class level alcohol intervention for college students: …...participation in Alcohol-Wise, a computerized intervention, is associated with changes in alcohol drinking behavior and

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=iada20

Download by: [72.177.229.198] Date: 23 February 2016, At: 13:28

The American Journal of Drug and Alcohol AbuseEncompassing All Addictive Disorders

ISSN: 0095-2990 (Print) 1097-9891 (Online) Journal homepage: http://www.tandfonline.com/loi/iada20

Randomized controlled trial of computerizedalcohol intervention for college students: role ofclass level

Ashleigh Sweet Strohman PhD, Sopagna Eap Braje PhD, Omar M. AlhassoonPhD, Sylvie Shuttleworth PhD, Jenn Van Slyke MS & Sharareh Gandy PhD

To cite this article: Ashleigh Sweet Strohman PhD, Sopagna Eap Braje PhD, Omar M. AlhassoonPhD, Sylvie Shuttleworth PhD, Jenn Van Slyke MS & Sharareh Gandy PhD (2015): Randomizedcontrolled trial of computerized alcohol intervention for college students: role of class level,The American Journal of Drug and Alcohol Abuse, DOI: 10.3109/00952990.2015.1105241

To link to this article: http://dx.doi.org/10.3109/00952990.2015.1105241

Published online: 18 Dec 2015.

Submit your article to this journal

Article views: 57

View related articles

View Crossmark data

ORIGINAL ARTICLE

Randomized controlled trial of computerized alcohol intervention for collegestudents: role of class levelAshleigh Sweet Strohman, PhDa, Sopagna Eap Braje, PhDa, Omar M. Alhassoon, PhD a, Sylvie Shuttleworth,PhDb, Jenna Van Slyke, MSa, and Sharareh Gandy, PhDa

aClinical Psychology PhD Program, California School of Professional Psychology, San Diego, CA, USA; bCounseling & Health PsychologyDepartment, Bastyr University California, San Diego, CA, USA

ABSTRACTBackground: Because of their ability to reach a much wider audience than face-to-face counselingor psychoeducation, computer-delivered interventions for risky or potentially problematic usehave been increasing on college campuses. However, there are very few studies that examine whobenefits most from such interventions. Objectives: The purpose of this study was to determine ifparticipation in Alcohol-Wise, a computerized intervention, is associated with changes in alcoholdrinking behavior and its consequences, perceptions of college drinking norms, and expectancies.It was hypothesized that class level (i.e. freshman/sophomore versus junior/senior) would mod-erate the effectiveness of Alcohol-Wise. Method: College students (n = 58) were randomly assignedto one of two conditions: (i) the computer-delivered intervention or (ii) wait-list control. Measureswere completed at baseline and approximately 30-days later. Results: At follow-up, freshman andsophomore students in the intervention group showed significant reduction in peak number ofstandard drinks and blood alcohol concentration, but the effect was not observed for juniors andseniors. The intervention group reported more accurate estimates of drinking norms at follow-uprelative to controls. There were no significant changes over time in alcohol expectancies in eithergroup. Conclusion: This study provides support for the potential usefulness of Alcohol-Wiseintervention at reducing short-term drinking among underclassmen but not upperclassmen in a4-year college setting. These findings suggest that computerized interventions may be moreeffective when provided early, but not later, in a student’s college career.

ARTICLE HISTORYReceived 30 May 2015Revised 5 October 2015Accepted 5 October 2015

KEYWORDSComputer deliveredintervention; collegedrinking; randomizedcontrolled trial; alcohol use;alcohol abuse treatment

Introduction

Excessive drinking in college can lead to consequencessuch as academic difficulties, engagement in risky sex-ual behavior, blackouts, alcohol poisoning, and evendeath from driving under the influence or being apassenger accompanying an intoxicated driver (1).According to the Centers for Disease Control andPrevention (2), excessive alcohol use is a leading causeof preventable death among college students.Additionally, alcohol-related physical and sexualassaults are a major problem on college campuses (3).In response to these public health concerns, the USDepartment of Education has awarded over $3.5 mil-lion to alcohol and drug prevention programs on col-lege campuses since 1998 (4). Many of these programsare in the form of computer-delivered interventions,also known as electronic or web-based online interven-tions. These programs are cost-effective and easily dis-seminated. Students have reported that they prefercomputer-delivered interventions over face-to-face

interventions (5). Given potential differences in treat-ment-seeking behavior among college students (6), thisintervention modality may increase the likelihood thatstudents who need services will utilize them.

Cost-efficiency and ease of dissemination notwith-standing, the efficacy of computer-delivered interven-tions is equivocal. A recent meta-analyses found thatthese type of interventions produce similar treatmentoutcomes as face-to-face ones at short-term follow-upbut not at long term-term follow-up, with face-to-faceoutperforming computer-delivered interventions (7).However, not all computer-delivered interventions per-form the same and the lack of benefit over other mod-alities must be considered in light of the fact that suchinterventions can reach a greater number of people sincemany may prefer the privacy and flexibility that theyoffer. To date, there are eight commercial computer-delivered interventions targeting alcohol use among col-lege students: (1) Alcohol 101, (2) Under the Influence,(3) Alcohol Response-Ability, (4) myStudentBody, (5)

CONTACT Omar M. Alhassoon, PhD [email protected] Clinical Psychology PhD Program, California School of Professional Psychology, 10455Pomerado Road, San Diego, CA 92131, USA.

AM J DRUG ALCOHOL ABUSEhttp://dx.doi.org/10.3109/00952990.2015.1105241

© 2015 Taylor & Francis

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

College Alc, (6) AlcoholEdu, (7) The AlcoholeCHECKUP TO GO (e-CHUG), and (8) Alcohol-Wise.Programs vary in content and length, but all of theseavailable interventions provide normative drinkinginformation, harm reduction strategies (e.g. tips forsafer drinking), and alcohol education. The body ofresearch supporting the efficacy of computer-deliveredinterventions for alcohol and other drug use is growing(7–10). However, some programs are being developedand implemented without being evaluated on their abil-ity to reduce alcohol consumption and related problemsamong college students. In addition, further research isneeded to identify specific characteristics of respondersversus non-responders (11).

Third Millennium Classroom’s program, Alcohol-Wise, was created in 2007 and is used on over 300college campuses. Despite the wide-spread usage (3rd

Millennium Classrooms, 2007), there are currently nopublished studies on its effectiveness. Alcohol-Wiseincludes elements of several existing alcohol compu-ter-delivered interventions, such as: alcohol educa-tion, personalized feedback, expectancy challenge,and skills based activities. Additionally, Alcohol-Wise utilizes motivation enhancement strategies toreduce students’ likelihood of engaging in riskydrinking behaviors in the future. The eCHECKUPTO GO (e-Chug), which is incorporated intoAlcohol-Wise, has garnered extensive empirical sup-port as a stand-alone treatment for alcohol treatmentand prevention (9,10,12–15). However, there are cur-rently no published studies that have examined theeffectiveness of the e-Chug when used in conjunctionwith other alcohol-related programs. Moreover, themajority of the studies on the e-Chug have focusedon mandated populations or incoming freshman.Therefore, it would be useful to examine how upper-classmen respond to the intervention.

In addition to the eCHECKUP TO GO brief inter-vention, Alcohol-Wise is an alcohol education coursethat is designed as a tool for prevention as well asintervention. Alcohol-Wise includes: social normsinformation, audio narration, student interviews, inter-active journaling reflections, lesson quizzes, and a finalexam to check for attention and retention ofinformation.

The current study is the first randomized controltrial to examine Alcohol-Wise’s effectiveness at redu-cing drinking behavior and to include class level (i.e.freshman/sophomore versus junior/senior) as a moder-ating variable. Alcohol-Wise specifies that the programshould be used as a prevention tool for incoming stu-dents as well as an intervention tool. It is not clear,however, how effective Alcohol-Wise is with older

students. The specific purpose of this pilot study is toexamine how upperclassmen (juniors and seniors) ver-sus underclassmen (freshmen and sophomores) parti-cipating in Alcohol-Wise reduced their drinkingfrequency, quantity, and alcohol-related negative con-sequences 30-days later relative to a control group. Thefirst hypothesis was that the effect of Alcohol-Wise onreducing drinking behavior would be moderated byclass level. The second hypothesis examined whetherAlcohol-Wise would result in a reduction in positivealcohol-related expectancies and an increase in accurateperceptions of drinking norms (normative perceptions)among students. Finally, the third hypothesis was thatsuch cognitive changes would also be moderated byclass level.

Method

Participants

Participants in the study were students enrolled at apublic university in Southern California. Participantswere treated according to the established ethical stan-dards of the American Psychological Association. Themethods and procedures used in this study wereapproved by the Institutional Review Board.Participants in the Alcohol-Wise experimental condi-tion and wait-list control condition were compensatedequally. Each participant was given $15 after complet-ing the initial phase of the study and $30 after com-pleting the follow-up surveys. In response to campusadvertising, 106 participants contacted the researcherby email expressing interest in the study. Theresearcher matched potential participants on sex,class level, and Greek affiliation and randomlyassigned each participant to one of two conditions:(i) the Alcohol-Wise experimental condition or (ii)the wait-list control condition. The researcher sentdifferent links directing the potential participants tothe appropriate condition. However, 29 students didnot enroll in the study for unknown reasons. A totalof 77 participants enrolled in the study. Seventy-sixparticipants started the study with 40 randomlyassigned to the experimental condition and 36 to thecontrol condition. Fifty-eight participants completedpre and post measures, 29 in the control conditionand 29 in the experimental condition. Only partici-pants who completed both the baseline and follow-upself-report measures were included in statistical ana-lyses (See Figure 1). Comparisons using t-tests andChi-squares showed no statistically significant differ-ences on demographic variables (i.e. sex, age, classlevel, ethnicity, housing, Greek affiliation, athletic

2 A. SWEET STROHMAN ET AL.

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

affiliation, and GPA) between participants who didnot complete self-report measures and those whocompleted all self-report measures.

A series of t-tests showed no statistically significantdifferences between the experimental and control condi-tion on research variables at baseline. Additional t-testsshowed no statistically significant differences betweenupperclassmen and underclassmen on research variablesat baseline. There were no statistically significant differ-ences between the control and experimental conditionson demographic variables including: sex, age, ethnicity,class level, GPA, housing, or Greek and athletic affilia-tion. The demographic information described in thissection is reported for the entire sample (n = 58).Participants ranged in age from 18–22 years (M = 20;SD = 1.22). Most participants were female (n = 43,79.3%) and white (n = 40, 69%) (See Table 1).

Procedures

After giving informed consent, the participants wererandomly assigned to either the wait-list control con-dition or experimental condition stratified by genderand class level. Participants completed the assessmentmeasures and the Alcohol-Wise intervention online.The total time for participation was approximately 3 hfor participants in the experimental condition andapproximately 1.5 h for participants in the wait-listcontrol condition. Alcohol-Wise is organized into sixmodules and each module took approximately 15 min

to complete, making the total intervention timeapproximately 90 min.

Participants in the experimental condition were askedto complete the eCHECKUP TO GO assessment fol-lowed by six modules of Alcohol-Wise. Approximately30 days after completing the intervention, participantsreceived an email and were asked to complete theeCHECKUP TO GO assessment. Each module ofAlcohol-Wise ended with a quiz regarding the informa-tion covered in the module. Participants in the wait-listcontrol condition completed the same assessments asparticipants in the experimental group but did not com-plete the Alcohol-Wise program. Participants in thewait-list control condition were given access to volunta-rily complete the intervention after the completion of thestudy.

Measures

Demographics informationDemographic information was collected including: race/ethnicity, sex, age in years, weight in pounds, Greek affilia-tion (i.e. membership in a fraternity or sorority), member-ship in a college athletic team, grade point average (GPA),and class level. Participants were also asked whether or notthey were taking any medications and whether or not theylived on-campus or in a residence hall.

eCHECKUP TO GOThe eCHECKUP TO GO assessment is comprised ofseveral different questionnaires, some of which have

Figure 1. Participant attrition flow chart.

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 3

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

been modified from originals to better serve a collegestudent population. The questionnaires incorporatedare well established for use in alcohol research. Thefollowing section will describe the different topicsaddressed by the survey that were examined in thepresent study.

Alcohol useThe first 13 questions of the eCHECKUP TO GOassessment include questions about demographicinformation and alcohol use. Participants wereasked at what age they first started drinking alcoholand to provide the frequency and quantity of theiralcohol use in a typical month. First, they were askedhow many weeks out of the month they drink alco-hol. Participants were then shown a week-long calen-dar (i.e. Sunday through Saturday) and asked to fillin how many standard drinks (i.e. 12 oz. of beer orwine cooler, 4.5 oz. of table wine, and 1.5 oz. ofspirits) they would typically consume each day. Thecalendar format was based on the timeline follow-back method (16). Participants were also asked toprovide the number of hours they drink to allowfor BAC estimations. Additionally, they were alsoasked to recall the time they drank the most in thepast month to allow for peak alcohol consumptionand BAC estimations. The participants were providedboth audio and visual descriptions of a standarddrink.

ExpectanciesExpectancies were measured utilizing an adapted ver-sion of the Alcohol Expectancy Questionnaire (AEQ)(17). The AEQ measures anticipated experiences asso-ciated with alcohol use. Participants were asked ques-tions regarding positive expectancies about the effectsof alcohol using a dichotomous scale (i.e. Yes or No).The revised version used in the current study is com-prised of 18 statements that assess alcohol reinforce-ment expectancies relevant to college population.Higher scores indicate higher positive expectancies inthe areas of general experience, sexual enhancement,social/physical pleasure, assertiveness, relaxation/ten-sion reduction, and arousal/interpersonal power as aresult of consuming alcohol. The AEQ scores are pre-dictive of current and future drinking behaviors, parti-cipation in treatment, and relapse (17). In the currentstudy, the Cronbach’s alpha coefficient was 0.88 for theadapted AEQ.

Perceptions of drinking normsPerceptions of drinking norms were measured byasking participants what percent of university stu-dents do not drink at all in a typical week. Thisquestion is similar to questions asked on TheDrinking Norms Rating Form (DNRF) (18). Duringthe Alcohol-Wise intervention, participants are giventhe normative data in comparison to theirestimations.

Table 1. Demographic characteristics of experimental and control group and the drop-outs.Characteristic Experimental group Drop-outs from experimental group Control group Drop-outs from control group

GenderFemale 82.8 (n = 24) 85.7 (n = 6) 75.9 (n = 22) 100.0 (n = 4)Male 17.2 (n = 5) 14.3 (n = 1) 24.1 (n = 7) 0.0 (n = 0)

Race/ethnicityWhite 72.4 (n = 21) 36.4 (n = 4) 65.5 (n = 19) 71.4 (n =5)Black 0.0 (n = 0) 9.1 (n = 1) 6.9 (n = 2) 0.0 (n = 0)Spanish/Hispanic/Latino 17.2 (n = 5) 45.5 (n = 5) 10.3 (n = 3) 14.3 (n = 1)Asian or Pacific Islander 10.3 (n = 3) 9.1 (n = 1) 13.8 (n = 4) 0.0 (n = 0)Other 0.0 (n = 0) 0.0 (n = 0) 3.4 (n = 1) 14.3 (n = 1)

Class levelFreshman 27.6 (n = 8) 14.3 (n = 1) 27.6 (n = 8) 25.0 (n = 1)Sophomore 24.1 (n = 7) 28.6 (n = 2) 27.6 (n = 8) 25.0 (n = 1)Junior 17.2 (n = 5) 0.0 (n = 0) 20.7 (n = 6) 0.0 (n = 0)Senior 31.0 (n = 9) 57.1 (n = 4) 24.1 (n = 7) 50.0 (n = 2)

HousingOn-campus 27.6 (n = 8) 28.6 (n = 2) 31.0 (n = 9) 25.0 (n = 1)Off-campus 72.4 (n = 21) 71.4 (n = 5) 69.0 (n = 20) 75.0 (n = 3)

Fraternity or Sorority MemberGreek affiliated 79.3 (n = 23) 71.4 (n = 5) 75.9 (n = 22) 100.0 (n = 4)Non-Greek 20.7 (n = 6) 28.6 (n = 2) 24.1 (n = 7) 0.0 (n = 0)

Athletic team memberAthlete 6.9 (n = 2) 0.0 (n = 0) 17.2 (n = 5) 0.0 (n = 0)Non-athlete 93.1 (n = 27) 100.0 (n = 7) 82.8 (n = 24) 100.0 (n = 4)

Grade point average3.0–4.0 89.7 (n = 26) 81.8 (n = 9) 72.4 (n = 21) 85.7 (n = 6)2.0–2.9 10.3 (n = 3) 18.2 (n =2) 27.6 (n = 8) 14.3 (n = 1)

Note. Data is reported as percentages.

4 A. SWEET STROHMAN ET AL.

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

Negative consequencesThe Alcohol Use Disorders Identification Test (AUDIT)(19) was developed as part of a six-nation WorldHealth Organization (WHO) project. The AUDIT is a10-item screening to identify excessive (i.e. hazardous)drinking by measuring consumption, behavior, andconsequences associated with consumption measuredon a 5-point Likert scale. In general, cut-off scoresranging from 6–11 may indicate hazardous drinking,but should be interpreted with caution among a collegepopulation (20,21). Lower specificity may result in overidentification of hazardous drinkers; however, given theprevalence of at-risk drinkers in college, over identifi-cation (false-positives) may be of less concern thanfalse-negatives; especially when evaluating the effective-ness of an intervention at reducing risky drinking beha-vior. The AUDIT assesses alcohol use over the past yearand is incorporated in the eCHECKUP TO GO assess-ment. In the current study, the Cronbach’s alpha coef-ficient was 0.97 for the adapted AUDIT.

Results

Statistical analyses

Prior to statistical analyses, the data was screened foroutliers, skewness, kurtosis, and restriction of range.Review of skewness and kurtosis revealed that thedata for the number of standard drinks at heaviestdrinking occasion were positively skewed at the base-line and follow-up assessments. Additionally, the nega-tive consequences variable was positively skewed at the

follow-up assessment. Outliers that were two standarddeviations above the mean were transformed usingWinsorization, which is the process of replacing out-liers with the next closest value. Winsorization has beenshown to provide a more accurate representative viewof the data when compared to more classical methods(e.g. trimming) and, as such, was selected as themethod of choice when attending to outlier variance(22–24). After Winsorization, all research variables metnormality assumptions (see Table 2).

The study broadly hypothesized that participants inthe Alcohol-Wise experimental condition would reportfewer drinking days, fewer standard drinks at peakdrinking occasion, a smaller peak BAC, and fewernegative consequences in a typical week at the 30-dayfollow-up compared to participants in the wait-list con-trol condition. More specifically, it was hypothesizedthat this effect will be stronger for underclassmen thanupperclassmen. The study also hypothesized that parti-cipation in Alcohol-Wise would reduce alcohol-relatedexpectancies that would be moderated by class level.The effect of the Alcohol-Wise on social norms was notexpected to be moderated by class level. A mixedbetween-within subjects 2 (intervention vs. control) ×2 (under vs. upperclassmen) × 2 (Time 1 vs. Time 2)analyses of variance (ANOVA) was conducted sepa-rately for each dependent variable.

Drinking daysThere was no significant three-way interaction betweenclass level, experimental condition, and time on drink-ing days, F(1, 54) = 0.00, p = 0.98, η2p = 0.00. There was

Table 2. Mean and standard deviation of drinking behavior, expectancies, and perception of drinking norms.Experimental group n = 29 Control group n = 29

Variable Underclassmen n = 15 Upperclassmen n = 14 Underclassmen n = 15 Upperclassmen n = 14

Typical drinking Days M (SD) M (SD) M (SD) M (SD)Time 1 2.47 (0.92) 3.36 (1.65) 2.44 (1.32) 3.23 (1.69)Time 2 2.20 (0.56) 3.36 (2.02) 2.19 (1.28) 3.23 (2.17)Heaviest drinking occasionTime 1 9.80 (4.62) 10.50 (10.22) 10.75 (8.53) 10.62 (7.99)Time 2 6.87 (2.67) 12.07 (11.29) 10.50 (6.22) 8.46 (4.89)Peak BACTime 1 0.22 (0.12) 0.19 (0.18) 0.25 (0.21) 0.20 (0.16)Time 2 0.14 (0.08) 0.23 (0.20) 0.23 (0.15) 0.17 (0.12)Total negative consequencesTime 1 15.33 (9.13) 16.21 (10.86) 15.38 (11.06) 21.92 (17.05)Time 2 10.93 (8.18) 16.79 (22.10) 13.06 (10.63) 21.23 (14.93)Explicit alcohol expectanciesTime 1 10.47 (4.09) 9.71 (4.14) 9.31 (4.83) 10.62 (5.12)Time 2 11.27 (3.04) 10.14 (5.64) 8.69 (5.00) 11.15 (4.56)Implicit alcohol expectanciesTime 1 −0.05 (0.61) −0.41 (0.42) −0.18 (0.56) −0.11 (0.47)Time 2 −0.21 (0.56) −0.20 (0.43) −0.61 (0.36) −0.12 (0.44)Perception of abstinenceTime 1 −5.87 (21.26) −19.29 (13.40) −12.75 (18.71) −12.62 (18.03)Time 2 3.60 (19.61) −6.71 (21.32) −10.25 (18.03) −14.31 (17.00)

Note. Time 1 = baseline and Time 2 = 30-day follow-up assessment. The Valence Alcohol Implicit Association Test (IAT) n ranges from 21–29, N = 50. The IATis reported as a D score. Perception of Abstinence is reported as a difference score (the difference between participants’ estimates and the actualpercentage [39%] of students who abstain from drinking alcohol) and negative scores reflect an overestimation of how many students abstain.

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 5

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

no difference between the control and experimentalcondition or upper and underclassmen on number ofdrinking days, across the two time periods.

Peak number of drinks consumed in one sittingThere was a statistically significant three-way interac-tion between class level, group condition, and time onstandard drinks consumed on heaviest drinking occa-sion, F(1, 54) = 8.05, p = 0.01, η2p = 0.13. Follow-upanalyses were conducted to parse out the interactioneffect. Two mixed between-within subjects 2 × 2ANOVAs were conducted first. There was a statisticallysignificant interaction between time and group condi-tion on standard drinks consumed on heaviest drinkingoccasion for underclassmen, F(1, 29) = 7.15, p = 0.01,η2p = 0.20. The same interaction was not statisticallysignificant for upperclassmen, F(1, 25) = 2.12, p = 0.16,η2p = 0.08. There was a statistically significant interac-tion between class level and time on standard drinksconsumed on heaviest drinking occasion for partici-pants in the experimental condition, F(1, 27) = 9.17,p = 0.01, η2p = 0.25. The same interaction was notstatistically significant for participants in the controlcondition, F(1, 27) = 9.98, p = 0.28, η2p = 0.04.Finally, a two-way ANOVA was conducted to deter-mine if there were any differences on standard drinksconsumed on heaviest drinking occasion betweenunder and upperclassmen and participants in theexperimental or control condition at Time 1. The inter-action effect between group condition and class levelwas not statistically significant, F(1, 54) = 0.12, p = 0.74,η2p = 0.00. The simple main effects for group condition,F(1, 54) = 0.05, p = 0.83, η2p = 0.00, and class level, F(1,54) = 0.09, p = 0.77, η2p = 0.00, did not reach statisticalsignificance, suggesting the groups did not differ onstandard drinks consumed on heaviest drinking occa-sion at Time 1. Overall, underclassmen in the experi-mental condition reported a statistically significantlyreduction of standard drinks consumed from Time 1(M = 9.8, SD = 4.62) to Time 2 (M = 6.87, SD = 2.67),whereas underclassmen in the control conditionreported a non-significant increase in standard drinksconsumed from Time 1 to Time 2. There was nostatistically significant difference between the controland experimental condition among upperclassmen,from Time 1 to Time 2 (see Figure 2).

Average blood alcohol contentThere was a statistically significant three-way interac-tion between class level, experimental condition, andtime on peak BAC, F(1, 54) = 4.54, p = 0.04, η2p = 0.08.Follow-up analyses were conducted to disentangle theinteractions effect. Two mixed between-within subjects

2 × 2 ANOVAs were conducted first. There was nostatistically significant two-way interaction betweentime and group condition on peak BAC for underclass-men, F(1, 29) = 2.43, p = 0.13, η2p = 0.08 or upper-classmen, F(1, 25) = 2.11, p = 0.16, η2p = 0.08. Therewas a significant main effect for time for underclass-men only, F(1, 29) = 6.10, p = 0.02, η2p = 0.17. Therewas a statistically significant two-way interactionbetween class level and time on peak BAC for partici-pants in the experimental condition, F(1, 27) = 11.75,p < 0.01, η2p = 0.30, and a non-significant interactionfor participants in the control condition, F(1, 27) =0.05, p = 0.83, η2p = 0.00. Finally, a two-way ANOVAwas conducted to determine if there were any differ-ences on peak BAC between under and upperclassmenand participants in the experimental or control condi-tion at Time 1. The interaction effect between groupcondition and class level was not statistically significant,F(1, 54) = 0.05, p = 0.83, η2p = 0.00. The simple maineffects for group condition, F(1, 54) = 0.20, p = 0.66, η2p= 0.00 and class level, F(1, 54) = 0.69, p = 0.41, η2p =0.01, did not reach statistical significance, suggestingthe groups did not differ on peak BAC at Time 1.Overall, underclassmen in the experimental conditionreported a statistically significant reduction in peakBAC from Time 1 (M = 0.22, SD = 0.12) to Time 2(M = 0.14, SD = 0.08), whereas underclassmen in thecontrol condition reported a non-significant reductionin peak BAC from Time 1 to Time 2. There was nostatistically significant difference between the controland experimental condition among upperclassmen,from Time 1 to Time 2. However, upperclassmen inthe experimental condition reported a non-significantincrease in peak BAC from Time 1 to Time 2 (seeFigure 3).

Negative consequencesThe interaction effect between experimental conditionand class level was not statistically significant for nega-tive consequences, F(1, 54) = 0.71, p = 0.40, η2p = 0.01.The main effects for experimental condition, F(1, 54) =2.57, p = 0.12, η2p = 0.05 and class level, F(1, 54) = 2.58,p = 0.11, η2p = 0.05, did not reach statisticalsignificance.

Alcohol expectanciesThere was no statistically significant three-way interac-tion between class level, experimental condition, andtime on positive expectancies, when measured by theadapted AEQ, F(1, 54) = 0.91, p = 0.34, η2p = 0.02.There was no difference between the control andexperimental condition or upperclassmen and

6 A. SWEET STROHMAN ET AL.

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

underclassmen on positive alcohol expectancies, acrossthe two time periods.

Drinking normsWith regards to perceptions of abstinence (i.e. estimatesof how many students do not drink alcohol), there wasa statistically significant interaction between group andtime, F(1, 56) = 4.84, p = 0.03, η2p = 0.08. Participantsin the experimental condition reported more accurateestimates of abstinence at Time 2 (see Figure 4).Thedifference in estimates of abstinence was not moderatedby class level.

Discussion

The present study sought to evaluate the efficacy ofAlcohol-Wise; a widely used computer-delivered inter-vention for alcohol use among college students. Despite

its extensive usage, this is the first randomized controltrial to evaluate Alcohol-Wise. Research on computerdelivery of alcohol prevention has been inconsistent,potentially due to heterogeneity of such interventionsand the samples they are tested on. Given the use ofAlcohol-wise as a prevention tool among students of allclass levels, it was important to consider how juniorand seniors, who are legally ably to drink, respond tothe intervention compared to freshmen and sopho-mores. This study adds to a growing body of literatureon the efficacy of computer-delivered interventions forreducing the quantity of alcohol consumed by collegestudents.

Consistent with the first hypothesis, underclassmen inthe Alcohol-Wise experimental condition reporteddrinking fewer standard drinks on their heaviest drinkingoccasion at the 30-day follow-up, compared to under-classmen in the wait-list control condition. The lack of

Figure 3. Y-axis represents the estimated marginal means of peak BAC across Time 1 and Time 2 (X-axis), between underclassmenand upperclassmen. Figure is divided by group condition (i.e. experimental and control).

Figure 2. Y-axis represents the estimated marginal means of standard alcoholic drinks reported for heaviest drinking occasion acrossTime 1 and Time 2 (X-axis), between underclassmen and upperclassmen. Figure is divided by group condition (i.e. experimental andcontrol).

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 7

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

change in drinking behavior among upperclassmen, sug-gests that intervening early in a student’s college careerrather than later, when students have already establishedpatterns of alcohol use, may be an effective strategy tocombat alcohol misuse on campus. However, contrary toexpectations, there was no difference between the experi-mental condition and the control for either upperclass-men or underclassmen on the number of drinking daysthey reported, with both groups reporting drinking lessthan three days on average. Previous research (25,26)found that college students tend to drink less frequently,but in greater quantities than non-college students. Thelack of effect for number of drinking days might be dueto the restriction of range on how many days studentsdrink in a typical week.

Similar to the first finding, underclassmen in theexperimental condition reported statistically significantreductions from Time 1 to Time 2 on peak BAC. Thisreduction reflects a change from BAC levels that poten-tially produce an alcohol-related blackout, to less harm-ful BAC levels. Both findings on the intervention’seffectiveness at reducing quantity but not frequencysuggests that Alcohol-Wise may lead to lower rates ofbinge drinking, a particularly harmful drinking beha-vior. The reduction in BAC detected is both statisticallyand clinically significant, further supporting the poten-tial of this intervention to reduce the harmful effects ofheavy alcohol use.

Contrary to the second hypothesis, the interventionwas not successful at changing self-reported expectan-cies regarding alcohol use. Alcohol-related positiveexpectancies may be resistant to change because theydevelop over time and are maintained by peers.Alcohol-related positive expectancies may require amore extensive intervention than that offered byAlcohol-Wise. It is possible, however, that college stu-dents can modify their drinking habits without concur-rent change in expectancies.

Finally, the intervention appears to have increasedaccuracy of perceptions of normative drinking behavioramong college students. Students in the experimentalcondition reported more accurate perceptions of absti-nence after the intervention, compared to students inthe control condition. Interestingly, contrary to thethird hypothesis, this relationship was not moderatedby class level. After participating in Alcohol-Wise, stu-dents had a more realistic view of how many studentsabstain from drinking alcohol during college. Themajority of students underestimated the number ofpeople who do not drink, suggesting they perceivealcohol consumption as a normal part of the collegeexperience. The present study is consistent with find-ings that have found that college students consistentlyoverestimate rates of alcohol consumption among theirpeers (27).

The present findings also provide valuable informa-tion about tracking/self-monitoring approaches.Participants in the wait-list control conditions com-pleted assessment measures yet did not significantlyreduce their alcohol consumption relative to the experi-mental group. Thus, it appears that including alcoholeducation components does bestow additional benefitsbeyond self-monitoring.

Limitations and future research

Although the study utilized a randomized controldesign, the gold standard in treatment outcomeresearch, the pilot nature of the study limits its gen-eralizability and its use as comprehensive evidencefor clinical efficacy (28). Results from the currentstudy should also be interpreted in light of partici-pant attrition and sample size. In order to reduceattrition in future studies, in-person interventionmight be used instead of online. Future studiesshould also consider a longitudinal design, with mul-tiple assessments in fall and spring semesters. Thepresent study included a short-term follow-up assess-ment, which prevented it from detecting the long-term impact of the intervention. Furthermore, a 30-day follow-up may not be long enough to detect

Figure 4. Numbers on the Y-axis represent how far participants’estimates of abstinence are from the actual norms (i.e. 39%),with 0 representing an accurate response and negative num-bers representing underestimates of percent of students whoabstain from drinking at Time 1 and Time 2 (X-axis).

8 A. SWEET STROHMAN ET AL.

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

significant changes in alcohol-related negative conse-quences (e.g. GPA changes), regardless of the mea-sure. Future studies should consider tracking negativeconsequences between baseline and follow-up assess-ments, extending the follow-up period to at least onesemesters’ length and asking questions about morespecific consequences related to alcohol use.

According to Baer (29), there are few observationalor longitudinal studies of college student drinking.Instead most studies rely on self-report measures,which require insight and a willingness to truthfullydisclose thoughts, feelings, and behaviors. Due to thesensitive nature of the topic some individuals may findthis to be challenging. This study attempted to reducesuch experimental effects by delivering both the inter-vention and the assessment electronically. In addition,this study used implicit, as well as explicit, measures toavoid total reliance on self-report.

Although not statistically significant, it might bepossible that upperclassmen may increase their drink-ing as a result of participating in Alcohol-Wise. Thispossibility should be explored in future studies. This isespecially important since some colleges mandated allstudents to participate in Alcohol-Wise, regardless oftheir class level.

Conclusion

This study provides evidence for the short-term use-fulness of a computer-delivered intervention in redu-cing drinking among underclassmen. Moreinterestingly, it provides insight into when preventivemeasures are most effective in a college setting. Sinceonly freshmen and sophomores appeared to benefitfrom the intervention, the study suggests that pre-vention efforts should be concentrated in the firsttwo years of college. Moreover, the interventionincreased the accuracy of perceptions of drinkingnorms among college students, potentially ameliorat-ing peer pressure to increase drinking. These results,however, should be interpreted with caution in lightof the small sample size.

Acknowledgement

The authors would like to thank Dr. Alan Lincoln forhis feedback on an earlier version of this manuscript.

Declaration of interest

The authors report no conflicts of interest. The authorsalone are responsible for the content and writing of thispaper.

Funding

This study was supported by a grant from 3rdMillennium Classrooms producer of Alcohol-Wise.The grant was used by Ashleigh Sweet Strohman tosolely pay for the conduct of the study.

ORCID

Omar M. Alhassoon http://orcid.org/0000-0003-0596-6085

References

1. Jackson KM, Sher KJ, Park A. Drinking among collegestudents: consumption and consequences. In: GalanterM, Galanter M, editors. Alcohol problems in adoles-cents and young adults: epidemiology, neurobiology,prevention, and treatment. New York: SpringerScience & Business Media, 2006. pp 85–117.

2. National Center for Chronic Disease Prevention andHealth Promotion. Excessive alcohol use – addressing aleading risk for death, chronic disease and injury at aglance. Centers for Disease Control and Prevention,May 5, 2011.

3. Hingson R, White A. New research findings since the2007 Surgeon General’s call to action to prevent andreduce underage drinking: a review. J Stud AlcoholDrugs 2014;75:158–169.

4. United States Department of Education. Alcohol andOther Drug Prevention on College Campuses: ModelPrograms. Washington, DC; 2008.

5. Kypri K, Saunders JB, Gallagher SJ. Acceptability ofvarious brief intervention approaches for hazardousdrinking among university students. AlcoholAlcoholism 2003;38:626–628.

6. Epler AJ, Sher KJ, Loomis TB, O’Malley SS. Collegestudent receptiveness to various alcohol treatmentoptions. J Am Coll Health 2009;58:26–32.

7. Carey KB, Scott-Sheldon LAJ, Elliott JC, Garey L,Carey MP. Face-to-face versus computer-deliveredalcohol interventions for college drinkers: a meta-ana-lytic review, 1998 to 2010. Clin Psychol Rev2012;32:690–703.

8. Carey KB, Scott-Sheldon LAJ, Elliott JC, Bolles JR,Carey MP. Computer-delivered interventions to reducecollege student drinking: a meta-analysis. Addiction2009;104:1807–1819.

9. Doumas DM, Andersen LL. Reducing alcohol use infirst-year university students: evaluation of a web-based personalized feedback program. J Coll Couns2009;12:18–32.

10. Doumas DM, Kane CM, Navarro TB, Roman J.Decreasing heavy drinking in first-year students: eva-luation of a web-based personalized feedback programadministered during orientation. J Coll Couns2011;14:5–20.

11. Lane DJ, Lindemann DF, Schmidt JA. A comparison ofcomputer-assisted and self-management programs forreducing alcohol use among students in first yearexperience courses. J Drug Educ 2012;42:119–135.

THE AMERICAN JOURNAL OF DRUG AND ALCOHOL ABUSE 9

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6

12. Alfonso J. The role of social norms in personalizedalcohol feedback: a dismantling study with emergingadults. J Child Adolesc Subst Abuse 2015:1–8.

13. Alfonso J, Hall TV, Dunn ME. Feedback-based alcoholinterventions for mandated students: an effectivenessstudy of three modalities. Clin Psychol Psychot2013;20:411–423.

14. Hustad JTP, Barnett NP, Borsari B, Jackson KM. Web-based alcohol prevention for incoming college stu-dents: a randomized controlled trial. Addict Behav2010;35:183–189.

15. Walters ST, Vader AM, Harris TR. A controlled trial ofWeb-based feedback for heavy drinking college stu-dents. Preven Sci 2007;8:83–88.

16. Sobell LC, Sobell MB. Self-report issues in alcoholabuse: state of the art and future directions. BehavAssess 1990;12:77–90.

17. Brown SA, Christiansen BA, Goldman MS. TheAlcohol Expectancy Questionnaire: an instrument forthe assessment of adolescent and adult alcohol expec-tancies. J Stud Alcohol 1987;48:483–491.

18. Baer JS, Stacy A, Larimer M. Biases in the perception ofdrinking norms among college students. J Stud Alcohol1991;52:580–586.

19. Saunders JB, Aasland OG, Babor TF, de la Fuente JR,Grant M. Development of the Alcohol Use DisordersIdentification Test (AUDIT): WHO collaborative pro-ject on early detection of persons with harmful alcoholconsumption – II. Addiction 1993;88:791–804.

20. Aertgeerts B, Buntinx F, Bande-Knops J,Vandermeulen C, Roelants M, Ansoms S, Fevery J.The value of CAGE, CUGE, and AUDIT in screening

for alcohol abuse and dependence among college fresh-men. Alcohol Clin Exp Res 2000;24:53–57.

21. Fleming MF, Barry KL, MacDonald R. The alcohol usedisorders identification test (AUDIT) in a college sam-ple. Int J Addict 1991;26:1173–1185.

22. Lix LM, Keselman HJ. To trim or not to trim: tests oflocation equality under heteroscedasticity and nonnorm-ality. Educational Psycholog Measure 1998;58:409–429.

23. Keselman HJ, Algina J, Lix LM, Wilcox RR, DeeringKN. A generally robust approach for testing hypothesesand setting confidence intervals for effect sizes. PsycholMethods 2008;13:110–129.

24. Erceg-Hurn DM, Mirosevich VM. Modern robust statis-tical methods: an easy way to maximize the accuracy andpower of your research. Am Psychol 2008;63:591–601.

25. Johnston LD, O’Malley PM, Bachman JG, SchulenbergJE. Monitoring the future national survey results ondrug use, 1975–2010. Ann Arbor: Institute for SocialResearch; 2011.

26. O’Malley PM, Johnston LD. Epidemiology of alcoholand other drug use among American college students. JStud Alcohol 2002;(Suppl. 14):23–39.

27. Suls J, Green P. Pluralistic ignorance and college stu-dent perceptions of gender-specific alcohol norms.Health Psychol 2003;22:479–486.

28. Flay BR, Biglan A, Boruch RF, Castro FG, GottfredsonD, Kellam S, Mościcki EK, et al. Standards of evidence:criteria for efficacy, effectiveness and dissemination.Prev Sci 2005;6:151–175.

29. Baer JS. Student factors: understanding individual var-iation in college drinking. J Stud Alcohol 2002;(Suppl.14):40–53.

10 A. SWEET STROHMAN ET AL.

Dow

nloa

ded

by [

72.1

77.2

29.1

98]

at 1

3:28

23

Febr

uary

201

6