Guilt and Expressive Writing for Reducing Alcohol Use in College Students

Overview

This research seeks to evaluate expressive writing as a novel intervention for problem drinking among college students. The vast majority of individually focused brief interventions targeting college drinking have focused on personalized feedback approaches and recent innovations have largely been limited to finer distinctions of these, which require assessment and programming for implementation. The present research proposes expressive writing as a novel alternative, which has been used extensively in other domains but not as an alcohol intervention strategy. H1a: Participants writing about negative drinking events will show reduced drinking and drinking-related negative consequences relative to students in the neutral control group. H1b: Participants writing about distressing non-alcohol events will show increased psychological wellbeing relative to students in the neutral control group. H1c: Participants writing about negative drinking events will show reduced drinking and consequences compared with an empirically-supported brief intervention (i.e., PNF). This is an exploratory hypothesis. H2a: Alcohol narratives will have stronger effects on alcohol outcomes relative to distress narratives. H2b: Alcohol guilt narratives will have the strongest effect on alcohol outcomes relative to all other conditions. H3a: Expression of guilt, assessed by self-report and by content coding with LIWC, will mediate intervention effects on drinking outcomes. H3b: Change thought, assessed by LIWC coding, will mediate intervention effects on drinking.

The current research builds on previous research targeting heavy drinking among college students. A large volume of research has provided an impressive data base supporting one type of individually-focused alcohol intervention for this population (i.e., personalized feedback). The success of this paradigm has probably contributed to the dearth of consideration of alternative paradigms. Expressive writing is one such alternative, which has received extensive support in other domains but has only recently been considered as a potential intervention for heavy drinking. The preliminary data examining this approach is promising and provides a firm foundation for this efficacy trial. Further, this research incorporates novel theoretical constructs including the specific focus on guilt in expressive writing content as well as "change thought," as an analogue to the mechanism presumed to underlie motivational treatments for alcohol and other substance use disorders. The investigators plan to employ similar methods which have been used successfully in other large NIAAA funded trials evaluating brief interventions for heavy drinking college students. This research consists of an intervention study to evaluate expressive writing as a brief intervention in reducing drinking and improving psychological well-being among college students. Participation in the study involves completion of a screening assessment, a baseline assessment, the intervention procedure, post-intervention assessment, and follow-up assessments at 1-, 3-, 6-, and 12-months. Heavy drinking college students (N = 600) will be randomly assigned to one of six conditions based on the 2 (alcohol vs. distress topics) × 2 (guilt vs. no guilt focus) + 1 (neutral control) + 1 (personalized normative feedback) design. Before completing the baseline survey, students will be randomly assigned to one of six study conditions, five of which involve writing during three sessions over the course of one month. Specifically, participants will be assigned to write about a heavy drinking event, a heavy drinking event that elicited guilt, a distressing event, a distressing event that elicited guilt, or their first day of college (neutral control condition). Participants randomly assigned to the PNF condition will receive traditional personalized normative feedback regarding how their drinking compares with other students of the same gender at the university. The norms will come from a large recently completed alcohol survey conducted at the University of Houston examining social norms and alcohol prevention (R01AA014576). To maintain consistency across conditions, participants in the PNF condition will still come into the lab three times. They will receive feedback during the first intervention session and will be asked to complete the same narrative prompts as the neutral control condition for their second and third session. For individuals in the expressive writing conditions, there will be three narrative prompts to complete every week for three weeks, the first of which will occur following the baseline assessment. All baseline assessments, narrative intervention assignments, and immediate post-tests for all conditions will be conducted in-lab. All other assessments including screening and follow-up assessments will be completed remotely by web. The rationale for including a personalized normative feedback condition is to be able to compare the efficacy of expressive writing interventions with existing brief alcohol interventions. Thus, the present design allows not only for evaluation of efficacy relative to a control condition but also will evaluate comparative efficacy relative to an existing empirically-supported brief alcohol intervention. Aims will be evaluated using multi-level regression analyses, often referred to as Hierarchical Linear Modeling (HLM) or mixed-effects modeling. With respect to evaluating main effects of experimental conditions on drinking, each participant will provide baseline, post-intervention, 1-month, 3-month, 6-month, and 12-month follow-up data. Hypotheses will be tested using specific contrast vectors, using a general linear hypothesis framework. The study consists of a (2×2+1+1) design, represented as a factorial design with the addition of a control group that will write about their first day of school and a computer-based PNF comparison group. Hypotheses will be tested with contrasts corresponding to the questions of interest. The first two hypotheses represent contrasts between the alcohol narrative conditions and the neutral control condition (H1a) and between the guilt narrative conditions and the neutral control condition (H1b). In examining these hypotheses, the investigators will construct two dummy coded variables reflecting alcohol versus non-alcohol narratives conditions and between guilt and non-guilt narrative conditions with the reference group being the neutral control condition. Thus, the PNF group will not be included in the tests of these two hypotheses. Dependent variables will include alcohol outcomes for H1a and psychological well-being for H1b. For these analyses each participant will provide up to 5 repeated measures (i.e., baseline, 1-month, 3-months, 6-months, and 12-months), yielding up to 3000 Level 1 cases (repeated-measures) across 600 Level 2 cases. The investigators will also evaluate mediators of intervention effects. Additionally, the investigators will follow procedures to assess mediation. Mediation will test indirect effects using the AB products method where A will represent effects of intervention contrasts by time interactions on mediators (expression of guilt and change thought). B will represent the associations of mediators on subsequent drinking outcomes. Both A and B paths will control for baseline outcomes. Evaluation of hypotheses regarding the moderation effect will test whether individual differences in guilt-proneness interact with intervention contrasts. These will be tested by expanding the above model to add main effects and product terms of proposed moderators with intervention contrasts. Power analyses focus on estimating a sample size large enough to detect "true" effects, thereby avoiding Type II errors. Sample size estimates were obtained for intervention contrasts. Necessary sample sizes were assessed via sample size and power equations for normally distributed outcomes. Effect-sizes and variance components were based on preliminary studies, and power was set at 0.80 for all estimates. Power was estimated using the Optimal Design software program. The investigators anticipate intervention effects relative to the neutral control condition on drinking to be in the range of delta =.30-.40). Based on the proposed sample size of 500 (\~PNF not included in H1a and H1b), given five assessment points, the investigators anticipate the ability to detect main effects of intervention contrasts with power=.80. Considering maximum anticipated attrition rates of 20% the investigators will have .80 power to detect effects sizes of delta = .28 and greater.