Baseline and Attrition Analyses
Characteristics of participants at baseline by treatment group are shown in . No significant differences were found on any of the socio-demographic, substance use, or other health behavior measures between groups. Sixteen participants were lost to attrition (5%), with no differences in attrition between treatment groups. Significantly more students who dropped out of the study received mostly B grades (rather than A grades) on their last report card (X2(3)=18.83, p=.001), reported a family alcohol or drug problem (X2(3)=6.53, p=.01), and used marijuana in the past 30 days (X2(3)=4.07, p=.04), than those who did not drop out.
Response and Intervention Implementation Fidelity
To determine the likelihood of participants responding to questions on the outcome survey in a socially desirable manner, students were asked about their willingness to provide honest answers to questions about their alcohol and drug use and other health habits. At baseline, 92.6% strongly agreed and 7.4% agreed that they were willing to give honest answers to questions on the survey, with none disagreeing or strongly disagreeing, indicating little probable influence of social desirability. In addition, to estimate the extent to which responses may have been unreliable due to participant lying or other factors, we included a bogus/fake drug (i.e., zanatel) among the list of substances that students were asked whether they used in the past 30-days. No one reported using the bogus drug, suggesting that widespread error due to lying or sloppy completion of the data collection instrument was unlikely.
To assess implementation fidelity, we collected feedback from participants immediately after administration of each intervention using a computer based, self-administered questionnaire. These data showed that participants who received the consultation and goal plan rated the intervention significantly better than those who received the standard care control on eight of nine measures of acceptability and potential efficacy, p's<.05.
Estimated marginal means and standard errors of health behavior measures are shown by group and time in . Omnibus repeated measures MANOVAs were performed for six groupings of health behavior measures. These analyses were significant for group by time interaction on alcohol consumption, F(4,278)=3.42, p=.01, marijuana use, F(5,277)=2.75, p=.02, and health-related quality of life, F(5,277)=2.87, p=.02.
Repeated measures MANOVAs of health behavior measures for time by group
Univariate repeated measures tests showed college students exposed to the brief intervention drank alcohol less frequently, F(1,281)=7.47, p=.01, as well as drank heavily less often, F(1,281)=9.54, p=.00, whereas students receiving the standard care control increased their alcohol use frequency and heavy use over time. The intervention group also used marijuana for a shorter length of time, F(1,281)=5.67, p=.02, used less quantity of marijuana, F(1,281)=4.97, p=.03, and used marijuana heavily less often, F(1,281)=5.98, p=.02, while the control group showed increases in these three measures of marijuana use over time. In addition, brief intervention participants experienced fewer days in which their spiritual health was not good, F(1,281)=6.90, p=.01, and fewer days in which their social health was not good, F(1,281)=9.55, p=.00, compared to control participants. While no omnibus group by time interactions were found for exercise, a univariate group by time interaction was found for 30-day moderate exercise, with brief intervention participants showing an increase and control participants a decrease in moderate exercise in the past 30 days, F(1,281)=4.73, p=.03. In addition, ANOVAs were performed for another four single measure health behaviors. Participants in the intervention group got more sleep, F(1,281)=9.49, p=.00, and drove less after drinking alcohol, F(1,266)=5.25, p=.02 than those in the control group.
MANOVAs also indicated significant time effects for exercise, F(6,276)=2.94, p=.01, nutrition, F(3,279)=3.97, p=.01, and health-related quality of life, F(5,277)=4.43, p=.00. Univariate tests showed increases in 30-day vigorous exercise for participants in both groups, F(1,281)=5.96, p=.02, and increases in the consumption of healthy carbohydrates, but just among control participants, F(1,281)=9.88, p=.00, and healthy fats, but primarily among those participants receiving the intervention, F(1,281)=4.14, p=.04. In addition, students in both treatment groups had fewer days in which their physical health was not good, F(1,281)=8.30, p=.00, fewer days in which their mental health was not good, F(1,281)=16.64, p=.00, and fewer days in which poor health kept them from conducting their usual activities, F(1,281)=4.63, p=.03. Finally, ANOVAs for single measure behaviors indicated significant time effects with fewer alcohol/drug problems among participants in both groups, F(1,179)=15.22, p=.00, increased number of hours of sleep each night for participants in both groups, F(1,281)=27.51, p=.00, and increased use of stress management techniques for participants in both groups, F(1,281)=63.25, p=.00.
Effect sizes were calculated for univariate tests within treatment groups. These effect sizes were generally small, with some approaching medium size. Small effects were found for the brief intervention on alcohol and marijuana behaviors, with reductions on alcohol and marijuana among brief intervention participants that paralleled equal size increases in consumption in the control group. Small effects were also found for the intervention group on reduced driving after drinking, and increased vigorous and moderate exercise. Small effect sizes were found for brief intervention participants on increasing two nutrition habits (i.e., eating healthy carbohydrates and fats), and improving all five measures of health-related quality of life, with the greatest improvements on spiritual, social, and mental health. Larger effects were also found for the intervention group on reductions in alcohol/drug problems, with effects approaching medium size on increases in sleep and stress management. On measures shown to significantly improve over time for both treatment groups, effect sizes were generally two to four times larger for brief intervention participants than for control students.