i

Brief Intervention for Heavy-Drinking College Students:4-Year Follow-Up and Natural History

John S. Baer, PhD. Daniel R. Kivlahan, PhD, Arthur W. Blume, MS, Patrick McKnight, PhD, and G. Alan Marlatt, PhD

Lifetime consumption of alcohol typicallyreaches its highest levels during an individ-ual's late teens and early 20s.' Numerousstudies document a range of negative con-sequences of high levels of alcohol con-sumption, including violence, date rape, ac-cidents, academic problems, and familyconflict.''^ The college campus is one set-ting where the pattern of youthful heavydrinking is felt acutely. College students, onaverage, drink more than their noncoUegepeers of the same age^ and routinely reportnegative consequences from both their ownand others' drinking."* While heavy drinkingin college is associated with personality fac-tors such as impulsivity,^ noncomformity,^and depression, • •'' contextual factors suchas distance from parents, close associationwith peers, dormitory residence, associationwith fraternities and sororities, large-groupsocial events, and athletics appear to sup-port and exacerbate heavy drinking as

Heavy drinking in college should not beconilised with dependent drinking in laterlife, however. Several studies note significantreductions in heavy drinking in the 20s.'^~'^At the individual level, Schulenberg and col-leagues'^ followed the drinking patterns ofhigh school seniors for at least 6 years andnoted that a subset of individuals (17%) re-ported more than isolated patterns of heavydrinking over time. Thus, college administra*Uon and health officials, who are under in-creasing pressure to provide both preventiveand treatment services as a public healthservice for college students, are faced with amultifaceted social problem that is common,risky, and limited in time for most but chronicfor some.

Unfortunately, few interventions have adocumented positive impaet in changing col-lege drinking behavior In particular, com-monly offered educational programs have lit-tle impact,'" We recently reported on one

Objectives. This study examined long-term response to an individual preventive intervention for tiigh-risk college drinkers relative to the natural histoiy of college drinking.

Methods. A single-session, individualized preventive inteivention was evaluated within a random-ized controlled triai with college freshmen who reported drinking heavily while in high school. An addi-tional group randomly selected from the entire screening pooi provided a normative comparison. Par-ticipant self-report was assessed annually tor 4 years.

Results. High-risk controls showed secular trends for reduced drinking quantity and negative con-sequences without changes in drinking frequency. Those receiving the brief preventive intervention re-ported significant additional reductions, particularly with respect to negative consequences. Categori-cal individual change analyses show that remission is normative, and they suggest that participantsreceiving the brief intervention are more likely to improve and less likely to worsen regarding negativedrinking consequences.

Conclusions, Brief individual preventive interventions for high-risk college drinkers can achieve long-term benefits even in the context of maturational trends, (-4m ; Public Health. 2001;91:1310-1306)

model of indicated prevention, which involvespersonalized individual feedback and briefmotivational interventions for high-risk stu-dents during a single brief, nonconfrontationalcounseling session. "*~' ' Students who receivedthis preventive intervention reported signifi-cantly greater reductions in alcohol-relatedproblems at the 2-year foUow-up comparedwith a randomly assigned control group, '' ^

In the current analysis, we examine thenatural history of drinking patterns and re-lated problems over 4 years, within bothhigh-risk and normative samples. The preven-tive intervention is evaluated over this ex-tended period of time. Finally, using individ-ual classification of reliable change,^^ wedescribe developmental trajectories of drink-ing among college students who drank heav-ily in high school and describe rates of clini-cally significant change.

METHODS

Participants and RecruitmentAll students younger than 19 years who

had committed themselves to attending theUniversity of Washington in the fall of 1990

(n —4000) were mailed a questionnaire dur-ing the spring before matriculation. Forcompleting the questionnaire, students werepaid $5 and entered into a drawing for aprize. The questionnaire included itemsabout quantity and frequency of alcoholconsumed during a typical week and peakalcohol consumption in the past 3 months.^'The questionnaire also included the RutgersAlcohol Problem Inventory^ to assessdrinking-related consequences during theprevious 3 years. A total of 2041 students(51%) provided complete questionnairesand indicated a willingness to participate ina future study.

From the completed questionnaires, 508individuals were identified as being at "highrisk" by the following criteria; drinking atleast once a month and consuming 5 to 6drinks on at least 1 occasion in the lastmonth, or experiencing at least 3 negativeconsequences from drinking (Rutgers AlcoholProblem Inventory items) on 3 to 5 differentoccasions in the previous 3 years. An addi-tional normative comparison sample was se-lected randomly from the entire pool of re-sponders (n=151), including 33 persons who

1310 I Research Articies I Peer Reviewed | Baer et al. American Journal of Public Health | August 2001, Vol 91 , No. 8

met high-risk criteria, to track the natural his-tory ol' changes in drinking behavior withinthe cohort over time.

Potential participants were invited by letterand by phone. Participants were paid $25 forentering the study, completing the baselinequestionnaire and interview, and providing 2collateral contacts who could confirm the pai"-ticipant's rates of alcohol consumption. Ofthose contacted, 348 (68%) who were identi-fied as being at high risk agreed to participateand were randomized into either the inter-vention or no-intervention control groups;113 (75%) of those randomly selected for thenatural history compaiison agreed to partici-pate (28 of the 113 also met high-risk crite-ria). At baseline assessment in the autumn oftheir freshman year, high-risk students re-port:ed, on average, drinking more than 10drinks per week during Tewer than 2 drinkingoccasions, reaching a typical estimated peakblood alcohol level of 0.12%. Nonnative com-parison students reported diinking about 5V2drinks per week during 1 drinking occasion,reaching a typical estimated peak blood alco-hol level of O.OB'Vo. The high-risk sample was55% female and 84*'/(i White. The normativecomparison sample was 54"/(i female and78% White.

Confidence in the representativeness of thesample is potentially limited by the 50% re-sponse rate to the initial mailing. However,drinking rates within the normative compari-son sample are quite consistent with averagesnoted in national databases (e.g., Monitoringthe Future"'*). The level of drinking amonghigh-risk participants is more extreme than innafional averages; 79.5% reported drinkingat least 5 to 6 drinks on 1 occasion in thepast month at baseline, compared with about40"/() within the Monitoring the Future Study.

Measures/Ml participants used 6-point scales to rate

the quantity, frequency, and peak occasionsof their drinking (QFF ') and completed tlieDaily Drinking Questionnaire,^^ which asksabout the actual number of drinks fbr eachday of a typical drinking week, yielding drink-ing days per average week and averagedrijiks per' drinking day.

Ib assess negative drinking consequences,we asked partidpants to complete the Rutgers

Alcohol Problem hiventory.^^ a 23-item in-stmment designed to assess problem drinkingamong adolescents, and the Alcohol Depen-dency Scale," a well-established assessmentfor severity of drinking problems. At baselineonly, trained interviewers administered the al-cohol dependence questions from the Diag-nostic Interview Schedule,^' questions ondrinking patteiTis and consequenees torn theBrief Drinker Profile, ^ and interview sectionsto assess family histor}- of alcoholism and per-sonal histoiy of condnct problems. Otherquestionnaires assessed alcohol expectancies,psychiatric symptomatology, sti'ess, perceiveddrinking nonns, and sexual behavior.^' Re-sults from these assessments are not includedin this report.

Baseline and Follow-Up ProcedureAt baseline, all participants completed the

interview and questionnaire packet descrtbedabove. Follow-up assessments, which repeatedthe questionnaire measures included at base-line, were completed by mail annually (witlithe excepfion of an initial 6-monlh follow-updescribed previously^'). Assessments werecompleted, whenever possible, within themiddle of the autumn academic term to mini-mize the impact of well-known seasonal varia-tions (football season, spruig break, final ex-aminations, etc.). Participants were paid $20for each annual assessment. All collateral re-porters were phoned after each participant as-sessment, and collateral interviews were suc-cessfully completed for approximately 50* /0 ofsubjects at each assessment point These dataare not analyzeti I'or this report; previousanalyses at the 2-year follow-up demon-sti'atcd reasonable reliability with subjeets'self-reports, with no evidence of systematicover- or imderreporting of drinking.^' Collect-ing collateral data is thought to enhance thevalidity of participants' self-report in condi-tions of confidentiality.^"

Preventive InterventionFreshman participants who had been ran-

domly assigned to the intervention gi'oupwere contacted during the winter term andscheduled for an individualized feedback ses-sion. Participants were instructed to self-moni-tor drinking patterns 2 weeks before thescheduled session. Eight interviewers using a

written manual^" provided personal feedbackconcerning tlie consumption patterns as re-ported on diary cards and at baseline assess-ment. Rates of drinking were compared withnorms for same-age peers. Individualizedfeedback also ineluded informafion aboutperceived risks and benefits of drinking,mythology concerning ddnkijig behavior, thebiphasic effects of alcohol, and placebo andtolerance effects. The style of the interventionwas consistent with motivational inter-viewing ': client-centered in tone, but never-theless seeking to highlight and explore dis-crepancies between current behavior andplans, goals, and aspirations. Bach participantwas aiso given a 1 -page list of tips for reduc-ing risks associated with drinking.'"

During the winter term of their secondyear in college, participants randomized intothe prevention condition were mailed feed-back results comparing their drinking and itsconsequences with the norms of their coiiegepeers. The summary sheets contained bargraph results fi-om 3 assessments (baseline,spiing of the freshman year, and fall quarterof the sophomore year) and concluded with aparagraph that personalized the feedback foreach participant. After the mailing, we alsophoned prevention group participants in theliighest-risk categories (n = 56) to express con-cern about risk and offer additional feedbacksessions. Thirty-four motivational intei-viewswere conducted in the second year, most overthe phone (n = 26).

Missing DataTwo methods of processing missing data

vvere compared: lLstwise deletion and mulfipleimputation. Participants with missing data atthe 4-year follow-up were eliminated fromdata analyses. Missing data occurring at the1 -, 2-, or 3-yeai- assessments were replacedby a multiple imputation method using maxi-mum likelihood estimation.'"'''^ Sensitivityanalyses indicated no dilferences in parame-ter estimates between imputation of missingvalnes and list-wise deletion of cases withmissing data: therefore, data with imputedvalues were nsed for all subsecjuent analysesto preserve sample degrees of freedom.

Data PreparationAll outcome measures were standardized

with the nonnative comparison sample base-

August 2001, Vol 91. No. 8 I American Journal of Public Health Baeretal. I Peer Reviewed i Research Articles I 1311

line mean and standard deviation, yieldingrelative scores from a normative baselinewhose meaji equals 0 and whose standarddeviation equals 1. We also converted thestandardized outcome measures in eachdrinking domain to 3 unit-weighted factorscores by computing the mean of the stan-dardized relevant measured indicators'''':drinking frequency (a 6-point Likert scale ofdrinking frequency and Daily DrinkingQuestionnaire estimates of drinking days perweek), drinking quantity (2 6-point Likertscales measuring average and peak dnnkmgquantity, respectively, and the Daily Drink-ing Questionnaire measure of average drinksper drinking day), and negative conse-quences (Rutgers Alcohol Problem Inventoiyand Alcohol Dependency Scale). Althoughthe metnc of the origmal variables is not re-tained, these unit-weighted factors enhancethe reliability and validity of the dependentvai'iables and avoid conclusions based onsingle variables. "*

RESULTS

Success of RandomizationNo significant baseline differences were ob-

served between prevention and control condi-tions for alcohol consumption, related conse-

quences, or demographic and individual dif-ference factors. '

Longitudinai AttritionAt the 4-year follow-up, 363 of the 433

unique participants (84%) from the 2 high-risk groups and the normative comparisongroup completed assessments. Complete datasets at baseline and all 4 follow-up periodswere provided by 328 participants (76'^/o).and 346 participants (80%) provided data at4 of 5 time points, including year 4. Consis-tent with universi^ norms, 53% of the sam-ple reported having graduated at the 4-yearfollow-up.

Attrition rates at the 4-year assessment didnot differ significantly between high-risk andnonnative comparison groups, or betweenrandomized high-risk group conditions (pre-vention vs control). Attiition was not signifi-cantly related to participant's sex or to base-line quantity, frequency, or consequences ofdrinking.

Changes in Drinking and NegativeConsequences by Prevention Condition

Three distinct mixed model analyses"'provided hypothesis tests witb restrictedmeiximum likelihood parameter estimatesfor the effects of group (prevention and con-

trol) and time (baseline and 1-, 2-, 3-, and4-year follow-up) on the 3 major outcomefactor scores: frequency, quantity, and nega-tive consequences. 1 he mixed models in-cluded specifications for a priori contrastsfor adjacent time points (baseline vs 1 -yearfollow-up, 1-year vs 2-year follow-up, etc.),repeated-measures effects for time, time-by-treatment interactions, and random-effectsestimates for subjects. A priori contrasts oftime-by-treatment effects allow evaluation ofincreasing or decreasing differences be-tween experimental groups at the 3- or 4-year assessments.

Standardized mean differences from nor-mative baselines among high-risk participantsfor drinking frequency, drinking quantity, andnegative drinking consequences over tbe 4-year follow-up are listed In Table 1. Figure 1displays drinking rates and consequences forhigh-risk participants and normative compari-son samples. Significant main effects overtime were noted in all 3 analyses. Over 4years, the magnitude of change was greatestfor measures of negative drinking conse-quences (F^32,^45.65, P<.001), comparedwith tbose of drinking quantity (F^^zi ~28.22, P<.00\) and drinking frequency(F432j=7.58, P<.001), which demonstratedthe smallest effect.

TABLE l-Mean Standardized Factor Scores (SD) and Differences (SEiVi) Between High-Risk Prevention

Group and High-Risk Controi Group Over 4 Years: Preventive Intervention for Coliege Drinkers

Follow-Up

Drinking Factor

Frequency

Control

Prevention

Difference

Quantity

Control

Prevention

Difference

Negative consequences

Control

Prevention

Difference

Baseline

0.74 (0.88}

0,78 (0.88)

-0.04 (0.10)

0.73 (0.90)

0.91 (0.92)

-0.17 (0.10)

1.46 (1.27)

1.39(1.26)

0,06 (0.14)

lYear

0.80 (0.92)

0,60 (0,89)

0.20 (0.10)

0,76 (0.82)

0.60 (0.89)

0,15(0,10)

1,23 (1,37)

0,79 (1,24)

0,44 (0,15)**

2 Year

0.62 (0,84)

0.52 (0,88)

0.09 (0.10)

0.59 (0.85)

0.46 (0,93)

0,12 (0,10)

0.95(1.22)

0,53 (1.06)

0.41(0.13)**

3 Year

0.88(1.02)

0.75 (0.93)

0.13(0,11)

0.51 (0.81)

0.48 (0.84)

0.03 (0.09)

0.80 (1.26)

0.52 (1.10)

0.28(0.13)*

4 Year

0.71 (0.99)

0.64 (1.04)

0.06(0.12)

0,38(0.77)

0.27 (0.78)

0.10 (0.09)

0,72 (1.25)

0.40 (1.06)

0,31 (0.13)*'

Time Test F,,., GroupxTimeTestF,.

7.58'

28,22*

45.65'

1.26

4,33'

2.38*

Me. Factor scores reflect eievations in SO units relative to baseiine mean for the normative comparison grojp. All mean vaiues are significantly greater than 0.0. Difference score- controiscore minus prevention score, and difference deviations are standard errors as computed by SAS PROC iWIXED, ^ Rounding of factor scores creates siignt discrepancies in difference scores.Statisticai tests of difference scores represent bivariate tests where H^: T jn,, , "Tpi ^ ini ^ .*P< ,05 ; ' *P< ,01 ; ' * *P< .001 for aii tests.

1312 Research Articles | Peer Reviewed I Baeretal. American Journal of Public Health i August 2 0 0 1 , Vol 9 1 , No. 8

FIGURE 1-Standardized factor scores for 3 dimensions of drinking behavior over 4 years,

shown separateiy for the high-risk control ^oup ( • ) , the high-risk prevention group ( A ) ,

and the normative comparison group ( 4 ) .

Significant prevention group-by-time in-teractions over 4 years were observed withrespect to negative drinking consequences(F 32i ~2.38, P<.05) ajid for drinldng quan-tity (F432,=4.33, P<.00\). but not with re-spect to drinking frequency. Thus, from amultivariate perspective, drinking problemsdeclined significantly over time, and the pre-ventive intervention produced significant dif-ferences in alcohol use and related problemsover 4 years. Group-by-time effects (differen-tia) change) are shown in Table 1 to result insignificant group differences in negative con-sequences at all follow-up assessments.

A priori contrasts of belween-group differ-ences in change scores ( J, , complete caseanalysis, 2-tailed test) represent group-by-time interactions for each interval and thusreveal when treatment effects occur. Differ-ences in the magnitude of change betweenthe high-risk prevention and high-risk ccmtrolgroups from baseline to 1 -year follow-upwere evident for frequency (P—.03), quantity(P=.OOO2), and negative consequences(P=.OO95). All other adjacent time changecontrasts between groups were nonsignificant(P>.05). Thus, the prevention program ap-pears to have its primary effect between thebaseline and 1 -year assessments. Differencesnoted at the 3- and 4-year assessments canbe interpreted as a continuation of effectsnoted earlier in the follow-up period.

Trajectories of the normative comparisonsubjects over time are aiso evident in Figure 1.For normative participants (separate analysison all normative control group subjects),drinking ft'equency was significantly increasedrelative to baseline at the 3-year follow-up{(,,^-2.18, P<.05) and 4-year follow-up((j,2 = 2.51, P<.05). Drinking quantity showedminor vaiiation overtime, with the 2-yearassessment being significantly above baseline(/ii —2.18, P<.05). Mean negafive conse-quences of drinldng remained quite stablethrough the college years.

We reanalyzed the drinking patterns de-scribed above to evaluate the possible moder-ating effects of the participant's sex. Sex didnot significantly moderate the prevention ef-fects previously noted for drinking quantity ornegative drinking consequences. A 3-way in-teraction for drinldng Frequency (Fy ,, —2.22,P<.05) proved difficoilt to interpret. F xamina-

August 2001. Veil 91, No. 8 • American Journal of Public Health Baeretal. I Peer Reviewed I Research Articles ) 1313

tion of means suggested that there was greatervariabili^ over time among control groupn\ales and that, relative to the contj'ol condi-tion, the preventive intervention may have re-duced drinking frequency among women butnot among men.

Anaiysis of Individuai ChangeAverage changes in drinldng rates and prob-

lems may mask considerable variability in indi-vidual responses, and thus they do not providea measure of individual risk status and do notindicate how many cases of problem alcoholuse might be prevented as a result of this pre-vention program. To describe outcomes andnatural history in tenns reflective of intiividualrisk status for coUege studenis, we computedrisk cutpoints and measures of individualchange by using aii algoritlim reported else-where.' ' We chose to examine the Alcohol De-pendency Scale for the current analysis be-cause it best reflects the negativeconsequences of drinking (where the preven-tion program had Die greatest impact) anddoes not contain coUege-spedfic items {e.g.,slept late for a class) that are less relevant tothose who have quit school or graduated bythe 4-year follow-up. TTie risk cutpoint of 5 Forthe Alcohol Dependency Scale was establishedempirically to reflect the point at which a sub-ject is equally likely to be a member of thehigh-risk population or of tiie normative popu-lation containing no high-iisk participants (a"functional" comparison^^). As described in anearlier report,'^ we used an empirical-per-centiles approach to determine cutpointe andconfirmed the stability of estimates throughbootstrapping.'"' A reliable change index ' wasdefined as a difference from baseline scoresthat meets or exceeds 2 standard errors of theestimate of difference seorcs.

Participants were then categorized on thebasis of their baseline scores and 4-year fol-low-up scores. Owing to the multidimensionaldefinition of "high risk," not all high-risk par-ticipants were above the risk cutpoint for theAlcohol Dependency Scale at baseline. Simi-larly, the normative comparison group in-cluded some Individuals with scores aboveiTsk cutpoints at baseline. Participants wereclassified as "resolved" if their score beganabove the risk cutpoint, changed reliably(more than tlie reliable change index), and

TABLE 2-individual Change (%), Based on the Alcohol Dependency Scaie (ADS), for High-

Risk Participants and Normative Comparison Participants From Baseline to 4-Year

Assessment: Preventive Intervention for College Drinkers

ADS Score

Above cutpaint at baseline

Normafive ccmparison (n = 35)

High-risk control (n = 116)

High-risk prevention (n = 115)

Beicw CLtpoint at baseline

Normative ccmparison [n 61)

High-risk control (n = 27)

High-risk prevention (n-30)

New Case

22.9

14.8

10,t]

Reliably Worse

8.6

9.5

2.6

9.8

3,7

0,0

No Change

51.4

40.5

36.5

55.7

74.1

70.0

Reliably Improved

8.6

17.2

18.3

11.5

7.4

20.0

Resoived

31.4

32.8

42.6

ended below the cutpoint. Paitidpants wereclassified as "reliably improved" if their scorechanged reliably in a direction of fewer de-pendence symptoms but did not cross the cut-point. Participants were classified as "reliablyworse" it' their score moved reliably in the di-rection of more dependence symptoms butdici not cross from below to above the riskcutpoint. "New cases" represents those indi-viduals who became reliably worse over timeand crossed from below to above the risk cut-point. If change did not exceed the reliablechange index, participants were classified as"no change."

Table 2 presents change categories, basedon the Alcohol Dependency Scaie, from base-line to 4-year follow-up for all participants.Within the normative comparison group,which represents tlie population of studentsa( this university and thus includes somehigh-risk participants, 22.9% of students withscores below the risk cutpoints at baselinebecame new cases; the modal trajectory wasno change (SS.V/ii). Among the 35 nonna-tive comparison students whose scores wereabove a risk cutpoint at baseline, 8.6'Vo reli-ably improved and 31.4% were resolved atthe 4-year follow-up. Trends for high-riskparticipants as well as preventive interven-tion effects can also be observed in Table 2.For example, aniong high-risk participantswith Alcohol Dependency Scale scores abovethe risk cutjioint at baseline, 32.8% of thosein the control condition were resolved at the4-year follow-up, indicating that roughly athird of those who drink heavily in highschool resolve drinking risk over the course

of 4 yeai s. Tliis rate improved to 42.6%among those in the prevention condition.Smprisingly, among high-risk participantswith Alcohol Dependency Scale scores belowthe risk cutpoint at baseline, 18.5% of thecontrol gi oup were reliably worse or werenew cases 4 years later, compared with 10%of the prevention gi oup. More generalizedcomparisons also follow from Table 2.Among high-risk participants, 67% of theprevention group had good outcomes over 4years [resolved, reliably improved, or nochange from a baseline score below lhe riskcutpoint), compared with 55%i of controls.

DISCUSSION

Cenh"al findings suggest that much heavydrinking among college students is ti"ansitory,despite some students who report a pattern ofcontinued or worsening consequences overtime. Changes in drinldng ai"e reflected in spe-cific dimensions of drinking behavior. Com-pared with a high-risk control sample, in thisrandomized triai, paiticipants receiving a briefindividual preventive intervention had signifi-cantly greater I'cductions in negative conse-quences that persisted over a 4-year period.Individual change analyses suggest that thedependence symptoms of those receiving thebrief intervention are more likely to decreaseand less likely to increase.

To understand the impact of the preventiveintervention in a broader context, we reportthe natural history of different dimensions ofdiinking through the paiticipants' 4 years ofcoUege. The frequency of drinking did not

1314 I Research Articles j Peer Reviewed : Baeretal. American Journal of Public Health | August 2001, Vol 91 , No. 8

change dramatically over 4 years for eitherhigh-risk or normative samples. Normativecomparison participants, representing the gen-eral student body, reported slight increases indrinking frequency over time, particularly atthe 3-year follow-up, when many in the sam-ple bad reached 21 years of age."*® Drinkingfrequency among bigb-risk students declinedonly minimally over the 4-year period.

Meati drinking quantity and negative con-sequences increased only marginally withinour normative comparison group, suggestingtbat students in general do not commonly orroutinely develop drinking problems duringthe college years.'''^'' However, drinking quan-tity and associated problems declined steadilyover time for high-risk students who enteredcollege witb a histoiy of heavy drinking. Re-cruitment of high-risk students was based ondrinking during high school, and thus bothdevelopmental and statistical trends probablymove toward less extreme behavior (but thesesame individuals reported mean increases indrinking upon entry into college"'' ). Fouryears after matriculation, our high-risk controlstudents continued to drink more frequentlytlian tbe normative comparison students, buttheir problem scores had markedly decreasedand, although still above those of the compar-ison group, were much less elevated (stan-dardized difference about 0.62) than theywere at baseline (standardized differenceabout 1.40). This pattern of data is consistentwith other longitudinal studies showing tbatadolescents with problem behaviors (includ-ing heav - drinking) remain less conventionalthan others as they age into adulthood, butdo not have worse psychosodal adjustment.'^

Prevention effects were observed for onlysome dimensions of drinking behavior. Thedimension of greatest interest, negative conse-quences of drinking, shows the greatest ef-fects. This is important not only because neg-ative consequences measure the degree towhich indi\'iduais may be harmed as a resultof drinking, but also because the preventiveintervention targeted individual choices andreduction of risk, rather than drinking ratesper se. Our fmdings are consistent with tbegoals of harm reduction interventions,"' ap-proaches that fociLs on minimizing tbe harm-ful cflects of high-risk drinking. The durationof our prevention effects is also noteworthy.

Modest dilTerential changes in drinking quan-tity and frequency, described in our earlierreport of 2-year outcomes,^' do not appear topersist for longer periods of time, yet wefound significantly reduced negative conse-quences 3 '/2 years after the preventive inter-vention. We are unaware of other studies(much less randomized trials) of preventionefforts among college students that demon-strate such long-lasting effects.

Our analyses of individual change suggestthat, regardless of baseline risk status, about 1in 3 college students, as represented by thenormative comparison participants, reliablyworsen during tbe college years. Anotherthird do not change, and a third reliably im-prove. Among bigh-risk samples, many morestudents reliably improve than worsen, andtbose with scores beginning below risk cut-points most often report no change in drink-ing status. Group differences in rates of reli-able worsening and improvement amonghigh-risk participants suggest that feedbackand advice may function by accelerating anormative developmental process of reducingdrinking, as well as by slowing a less typicaldevelopmental process that may otherwiselead to an escalation of drinking for some in-dividuals during college years.

The results described in this report should,of course, be interpreted with attention to tbeinberent limitations of the research method.The study included students fT"om only onelarge public university. Given variability indrinking across educational settings poten-tially based on size, private fimding, and entrycriteria,"" our results may not be generalizableto all other student populations. Assessmentwas based on self-report, which could result ininaccurate or socially desirable reporting. Self-report of alcohol use, however, has beenshown to be quite accurate in many contextswhere no penalties exist for specific re-sponses. ^ Self-report indices could also reflectstudents' desire to please researchers, andthose receiving the feedback may experiencea greater demand for such reporting. Our ex-perimental design did not include an atten-tion-only control to test tbis effect. However, itis difficult to conclude that response biaseswould result from such a limited contact (1hour during freshman year and mailed feed-back during sophomore year), would extend

for 4 years, and would produce treatment ef-fects in one dimension but not another. Also,participants were reminded that collateral re-ports were obtained at each assessment pointto encourage honest responding.

Although our data are consistent with abroad literature showing that brief interven-tions are effective in reducing alcohol anddnjg use,"*" to date, we do not know moreprecisely how these interventions work (ourdata suggest that developmental processes un-derlying drinking reduction and drinking es-calation might be affected). Nor do we knowthe critical components for content and deliv-ery of the preventive intervention. Is the pri-mary component merely increased attentionto the issue? Can interventions based on thiscomponent be conducted by peers in groupsor witb only mailed feedback? Continued re-search on these types of issues vrill facilitatethe adoption of effective prevention program-ming and the continued reduction in harmbased on youthful heavy drinking. •

About the AuthorsJohn S. Beier and Daniel R. Kivlahan are, and at the timeof the study Patrick McKnighl was, with the Center of Ex-cellence in Substance Abuse Treatment and Education, VAPuget Sound Health Care System, Seattle, Wash, fohn S.Boer, Arthur W. Blume, and G. Alan Marlatt are with theDepartment of Psychology, and Daniel R. Kivlahan is withthe Department of Psychiatry and Behavioral Sciences,University of Washington, Seattle.

Requests for reprints should be sent to fohn S. Baer. PhD,SJ16-ATC VA Puget Sound Health Core System, 1660 SColumbian Way, Seattle, WA 98108 (e-mail: jshaer@u. wash ington. edu).

This article was accepted October 4, 2000.

ContributorsJS. IJaer. D.R. Kivlahan, and G. A. Marlatt designed thestudy, obtained funding, and .supervised data collection.AW. Blume assisted in data inanagenient, conducteddata analyses, and \\Tote parts of the "Melliods"' section.P. McKnight conducted mixed model and individualchange data analyses and wrote parts of the "ResiJts"section. J. S. Baer wrote the initial draft of all other sec-tions. D.R. Kivlahan served as primary editor. All au-thors contributed to editing and revision.

AcknowledgmentsThis rpsearch was supported by grant R37 AAO5591from the National Institiitr: on /Mcoliol Abuse and Alco-holism.

We gratefully acknowledge the suppott and assis-tance in the preparation of this report of Ihad Leffing-well, Maxine Pollock, Dan Neal, and Lisa Roberts. Othersproviding essential support for the ongoing managementof the research project include Maiy Lariiner, JasonKilmer, Lori Quigley, £ind Ken Weingardt.

August 2001, Vol 91, No. 8 I American Journal of Public Health Baeretal. I Peer Reviewed ( Research Articles I 1315

References

1. Ninth Special Report to the US Confess on Alcoholand Health. Washiti^on, DC: US Dept of Heallh aridHiiniaii Services; 1997.

2. Institute of Medicine. Broadening tke Base of Treat-ment for Alcohol Problems. Washington, DC: NatitmalAcademy Press; 1990.

3. Schulenberg J, Maggs JL, Long SW et al. 'ITiepioblem of college drinking: irisightK from a develop-nieiital perspective. Aicohoi Clin and Exp Res.2001;25:473-477.

4. Weclisler H, Davenport A, Dowdall G, M(jeykensB. Castillo S. Health and behavioral consequences olbinge drinking in college. A national sun'ey of studentsat 140 campuses.yAM4. iy94;272:lf)72-ie77.

5. Camalta CD, Nagoshi CT. Siress, depression, iira-tioiial beliefs, and alcohol use and problems in a col-lege student sample. Alcohol Clin Exp Res. 1995:19:142-146,

6. Havey JM, DiKld DK. Variahles associated with al-cohol abuse among sclf-idcTitified collegiate COAs andtheir peers. Addict Behiw. 1993;18:567-575.

7 Brennaii A!-, VValfish S, AuBiiclioii P. Alcohol useanci ahuse in college students, I: a r'eview ol' individualand personality correlates. IntJAddict. I986;21:449-474.

8. Kushiier MG, Shcr KJ, [•;ricl(son DJ. Prospectiveanalysis ol' die reladon between DSM-Ill aiixiely disor-ders and alcohol uf;e disorders. AmJ Psychiatry. 1999;

9. Cashin JR, Presley CA, Meihnan PW. Alcohol usein the Greek system: follow the leadef! J Stud Alcohol.1998;59:63~70.

10. Gfi-uerer JC, Greenblatt JC, Wright DA. Suhstaiicetise in tlie US college-age populatioii: differences ac-coiding to educational statiLS and living arrangement.AmJPuhlicIlealth. lW97;87:fj2-B:i.

tt. Leichtiter JS, Meilman PW, Presley CA, Cashin JR.Alcohol use and related consequences among studentswiili varying levels of itivolvemenl in college athletics.JAm CollHealth. 1998;46:257-262.

12. Rosenhluth J, Nathan PE, Lawsori DM. linvirovi-[nental influences on drinking hy college students in acollege pub: a behavioral ohaervation in the natural en-vironment. Addict Behav. I978;3:117-]21.

KJ. Wechsler 11. Dowdall GW, Davenport A, CastilloS. Correlates of coliege sttident binge drinking. AmfPublic Health, 1995;85:921-926.

14. Johnston LD, O'Malley PM, Bachman JG, eds. Na-tional Survey Results on Drug Use from the Monitoringthe Future. Study, 1975-1995. Vol I: Secondary SchoolStudents. Kockville, Mel: National Institute on DrugAbiLse; 1995. NIH puhtjcation 95-402(S.

15. Donovan JE, Jessor K, Jessor L. Problem drinkingin adolesceTiee and young adtslthootl; a follow-upstudy. J Stud AktJhol. 1983;44:tO9-37

16. Jessor RJ, Donovan JE, Costa EM. Beyond Adoles-cence: Problem Behavior and Young Adult Development.Cambridge, England: Cambiidge University I'ress,t99t.

17 Sdiulenheig J, O'Malley PM, BadnnaiiJC.Wacisworth KN. Johnston LD. Getting drunk and grow-

ing up: Oajertories of frequent binge drinking tluringtlie transition to young adulthood. / Stud Aicohoi.1996;57:289-304.

18. Moskowitz JM. The primaiy prevention of alcoholproblems: a critical review of the I'esearch liferaltne.J Stud Alcohol. 1989;50:54-88.

t9. I3aer' JS. Etiology ajid secondary prevention of alco-hol piobleuis with votmg adults, tn: Baer JS, Mariatt GA,McMahon RJ, eds. Addictive Behaviors Across the Lifespan.Newbtiry Park, Calif: Sage Publicadons; 1993:111-137

20. Dimelf LA, Baer JS, Kivlahan DR. Mariatt GA.Brief Alcohol Screening and Intervention for College Stu-dents: A Harm Reduction Approach. New York, NY:Otiilford Press; 1999.

21. Mariatt GA, Baer JS, Kivlalian Dfi, et al. Sc:reeningand brief intervention for liigh-risk college studentdrinkers: results from a two-year follow-up assessment.J Comult Clin Psychol. I998;66:(JO4-615.

22. Roberts LJ, Neal DJ, Kivlalian DR, Baer JS. Mar-iatt GA. Individual drinking changes following a briefintervention among college students: clinical signiii-cance in an indicated preventive context. / Consult ClinPsychol. 2OOO;68:5OO-5O5.

23. White HR, Lahouvie EW. Towards the assessmentof adolescent pioblem drinking. J Stud Alcohol. 1989;50:30-37

24. Johnston LD, O'Malley PM, Bachman JG, eds. Na-tional Survey Results on Drug Use From the Monitoringthe Future Study, 1975-1995, Vol 2: College Studentsand Young AdulU. Rockville, Md: National instutite onDrug Abuse; 1999. NIH publication 99-4661.

25. Collins RL, Parks GA, Mariatt GA. Social detemii-riants of alcohol consumption: tlie elfects of social in-teraclion and model stattis on the selt-adniinistradon ofalcohol. J Consult Clin Psychol. 1985;53: t89-200.

26. Skimier HA, Honi JL. Alcohol Dependency Scale(ADS). Ibronto, Ontario: Addiction Research iTOnda-tion; 1984.

27. HelzerJIi, Robins LN. The Diagnostic InterviewSchedule: its development, evoludon, anti use. SocPsy-chiatry Psychiatr Epidemiol. 1988;23:6-16.

28. Miller WR, Mariatt GA. Brief Drinher Profile.Odessa, Fla: Psychological Assessment Resources;1984.

29. Murray DM, Peiry CL. Measurement of substanceuse among adolescents; when is the "bogus pipeline"method needed? AtWirt .Seteu. 1987;12:225-233.

30. Miller WR, Rollnick S. Motivational Interviewing:Preparing People for Change. New York, NY: GtiilfordPress; 1991.

31. Litde RJA, Rubin DB. Statistical Aimit/sis WithMissing Data. New York, NY: John Wiley & Sons; 1987

32. Schaefer JL. Analysis of Incomplete MultivariateData. London, England; Chapman & Hall; 1997.

33. Gorsuch R, Factor Analysis. 2nd ed. Hillsdale, NJ;LawTence t'^rlbaum; t983.

37. Jacobson NS, 'I'raux P. Clinical signiticance: a sta-tistical approach to definiiig meanlngftil change In psy-chotherapy research, j Consult Clin Psychol, 1991 ;59:12-19.

38. Baer JS, Mariatt GA, Kivlahan DR, Fnmime K,Larimer M, Williants E. An experimental test of threemethods of alcohol (isk (eduction witli young adults.J Consult Clin Psychol. 1992;60;974-979.

3 9. Uaer JS, Kivlahan DR, Mariatt GA. High-riskdrinking across tlie transition Trorn high school to col-lege, Alcohol Clin Exp Res. 1995;I9:54-61.

40. Mariatt GA. Harm Reduction: Pragmatic Strategiesfor Managing High-Risk Behaviors, New York, NY: Guil-foid Press; 1998.

41. Wechsler H. Moinar BE. Davenport AE, Baer JS.College alcohol use: a liill or empty glass?y.^m C(dlHealth. 1999;47;247-252.

42. Bahor TF, Stephens RS, Mailatt GA. Verbal reportmethods in ciinicai research (jn alcoholism: responsebias and its minimization./Sturf/lfcoAo;. 1987;48;410-424.

43. Wilk A!. Jensen NM. Havighiust "!C. Meta-analysisoi' randomized control trials addressing brief interven-dons in heavy alcohol drinkei^s.y Gen Intern Med.1997;12;274-28:i.

34. Shadish WR. Pkinned cridcal muMplism; someelaboradons. Behav ,'lssess. 1986;8;75-103.

35. Litde RC, Miilken GA, Stoup WW, Woffinger RO.SAS System fm- Mixed Models. Cary. NC: SAS Institute;1996.

36. Efron B, Tibshiiasii RJ. An introduction to Boot-strap. London. England; Chapman & Hail; 1993.

1316 I Research Articles I Peer Reviewed I Baeretal. American Journal of Public Health | August 2U01, Vol 91 , No. 8