnoses in the range AUC = 0.70–0.78. Items from the Self-Injurious Thoughts and Behaviors Interview (SITBI – see [41]) assessed 12-month non-suicidal self-injury, suicidal ideation, suicide plans, suicide attempts, and non-suicidal self-injury. We also assessed risk for other mental disorders or symptoms, including lifetime intermittent explosive disorder symptoms, lifetime post-traumatic stress disorder symptoms, and lifetime eating disorder symptoms (using MINI items – see [42]).

Analyses

Nonresponse propensity weights [43] were created to adjust for potential non-response bias. Multiple imputation by chained equations [44] was used to adjust for survey attrition and within-survey item nonresponse. One case was eliminated for analysis due to missing information on auxiliary variables necessary for calculating non-response weights, resulting in a final sample for analysis of n=5,843. Logistic regression analysis examined the strength of individual-level associations (i.e., odds ratios [OR]) between baseline predictor variables and 12-month hazardous drinking or AUD at 12-month follow-up. Two series of models were constructed. A first series predicted AUD at 12-month follow-up among those 5,590 students without AUD at baseline; a second series predicted hazardous drinking or AUD at 12-month follow-up among those 4,381 students without hazardous drinking or AUD at baseline. Baseline predictor blocks in the multivariate models included the three AUDIT subscales (including a dummy variable for baseline hazardous drinking in the first series of models), sociodemographic variables, (number of) traumatic experiences in childhood-adolescence, (number of) stressful events experienced in the past 12-months, (number of) lifetime and 12-month mental disorders, and 12-month self-injurious thoughts and behaviors. Individual-level predicted probabilities based on the multivariate equations were created, and area under the curve (AUC) values calculated. The multivariate model with the highest AUC was selected for further evaluation of predictive accuracy. Predicted probabilities were discretized into deciles and cross-classified with observed cases to visualize the concentration of risk associated with high composite predicted probabilities. Sensitivity was defined as the proportion of cases found among pre-defined proportions (e.g., 10%) of respondents with highest predicted probabilities. Positive Predictive Value (PPV) was defined as the probability of effectively developing the outcome when being among pre-defined proportions (e.g., 10%) of respondents with highest predicted probabilities. We used the method of leave-one-out cross-validation [45] to correct for the over-estimation of prediction accuracy when both estimating and evaluating model fit in a single sample. Using summary measures of predicted probabilities calculated using coefficients from the final model, we estimated Potential Impact Fractions (PIF), representing the proportion of outcome cases potentially reduced after a change in the exposure of a related ordinal categorical predictor [46]. We use PIFs instead of the more commonly used Population Attributable Risk Proportion because PIFs are indicated in data where the lowest exposure of a risk factor (i.e. the use of alcohol) is non-zero.

RESULTS

Description of the sample

The majority of the sample (n=5,843) was female (57.0%), only few participants (4.3%) were of non-Belgian nationality and 17.2% of the students indicated that they were raised in households with a difficult financial situation. For most students (62.0%) both parents had a college education, only a small proportion of students (15.4%) indicated that neither of their parents had a college education. More than half of all students met criteria for at least one of the three lifetime or five 12-month disorders (57.7%), and 25.3% reported exactly one, 15.0% exactly two, 8.7% exactly three, and 8.7% four or more mental disorders. More than half of the sample (58.4%) reported at least one traumatic experience prior to the age of 17, with 34.4% experiencing parental psychopathology as the most reported one, followed by bully victimization (32.4%). Every second student (57.6%) also reported at least one stressful life event in the past year, with the experience of life-threatening illness or injury of a close friend or family member most frequently reported (i.e. 20.6%).

Incidence of AUD

Prevalence of hazardous drinking and AUD at baseline and follow-up are shown in table 1. Three findings stand out. First, 12-month prevalence of hazardous drinking and AUD at baseline was 21.6% and 4.7%, respectively. Second, persistence of hazardous drinking/AUD among college students (i.e. the proportion of those who meet criteria for hazardous drinking/AUD both at baseline and follow-up) was 60.4% and 51.5%, respectively. Third, the incidence of AUD among college students is estimated at 3.9% (SE=0.4): an estimated 206 out of the 5,590 college students met criteria for AUD in follow-up while they did not meet criteria for AUD at baseline. More specifically, among those 4,381 students without 12-month hazardous drinking or AUD at baseline, only 1.1% made the transition to AUD one year later. By comparison, this was 13.6% among those 1,209 students with 12-month hazardous drinking (but no AUD) at baseline.

Table 1.

Twelve-month hazardous drinking and alcohol use disorders at follow-up versus baseline

Baseline 12-month follow-up
n % (SE) n % (SE)
no hazardous drinking or AUD 4381 73.7 (0.6) no hazardous drinking or AUD 3675 83.4 (0.8)
hazardous drinking, no AUD 661 15.5 (0.8)
AUD 46 1.1 (0.2)
hazardous drinking, no AUD 1209 21.6 (0.6) no hazardous drinking or AUD 319 26.0 (1.8)
hazardous drinking, no AUD 730 60.4 (2.0)
AUD 160 13.6 (1.4)
AUD 253 4.7 (0.3) no hazardous drinking or AUD 20 7.8 (2.3)
hazardous drinking, no AUD 104 40.8 (4.2)
AUD 130 51.5 (4.3)
Open in a new tab

AUD = Alcohol use disorder; SE = Standard error

Bivariate and multivariate predictors of AUD incidence

Table 2 presents a summary of the baseline variables that significantly predicted AUD at follow-up first among the subsample of students without AUD at baseline and then among only students without hazardous drinking or AUD at baseline. First, among those without AUD at baseline, results from the bivariate analyses show that incidence of AUD at follow-up was associated with baseline hazardous drinking (OR=14.20), alcohol use characteristics (ORs 1.76-2.35), being male (OR=2.67), break-up with a romantic partner in the past year (OR=2.05), serious betrayal by someone else than a romantic partner (OR=1.53), and other stressful events in the past year (OR=1.89). Especially those with two or more stressful events in the past year had higher odds for subsequent onset of AUD (ORs 1.54-2.17). Past 12-month drug abuse/dependence and a lifetime eating disorder were also associated with incident AUD (ORs=5.27 and 1.54 respectively). Second, among students without hazardous drinking or AUD at baseline, bivariate analyses to predict hazardous drinking or AUD at 12-month follow-up revealed similar results, though there were a few additional predictors, such as studying biomedical sciences (OR=1.21) or screening positive for 12-month broad mania (OR=2.31).

Table 2.

Summary of baseline predictors being significantly associated with 12-month hazardous drinking or AUD at follow-up

12-month AUD at follow-up among students with and without hazardous drinking (no AUD) at baseline (n = 5,590) 12-month hazardous drinking or AUD at follow-up among students without hazardous drinking or AUD at baseline (n = 4,381)
Baseline predictors Median Med (SE) [IQR] Prevalence % (SE) Bivariate OR (95%CI) Multivariateb aOR (95%CI) Median Med (SE) [IQR] Prevalence % (SE) Bivariate OR (95%CI) Multivariatec aOR (95%CI)
AUDIT subscales
   AUDIT consumption score (0-12) 3.1 (0.0) [2.0-4.9] 1.76 (1.07-2.90) 1.29 (1.07-1.56) 2.5 (0.0) [1.8-3.8] 1.65 (1.05-2.60) 1.48 (1.05-2.09)
   AUDIT dependence score (0-12) 0.0 (0.0) [0.0-0.5] 2.35 (1.11-4.96) 1.38 (1.09-1.75) 0.0 (0.0) [0.0-0.0] 2.44 (1.11-5.37) 1.51 (1.09-2.09)
   AUDIT alcohol-related problems score (0-16) 0.0 (0.0) [0.0-1.2] 1.78 (1.06-2.99) 1.31 (1.06-1.61) 0.0 (0.0) [0.0-0.4] 1.84 (1.07-3.16) 1.43 (1.07-1.91)
Hazardous drinking (vs. no hazardous drinking) 22.7 (0.6) 14.20 (1.34-150.25) 3.36 (1.31-8.63) 0.0 (0.0) / /
Sociodemographic variables
   being male 41.9 (0.7) 2.67 (1.27-5.62) 1.53 (1.12-2.10) 36.9 (0.8) 1.66 (1.14-2.43) 1.38 (1.08-1.76)
   Human Sciences 53.8 (0.7) (ref) (ref) 52.6 (0.8) (ref) (ref)
   Science and Technology 26.2 (0.6) 0.84 (0.46-1.54) 0.85 (0.44-1.64) 26.5 (0.7) 0.97 (0.73-1.29) 0.84 (0.52-1.34)
   Biomedical Sciences 20.0 (0.5) 1.00 (0.61-1.63) 1.18 (0.75-1.88) 20.9 (0.6) 1.21 (1.01-1.45) 1.28 (1.01-1.61)
Twelve-month stressful experiences
   break-up romantic partner 17.9 (0.6) 2.05 (1.20-3.48) 1.67 (1.08-2.59) 16.4 (0.6) 1.65 (1.14-2.39) 1.58 (1.09-2.30)
   serious betrayal someone else 12.0 (0.5) 1.53 (1.03-2.27) 1.40 (0.80-2.45) 11.7 (0.5) 1.19 (0.93-1.52) 1.09 (0.71-1.65)
   any other stressful experiences 15.7 (0.5) 1.89 (1.17-3.04) 1.46 (0.95-2.24) 14.6 (0.6) 1.21 (0.97-1.52) 1.11 (0.78-1.57)
no stressful experience 43.2 (0.7) (ref) (ref) (ref) (ref)
exactly 1 stressful experience 28.5 (0.7) 1.29 (0.91-1.82) / 28.1 (0.8) 1.11 (0.91-1.34) /
exactly 2 stressful experiences 16.8 (0.6) 1.54 (1.01-2.37) / 16.2 (0.6) 1.17 (0.91-1.51) /
exactly 3 stressful experiences 7.1 (0.4) 1.98 (1.13-3.46) / 6.7 (0.4) 1.49 (1.10-2.02) /
4+ stressful experiences 4.5 (0.3) 2.17 (1.20-3.91) / 4.1 (0.3) 1.24 (0.80-1.91) /
Mental disorders
   12-month panic attacks 25.7 (0.6) 1.26 (0.94-1.71) / 25.2 (0.7) 1.14 (0.96-1.36) 1.33 (1.05-1.68)
   12-month broad mania 1.3 (0.2) 1.47 (0.37-5.75) / 1.3 (0.2) 2.31 (1.34-3.99) 2.33 (1.05-5.17)
   12-month drug abuse/dependence 0.4 (0.1) 5.27 (1.59-17.53) / 0.2 (0.1) 2.78 (0.46-16.99) 0.93 (0.01-86.40)
   lifetime eating disorder 12.7 (0.4) 1.54 (1.06-2.23) / 12.1 (0.5) 1.26 (1.04-1.53) 1.24 (0.93-1.66)
AUC a 0.887 0.785
Open in a new tab

aOR = adjusted odds ratio; AUD = Alcohol Use Disorder; AUDIT= Alcohol Use Disorders Identification Test; CI = Confidence interval; IQR= Interquartil range; OR = Odds ratio; SE = Standard error

a

AUC values corrected for potential over-estimation of prediction accuracy using the method of leave-one-out cross-validation.

b

The final selected multivariate model includes the three AUDIT subscales, the hazardous drinking dummy variable, all nine sociodemographic variables, and all seven 12-month stressful experiences under study.

c

The final selected multivariate model includes the three AUDIT subscales, all nine sociodemographic variables, all seven childhood-adolescent traumatic experiences, all seven 12-month stressful experiences, all eight mental disorders, and 12-month self-injurious thoughts and behaviours.

Table 2 also shows the final selected multivariate models, adjusting for all other risk domains included in those models. AUD at follow-up among those without AUD at baseline (cross-validated AUC=0.887) was predicted by being male (OR=1.53), a break-up with a romantic partner in the past year (OR=1.67), hazardous drinking (OR=3.36) and alcohol use characteristics at baseline (OR between 1.29 and 1.38). For the more restricted subset of only students without hazardous drinking or AUD at baseline (AUC=0.785), predictors for hazardous drinking or AUD at 12-month follow-up were similar with a few additional predictors (i.e., studying biomedical sciences 12-month panic attacks and 12-month broad mania; ORs in the 1.28-2.33 range).

Prediction accuracy

Table 3 shows cross-validated sensitivity and PPV for different proportions of students at highest predicted risk based on the final multivariate models described above. Multivariate cross-validated prediction (cross-validated AUC=0.887) shows that an estimated 55.5% of incident AUD cases would occur among the 10% of students at the highest predicted risk and that an estimated 20.1% of these high-risk students would go on to meet criteria for AUD at follow-up compared to only 3.9% in the lowest risk subgroup. Among the subset without hazardous drinking or AUD, sensitivity among the 10% of students at the highest predicted risk was 33.1% and the positive predicted value 51.3% versus 16.6% in the lowest risk subgroup.

Table 3.

Concentration of risk for 12-month hazardous drinking or alcohol use disorder at follow-up in different proportions of incoming freshmen at highest predicted risk at baseline based on the final multivariate model

12-month AUD at follow-up among students with and without hazardous drinking (no AUD) at baseline (n = 5,590) 12-month hazardous drinking or AUD at follow-up among students without hazardous drinking or AUD at baseline (n = 4,381)
% Highest Risk Sensitivity SE PPV SE Sensitivity SE PPV SE
100 100.0 0.0 3.9 0.4 100.0 0.0 16.6 0.8
90 99.9 0.2 4.3 0.4 98.7 0.6 18.1 0.8
80 99.8 0.4 4.8 0.5 96.6 1.0 19.9 0.9
70 99.4 0.7 5.5 0.5 93.8 1.2 22.0 1.0
60 98.9 0.9 6.3 0.6 89.9 1.6 24.4 1.2
50 98.1 1.3 7.4 0.7 84.2 1.9 27.3 1.3
40 96.3 1.8 9.0 0.9 76.6 2.2 30.9 1.5
30 91.7 2.7 11.4 1.1 66.2 2.4 35.3 1.9
20 80.9 3.9 15.0 1.5 52.5 2.5 41.4 2.4
10 55.5 4.9 20.1 2.5 33.1 2.2 51.3 3.3
Open in a new tab

AUD = Alcohol use disorder; SE = Standard error

Sensitivity = proportion of risk for AUD disorder cases found among the row % of responders at highest predicted risk, based on cross-validated predicted probabilities.

Positive predictive value (PPV) = probability of effectively developing risk for AUD when being among the row % at highest predicted risk, based on cross-validated predicted probabilities.

Population impact

In Table 4 we show adjusted PIFs of baseline predictors based on the final multivariate models described above. Four out of five students with incident AUD would hypothetically be preventable if baseline hazardous drinking were to be eliminated along with a reduction of one standard deviation in alcohol use characteristics scores, and another 15.0% would potentially be preventable if all 12-month stressful events were eliminated. Among the subset without hazardous drinking or AUD at baseline, alcohol use characteristics is attributable to 46.5% of new onset hazardous drinking/AUD, and eliminating all 12-month stressful events would result in another 5.5% reduction, assuming a full causal relationship.

Table 4.

Potential impact fractions of baseline predictors on risk for 12-month hazardous drinking or AUD at follow-up

Difference distribution 12-month AUD at follow-up among students with and without hazardous drinking (no AUD) at baselinea 12-month hazardous drinking or AUD at follow-up among students without hazardous drinking or AUD at baselineb
aPIF%c aPIF%c aPIF%c aPIF%c
AUDIT consumption score - 1 Standard Deviation 36.1 80.0 34.0 46.5
AUDIT dependence score - 1 Standard Deviation 15.6 6.4
AUDIT alcohol-related problems score - 1 Standard Deviation 23.4 11.2
Hazardous drinking no hazardous drinking 50.8 / /
All traumatic experiences no traumatic experiences / / 0.0d 0.0d
All 12-month stressful events no 12-month stressful events 15.0 15.0 5.5 5.5
All mental disorders no mental disorders / / 0.0d 0.0d
Open in a new tab

AUD = Alcohol use disorder; AUDIT= Alcohol Use Disorders Identification Test

a

The final selected multivariate model includes the three AUDIT subscales, the hazardous drinking dummy variable, all nine sociodemographic variables, and all seven 12-month stressful experiences under study.

b

The final selected multivariate model includes the three AUDIT subscales, all nine sociodemographic variables, all seven childhood-adolescent traumatic experiences, all seven 12-month stressful experiences, all eight mental disorders, and 12-month self-injurious thoughts and behaviours.

c

aPIF = adjusted Potential Impact Fraction. The PIF represents the number of outcome cases that are potentially impacted (reduced) after a change in the exposure of a related continuous of categorical predictor.

DISCUSSION

Main Findings

The aim of this study was to develop and evaluate the performance of a risk algorithm that aims to identify students at high risk of new onset AUD during the first year of college. We found a 3.9% one-year incidence of AUD (in line with the 4.0% reported for 20-29 year olds in the general population [13]), that we could predict with a cross-validated AUC of 0.887 with three baseline variables: alcohol consumption scores, male gender, and having broken up with a romantic partner in the prior 12 months. Using this algorithm, more than half of incident AUD cases would occur among the 10% of students at the highest predicted risk. If colleges were to evaluate incoming students with this algorithm and focus intervention on the students in the highest 10% of predicted risk, they would be targeting more than half of the incident cases. Additionally, a high proportion of AUD incidence was attributable to baseline consumption patterns, which, if reduced by an intervention program during the first year of college, could conceivably decrease AUD incidence by as much as 80.0%. Preventing and/or helping students to cope effectively with stressful life events might also reduce AUD incidence by an additional 15.0%. This is consistent with the findings of Prince, Read and Colder [47] in which relatively small absolute differences in alcohol consumption in the first semester of college predicted large differences in alcohol-related consequences post-graduation. Similarly, Read et al. [48] observed that trauma and post-traumatic stress at matriculation predicted alcohol consequences at the end of the school year. Among the past-year stressful life events assessed, break-up with a romantic partner was the most predictive of increasing risk for AUD. This may be a particularly stressful event for incoming college students given that the transition to college already involves social network changes as many new relationships are formed and need to be balanced with older relationships [49]. In another prospective longitudinal study of young adults over 18 months, romantic relationship dissolution was associated with increased substance use, including heavy alcohol use [50].

Strengths and limitations

While we present a novel approach to risk prediction by estimating the concentration of risk in different proportions of incoming students at highest predicted risk based on a multivariate model of baseline predictors, and by using PIF to simulate population impact of the change of alcohol consumption with a large longitudinal sample of college students, there are several limitations that should be considered when interpreting the results. First, the response rates were modest (51.8% at baseline and 41.6 at follow-up) but consistently higher to those reported in other recent large-scale surveys of college students (39-44%) [3, 51]. Additionally, we used cutting edge missing data techniques [44] to increase the representativeness of the data. Because the sample was drawn from one university in Belgium, replicating the findings in other universities represents an important goal for future research. The sample size lacked power to predict AUD exclusively at follow-up among students without hazardous drinking or AUD at baseline. A further limitation is the self-administered self-report assessment of AUD and other mental disorders, rather than a clinician diagnosis based on face-to-face interviews. However, our measures of AUD and other mental disorders were well-validated screening scales used in many prior general populations surveys, and have shown high diagnostic concordance with clinical diagnoses [26, 52]. Finally, while we included a range of baseline predictors (socio-demographic, consumption patterns, life events and mental disorders) there are other predictors which might increase predictive accuracy in the future such as personality traits like sensation-seeking, urgency, and low constraint [11], family history of problem drinking [53], and protective factors (e.g., emotion regulation competencies [54]). However, it will be important for future research to identify the fewest number of predictors possible that provides the optimal level of accuracy to reduce respondent burden and increase the feasibility of evaluating all incoming students.

Clinical and policy implications

Alcohol use is a large problem across college campuses worldwide. Screening algorithms, such as the one in this study based on integrative multivariate prediction models, may be a useful resource for detecting high-risk students and tailing interventions to those students based on population-level estimates of the factors that contribute the most to overall incidence, namely consumption patterns and effects of romantic relationship break-up. Our findings, along with those of others [47], suggest that future AUD can be predicted in the first year of college with reasonable precision and this early detection could be beneficial for college counselors to implement timely preventive strategies. Several promising interventions have been evaluated in this regard for college students [55–57]. However it may be particularly challenging to get students with alcohol use problems into treatment as prior research has shown that students with AUD are less willing to seek treatment [15] and do not perceive their symptoms to be a problem [11]. Digital risk screeners with subsequent normative feedback, including information about potential preventive options, might be a promising approach to motivate at risk students for preventive interventions [17]. Perhaps interventions that stem from student orientation or activities, that address recognition of the problem and are presented less as traditional modes of treatment delivery would be more acceptable for these students [58]. Increasingly, online interventions, which have the potential to reach a greater number of students at a low cost to university administrators, have shown promise in general community and healthcare settings [59–60] though initial results in college students have been mixed [61–63]. The current study provides data to suggest who and what to target in such interventions and the importance of targeting those students during their first year of college.

Acknowledgements

The Leuven College Survey was carried out in conjunction with the World Health Organization World Mental Health (WMH) survey initiative and is a part of the World Mental Health International College Student project. The WMH survey is supported by the National Institute of Mental Health (NIMH; R01MH070884), the John D. and Catherine T. MacArthur Foundation, the Pfizer Foundation, the US Public Health Service (R13-MH066849, R01-MH069864, and R01 DA016558), the Fogarty International Center (FIRCA R03-TW006481), the Pan American Health Organization, Eli Lilly and Company, Ortho-McNeil Pharmaceutical, GlaxoSmithKline, and Bristol-Myers Squibb. A complete list of all within-country and cross-national WMH publications can be found at http://www.hcp.med.harvard.edu/wmh/. In Belgium specifically, these activities were supported by the Belgian Fund for Scientific Research (11N0514N/11N0516N/1114717N/1114719N), the King Baudouin Foundation (2014-J2140150-102905), Eli Lilly (IIT-H6U-BX-I002), and Fonds GaVoorGeluk (2018-LUF-0070).

Declaration of interest

In the past 3 years, Dr. Kessler received support for his epidemiological studies from Sanofi Aventis, he was a consultant for Johnson & Johnson Wellness and Prevention, Shire and Takeda, and served on an advisory board for the Johnson & Johnson Services Inc. Lake Nona Life Project. Dr. Kessler is a co-owner of DataStat, Inc., a market research firm that carries out healthcare research. The other authors have no interests to declare.

REFERENCES

  • 1.Arnett JJ Emerging Adulthood: the winding road from the late teens through the twenties. 2nd ed. New York: Oxford University Press; 2015. [Google Scholar]
  • 2.Blanco C,Okuda M,Wright C,Hasin DS,Grant BF, Liu SM et al. Mental health of college students and their non-college-attending peers: results from the National Epidemiologic Study on Alcohol and Related Conditions. Arch Gen Psychiatry 2008; 65(12): 1429–37. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 3.Eisenberg D, Hunt J, Speer N Mental health in American colleges and universities: variation across student subgroups and across campuses. J Nerv Ment Dis 2013; 201(1): 60–67. [DOI] [PubMed] [Google Scholar]
  • 4.Auerbach RP, Mortier P, Bruffaerts R, Alonso J, Benjet C, Cuijpers P et al. WHO World Mental Health Surveys International College Student Project: Prevalence and distribution of mental disorders. J Abnorm Psychol 2018; 127(7): 623–638. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 5.Bruffaerts R, Mortier P, Kiekens G, Auerbach RP, Cuijpers P, Demyttenaere K. et al. Mental Health Problems in College Freshmen: Prevalence And Academic Functioning. J Affect Disord 2018; 225: 97–103. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 6.Auerbach RP, Alonso J, Axinn WG, Cuijpers P, Ebet DD, Green JG et al. Mental disorders among college students in the World Health Organization World Mental Health Surveys. Psychol Med 2016; 46(14): 2955– 2970. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 7.GBD 2016 Alcohol and Drug Use Collaborators. The global burden of disease attributable to alcohol and drug use in 195 countries and territories, 1990-2016: a systematic analysis for the Global Burden of Disease Study 2016. Lancet Psychiatry 2018; 5(12): 987–1012. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 8.Pulay AJ, Dawson DA, Hasin DS, Goldstein RB, Ruan WJ, Pickering RP et al. Violent behavior and DSM-IV psychiatric disorders: results from the national epidemiologic survey on alcohol and related conditions. J Clin Psychiatry 2008; 69(1): 12–22. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 9.Hingson RW, Zha W, Weitzman ER Magnitude of and trends in alcohol-related mortality and morbidity among U.S. college students ages 18-24, 1998-2005. J Stud Alcohol Drugs Suppl 2009; 16: 12–20. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 10.Jackson C, Geddes R, Haw S, Frank J Interventions to prevent substance use and risky sexual behavior in young people: A systematic review. Addiction 2012; 107; 733–747. [DOI] [PubMed] [Google Scholar]
  • 11.Nichols LR, Samek DR, McConnell L Key personality traits and alcohol use disorder symptoms in first and second year college students: detangling antecedent from consequence. Addict Behav 2019; 89: 178–187. [DOI] [PubMed] [Google Scholar]
  • 12.van Zoonen K, Buntrock C, Ebert DD, Smit F, Reynolds CF 3rd, Beekman AT et al. Preventing the onset of major depressive disorder: a meta-analytic review of psychological interventions. Int J Epidemiol 2014;43(2): 318–29. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 13.Grant BF, Goldstein RB, Chou SP, Huang B, Stinson FS, Dawson DA et al. Sociodemographic and psychopathologic predictors of first incidence of DSM-IV substance use, mood and anxiety disorders: results from the Wave 2 National Epidemiologic Survey on Alcohol and Related Conditions. Mol Psychiatry 2009; 14(11): 1051–66. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 14.Harrer M, Adam SH, Baumeister H, Cuijpers P, Karyotaki E, Auerbach RP et al. Internet interventions for mental health in university students: A systematic review and meta-analysis. Int J Methods Psychiatr Res 2019;28(2): e1759. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 15.Ebert DD, Mortier P, Kaehlke F, Bruffaerts R, Baumeister H, Auerbach RP et al. Barriers of mental health treatment utilization among first-year college students: First cross-national results from the WHO World Mental Health International College Student Initiative. Int J Methods Psychiatr Res 2019; 28(2): e1782. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 16.Ganz T, Braun M, Laging M, Schermelleh-Engel K, Michalak J, Heidenreich T Effects of a stand-alone web-based electronic screening and brief intervention targeting alcohol use in university students of legal drinking age: A randomized controlled trial. Addict Behav 2018;77: 81–88. [DOI] [PubMed] [Google Scholar]
  • 17.Ebert DD, Franke M, Kählke F, Küchler AM, Bruffaerts R, Mortier P et al. Increasing intentions to use mental health services among university students. Results of a pilot randomized controlled trial within the World Health Organization’s World Mental Health International College Student Initiative. Int J Methods Psychiatr Res 2019; 28(2): e1754. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 18.Hallfors D, Brodish PH, Khatapoush S, Sanchez V, Cho H, Steckler A Feasibility of screening adolescents for suicide risk in “real-world” high school settings. Am J Public Health 2006; 96(2): 282–287. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 19.Kiekens G, Hasking P, Claes L, Boyes M, Mortier P, Auerbach RP et al. Predicting the incidence of non-suicidal self-injury in college students. Eur Psychiatry 2019; 59: 44–51. [DOI] [PubMed] [Google Scholar]
  • 20.Ebert DD, Buntrock C, Mortier P, Auerbach R, Weisel KK, Kessler RC et al. Prediction of major depressive disorder onset in college students. Depress Anxiety 2019; 36(4): 294–304. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 21.Mortier P, Demyttenaere K, Auerbach RP, Cuijpers P, Green JG , Kiekens G et al. First onset of suicidal thoughts and behaviours in college. J Affect Disord. 2017;207: 291–299. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 22.Ebert DD, Cuijpers P, Muñoz RF, & Baumeister H Prevention of Mental Health Disorders using Internet and mobile-based Interventions: a narrative review and recommendations for future research. Front Psychiatry 2017; 8: 116. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 23.Ribeiro JD, Franklin JC, Fox KR, Bentley KH, Kleiman EM, Chang BP et al. Self-injurious thoughts and behaviors as risk factors for future suicide ideation, attempts, and death: a meta-analysis of longitudinal studies. Psychol Med 2016; 46(2): 225–36. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 24.Saunders JB, Aasland OG, Babor TF, de la Fuente JR, Grant M Development of the alcohol use disorders identification test (AUDIT): WHO collaborative project on early detection of persons with harmful alcohol consumption–II. Addiction 1993: 88(6): 791–804. [DOI] [PubMed] [Google Scholar]
  • 25.Babor T, Higgins-Biddle J, Saunders J, Monteiro M The Alcohol Use Disorders Identification Test. Guidelines for use in primary care. 2001. Available from http://apps.who.int/iris/handle/10665/67205 [Google Scholar]
  • 26.Reinert DF, Allen JP (2002). The Alcohol Use Disorders Identification Test (AUDIT): A review of recent research. Alcohol Clin Exp Res 2002; 26(2): 272–279. [PubMed] [Google Scholar]
  • 27.Kessler RC, Ustun TB The World Mental Health (WMH) Survey Initiative Version of the World Health Organization (WHO) Composite International Diagnostic Interview (CIDI) Int J Methods Psychiatr Res 2004; 13(2): 93–121. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 28.Felitti VJ, Anda RF, Nordenberg D, Williams DF, Spitz AM, Edwards V et al. Relationship of childhood abuse and household dysfunction to many of the leading causes of death in adults. The Adverse Childhood Experiences (ACE) Study. Am J Prev Med 1998; 14(4): 245–258. [DOI] [PubMed] [Google Scholar]
  • 29.Swearer S, Cary P Perceptions and attitudes toward bullying in middle school youth. J Appl Sch Psych 2003;19(2): 63–79. [Google Scholar]
  • 30.Nansel TR, Overpeck M, Pilla RS, Ruan WJ, Somins-Morton B, Scheidt P Bullying behaviors among US youth: prevalence and association with psychosocial adjustment. JAMA 2001; 285: 2094–100. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 31.Brugha TS, Cragg D The List of Threatening Experiences: the reliability and validity of a brief life events questionnaire. Acta Psychiatr Scand 1990; 82(1): 77–81. [DOI] [PubMed] [Google Scholar]
  • 32.Bray RM, Hourani LL Substance use trends among active duty military personnel: findings from the United States Department of Defense Health Related Behavior Surveys, 1980-2005. Addiction 2007; 102(7): 1092–1101. [DOI] [PubMed] [Google Scholar]
  • 33.Vogt DS, Proctor SP, King DW, King LA, Vasterling JJ Validation of scales from the Deployment Risk and Resilience Inventory in a sample of Operation Iraqi Freedom veterans. Assessment 2008; 15(4): 391–403. [DOI] [PubMed] [Google Scholar]
  • 34.Stoeckel M, Weissbrod C Growing up with an ill parent: An examination of family characteristics and parental illness features. Fam Syst Health 2015; 33(4): 356–362. [DOI] [PubMed] [Google Scholar]
  • 35.Rostila M, Berg L, Arat A, Vinnerljung B, Hjern A Parental death in childhood and self-inflicted injuries in young adults-a national cohort study from Sweden. Eur Child Adolesc Psychiatry 2016; 25(10): 1103–1111. [DOI] [PubMed] [Google Scholar]
  • 36.Buitron V, Hill RM, Pettit JW, Green KL, Hatkevich C, Sharp C Interpersonal stress and suicidal ideation in adolescence: An indirect association through perceived burdensomeness toward others. J Affect Disord 2016; 190: 143–149. [DOI] [PubMed] [Google Scholar]
  • 37.Trotman GE, Young-Anderson C, Deye KP. Acute sexual assault in the pediatric and adolescent population. J Pediatr Adolesc Gynecol 2016; 29(6): 518–526. [DOI] [PubMed] [Google Scholar]
  • 38.Viswanathan S, Datta S, Sheridan P, Lax-Pericall T “Too young to be worried!” Psychiatric assessment and follow-up of young people after severe physical assault in an inner city hospital of South London. Ann Med Health Sci Res 2014; 4(1): 85–89. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 39.Salekin RT Psychopathy and recidivism from mid-adolescence to young adulthood: cumulating legal problems and limiting life opportunities. J Abnorm Psychol 2008; 117(2): 386–395. [DOI] [PubMed] [Google Scholar]
  • 40.Kessler RC, Santiago PN, Colpe LJ, Dempsey CL, First MB, Heeringa SG et al. Clinical reappraisal of the Com- posite International Diagnostic Interview Screening Scales (CIDI-SC) in the army study to assess risk and resilience in service members (army STARRS). Int J Methods Psychiatr Res 2013: 22(4): 303–321. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 41.Nock MK, Holmberg EB, Photos VI, Michel BD Self-Injurious Thoughts and Behaviors Interview: development, reliability, and validity in an adolescent sample. Psychol Assess 2007; 19(3): 309–317. [DOI] [PubMed] [Google Scholar]
  • 42.Sheehan DV, Lecrubier Y, Sheehan KH, Amorim P, Janavs J, Weller E et al. The Mini-International Neuropsychiatric Interview (M.I.N.I.): the development and validation of a structured diagnostic psychiatric interview for DSM-IV and ICD-10. J Clin Psychiatry 1998; 59 Suppl 20: 22–33;quiz 34-57. [PubMed] [Google Scholar]
  • 43.Rosenbaum PR, Rubin DB The Central Role of the Propensity Score in Observational Studies of Causal Effects. Biometrica 1983; 70: 41–55. [Google Scholar]
  • 44.van Buuren S Multiple imputation of discrete and continuous data by fully conditional specification. Stat Methods Med Res 2007; 16(3): 219–242. [DOI] [PubMed] [Google Scholar]
  • 45.Efron B, Gong G A leisurely look at the bootstrap, the jackknife, and cross-validation. Am Stat 1983; 37 (1): 36–48. [Google Scholar]
  • 46.Barendregt JJ, Veerman JL Categorical versus continuous risk factors and the calculation of potential impact fractions. J Epidemiol Community Health 2010; 64(3): 209–12. [DOI] [PubMed] [Google Scholar]
  • 47.Prince MA, Read JP, Colder CR Trajectories of college alcohol involvement and their associations with later alcohol use disorder symptoms. Prev Sci 2019; 20(5): 741–752. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 48.Read JP, Colder CR, Merrill JE, Ouimette P, White J, Swartout A Trauma and posttraumatic stress symptoms predict alcohol and other drug consequence trajectories in the first year of college. J Consult Clin Psychol 2012; 80(3): 426–439. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 49.Meisel MK, Barnett NP Protective and Risky Social Network Factors for Drinking During the Transition From High School to College. J Stud Alcohol Drugs 2017; 78(6): 922–929. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 50.Fleming CB, White HR, Oesterle S, Haggerty KP, Catalano RF Romantic Relationship Status Changes and Substance Use Among 18- to 20-Year-Olds. J Stud Alcohol Drugs 2010;71(6): 847–56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 51.Paul E, Tsypes A, Eidlitz L, Ernhout C, Whitlock J Frequency and functions of non-suicidal self-injury: associations with suicidal thoughts and behaviors. Psychiatry Res. 2015; 225(3): 276–82. [DOI] [PubMed] [Google Scholar]
  • 52.Kessler R, Farley P, Gruber M, Harshaw Q, Jewell M, Sampson N, et al. Concordance of computerized self-report measures of DSM-IV-Tr mood and anxiety disorders compared to gold standard clinical assessments in primary care. International Society for Pharmacoeconomics and Outcomes Research 15th Annual International Meeting; 2010; Atlanta, Georgia. [Google Scholar]
  • 53.Meque I, Salom C, Betts KS, Alati R Predictors of Alcohol Use Disorders Among Young Adults: A Systematic Review of Longitudinal Studies. Alcohol Alcohol. 2019; 54(3): 310–324. [DOI] [PubMed] [Google Scholar]
  • 54.Berking M, Margraf M, Ebert D, Wuppermann P, Hofmann S, Junghanns K Emotion regulation skills as a predictor of relapse during and after treatment of alcohol dependence. J Consult Clin Psychol 2011; 79(3); 307–318. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 55.Scott-Sheldon LA, Carey KB, Elliott JC, Garey L, Carey MP Efficacy of alcohol interventions for first-year college students: a meta-analytic review of randomized controlled trials. J Consult Clin Psychol 2014; 82(2): 177–88. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 56.Hennessy EA, Tanner-Smith EE, Mavridis D, Grant SP Comparative Effectiveness of Brief Alcohol Interventions for College Students: Results from a Network Meta-Analysis. Prev Sci 2019;20(5):715–740. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 57.Turrisi R, Larimer M, Mallett K, Kilmer J, Ray A, Mastroleo, et al. A randomized clinical trial evaluating a combined alcohol intervention for high-risk college students. J Stud Alcohol Drugs 2009; 70(4): 555–567. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 58.Kazdin AE Innovations in psychosocial interventions and their delivery: Leveraging cutting-edge science to improve the world’s mental health. New York: Oxford University Press; 2018. [Google Scholar]
  • 59.Riper H, Hoogendoorn A, Cuijpers P, Karyotaki E, Boumparis N, Mira A Effectiveness and treatment moderators of internet interventions for adult problem drinking: An individual patient data meta-analysis of 19 randomised controlled trials. PLOS Medicine 2018; 15(12): e1002714. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 60.Sundström C, Blankers M, Khadjesari Z Computer-Based Interventions for Problematic Alcohol Use: a Review of Systematic Reviews. Int J Behav Med 2017; 24(5): 646–658. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 61.Bhochhibhoya A, Hayes L, Branscum P, Taylor L The Use of the Internet for Prevention of Binge Drinking Among the College Population: A Systematic Review of Evidence. Alcohol Alcohol 2015; 50(5): 526–35. [DOI] [PubMed] [Google Scholar]
  • 62.Carey KB, Scott-Sheldon LA, Elliott JC, Garey L Carey MP Face-to-Face Versus Computer-Delivered Alcohol Interventions for College Drinkers: A Meta-Analytic Review, 1998 to 2010. Clin Psychol Rev 2012; 32(8): 690–703. [DOI] [PMC free article] [PubMed] [Google Scholar]
  • 63.Kypri K, Vater T, Bowe SJ, Saunders JB, Cunningham JA, Horton NJ et al. Web-Based Alcohol Screening and Brief Intervention for University Students: A Randomized Trial. JAMA 2014; 311(12): 1218–1224. [DOI] [PMC free article] [PubMed] [Google Scholar]

RESOURCES