Delivery of a Prevention Program in Large College Classes: Effectiveness of the Expectancy Challenge Alcohol Literacy Curriculum

Abstract

Despite modest reductions in alcohol use among college students, drinking-related harms continue to be prevalent. Group-delivered programs have had little impact on drinking except for experiential expectancy challenge interventions that are impractical because they rely on alcohol administration. Expectancy Challenge Alcohol Literacy Curriculum (ECALC), however, offers a non-experiential alternative suitable for widespread implementation for universal, selective, or targeted prevention. ECALC has been effective with mandated students, fraternity members, and small classes of 30 or fewer first-year college students. Larger universities, however, typically have classes with 100 students or more, and ECALC has not yet been tested with groups of this size. To fill this gap, we conducted a group randomized trial in which five class sections with over 100 college students received either ECALC or an attention-matched control presentation and completed follow-up at four weeks. ECALC was associated with significant changes on six subscales of the Comprehensive Effects of Alcohol Scale (CEOA), post-intervention expectancies predicted drinking at four-week follow-up, and there were significant expectancy differences between groups. Compared to the control group, students who received ECALC demonstrated significant expectancy changes and reported less alcohol use at follow-up. Findings suggest ECALC is an effective, single session group-delivered intervention program that can be successfully implemented in large classes.

Alcohol use continues to be a public health concern for colleges with students consuming more alcohol than their non-college attending peers (Ahrnsbrak et al., 2017; Harding et al., 2016; Substance Abuse and Mental Health Services Administration [SAMHSA], 2020). Approximately one-third of college students report engaging in risky binge drinking, defined as consuming five or more drinks on an occasion for men and four or more drinks on an occasion for women, in the last month. Twelve percent report five or more binge drinking episodes in the last month, and 9% meet criteria for alcohol use disorder (SAMHSA, 2020; Schulenberg et al., 2018). High rates of binge drinking among college students put them at greater risk for experiencing a myriad of negative consequences including academic problems, interpersonal problems, accidental injury, sexual violence, unsafe/unprotected sex, development of an alcohol use disorder, and death (Ansari et al., 2013; Hingson et al., 2017; Hingson et al., 2002; Merrill & Carey, 2016; White et al., 2020; White & Hingson, 2013). As a result, universal, selective, and indicated prevention strategies focused on alcohol use have been implemented on many college campuses (Hennessy et al., 2019; Malloy et al., 2002; Prosser et al., 2018; Scott-Sheldon et al., 2014). A few programs have been found to be effective when delivered to individuals (e.g., Brief Alcohol Screening & Intervention for College Students; Hennessy et al., 2019; Scott-Sheldon et al., 2014), but effective group-delivered programs are scarce (Malloy et al., 2002; Prosser et al., 2018; Scott-Sheldon et al., 2014). In 2002, the National Institute on Alcohol Abuse and Alcoholism identified expectancy challenge (EC) as the only group-delivered intervention strategy with evidence of effectiveness with college students (Malloy et al., 2002). Subsequent meta-analytic reviews have supported EC interventions as effective in changing expectancies and reducing drinking (Scott-Sheldon et al., 2012; Gesualdo & Pinquart, 2021).

A simple definition of alcohol expectancies is that they represent learned information about the potential effects of alcohol and are not limited to outcomes that are pharmacologically related to alcohol (Goldman, 1999). A more comprehensive understanding describes expectancies as information stored in memory that becomes part of “templates” that are “anatomically instantiated” and “may be seen not just as coded patterns of neural activation, but also as actual changes in the physical neural substrate (Goldman, 2002, p.741).” Expectancy effects can include believing one will become more social, experience sexual enhancement, and become more relaxed after drinking, among other outcomes (Fromme et al., 1993). Expectancies can be examined at the item level, or via subscales that are groups of related expectancy concepts (Dunn & Goldman, 2000; Fromme et al., 1993). Expectancies mediate the influence of other biopsychosocial antecedents on later drinking behaviors (Miller et al., 1990; Stacy et al., 1991; Zucker et al., 1996), and evidence suggests a biological substrate for expectancy mediation of other influences (Goldman, 2002). Expectancies also predict drinking initiation and alcohol use patterns (e.g., Christiansen & Goldman, 1983; Christiansen et al., 1989; Magri et al., 2020; Stacy, 1997) and when challenged can result in predictable changes in drinking (Darkes & Goldman, 1993, 1998; Dunn et al., 2000). Expectancy changes have been demonstrated with experiential and non-experiential EC interventions, and both approaches have been effective in reducing alcohol consumption (Merrill & Carey, 2016; Scott-Sheldon et al., 2012). “Experiential” EC methods are those requiring a drinking exercise involving the consumption of alcoholic and placebo beverages while “non-experiential” methods have no drinking component (Gesualdo & Pinquart, 2021; Scott-Sheldon et al., 2012; Scott-Sheldon et al., 2014; Scott-Sheldon et al., 2016). Almost all EC programs found to be effective have been experiential and therefore impractical for widespread use on college campuses because of the reliance on administration of alcohol. Several non-experiential EC programs have been investigated without success in reducing alcohol use (Gesualdo & Piquart, 2021; Scott-Sheldon et al., 2012). Expectancy Challenge Alcohol Literacy Curriculum (ECALC), however, is currently the only non-experiential EC intervention with evidence of effectiveness when delivered to groups or individuals (Dunn et al., 2020; Dunn et al., 2022; Fried & Dunn, 2012).

ECALC is a single session 45-minute, interactive, non-experiential EC program designed to modify alcohol expectancy processes through a series of engaging exercises focusing on media literacy and the pharmacological effects of alcohol (Fried & Dunn, 2012). ECALC is like a simple computer game. A trained facilitator with internet access operates the program and leads a group or individual through content divided into 11 modules. ECALC differs from didactic interventions because it involves interactive exercises that encourage participants to examine their own expectancy processes, while simultaneously aligning expectancies with the pharmacological effects of alcohol. Rather than simply presenting educational information, ECALC engages participants in challenging the cognitive process by which expectancies influence alcohol use (Fried & Dunn, 2012; Dunn et al., 2022; Dunn et al. 2020).

Effectiveness of ECALC has been examined in several randomized controlled trials with various populations of college students. When compared to a widely used brief motivational intervention (BMI), individually administered ECALC was superior or non-inferior to BMI on all measures of alcohol use at four-week follow-up with a sample of 121 mandated students (Dunn et al., 2020). In comparison to an attention-matched control group, receipt of ECALC resulted in significant changes in expectancies and decreases in alcohol use at four-week follow-up in a sample of 250 fraternity members (Fried & Dunn, 2012). Results from an ECALC study with a sample of 991 first-year college students found significant changes in expectancies following the intervention, and that these changes mediated the effects of the intervention on subsequent alcohol-related harms and alcohol use (Dunn et al., 2022). Classes for this study were a maximum of 30 students, and as a result, impact of ECALC when delivered to larger college classes (≥100) often required for newer students at large universities is unknown. Implementing intervention programming in large classes can be challenging due to the limited attention span of young adults, especially with increased external stimuli found in large rooms with many other people (Cooper & Robinson, 2000). Students are more likely to retain information in smaller classroom settings because student engagement is more difficult with larger class sizes (Bradbury et al., 2016; Marx et al., 2016). Although ECALC has been validated for use with smaller groups or classes (≤30), effectiveness when administered to large groups typical of many introductory college classes is unknown. An effective single session intervention that can be reliably administered to large classes would decrease burden on intervention facilitators and academic institutions and reduce intervention cost.

The purpose of our study was to evaluate the effectiveness of ECALC in altering expectancy processes, reducing alcohol consumption, and reducing alcohol-related harms among student drinkers in large classes compared to an attention-matched control group. We hypothesized that those who received ECALC would exhibit changes in alcohol expectancies, and these changes would mediate the relationship between completion of ECALC and reductions in alcohol use and related consequences during the four weeks following program delivery.

Method

Participants

Participants were undergraduate students in psychology classes at a large state university in the southeastern United States, all of whom received extra credit for participation. During one semester, five classes with more than 150 students were randomly assigned to condition (ECALC=3 classes, control=2 classes). To participate in the study, students had to be enrolled and present in one of the classes recruited for participation. Students who self-identified as having consumed alcohol in the past month (n=678) completed baseline measures and received either ECALC or the attention-matched control presentation. Follow-up surveys were completed online by 357 students (see Table 1). All students present in class on program delivery day participated, however, only students 18 years of age or older were permitted to complete informed consent and follow-up assessment measures. Only students who consumed alcohol in the past month were retained for analyses.

Table 1.

Demographic Characteristics by Condition

Variable	Sample (N=678)	Control (N=302) n (% of N)	Experimental (N=376)
Biological Sex
Male	246 (36.3%)	91(30.1%)*	155 (41.2%)*
Female	432 (63.7%)	211(69.9%)*	221 (58.8%)*
Class Standing
Freshman	282(41.6%)	51 (16.9%)*	233 (62.0%)*
Sophomore	132 (19.4%)	58 (19.2%)	74 (19.7%)
Junior	169 (24.9%)	120 (39.7%)*	49 (13.0%)*
Senior	79 (11.6%)	66 (21.9%)*	13 (3.5%)*
Post-Bac	3 (0.4%)	2(0.6%)	1 (0.3%)
Race/Ethnicity
Caucasian	479 (70.4%)	208 (68.9%)	269 (71.5%)
Black/African-American	37 (5.4%)	19 (6.3%)	18 (4.8%)
Asian American	25 (3.7%)	9 (3.0%)	16 (4.3%)
Other	32 (4.7%)	15 (5.0%)	17 (4.5%)
Hispanic	95 (14.0%)	45 (14.9%)	50 (13.3%)
Non-Hispanic	585 (86.0%)	251 (83.1%)	334 (88.4%)
	Mean (SD)	Mean (SD)	Mean (SD)
Age (years)	20.00 (2.58)	20.78 (2.77)	19.39 (2.23)

Note.

^*significant difference between groups for this variable, p <.05.

The majority of participants were female (63.7%), White (70.4%), and Non-Hispanic (86.0%), and 41.6% were freshman. Additional demographic information for the entire sample and by randomized group can be found in Table 1. There were no statistically significant baseline differences between the experimental and control groups’ race/ethnicity. The experimental group had a higher proportion of male participants (155 experimental, 41.2%; 91 control, 30.1%), whereas the control group had a higher proportion of females (221 experimental, 69.9%; 211 control, 58.8%, χ²(1) = 8.91, p =.003). The experimental group was slightly younger than the control group (experimental: M = 19.93 years [SD = 2.23]; control: M = 20.78 years [SD = 2.78]; t [665] = 7.17, p < .001). The experimental group also had a greater proportion of freshman (233 experimental, 62.0%; 51 control, 16.9%), whereas the control group had more junior (49 experimental, 13.0%; 120 control, 39.7%) and senior students (13 experimental, 3.5%; 66 control, 21.9%, χ²(4) = 178.44, p <.001). It should be noted that analyses included students ranging from freshman to senior year and that juniors and seniors may have started to mature out of risky alcohol use (Johnston et al., 2015). To account for these differences, additional analyses were conducted to compare education status on key variables. Results revealed no significant differences in relation to educational status for baseline drinking variables, follow up drinking variables, or alcohol-related harms. Educational status was significantly related to expectancies (B = −0.54, p = 0.02). This is consistent with research that suggests expectancies change across the developmental trajectory (Bekman et al., 2011; Nicolai et al., 2012). ECALC has been shown to be effective for all college age groups and individuals in all class years were retained for analyses (Dunn et al., 2020; Fried & Dunn, 2012).

Procedure

Participants completed a demographic survey (e.g., biological sex, race, ethnicity, class standing) and baseline measures of expectancies, alcohol use, and alcohol-related consequences immediately before receiving either ECALC or control presentation during their regularly scheduled class time instead of usual course content. They completed the expectancy measure a second time at the end of both presentations. Graduate students trained as program facilitators followed scripted protocols for both ECALC and control presentations. All data collection was anonymous and we used an innocuous code identifier to connect baseline and follow-up data.

We informed participants during baseline assessment that they would be eligible to engage in follow-up study participation online in four weeks, and we instructed them to login to the research portal to complete follow-up four weeks after the baseline period. When their follow-up window opened, participants had one week to complete the follow-up survey to ensure follow-up data collection was only capturing the month in between the baseline and follow-up period. Participants who completed follow-up were provided with additional course credit.

Measures

Timeline Follow-back (TLFB)

We obtained alcohol consumption data using a modified version of the TLFB (Sobell & Sobell, 1992), previously used in ECALC studies where TLFB was self-administered rather than by an interviewer (e.g., Dunn et al., 2022; Dunn et al., 2020; Fried & Dunn, 2012). TLFB is a reliable (r = 0.76– 0.98) and valid measure of drinking behavior (Sobell & Sobell, 1992). Five measures of alcohol use over the past 30 days were computed including number of drinking days per month, mean drinks per week, mean drinks per sitting, number of binge drinking episodes in the past month, and peak drinks per sitting. These variables were used as indicators of a latent variable of alcohol use.

Comprehensive Effects of Alcohol Scale (CEOA)

We assessed alcohol expectancies using the Comprehensive Effects of Alcohol Scale (CEOA: Fromme et al., 1993), a measure used in previous ECALC studies (e.g., Dunn et al., 2000; Fried & Dunn, 2012; Dunn et al., 2020; Dunn et al., 2022). The CEOA consists of 38 items rated on a 4-point Likert scale, four subscales described as positive by the creators of the measure (Sociability, Tension Reduction, Liquid Courage, and Sexuality) and three subscales described as negative (Cognitive and Behavioral Impairment, Risk and Aggression, and Self-Perception). CEOA items were summed to create subscale scores. Example items from the CEOA include “I would be outgoing,” “I would feel unafraid,” and “I would feel creative.” All subscales have appropriate internal consistency and temporal stability (r = 0.53–0.81; Fromme et al., 1993). Internal consistency in the current study was acceptable at baseline, post-intervention and four-week follow-up (Sociability α = .86 to α = .94; Tension Reduction α = .74 to α = .86; Liquid Courage α = .83 to α = .90; Sexuality α = .76 to α = .83; CBI α = .85 to α = .87; RA α = .83 to α = .77; Self-Perception α = .68 to α = .75.)

Brief Young Adult Alcohol Consequences Questionnaire (BYAACQ)

The BYAACQ is a 24-item measure that assesses alcohol-related consequences that have occurred in the previous 30 days (Kahler et al., 2005). Internal consistency for this measure was excellent in our study (α = .90 to α =.92). The BYAACQ has good test-retest reliability (r = 0.86), minimal item redundancy, and covers a range of outcomes relevant to college student drinking such as saying or doing embarrassing things, missing class, getting sick, and engaging in risky behaviors (e.g., driving under the influence; Kahler et al., 2005; Kahler et al., 2008). Example items include “While drinking, I have said or done embarrassing things,” “I have passed out from drinking,” and “My drinking has gotten me into sexual situations I later regretted.” All items were summed to create a harms variable for analyses.

Intervention

ECALC

ECALC is a 45-minute non-experiential expectancy challenge intervention. The format is similar to a simple computer game and it is accessed through the internet. ECALC content includes information about the pharmacological effects of alcohol and media literacy, and it teaches participants to deconstruct alcohol advertisements to distinguish between pharmacological effects of alcohol and expectancy effects portrayed in media. Trained presenters advance the computer program from screen to screen, encourage participation in a series of exercises and answer questions. Training of presenters includes practicing short statements for transitioning from screen to screen and learning frequently asked questions and responses consistent with the goals of each exercise. Key content is all delivered through automated narration to standardize administration. We did not assess fidelity because of the automated nature of ECALC and content being delivered the same way with each administration.

Attention Control

Participants in the attention control group received a body image media literacy presentation similar in length and style of presentation to ECALC but focused on deconstructing advertisements in relation to body image. This attention control group has been used in previous ECALC studies (e.g., Fried & Dunn, 2012; Dunn et al., 2022).

Data Analysis Plan

Chi-square analyses were conducted to examine participant differences across group. MANOVA was used to test the effect of ECALC on individual expectancy subscales as captured by the CEOA with the subscales as dependent variables. Structural equation modeling (SEM) was used to test differences in alcohol expectancies via intervention, as well as a model of changes in alcohol expectancies mediating the relationship between condition and reductions in alcohol use and related consequences at four-week follow-up. Pre-intervention and post-intervention alcohol use were specified as latent variables. Pre-intervention and post-intervention harms were defined using summed items of the B-YAACQ, and we controlled for pre-intervention alcohol use and harms in analyses.

There were 678 individuals who endorsed drinking at T1 (prior to the intervention). Immediately following ECALC or control presentation, 659 students completed measures of expectancies; however, only 357 completed measures of alcohol use and alcohol-related harms at T2 (four weeks post-intervention). Thus, there was considerable missing data on alcohol use and alcohol-related harms at T2. Full information maximum likelihood estimation was used to account for missingness to allow for the use of all available data for the primary analysis (n = 678 observations). As is noted below, missingness was greater in the control condition than experimental condition and was also correlated with pre-intervention drinking levels. We do not believe we have missing completely at random (MCAR), but cannot rule out missing at random (MAR) or missing not at random (MNAR). Consequently, sensitivity analysis was also completed. We used 95% confidence intervals derived from 5000 random draws to calculate bias corrected indirect effects (MacKinnon, 2008).

A measurement model was estimated for the positive expectancy latent variable and the pre-intervention and post-intervention alcohol use variables. Alcohol expectancies at immediate post-intervention were specified as a latent mediator variable with the positive expectancy subscales of Tension Reduction, Sociability, Liquid Courage, and Sexuality as the indicators. The choice to only include positive expectancies in the model is based on theory and previous research. Changes in positive expectancies are more likely to predict treatment successes (Nielsen, 1992) and a previous ECALC study only demonstrated significant changes in positive expectancies that predicted significant reductions in drinking behaviors (Fried & Dunn, 2012). Modification indices of the latent variables were examined, and several correlated errors were identified. Correlated errors greater than 20 (n = 6) were iteratively freed. Results of the measurement model revealed good fit to the data [χ²(45) = 90.74, p < .001, CFI = .99, RMSEA = .04, SRMR = .04]. A structural model was specified with an expectancy latent variable as the mediator between condition (0 = control, 1 = ECALC) and alcohol outcomes at 30-day follow-up. Pre-intervention alcohol use and harms from the 30 days prior to intervention were added as model covariates (see Figure 1). Note that parameters in text are unstandardized while parameters in Figure 1 are standardized for ease of interpretation.

Figure 1. Structural equation model of the effects of ECALC on immediate post-intervention expectancies and 30-day post-intervention alcohol use and harms.

Note. All paths are standardized and statistically significant at p < .05.

Power Analysis

An a-priori power analysis was conducted using effect size data from a study conducted by Dunn et al. (2020) in which mandated college students completed ECALC (dw= .62). For mean-based analyses, a minimum of 84 participants are recommended with 42 participants in each group to have 80% power for detecting a medium-sized effect when employing .05 criterion of statistical significance.

Results

Chi-square analyses conducted to examine differences in follow-up completion between experimental and control groups revealed significant differences [χ2(1) = 26.05, p <.001] in proportion of individuals who completed follow-up for the experimental group (43.88% for experimental; 63.58% for control). Additional chi-square analyses conducted to examine demographic differences for follow-up completion found no differences pre and post intervention for race (p = .933) or class standing (p = .548). However, there was a significant difference for biological sex, such that a greater proportion of individuals who identified as female completed follow-up (59.02%), compared to those who identified as male [41.46%, χ2(1) = 19.40, p <.001]. Among those who did not complete the four-week follow-up, there were no differences in pre-intervention expectancies (p = .33), post-intervention expectancies (p = .84), or alcohol harms (p = .41). However, there was a difference in alcohol use such that those who drank more were more likely to be missing (p = .03, Cohen’s d = 0.17).¹ Chi-square analyses also indicated there were no statistically significant differences between groups at baseline regarding alcohol use, alcohol-related harms, or any expectancy subscale.

Effects of ECALC on expectancies

A MANCOVA was completed to examine effects of ECALC on post intervention expectancies with each expectancy subscale as a dependent variable and controlling for baseline expectancies. Results indicated significant effects of ECALC [Wilks λ = 0.864, F(7, 638) = 14.32, p < .001, Cohen’s d = 0.79], with significantly lower scores on Sociability (Cohen’s d = 0.64), Liquid Courage (Cohen’s d = 0.40), Sexuality (Cohen’s d = 0.36), and Tension Reduction (Cohen’s d = 0.30), and significantly higher scores on the Cognitive-Behavioral Impairment (Cohen’s d = 0.50) and Risk and Aggression (Cohen’s d = 0.25). There was no evidence of a significant effect on Self-Perception expectancies (Cohen’s d = 0.09). These results suggest that there were broad and statistically significant effects of the intervention on alcohol expectancies. See Table 2.

Table 2.

Alcohol Expectancy Changes Across Experimental and Control

Variable	Baseline	Post-Test	F (df)	Cohen’s d	Follow-Up	F (df)	Cohen’s d
Sociability			66.80 (1, 644)*	0.64		10.68 (1, 342)*	0.38
Experimental	27.01 (4.76)	24.15 (6.83)			26.04 (5.80)
Control	27.13 (4.16)	26.89 (4.46)			27.37 (4.44)
Cognitive Behavior Impairment			39.45 (1, 644)*	0.50		26.47 (1, 342)*	0.47
Experimental	25.65 (5.17)	26.33 (5.55)			28.80 (5.15)
Control	26.11 (5.35)	25.04(5.60)			26.85 (5.45)
Liquid Courage			26.05 (1, 644)*	0.40		3.30 (1, 342)	0.29
Experimental	14.03 (3.53)	12.76 (3.53)			13.35 (4.20)
Control	13.24 (3.54)	13.13 (3.91)			13.69 (4.09)
Risk & Aggression			9.92 (1, 644)*	0.25		1.16 (1, 342)	0.09
Experimental	12.90 (3.40)	11.94 (3.92)			12.77 (3.70)
Control	12.23 (3.46)	11.95 (3.73)			12.80 (3.77)
Sexuality			20.37 (1, 644)*	0.36		0.63 (1, 342)	0.20
Experimental	10.32 (3.01)	9.34 (3.46)			10.02 (3.71)
Control	9.93 (2.95)	9.69 (3.16)			10.03 (3.55)
Self-Perception			1.03 (1, 644)	0.09		1.68 (1, 342)	0.06
Experimental	5.42 (1.88)	5.64 (2.19)			5.80 (2.07)
Control	5.40 (2.06)	5.47 (2.10)			5.99 (2.19)
Tension Reduction			14.42 (1, 644)*	0.30		3.76 (1, 342)	0.20
Experimental	8.71 (2.06)	8.22 (2.51)			8.65 (2.36)
Control	8.75 (2.03)	8.76 (2.33)			8.86 (2.28)

Note:

^*<.001

MANCOVA was also used to examine effects of ECALC on follow-up expectancies (30-days post-intervention) with each expectancy subscale as a dependent variable, while controlling for baseline expectancies. Results indicated significant effects of ECALC [Wilks λ = 0.87, F(7, 366) = 7.11, p < .001, Cohen’s d = 0.77], with significantly lower scores on Sociability (Cohen’s d = 0.35) and significantly higher scores on Cognitive-Behavioral Impairment (Cohen’s d = 0.56). There was no evidence of significant effects on Liquid Courage (Cohen’s d = 0.20), Self-Perception (Cohen’s d = 0.14), Risk and Aggression (Cohen’s d = 0.11), Tension Reduction (Cohen’s d = 0.11), or Sexuality expectancies (Cohen’s d = 0.09, see Table 2).

Finally, a repeated measures MANOVA was conducted to examine effects of ECALC over time and to examine within subject changes. Results demonstrated significant between subject effects from ECALC [Wilks λ = 0.93, F(7, 336) = 3.72, p < .001, Cohen’s d = 0.56]. Results also showed a significant effect within subjects for time on expectancies [Wilks λ = 0.68, F(14, 329) = 11.25, p < .001, Cohen’s d = 1.38], and a group by time interaction [Wilks λ = 0.80, F(14, 329) = 6.01, p < .001, Cohen’s d = 1.01]. In the examination of group by time interaction within subjects, there were significantly lower scores for Sociability (Cohen’s d = 0.54) Tension Reduction (Cohen’s d = 0.29), Sexuality (Cohen’s d = 0.29), and Liquid Courage (Cohen’s d = 0.29), and significantly higher scores across time for Risk and Aggression (Cohen’s d = 0.32). There was no significant quadratic effect for Self-Perception (Cohen’s d = 0.09) and Cognitive Behavioral Impairment (Cohen’s d = 0.14).

Mediated Effects of ECALC on Alcohol Use and Alcohol Harms via Expectancies

The structural model (see Figure 1) showed good fit to the data [χ²(80) = 187.29, p < .001, CFI = .95, RMSEA = .04, SRMR = .08]. Biological sex and age were initially added as covariates, however, neither were significantly associated with expectancies, alcohol use, or alcohol harms, and were removed from the analysis for parsimony. Condition was a robust predictor of immediate post-intervention expectancies (B = −2.72, p < .001; Cohen’s d = −0.45). Expectancies post-intervention predicted drinking at follow-up (B = 0.02, p = .029). This effect was more robust if we did not control for alcohol use over the 30-days prior to the intervention (B = 0.21, p < .001), therefore we controlled for pre-intervention drinking to provide the most conservative estimate. Results also suggest there was a small but significant indirect effect from group to alcohol use through expectancies [IND = −0.04, CI 95% (−0.09, −0.01)] and from group to harms through expectancies and alcohol use [IND = −0.06, CI 95% (−0.14, −0.01)]. This model accounted for 6% of the variance in expectancies, 46% of the variance in post-intervention alcohol use, and 40% of the variance in alcohol-related harms (see Table 3 for drinking means).

Table 3.

Drinking Variables

Variable	Baseline	One Month Follow Up
	Mean (SD)	Skewness Static (SE)	Mean (SD)	Skewness Static (SE)
Peak Drinks Per Sitting
Experimental	8.01(6.49)	2.70 (0.13)	5.40 (4.96)	1.80 (0.19)
Control	7.27 (5.71)	2.57 (0.14)	5.95 (5.13)	1.43 (0.17)
Mean Drinks Per Week
Experimental	1.81 (2.64)	2.67 (0.13)	0.75 (0.96)	2.17 (0.19)
Control	1.37 (1.91)	3.16 (0.14)	0.87 (1.07)	1.72 (0.17)
Drinking Days Per Month
Experimental	4.74 (4.11)	2.03 (0.13)	4.91 (4.00)	1.61 (0.21)
Control	4.94 (3.95)	1.36 (0.14)	5.42 (3.95)	1.30 (0.19)
Binge Episodes Per Month
Experimental	3.51 (3.46)	2.56 (0.14)	3.93 (3.21)	1.54 (0.24)
Control	3.44 (3.34)	1.68 (0.16)	4.14 (3.34)	1.13 (0.22)
Harms
Experimental	5.82 (4.97)	1.05 (0.13)	6.73 (6.52)	1.96 (0.19)
Control	5.17 (4.42)	0.93 (0.14)	6.17 (5.38)	1.01 (0.18)

Discussion

The primary purpose of our study was to examine the effectiveness of ECALC on alcohol use and related harms when delivered to large college classes with over 100 students. As hypothesized, when compared to the control group, individuals who received ECALC demonstrated changes in alcohol expectancies after receiving the program, with moderate to large effects on all CEOA subscales except self-perception. The ECALC group also showed within subject changes for expectancies across time with medium to small effects. Our findings are consistent with previous studies and add evidence to the conclusion that ECALC is effective in altering alcohol expectancies associated with risky alcohol use (Dunn et al., 2022; Patrick et al., 2010). Like results found in Dunn et al. (2022), changes in expectancies mediated the relationship between group (ECALC or control) and reductions in alcohol use and harms at four-week follow-up. In the current study, there was no significant direct effect of expectancies on harms at four-week follow-up. One of our previous studies using ECALC with smaller groups of college students yielded larger effects on expectancies (Cohen’s d =1.23–0.30) with intervention models accounting for 54% of variance in drinking and 44% of variance in alcohol-related harms (Dunn et al., 2022). Although results of the current study suggest there may be some decrease in benefit when delivering the intervention to larger groups, significant positive outcomes were still achieved. When delivered to groups of more than 100 students, ECALC was associated with changes in expectancies and reductions in alcohol use and related harms.

Our previous studies have found ECALC to be effective when group-delivered to first-year college students and male fraternity members, and when individually delivered to mandated students of both sexes (Dunn et al., 2020; Dunn et al., 2022; Fried & Dunn, 2012). In addition to being effective with a wide variety of high-risk populations, ECALC has shown superior effectiveness on several alcohol use variables at four-week follow-up when compared to a well-validated brief motivational intervention (Dunn et al., 2020). Results from the present study add to the growing literature supporting ECALC as an effective, non-experiential EC intervention for targeting alcohol consumption through expectancies. In combination with results from previous studies, the present study suggests that ECALC could be effectively integrated into college classrooms as part of a standard curriculum. Further, ECALC could be an effective adjunct or alternative to other widely used programs (e.g., AlcoholEdu) that have not produced consistent reductions in alcohol consumption (Cronce & Larimer, 2011; Croom et al., 2008; Dunn et al., 2022).

Sociability, Liquid Courage, Sexuality, and Tension Reduction expectancies have been shown to have a larger influence on alcohol use and alcohol-related harms than other types of expectancies (Dunn et al., 2022; Patrick et al., 2010). Results of the current study found continued support for the significant relationship between these expectancies, alcohol use, and alcohol-related harms. Additionally, when considered as a latent variable in the full model, Sociability, Sexuality, Tension Reduction, and Liquid Courage expectancies significantly mediated the relationship between ECALC and alcohol use and alcohol-related harms. Significant changes were found in all expectancy subscales excluding self-perception following exposure to ECALC. These findings add to the evidence that ECALC is effective at targeting expectancies, in line with past research (Dunn et al., 2022).

There are several limitations to the current study. First, alcohol use and harms were only assessed at baseline for the four weeks preceding intervention, and at four-week follow-up, and this study does not provide evidence of duration of effects beyond four weeks. Few studies have investigated the longitudinal effects of EC interventions (Scott-Sheldon et al., 2012), and there is clearly a need for research examining the durability of ECALC effects as it is possible these effects degrade overtime. We currently have an NIAAA-funded study underway examining ECALC outcomes up to 6 months post-intervention using ecological momentary assessment (EMA), and measuring the effect of weekly boosters. Second, monthly recall of alcohol use and harms may result in a retrospective recall bias affecting the accuracy of reporting (Dulin et al., 2017), but this bias would apply to both baseline and follow-up data in the present study. Future studies should examine the effects of expectancies and alcohol use via a daily event level assessment to determine deterioration of intervention effects or lack thereof (also being addressed in the NIAAA-funded study noted above). Third, we conducted this study at a large state university in a large metropolitan area with a relatively diverse population of students. Therefore, results may not be generalizable to rural or small communities, meriting future research across campuses with different characteristics. Although the sample is representative of the university where the study was conducted, additional studies should be conducted to discern the impact of ECALC on diverse populations. Fourth, there was significant follow-up attrition because of the use of online data collection and restrictions imposed by the university on reminding participants to complete follow-up measures. Previous ECALC studies have completed follow-up assessments in-person to minimize attrition rates (Dunn et al., 2020; Dunn et al., 2022; Fried & Dunn, 2012). The choice to collect follow-up data online in the present study, however, was intended to make it easier for students to complete measures and minimize the use of class time. Future studies should provide greater incentives beyond course credit to complete follow-up measures and minimize attrition. Additionally in relation to attrition rates between the experimental and control group, low follow-up completion rates could reflect lower acceptability/feasibility of the intervention itself. previous ECALC studies, however, have demonstrated both the acceptability and feasibility of the intervention for use with college students (Dunn et al., 2022; Dunn et al., 2020; Fried & Dunn, 2012). Regardless, future studies should continue to monitor and aim to improve the acceptability and feasibility of ECALC. Fifth, there was variance in the method of data collection at baseline and follow-up (i.e., in person versus online). Although previous studies have suggested that data quality is unaffected by mode of data collection, future studies should provide the option to complete all surveys via the same method to minimize potential effects of method variance (Shapka et al., 2016). Sixth, there were several differences between variables in relation to follow-up completion. There was a lower proportion of individuals who received ECALC and completed follow-up compared to control participants, an artifact for which we can only conjecture. It is possible that because we did not sample from a specific type of class, individuals enrolled in certain courses may have been more or less motivated to complete follow-up based on rigor of the course. Although this artifact might reflect acceptability and feasibility of follow-up after intervention, in a naturalistic setting for intervention implementation, there would be no follow-up involved. A greater proportion of students who identified as female completed follow-up than those who identified as male. This is consistent with literature suggesting there are differential predictors for treatment retention across genders (Green et al., 2002). It should also be noted that those who drank more alcohol were less likely to complete follow-up. This could be because those who drink more are more likely to experience academic problems, as 25% of college students experience academic problems because of alcohol use (Wechsler et al., 2002). Although students were ensured they could participate in follow-up regardless of class standing, future studies should ensure students’ participation is voluntary and not connected to academic engagement in class. Incentives beyond course credit may be beneficial to enhance retention rates. Seventh, we did not assess intervention fidelity. Variability in program delivery was identified as a potential problem in an early version of ECALC, and the web-based version used in the present study was developed to ensure consistency (the same web-based version was also used in three previous studies, Dunn et al., 2022; Dunn et al., 2020; Fried & Dunn, 2012). All substantive aspects of ECALC are delivered automatically by a recorded narrator, and we train intervention facilitators to encourage engagement with the material and promote an interactive intervention environment.

Our findings indicate ECALC is effective when delivered to large groups of college students, and add to several previous studies demonstrating the effectiveness of ECALC in reducing alcohol consumption. Our results also provide further support for the conclusion that changes in expectancies, particularly sociability expectancies, facilitate changes in alcohol use. The availability of ECALC as an effective program delivered in under an hour is an attractive, low-cost approach to reducing negative consequences associated with alcohol use.