A Randomized Pilot Trial of a Mobile Delivered Brief Motivational Interviewing and Behavioral Economic Alcohol Intervention for Emerging Adults

Abstract

Objective:

Mobile health, or mHealth, interventions show potential to broaden the reach of efficacious alcohol brief motivational interventions (BMIs). However, efficacy is mixed and may be limited by low participant attention and engagement. The current study examined the feasibility, acceptability, and preliminary efficacy of a live text-message delivered BMI in a pilot randomized clinical trial.

Methods:

Participants were 66 college students (63.6% women; 61.9% white; M_age = 19.95, SD = 1.66) reporting an average of 11.88 (SD = 8.74) drinks per week, 4.42 (SD = 3.59) heavy drinking episodes, and 8.44 (SD = 5.62) alcohol-related problems in the past month. Participants were randomized to receive either 1) education or 2) an alcohol BMI plus behavioral economic Substance-Free Activity Session (SFAS), each followed by four weeks of mini sessions. All sessions were administered via live text-message. Participants completed assessments post intervention (after the 4^th mini session) and at 3-month follow-up.

Results:

90.9% completed both initial full-length sessions and at least two of the four mini sessions with 87.9% retention at 3-month follow-up. Participants found the interventions useful, interesting, relevant, and effective, with no between-group differences. There were no statistically significant group differences in drinks per week or alcohol-related problems at follow-up, but BMI + SFAS participants reported fewer past-month heavy drinking episodes than those who received education.

Conclusions:

Live text-messaging to deliver the BMI + SFAS is feasible and well-received. The preliminary efficacy results should be interpreted cautiously due to the small sample size but support further investigation.

Heavy episodic alcohol use (4/5+ standard drinks for women/men respectively) is a significant public health concern that typically peaks between the ages of 18 and 25 (Johnston et al., 2016; Schulenberg et al., 2018). Heavy episodic drinking among emerging adults is associated with a range of problems (e.g., hangovers, blackouts, repercussions of a DUI; Barnett et al., 2014; Hingson et al., 2017; Wilhite & Fromme, 2015), and frequent heavy episodic drinking can interfere with achieving important developmental milestones such as graduating college and developing a career (Jennison, 2004; Ligouri & Lonbaken, 2015). Further, national data show that emerging adults between the ages of 18 and 25 have the highest prevalence of a past-year alcohol use disorder (AUD; 15.6%) (Substance Abuse and Mental Health Services Administration, 2021).

In-Person Interventions for Emerging Adult Alcohol Misuse

In-person brief motivational interventions (BMIs) are a widely used approach to address risky alcohol use (Cronce et al., 2018), and are recognized as a Tier 1¹ approach for prevention of harmful alcohol use in college students by the National Institute of Alcohol Abuse and Alcoholism (NIAAA, 2015, 2019). There have also been efforts to enhance the efficacy of BMIs by adding supplemental content (Murphy et al., 2022), such as the in-person Substance-Free Activity Session (SFAS; Dennhardt et al., 2015; Murphy et al., 2019). The SFAS is a 45–50 minute behavioral economic-informed (Bickel et al., 2014) intervention supplement to a standard in-person alcohol BMI (e.g., personalized normative feedback on use and problems and provision of strategies for safer drinking). Whereas alcohol BMIs focus on motivating reductions of harmful alcohol use by highlighting the risks associated with drinking, the SFAS attempts to indirectly motivate reductions in alcohol use by increasing future orientation and motivation to engage in goal-directed and enjoyable substance-free activities that are associated with future goals. Two randomized controlled trials with college students who reported heavy drinking suggest that the in-person BMI + SFAS is associated with significant and enduring reductions in alcohol use and problems relative to control, with changes in drinking mediated by increased engagement in substance-free activities and use of protective behavioral strategies (Murphy, Dennhardt, et al., 2012; Murphy et al., 2019). Thus, there is a need to identify ways to expand the reach of BMIs including the SFAS.

Mobile Health Interventions for Emerging Adult Alcohol Misuse

Despite substantial growth in the use of mobile health (mHealth) in recent years, evidence for the efficacy of alcohol mHealth interventions is mixed (Hutton et al., 2020; Kaner et al., 2017; Kazemi et al., 2017). Only half of the studies in a recent systematic review of alcohol mHealth interventions for young people (ages 12–26) by Hutton et al. (2020) reported some effectiveness in reducing alcohol consumption. Additionally, studies that have evaluated “event-specific” automated text-message interventions have shown efficacy in altering perceptions about normative drinking at events such as 21^st birthdays or football tailgates, but have not been associated with reductions in alcohol use or problems (Bernstein et al., 2018; Cadigan et al., 2019).

mHealth interventions for alcohol use have largely taken the form of automated unidirectional text-messaging (Bock et al., 2016) or automated bidirectional messaging with a bank of pre-formulated, semi-tailored responses (Suffoletto et al., 2015, 2016; Witkiewitz et al., 2014). Many of these studies report good feasibility and acceptability, suggesting the substantial promise of this low-cost and convenient modality in reaching heavy drinking emerging adults. However, fully automated approaches do not allow for interpersonal dialogue with a therapist (e.g., unique reflections, open-ended questions, tailored suggestions, or responses to participant questions) that might facilitate greater processing of the personalized feedback as well as more effective motivational enhancement and goal setting.

Although non-treatment-seeking emerging adults are more likely to agree to participate in mHealth interventions than face-to-face interventions (Neighbors et al., 2018), many intervention participants report attending to other activities when attention to intervention content was required (Lewis & Neighbors, 2015), suggesting that the intervention dose may have been diluted. Indeed, a review of six studies examining in-person and mHealth brief interventions found that automated mHealth interventions were less effective in reducing alcohol-related outcomes than in-person delivered interventions (Rodriguez et al., 2015). Several factors, such as participant distractibility (Lewis & Neighbors, 2015), low participant diligence and conscientiousness in responding due to lower levels of perceived intervention “monitoring” (Rodriguez et al., 2015) and trust or rapport (Elliott et al., 2008), are hypothesized to contribute to this relative lack of efficacy. Importantly, one study found that a computerized BMI that required participants to actively recall intervention content was more efficacious than a computerized BMI that did not require this extra processing in part because of greater information retention at follow-up (Jouriles et al., 2010). Thus, continuous and stimulating interactivity with a session facilitator may enhance mHealth intervention efficacy.

Counselor Administered Synchronous Text Messaging (CoAST)

We define counselor-administered synchronous text-messaging (CoAST) as real-time, guided delivery of a circumscribed intervention by a counselor to a participant through a text-messaging or instant-messaging platform within a specified timeframe or session, similar to an in-person counseling session except delivered via text. A systematic review by Hoermann et al. (2017) found that synchronous text-based dialogue systems are generally associated with positive treatment gains, equivalent but not superior to treatment-as-usual (e.g., face-to-face, telephone), for various mental/behavioral health conditions, including AUD in adults. To our knowledge, however, only one other published study has examined a CoAST delivered brief intervention for alcohol-related outcomes in a non-treatment-seeking sample. In a small sample randomized pilot trial, Teeters et al. (2018) found that a CoAST alcohol-impaired driving intervention for college students was efficacious in reducing drinking and driving incidents at 3-month follow-up compared to an active education control. The CoAST approach includes real-time dialogue with a therapist, allowing for reflections, open-ended questions, or suggestions tailored to both the participant’s personalized feedback and their (text-based) responses. This degree of personalization exceeds what is possible with many algorithm-based automated approaches and might facilitate greater attention and processing of the personalized feedback and more effective motivational enhancement and goal setting. Thus, this potentially promising approach to administering BMIs requires further research.

The Current Study

The current study adapted the efficacious in-person alcohol BMI plus Substance-Free Activity Session supplement (SFAS; Murphy, Dennhardt, et al., 2012; Murphy et al., 2019) for CoAST delivery and examined the feasibility, acceptability, and preliminary efficacy of this adapted brief intervention. Using a randomized pilot trial design with a target sample size of 100, the current study compared a CoAST delivered alcohol BMI + SFAS (Murphy et al., 2019) to a CoAST delivered Alcohol + Nutrition Education (“education”) active control.² We hypothesized that the BMI + SFAS would be (1) feasible, as indicated by high enrollment and randomization and retention rates (≥70%) as well as high rates of intervention fidelity and Motivational Interviewing (MI) integrity, and (2) acceptable, as indicated by high participant satisfaction ratings (≥70%). This pilot trial was not sufficiently powered to detect all intervention effects. Nonetheless, our primary goal was to establish the feasibility and acceptability of the approach, and a secondary aim was to evaluate the hypotheses that participants who received the BMI + SFAS would report fewer past-month drinks per week, heavy drinking episodes (HDEs), and alcohol-related problems at follow-up, compared to those who received education.

Method

Participants and Procedures

Participants were 66 undergraduate college students recruited from multiple colleges and universities across multiple states in the southeastern United States. See Table 1 for sample demographics and characteristics. Students were eligible to participate in the study if they: (a) were between 18 and 25 years old; (b) were an undergraduate student working towards a bachelor’s or associate’s degree; (c) were working fewer than 30 hours per week if employed; (d) reported two or more HDEs (4/5 or more standard drinks for women/men respectively in an occasion) in the past month; (e) could speak, read, and write in English; (f) had access to the Internet via smartphone, tablet, laptop, or desktop computer (in order to download and access the encrypted messaging app, Babelapp); and (g) provided contact information for study personnel to reach them if deemed eligible. Students were excluded from participation if they were currently in treatment for substance use. Participants were screened for study eligibility in the Spring and Fall semesters of 2019 using a brief online survey.³ A secure web link/QR code to the survey was included on flyers that were posted on university and college campuses and in locations frequented by students, in research participation solicitation emails to consenting and publicly available student email addresses, and in social media postings on Facebook, Instagram, and Reddit. The survey was also listed in the psychology research subject pool (SONA) at the first author’s graduate institution.

Table 1.

Baseline Descriptive Statistics in the Full Sample and by Intervention Condition

	Full Sample (N = 66)	BMI + SFAS (n = 37)	Education (n = 29)	t / χ2 (df)	p
	M (SD) / % (n)	M (SD) / % (n)	M (SD) / % (n)
Age	19.95 (1.66)	20.19 (1.63)	19.66 (1.67)	−1.31 (64)	0.20
Gender/Birth Sex (Woman/Female)	63.6% (n = 42)	62.2% (n = 23)	65.5% (n =19)	0.08 (1)	0.78
Race/Ethnicity
Black/African American	20.6% (n = 13)	27.8% (n =10)	11.1% (n = 3)	2.62 (1)	0.11
White	61.9% (n = 39)	61.1% (n = 22)	63.0% (n = 17)	0.02 (1)	0.88
Asian	1.5% (n = 1)	2.7% (n = 1)	0% (n = 0)	0.80 (1)	0.37
Multiracial	7.6% (n = 5)	2.7% (n = 1)	13.8% (n = 4)	2.86 (1)	0.09
Hispanic/Latino/a/x	12.1% (n = 8)	8.1% (n = 3)	17.2% (n = 5)	1.27 (1)	0.26
Year in School
First-year	34.8% (n = 23)	27.0% (n = 10)	44.8% (n = 13)	2.27 (1)	0.13
Sophomore	18.2% (n = 12)	21.6% (n = 8)	13.8% (n = 4)
Junior	33.3% (n = 22)	29.7% (n = 11)	37.9% (n = 11)
Senior	13.6% (n = 9)	21.6% (n = 8)	3.4% (n = 1)
Alcohol Related Outcomes, past month
Drinks Per Week	11.88 (8.74)	11.86 (8.99)	11.90 (8.57)	0.02 (64)	0.99
Number of HDEs	4.42 (3.59)	4.30 (3.66)	4.59 (3.55)	0.32 (64)	0.75
Alcohol Problems	8.44 (5.62)	8.57 (5.63)	8.28 (5.69)	−0.21 (64)	0.84

Note. HDE = Heavy Drinking Episode. Results of t-tests presented for continuous variables of age, drinks per week, heavy drinking episodes, and alcohol problems. All others are chi-square tests for categorical variables. The Chi-square test for year in school has df = 1 because year in school was dichotomous: Freshman vs. all other classes.

All study procedures were conducted remotely, including screening, enrollment, randomization, intervention, and survey assessments. If eligible participants were interested in enrolling in the study, study personnel provided login information and helped connect the participant to the study on the encrypted messaging app, Babelapp (accessible at https://www.babelapp.com/en/ and in the Apple App Store and Google Play Store). To enroll, participants e-signed the informed consent form and completed the baseline survey. Upon survey completion, study personnel contacted the participant via Babelapp to schedule their first text-message session. Randomization occurred at the commencement of the first text-message session once the participant confirmed their availability to complete the session at that time and was stratified by sex (woman or man)⁴ and baseline past-month HDEs (2–3 or 4+).

Participants were randomized to either the BMI + SFAS or education condition. To better evaluate the feasibility of the BMI + SFAS (our primary aim), randomization was weighted to assign more participants to this condition. The first text message session was either the alcohol BMI or education, and the second text message session was the SFAS or nutrition education, respectively. Each session lasted between 45–60 minutes. Over the next four weeks following the second text message session, participants engaged in once weekly CoAST “mini sessions” for 20–30 minutes that discussed material related to their intervention assignment. Counselors collaborated with participants to determine a standing time each week (i.e., when the participant is not engaged in a regularly scheduled obligation such as class or work) to complete the mini sessions. All intervention contact, including mini sessions, was conducted through Babelapp to enhance participants’ privacy and security. Participants were also asked to complete a post-intervention survey immediately following the 4^th mini session and a 3-month follow-up survey approximately 12 weeks after the post-intervention survey. Participants could receive up to $65 (cash or Amazon gift card) or 2 SONA credit hours plus $20 at 3-month follow-up for their participation in the study. All procedures were approved by the primary author’s graduate university Institutional Review Board. This trial and the analyses presented below were planned as part of the first author’s dissertation (F31AA026486) and were registered with ClinicalTrials.gov (NCT04105725).

Measures

Feasibility.

Feasibility was assessed using four of the study components and feasibility quantification methods defined in Leon et al. (2011). Aspects of feasibility examined were recruitment (number enrolled and randomized per month), randomization (proportion of screened eligible who enroll), retention (proportion of participants who completed both full-length sessions and at least two of the four mini sessions), and treatment fidelity (ratings of counselor adherence to each of the intervention protocols and Motivational Interviewing Treatment Integrity [MITI] for the BMI + SFAS intervention only). Participants in either study condition who completed both full-length sessions and at least two mini sessions were considered “intervention completers.” Because mini sessions were designed to be reviews of the content previously discussed in the full-length sessions, we determined that completion of both full-length sessions and at least two mini sessions comprised a minimally sufficient “dose” of the intervention.

To quantify intervention fidelity, we randomly selected 20%t of de-identified intervention session transcriptions from each condition and session (n = 8 BMI, n = 8 SFAS; n = 6 alcohol education, n = 6 nutrition education) which were then reviewed by three trained independent coders for treatment fidelity using a brief intervention adherence protocol (see Murphy et al., 2019). Coders were aware of which intervention and session they were coding and rated each component of the intervention session from 0 = Didn’t do it, N/A to 3 = Above Expectations, with scores of two or higher indicating adherence to the protocol. Different adherence protocols were developed for each intervention to ensure that specific session components were delivered as prescribed, and also included three items intended to capture any cross-contamination of intervention content. Motivational Interviewing Treatment Integrity (MITI) coding of the BMI + SFAS followed Moyers et al. (2014). Because the education sessions were delivered in didactic rather than MI style, MITI coding was used in the evaluation of the BMI + SFAS intervention only and was completed by two of the independent coders mentioned above who were trained in MI and the SFAS but who were not counselors for the current trial. Four global scores (cultivating change talk, softening sustain talk, partnership, and empathy) were rated on a 5-point Likert-type scale with 1 denoting low and 5 denoting high MI adherence.

Acceptability.

Participants were asked to provide their feedback on the intervention elements, which included the two initial full-length sessions and the four weeks of mini sessions. Eight items were asked to gauge participants’ perception of intervention utility, interestingness, personal relevance, and effectiveness (in reducing alcohol use in other college students, in reducing their alcohol use, in increasing time spent studying or in other academic-related activities, and in improving nutrition and maintaining a balanced diet). Response options ranged from 0 = not at all to 10 = very.

Daily Drinking Questionnaire (DDQ).

Alcohol use was assessed using the Daily Drinking Questionnaire (DDQ; Collins et al., 1985). Participants were asked to estimate the number of drinks consumed each day in a typical week over the past month. Separate items were included to assess the frequency of heavy episodic drinking over the past month, defined as four or more (if a woman) or five or more (if a man) standard drinks on one occasion (Wechsler et al., 1995).

Brief-Young Adult Alcohol Consequences Questionnaire (BYAACQ).

Problems as a result of alcohol consumption were assessed using the 24-item Brief Young Adult Alcohol Consequences Questionnaire (B-YAACQ; Kahler et al., 2005). Participants were asked to indicate if they had had a particular experience or consequence related to their alcohol use in the past month (1 = Yes, 0 = No). Items were summed for a total score. Internal consistency for the current sample was 0.89.

Training of Counselors and Supervision

Intervention sessions were delivered by four clinical psychology doctoral-level graduate students who had completed at least 20 hours of training and supervision in MI, including readings, watching videos, and text message-based and in-person role-playing, and had prior experience delivering an alcohol BMI and/or the SFAS. De-identified text-message sessions were transcribed into Word document files from Babelapp and were reviewed during weekly group supervision with a licensed clinical psychologist with expertise in brief alcohol interventions for emerging adults, the SFAS, and MI.

Interventions

Brief Motivational Intervention and Substance-Free Activity Session (BMI + SFAS).

The BMI + SFAS is a text-message delivered brief intervention adapted from an in-person BMI for emerging adult heavy alcohol use (Murphy, Dennhardt, et al., 2012; Murphy, Skidmore, et al., 2012; Murphy et al., 2019). Each session lasted approximately 45–60 minutes from the first text message sent by the counselor to the intervention closing text message sent by the counselor. In the BMI, counselors presented via jpeg/PDF and discussed personalized normative feedback on drinking patterns, alcohol-related consequences, and protective behavioral strategies and set drinking and harm reduction-related goals if the participant indicated readiness to do so. In the SFAS, personalized feedback was presented on weekly time allocation, college/career-related goals were discussed, and an Episodic Future Thinking task was conducted (Atance & O’Neill, 2001; Peters & Büchel, 2011). The Episodic Future Thinking task asked participants to imagine as vividly as possible an important event related to their goals that they would like to see in the next three months (e.g., making the Dean’s list, getting all A’s in their classes). Counselors facilitated this task by encouraging participants to imagine salient aspects of this future event, such as how they will be feeling, what they will be thinking, and other contextual factors, in order to help motivate the participant towards achieving their stated college/career goals. Counselors utilized MI style (Miller & Rollnick, 2013) throughout the sessions, as well as in the four weeks of mini-sessions, and each mini-session followed a semi-structured protocol. See Supplemental Table 1 for BMI + SFAS content outline.⁵

Education.

Alcohol and nutrition education was selected as the active control condition because research indicates that providing alcohol education in a non-personalized and informational style has little to no effect on alcohol use (Barnett et al., 2004; Borsari & Carey, 2005; Murphy et al., 2001), and there is no research indicating that nutrition education may impact alcohol use, despite associations between poor nutrition and heavy alcohol use. Both education sessions were scripted and lasted approximately 45 to 50 minutes each. The alcohol education session presented information via jpeg/PDF on standard drink size, blood alcohol concentration (BAC), alcohol poisoning, and tolerance. The nutrition education session presented information on the major food groups, strategies to increase essential nutrients, and strategies to decrease certain food components that can contribute to poor health and disease (Meshesha et al., 2020). Mini sessions reviewed material from both full-length sessions. See Supplemental Table 2 for education content outline.⁶

Data Analysis Plan

Attrition effects were evaluated using the Mann-Whitney U test to determine if there were systematic differences in baseline alcohol outcomes by intervention completers and non-completers. Data from independent coders were used to calculate average ratings of counselor protocol adherence for both interventions and average MITI global score ratings (Moyers et al., 2014) for the BMI + SFAS. Independent samples t-tests were used to compare intervention feedback ratings by intervention condition. Data analysis to examine the feasibility and acceptability of interventions used SPSS (Statistics for Mac, Version 26.0, Released 2019; IBM Corp., Armonk, NY).

All variables included in models examining alcohol use outcomes were evaluated for outliers, as well as skew and kurtosis, and corrected per recommendations of Tabachnick and Fidell (2013). Values greater than 3.29 standard deviations were considered outliers and winsorized to one unit larger than the largest non-outlier. No variables were power transformed. Because our alcohol outcomes were count data (i.e., non-negative integers), we used negative binomial regression models in Mplus version 8.2 (Múthen & Múthen, 1998-2017) to assess intervention effect (0 = education, 1 = BMI + SFAS) on past-month drinks per week, HDEs, and alcohol-related problems at 3-month follow-up. Negative binomial regression models accommodate over-dispersed (i.e., variance exceeds the mean) count outcome data. All models included dichotomized covariates of sex/gender (0 = women, 1 = men), race/ethnicity (0 = non-white, 1 = white), counselor effect (0 = sessions conducted by counselor A, 1 = sessions conducted by counselors B, C, D)⁷, semester enrolled (0 = Spring, 1 = Fall), and class cohort (0 = First-year, 1 = Upperclass). All models were estimated using robust maximum likelihood (MLR) in Mplus.

Results

Evaluation of Feasibility

On average, we recruited 9.23 (SD = 5.38) participants per month for seven months in the spring 2019 and fall 2019 semesters. Our lowest recruitment month was March 2019 (n = 3), and our highest recruitment month was September 2019 (n = 18). The CONSORT diagram illustrating participant flow is presented in Figure 1. A total of 66 (22.1%; 66/299 eligible) participants were enrolled and randomized, and, among these, 100% (37/37) of those randomly assigned to the BMI + SFAS intervention completed both full-length sessions, whereas 89.7% (26/29) of those in the education intervention completed both full-length sessions. Thirty (81.1%; 30/37) participants who were randomly assigned to the BMI + SFAS intervention and 21 (72.4%; 21/29) to education were “intervention completers,” that is, they completed both full-length sessions and at least two of the four mini sessions.

Figure 1.

CONSORT Flowchart.

Although we had no a priori hypotheses regarding the 3-month follow-up rate, attrition in the BMI + SFAS condition was relatively low (33/37; 89.2% completed 3-month follow-up; see Figure 1). However, attrition within the education condition was slightly higher (22/29; 75.9% completed 3-month follow-up). One participant assigned to education was removed from longitudinal analyses because they erroneously received an SFAS mini session.

Intervention Fidelity.

Each of the BMI + SFAS sessions consisted of approximately 38–40 total text exchanges between counselor and participant, and the SFAS session typically followed within a week of the BMI session (M = 4.03 days, SD = 2.35, ranging from 1 to 8 days). One participant completed the SFAS 25 days after the BMI and was not included in these descriptive statistics. The average session rating for adherence to intervention protocol for the BMI + SFAS intervention was 1.87 (SD = 0.11), with 86.9% of the intervention elements meeting or exceeding expectations (≥ 2). MITI relational scores of partnership (M = 4.34, SD = 0.55) and empathy (M = 4.13, SD = 0.49), and technical scores of cultivating change talk (M = 4.13, SD = 0.61) and softening sustain talk (M = 3.97, SD = 0.47) were considered to meet or approximate the “Good” competency and proficiency thresholds (≥ 4) as specified by Moyers et al. (2014) for in-person MI.

Each of the education sessions consisted of approximately 44–48 total text exchanges between counselor and participant. Like the BMI + SFAS, the education session on nutrition typically followed within a week of the education session on alcohol (M = 3.58 days, SD = 2.28, ranging from 1 to 8 days). The average protocol rating for the education sessions was 1.93 (SD = 0.16), with 93.0% of the session elements meeting or exceeding expectations (≥ 2). There was not a statistically significant difference in adherence ratings between the two interventions (t(15.04) = 1.09, p = 0.29).

Evaluation of Acceptability

See Figure 2 for participant feedback ratings. Fifty-one participants (77.3%, 51/66) provided intervention feedback. Participants across both conditions rated the full-length text message sessions overall as highly useful and interesting with no group differences. Participants in the BMI + SFAS condition rated their text sessions as slightly more effective in increasing their time spent studying or participating in academic/career-related activities (M = 7.21, SD = 2.48) than did those in the education condition (M = 5.83, SD = 3.19), though this difference was not statistically significant (t(49) = −1.60, p = 0.12). Conversely, those in the education condition (M = 7.29, SD = 3.10) rated their text sessions as significantly more effective in modifying their eating and nutrition patterns than those in the BMI + SFAS (M = 5.24, SD = 3.48; t(48) = 2.05, p = 0.046).

Figure 2.

Participant Ratings of the CoAST Delivered Interventions

Note. Effective (other) = participant’s rating of intervention effectiveness in reducing alcohol use in other college students; Effective (self) = participant’s rating of intervention effectiveness in reducing their own alcohol use; Effective (study) = participant’s rating of intervention effectiveness in increasing time spent studying or in other academic related activities; Effective (nutrition) = participant’s rating of intervention effectiveness in improving nutrition and maintaining a balanced diet. * p < 0.05.

Baseline Descriptive Statistics and Differences by Intervention Completion

Table 1 presents demographic characteristics and baseline alcohol use descriptive statistics for the full sample and by condition. There were no significant differences in demographic or alcohol use variables at baseline by condition; however, there were statistically significant differences between first-year students and upperclass students (sophomores, juniors, seniors) in baseline drinks per week (First-year M = 7.89, SD = 5.79; upperclass M = 13.34, SD = 9.36; t(58) = −2.33, p = 0.02) and HDEs (First-year M = 2.89, SD = 3.18; upperclass M = 5.02, SD = 3.70; t(58) = −2.16, p = 0.04). Therefore, class was included as a dichotomous covariate in the respective preliminary efficacy models.

Overall, there were no significant baseline differences between intervention completers (n = 60) and intervention non-completers (n = 5) in past month drinks per week (U = 103.00, p = 0.25), HDEs (U = 122.00, p = 0.48), or alcohol-related problems (U = 133.00, p = 0.67).

Preliminary Efficacy: Alcohol Use Outcomes

Table 2 presents estimates of intervention effects on alcohol-related outcomes at the 3-month follow-up (see also Figure 3). There was a significant intervention effect on past-month HDEs with BMI + SFAS participants reporting fewer HDEs at follow-up compared to the education participants (B = −0.53, SE = 0.25, p = 0.03, 95% CI [−1.016, −0.046]). In other words, the BMI + SFAS reduced the mean number of HDEs by 41% compared to education (incidence rate ratio = 0.41, 95% CI [0.045, 0.638]). There was also a significant counselor effect on both HDEs (B = −0.61, SE = 0.28, p =0.03, 95% CI [−1.161, −0.065]) and alcohol-related problems (B = −0.67, SE = 0.29, p = 0.02, 95% CI [−1.231, −0.112]), such that receiving an intervention from counselor A, as compared to the three other counselors, was associated with more past month HDEs and alcohol-related problems at 3-month follow-up.

Table 2.

Estimated Marginal Means and Negative Binomial Regression Model Results for Intervention Effect on Alcohol Related Outcomes at 3-Month Follow-up

	BMI + SFAS EMM (SE)	Education EMM (SE)	B (SE)	[95% CI]	p
Drinks per Week (N = 51)	7.43 (0.85)	8.07 (1.53)
Intervention			0.05 (0.24)	[−0.421, 0.517]	0.84
Baseline drinks per week			0.05 (0.01)	[0.025, 0.068]	<0.001
Sex			−0.08 (0.21)	[−0.501, 0.337]	0.70
Race/ethnicity			0.17 (0.24)	[−0.294, 0.626]	0.48
Enroll			−0.24 (0.22)	[−0.665, 0.179]	0.26
Counselor			−0.10 (0.24)	[−0.562, 0.360]	0.67
Class			0.60 (0.30)	[0.017, 1.182]	0.04
Heavy Drinking Episodes (HDEs, N = 51)	2.73 (0.55)	4.05 (0.84)
Intervention			−0.53 (0.25)	[−1.016, −0.046]	0.03
Baseline HDEs			0.26 (0.04)	[0.172, 0.344]	<0.001
Sex			−0.02 (0.26)	[−0.532, 0.493]	0.94
Race/ethnicity			0.03 (0.26)	[−0.478, 0.541]	0.90
Enroll			−0.06 (0.27)	[−0.593, 0.475]	0.83
Counselor			−0.61 (0.28)	[−1.161, −0.065]	0.03
Class			0.28 (0.35)	[−0.412, 0.965]	0.43
Alcohol-Related Problems (N = 50)	5.79 (0.87)	5.16 (0.92)
Intervention			−0.07 (0.29)	[−0.629, 0.490]	0.81
Baseline problems			0.06 (0.03)	[0.008, 0.111]	0.02
Sex			−0.01 (0.31)	[−0.613, 0.600]	0.98
Race/ethnicity			0.33 (0.28)	[−0.224, 0.888]	0.24
Enroll			−0.19 (0.27)	[−0.724, 0.342]	0.48
Counselor			−0.67 (0.29)	[−1.231, −0.112]	0.02
Class			0.28 (0.31)	[−0.321, 0.886]	0.36
Baseline drinks per week			0.02 (0.01)	[−0.003, 0.050]	0.08

Note. EMM = Estimated marginal means and standard errors at 3-month follow-up, accounting for baseline levels and covariates (sex, race/ethnicity, counselor, class only), are presented as a function of intervention in the 2^nd and 3^rd columns. Intervention: 0 = Education, 1 = BMI + SFAS; Sex: 0 = women, 1 = men; Race/ethnicity: 0 = non-white participants, 1 = white participants; Enroll: 0 = Spring 2019, 1 = Fall 2019; Counselor: 0 = counselor A, 1 = counselors B, C, D; Class: 0 = first-year students, 1 = upperclass students.

Figure 3.

Change in past-month alcohol related outcomes by intervention condition.

Finally, Table 3 presents within-group means at baseline and 3-month follow-up by study condition for all alcohol use outcomes, as well as within-group effect sizes for changes from baseline to 3-month follow-up. Though both groups reported fewer drinks per week and alcohol-related problems at 3-month follow-up with no statistically significant group differences, within-group effect sizes for the BMI + SFAS condition were medium in size and statistically significant, whereas those for the education condition were not (see 95% CIs for Cohen’s d_rm in columns four and seven of Table 3).

Table 3.

Within-Group Change Effect Sizes and 95% Confidence Intervals by Intervention Condition

	BMI + SFAS			Education
	Baseline M (SD)	3-month M (SD)	Cohen’s drm (95% CI)	Baseline M (SD)	3-month M (SD)	Cohen’s drm (95% CI)
Drinks Per Week	12.16 (9.55)	8.00 (6.08)	0.46 (0.263, 1.041)	10.50 (8.83)	7.30 (8.14)	0.38 (−0.024, 0.896)
HDEs	4.26 (3.37)	2.58 (3.30)	0.50 (0.254, 1.029)	3.45 (2.48)	3.30 (4.09)	0.04 (−0.382, 0.495)
Alcohol-Related Problems	8.30 (5.52)	5.77 (5.59)	0.46 (0.136, 0.900)	7.50 (4.51)	4.80 (3.72)	0.65 (−0.006, 0.918)

Note. HDE = Heavy Drinking Episode. Cohen’s d_rm = mean effect size difference controlling for the correlation between repeated measures of the same variable. Effect sizes with confidence intervals that include 0 are considered non-significant. n = 51 for Drinks Per Week and HDEs, n = 50 for Alcohol-Related Problems.

Discussion

The current pilot trial evaluated the feasibility, acceptability, and preliminary efficacy of a CoAST delivered brief motivational intervention (BMI) plus behavioral economic-informed substance-free activity session (SFAS) for reducing heavy drinking in emerging adult college students. To our knowledge, this is the second study to utilize synchronous text-messaging for remote guided delivery of an alcohol-related brief intervention (see Teeters et al., 2018), and the first study to adapt an efficacious in-person alcohol BMI + SFAS (Murphy, Dennhardt, et al., 2012; Murphy et al., 2019) for CoAST delivery.

Overall, our results indicate that the BMI + SFAS intervention was feasible to deliver and acceptable and has the potential to reduce HDEs in a heavy drinking, non-treatment seeking, college student sample. Recruitment rate (enrollment per month), retention rate (proportion of enrolled participants completing the intervention), and treatment adherence and fidelity (including ratings of counselor adherence to the intervention protocols and motivational interviewing treatment integrity [MITI]) support study feasibility. An a priori power analysis to determine sample size indicated that an N = 62 would be adequate to detect main effects for our primary outcome variables, and we planned to overrecruit by enrolling 100 participants. However, the randomization rate (proportion of screened eligible who enrolled) was substantially lower than predicted (22% actual vs. 70% targeted and 66% of the target sample size N = 100), suggesting that there is room for improvement in recruitment and randomization into the BMI + SFAS intervention. Although we are unable to evaluate the potential role of treatment modality on recruitment, many research studies of in-person BMIs have reported low recruitment and retention rates yet these BMIs have been disseminated widely on college campuses (Huh et al., 2015; Mun et al., 2015). As such, rate of randomization, particularly in incentivized intervention research studies with non-treatment-seeking participants, may not be a highly useful or ecologically valid indicator of feasibility. Alternatively, it may be largely determined by advertising efforts and the nature of the incentive offered, rather than the specific appeal of the intervention approach. Indeed, few college students present for alcohol intervention services of any type, so widescale dissemination of these approaches will require incentives or mandates (Helle et al., 2021).

High participant satisfaction ratings are indicative of the acceptability of the CoAST BMI + SFAS. Interestingly, although the education sessions were similarly acceptable in terms of participant satisfaction ratings, the education condition had slightly lower rates of retention and follow-up assessment completion. In particular, a smaller proportion of participants in the education condition completed mini sessions and the corresponding weekly surveys, as compared to participants in the BMI + SFAS condition. Considering that participants in the education condition rated the sessions as slightly less personally relevant, though not significantly so, compared to participants in the BMI + SFAS condition, it is possible that these participants found the material less engaging and were, therefore, less motivated to complete the mini sessions. Additionally, because both interventions were delivered via CoAST, it could suggest that it is not simply interaction with another person that makes the intervention engaging. Rather, an interaction that is both conversational and includes personally relevant feedback or discussion of the participant’s beliefs and behaviors may be essential to successfully engage emerging adults in remotely delivered BMIs.

Although the efficacy results of this pilot trial should be interpreted cautiously due to the small sample size, a medium effect on past-month HDEs was observed at 3-month follow-up in the BMI + SFAS condition compared to the education condition, suggesting that the CoAST delivery approach for the BMI + SFAS merits further investigation. There were no significant treatment effects on weekly drinking or alcohol problems, but within person effect sizes indicated significant reductions only in the CoAST condition. All participants received psychoeducation about alcohol and completed alcohol-related assessments at several time points, which can generate assessment reactivity and could have contributed to the relatively small reductions in weekly drinking and alcohol problems observed in the education condition. The overall pattern of feasibility, acceptability, and preliminary efficacy data extend the promising preliminary results obtained by Teeters et al. (2018) with a brief drinking and driving focused CoAST intervention and suggest that this modality merits further investigation.

In-person, telephone, video conference, CoAST, automated text-messaging, and web-based brief interventions may all be evaluated on continua of costs and convenience features, as well as on continua of engagement and efficacy. Similar to automated text-messaging and web-based interventions, as well as telephone and video conferencing approaches, CoAST delivered interventions may be better able to reach individuals who are high-risk and less likely to attend an in-person session (Neighbors et al., 2018). Given that many high-risk college student drinkers are not interested in pursuing alcohol interventions (Helle et al., 2021; Wu et al., 2007) and that more informal or lower threshold options are generally preferred among college student drinkers (Buscemi et al., 2010; Helle et al., 2021), anything that reduces barriers to treatment engagement may have public health benefits. This may be especially important given that substance use and substance use treatment are stigmatizing and experiencing this stigma is a key contributor to individuals not seeking treatment (Livingston et al., 2012). Remote access to an efficacious intervention and live access to a trained provider is also valuable in reaching individuals who may be motivated to attend in-person sessions but do not have the resources to do so (e.g., time, transportation, finances), or who may simply prefer to engage with a counselor from the privacy of their home or dorm rather than attending an in-person session at a university counseling or health center. For example, remotely delivered interventions such as CoAST may be especially helpful for students who do not live on campus or who have life responsibilities or health issues that prevent them from accessing on-campus services and may have a particularly important role during the COVID-19 pandemic. Additionally, colleges that do not have the resources to provide on-site services may benefit from being able to partner with other schools or health care agencies to provide effective and engaging remotely delivered interventions.

Although CoAST delivery may reduce the participant burden associated with attending in-person sessions and provide counselors with flexibility to conduct sessions from a variety of locations, the actual session time required for participants and counselors was similar to in-person, despite the abbreviated intervention content. This issue is consistent with findings from a systematic review that observed an increased amount of time was required to deliver a comparable dose of an intervention through synchronous text-based systems compared with one-on-one telephone (voice call) delivery (Hoermann et al., 2017). Moreover, in contrast to many web-based “chats,” as it was designed in the current study, CoAST appointments were scheduled in advance to ensure that both counselor and participant were available to respond synchronously and to retain as much of the intervention “dose potency” as possible. Finally, unlike automated BMIs, which can be delivered more easily once an initial system is set up, CoAST interventions require additional and ongoing clinical resources, including some specific training in remote intervention delivery.

Overall, for clinicians, the effort to deliver CoAST interventions may be similar to what is required for in-person, one-on-one voice call, or video conference BMIs. Thus, there is still a potentially significant time/cost burden present that is alleviated by automated text-messaging and web-based interventions or automated interactive voice response (IVR) technology over telephone (Andersson, 2015). Indeed, although there is evidence to suggest higher cost-effectiveness of text therapies compared to traditional psychotherapy (see Schwatzmann & Boswell, 2020), some digital mental health intervention implementation studies indicate that cost-effectiveness may not be immediately realized and more research is needed to understand when this benefit occurs (see Schueller & Torous, 2020). However, considering that mHealth intervention efficacy is mixed (Hutton et al., 2020; Kaner et al., 2017; Kazemi et al., 2017) and that this may be due, at least in part, to reduced participant attention and engagement (Lewis & Neighbors, 2015), counselor time allocated to delivering synchronous text-based interventions may be a worthwhile investment. Future research is required to explicitly evaluate how CoAST delivery impacts participant engagement and retention compared to fully automated (text message or web-based), telephone (IVR or one-on-one), videoconference (Celio et al., 2017; King et al., 2020; Lee et al., 2021), and in-person BMIs. Lastly, future research should also seek to understand the impact of these various modes of delivery on intervention utilization, costs, and efficacy outcomes by comparing equivalent content across in-person, videoconference, guided synchronous text-based systems (e.g., CoAST), and fully automated intervention approaches.

Limitations

Several limitations should be considered. First, our sample size was small and not adequately powered to detect significant treatment effects or to estimate reliable effect sizes to be used later for sample size determination in a larger scale trial (Leon et al., 2011). Thus, our tests of preliminary efficacy should be considered within the context of this important limitation and interpreted with caution. However, in line with the purpose of pilot trials outlined by Leon et al. (2011), the current study is appropriately designed and sized for the assessment of feasibility and acceptability, which our results largely confirm. Our sample was also relatively homogenous, which limited both the racial and cultural diversity of the sample as well as the range and variability of drinking patterns and alcohol problem severity observed. Most notably, average typical drinks per week, HDEs, and alcohol-related problems in the current sample were lower compared to previous trials with similar inclusion criteria (Murphy et al., 2012, 2019). Our assessments of alcohol use outcomes were retrospective self-report measures which are reliable and valid with emerging adults but subject to recall bias. Further, the extent to which remote assessment may have influenced retrospective self-reporting is also unclear. Given prior research on participants’ low level of attentiveness to remote interventions (Lewis & Neighbors, 2015; Rodriguez et al., 2015), assessment or questionnaire attentiveness and consistency should be closely examined in future studies.

Summary and Conclusions

Overall, the current study demonstrated the feasibility and acceptability of a CoAST delivered BMI + SFAS for hazardous drinking among college students. Rates of recruitment, retention, and intervention adherence/fidelity indicated the feasibility of a CoAST delivered brief alcohol intervention with non-treatment-seeking emerging adult college students, and high ratings of participant satisfaction indicated its acceptability in the target population. Further, preliminary efficacy results indicated that the CoAST BMI + SFAS, compared to an education control condition, was associated with a medium effect on past-month heavy drinking episodes. These results provide additional support for behavioral economic- and reinforcement-based alcohol intervention approaches that attempt to motivate changes by increasing the salience of future outcomes and the association with patterns of drinking and patterns of engagement in goal-directed and enjoyable substance-free activities (Daughters et al., 2018; Murphy et al., 2019). Future research to address the significant limitations of this study, as well as to expand knowledge regarding the synchronous text-based delivery format, is necessary to evaluate the efficacy of CoAST delivered brief interventions on reducing hazardous alcohol use and related problems.

Public Health Significance:

Live text-messaging with a counselor is a feasible and well-received way to deliver a brief alcohol intervention to reduce harmful alcohol use in emerging adults. This study found that live text-messaging with a counselor can reduce the frequency of binge drinking, a significant risk factor for experiencing alcohol-related problems. Live text-messaging may provide the benefits of interpersonal contact while retaining a sense of privacy and minimizing some burden associated with in-person sessions (e.g., travel costs, discomfort).