Genetically Influenced Externalizing and Internalizing Risk Pathways as Novel Prevention Targets

Abstract

Objective:

Risky substance use among college students is widespread, and associated with numerous adverse consequences. Current interventions focus primarily on students’ current substance use; we hypothesize that shifting focus from current use to underlying risk factors is a complementary approach that may improve effectiveness of prevention/intervention programming. This approach aligns with the personalized medicine movement, which aims to harness knowledge about underlying etiological factors to provide individuals with specific information about their unique risk profiles and personalized recommendations, to motivate and enable individuals to better self-regulate their health.

Method:

Our group is building and evaluating an on-line Personalized Feedback Program (PFP) for college students that provides feedback about the individual’s underlying genetically-influenced externalizing and internalizing risk factors for substance use, along with personalized recommendations/resources. The project capitalizes on work from a university-wide research project (Spit for Science; S4S), in which >12,000 students (~70% of five years of incoming freshmen) are being followed longitudinally to assess substance use and related factors across the college years. In this paper, we describe our foundational work to develop the PFP.

Results:

From the S4S data we have identified risk factors across four domains (Sensation-Seeking, Impulsivity, Extraversion, and Neuroticism) that are correlated with college students’ substance use. We developed an on-line self-guided PFP, in collaboration with professionals from student affairs, and using feedback from students, with the ultimate goal of conducting a randomized clinical trial.

Conclusion:

The provision of personalized risk information represents a novel approach to complement and extend existing college substance use programming.

Risky substance use among college students is widespread. Data from the 2018 Monitoring the Future report indicate that 28% of college students report binge drinking, and marijuana use among college students has reached a historic high (Schulenberg et al., 2019). One of the most intensive longitudinal studies conducted with college students found that nearly half (47%) of all students meet criteria for an alcohol or marijuana use disorder at least once in the first three years of college (Caldeira et al., 2009). College students use alcohol at higher rates than their non-college-attending peers (Schulenberg et al., 2019). Importantly, problematic substance use is associated with serious consequences, including unwanted sexual encounters, legal consequences, assault, injury, and suicide (Arria et al., 2013; Hingson, Zha, & Weitzman, 2009).

College represents a unique opportunity for interventions to have positive life-course altering health benefits for a significant, rapidly growing (Dick et al., 2018), and increasingly diverse portion of the population (Dick, 2018; Dick & Hancock, 2015). Post-secondary enrollment is expected to grow to >24 million individuals by 2024, and college is now considered a “pervasive” American experience (Dick, 2018). College students are entering a critical developmental period for the establishment of health behaviors that persist into adulthood (Dick et al., 2018; Dick & Hancock, 2015), and college represents one of the few times in an individual’s life where all primary activities – social, career, health, and safety – are concentrated and controlled within a single setting. Accordingly, effective programs to address substance use in college have the potential for long-lasting positive impact.

The current “gold standard” for reducing risky substance use among college students is the use of brief motivational interventions (BMIs) (Larimer & Cronce, 2002; Lee, Neighbors, Kilmer, & Larimer, 2010). BMIs are widely used on college campuses, with one national study finding that 62% of schools report utilizing empirically supported alcohol prevention BMI programming (Nelson, Toomey, Lenk, Erickson, & Winters, 2010). BMIs have been adopted for both universal prevention programming intended for all college students, and targeted programming for mandated students (Barnett et al., 2004; Borsari & Carey, 2005; Hustad, Barnett, Borsari, & Jackson, 2010; A. M. White, Kraus, & Swartzwelder, 2006). They aim to reduce harmful drinking practices by providing students with information about how their drinking compares to others, recognizing possible consequences associated with excessive alcohol use, and encouraging students to undertake new strategies to monitor their drinking, set limits, and reduce risk. Delivered in-person and through web-based platforms, these interventions consistently yield significant, albeit small, effects for alcohol use outcomes (for reviews, see Carey, Scott-Sheldon, Carey, & DeMartini, 2007; Carey, Scott-Sheldon, Elliott, Garey, & Carey, 2012; Cronce & Larimer, 2011).

Despite their widespread use and current status as the most effective available resource, small effect sizes have led researchers to highlight the “need for the development of more effective intervention strategies” (Huh et al., 2015). A review of interventions for college student drinking noted that “significant enhancement of personalized feedback intervention efficacy has not been observed in over 15 years of study” (Miller, Meier, Lombardi, & Leffingwell, 2015). Newer directions of study include the application of behavioral economic theory to addressing college substance use (Murphy et al., 2019, 2012, 2015), as well as the use of pharmacotherapy (DeMartini et al., 2016; O’Malley et al., 2015), both of which have been demonstrated to enhance reductions in substance use associated with BMIs. However, these strategies have yet to be widely adopted on college campuses. Clearly, additional tools to address risky substance use among college students are needed.

Another limitation is that prevention and intervention programs for college students have historically focused exclusively on alcohol. As noted above, there have been significant changes in substance use patterns among college students, with illicit drug use rising in popularity (Schulenberg et al., 2019). Data from the Monitoring the Future study indicate that 52% of college students report lifetime use of marijuana, 42% report past year use, and 25% report past-month use (Schulenberg et al., 2019). Further, data indicate that most risky alcohol use among college students now occurs in the context of other drug use, most commonly marijuana (Cho et al., 2015). Although BMIs have been expanded to include items focused on marijuana use, there have been limited effects for marijuana use (Elliott, Carey, & Vanable, 2014; Lee et al., 2010; H. R. White et al., 2015).

What is striking about the literature on substance prevention/intervention is that it focuses almost entirely on current substance use behavior, and, for the most part, does not make use of the large body of etiological research on developmental pathways of risk for alcohol and other drug problems. Some BMIs include a brief section about family history of alcohol problems, which reflects both genetic and environmental risk; however, this is generally the only mention of the importance of underlying risk factors. We believe this represents an untapped opportunity.

We propose a new, complementary approach for expanding substance use programming options for emerging adults: the development of personalized feedback based on known underlying genetically-influenced risk factors. This approach capitalizes on the element of current prevention programming that appears to be most critical to effectiveness: the delivery of personalized feedback to stimulate change (Cronce, Bittinger, Liu, & Kilmer, 2014). However, rather than focusing feedback on current patterns of alcohol use, we propose providing personalized feedback on underlying risk factors known to contribute to patterns of substance use, along with personalized recommendations and resource information tailored to the individual’s risk profile(s). In this way, we utilize findings from the basic epidemiological literature about the key pathways of risk for substance use problems to create a personalized risk assessment for each student. Our intervention shifts the focus from the provision of information about relative levels of substance use (what is currently provided), and provides information about risk across major risk pathways that contribute to patterns of substance use.

The development of alcohol problems is often discussed in the context of two broad pathways of genetically-influenced risk: externalizing characteristics and internalizing characteristics; accordingly, these are the focus areas targeted by our Personalized Feedback Program (PFP). The externalizing pathway is characterized by a predisposition to behavioral undercontrol, poor response inhibition (Dick et al., 2010), and increased sensitivity to reward (Gatzke-Kopp et al., 2009; Robert F Krueger, Markon, Patrick, Benning, & Kramer, 2007; Zucker, Heitzeg, & Nigg, 2011; Zuckerman & Kuhlman, 2000). An extensive literature documents the role of externalizing characteristics in the development of substance use problems (Farmer et al., 2016; Meque, Dachew, Maravilla, Salom, & Alati, 2019; Steele, Forehand, Armistead, & Brody, 1995; Zucker et al., 2011). A number of personality traits that capture the various facets of externalizing have been associated with increased likelihood of harmful substance use, including extraversion/sociability, impulsivity, and sensation-seeking (Kendler et al., 2003).

The second widely established pathway of risk is through problems with emotion regulation as associated with internalizing symptomatology (Hussong, Jones, Stein, Baucom, & Boeding, 2011). Although the evidence for internalizing risk factors is not as robust as that for externalizing, multiple studies indicate that internalizing symptoms in childhood and adolescence predict subsequent substance use (Hussong et al., 2011; Kushner et al., 2012; Steele et al., 1995). Further, there is a consistent subset of drinkers who report drinking to reduce negative affect (Carpenter & Hasin, 1999; R. Tarter, Kirisci, Hegedus, Mezzich, & Vanyukov, 1994). These individuals have higher levels of depression, anxiety, and neuroticism (LaBrie, Kenney, Napper, & Miller, 2014).

These pathways also map onto decades of research on patterns of heterogeneity among individuals with alcohol use disorder, with studies consistently finding a dichotomous etiology, with Type A (Babor et al., 1992) / Type I (Cloninger, Sigvardsson, & Bohman, 1996), characterized by negative-affect-related drinking and elevated rates of depression/anxiety, and Type B (Babor et al., 1992) / Type II (Cloninger et al., 1996), characterized by impulsive and antisocial behaviors. These distinct etiological pathways have been shown to be invariant across cultures, gender, and age (Carragher et al., 2016). These literatures all indicate that individuals who misuse alcohol are a heterogeneous group.

Another advantage of focusing on underlying externalizing and internalizing risk factors, rather than current substance use per se, is that twin studies robustly demonstrate that different forms of substance use have a common genetic etiology, with the majority of the underlying risk for alcohol and other illicit drugs being shared (Kendler, Prescott, Myers, & Neale, 2003; Krueger et al., 2009), which likely contributes to their frequent co-occurrence (Williams, Liccardo Pacula, Chaloupka, & Wechsler, 2004). Because the underlying risk pathways appear to be shared across drugs of abuse, by shifting the focus to these underlying etiological factors, it is possible that a personalized risk assessment will be effective at reducing other drug use in addition to alcohol without the need to individually target multiple substances.

The provision of personalized feedback is in line with the current movement toward personalized medicine, moving beyond a “one-size fits all” approach to preventing and treating deleterious health outcomes (Chouchane, Mamtani, Dallol, & Sheikh, 2011). This movement has gained momentum in recent years with the 2015 White House launch of the Precision Medicine Initiative (Collins & Varmus, 2015). Although much attention surrounding precision medicine focuses on the incorporation of genomics into personalized medicine, it is also possible to use information about the known intermediary phenotypes by which genetic risk unfolds across development to provide personalized risk information, even absent specific genotypic information (Dick et al., 2018; Dick & Hancock, 2015).

Multiple papers have addressed the promise of personalized medicine as applied to alcohol use disorder (AUD) (Krystal & O’Malley, 2015; Litten et al., 2015; Sher, 2015). The heterogeneous nature of AUD (Babor et al., 1992; Cloninger, Bohman, & Sigvardsson, 1981) suggests that focusing on etiological mechanisms, rather than diagnoses, may provide novel, more effective targets for prevention and intervention. The Addictions Neuroclinical Assessment (ANA) framework (Kwako, Momenan, Litten, Koob, & Goldman, 2016) is one such effort to address the heterogeneity inherent in substance use problems and focus on core etiological processes. The domains indicated by the ANA, incentive salience, negative emotionality, and executive function, map onto personality and clinical features related to externalizing and internalizing characteristics (Kwako et al., 2019). The Research Domain Criteria (RdoC) program at the National Institute of Mental Health represents another effort aimed at focusing on underlying mechanisms of risk (Cuthbert, 2014; Cuthbert & Insel, 2013).

Although the field of personalized medicine as applied to complex behavioral outcomes is in its infancy, a literature is emerging that focuses on prevention programming tailored to individual risk profiles. Initial studies suggest that this personalized approach is likely to enhance the effectiveness of substance use intervention. Conrod et al. developed a school-based alcohol prevention program geared toward adolescents in their early to mid-teenage years that targets externalizing (impulsivity, sensation-seeking), and internalizing (anxiety sensitivity, hopelessness/depression) personality characteristics (Conrod et al., 2013; O’Leary-Barrett et al., 2013). In a cluster randomized controlled trial of >2600 students from secondary schools (9th grade) in London, United Kingdom, the personality targeted prevention intervention program focused on underlying risk dimensions rather than use, per se, was associated with a 29% reduced odds of drinking, 43% reduced odds of binge drinking, and 29% reduced odds of problem drinking relative to students in control schools (Conrod et al., 2013). Effects were maintained up to 24 months post intervention. The effectiveness of the personality-targeted prevention intervention on alcohol use among adolescents has been demonstrated across multiple studies and cultures at this point (Conrod, Castellanos-Ryan, & Mackie, 2011), though it requires an in-person format, limiting scalability, and has not yet been integrated into approaches to address substance use in college students.

Here we describe the development of a new on-line approach to college student substance use prevention that provides personalized feedback on underlying etiological risk factors and how they can contribute to substance use problems, rather than on current substance use. This approach builds upon basic etiological research on externalizing and internalizing pathways of risk to create an on-line, self-guided personalized feedback program, intended as a universal prevention. This work is on-going; below we review work done to date to build the program, as well as other design considerations that we believe have broad relevance for prevention programming in young adults. Finally, we discuss our work in the context of other on-going efforts to expand existing prevention and intervention programs by incorporating information about basic underlying etiological risk factors. Our goal in writing this paper is to stimulate discussion of creative ways that we can expand substance use prevention and intervention efforts by working across silos that traditionally exist between substance use researchers who work on basic etiological processes and those who work on prevention and intervention.

Methods

The Spit for Science Sample

Our work builds upon a unique, established registry for large-scale transdisciplinary research on substance use problems in college students, the Spit for Science project (Dick et al., 2014) (S4S; spit4science.vcu.edu). S4S is the largest genetically informative study of the factors that contribute to substance use outcomes among college students. S4S is a university-wide research registry with the goal of understanding pathways of risk (both genetic and environmental) for substance use and related mental health outcomes across the college years. For five incoming freshman cohorts (enrolled in 2011–2014 and 2017), all incoming freshmen age 18 or older were invited to participate in an on-line survey at the beginning of the fall semester of their first year, provide a saliva DNA sample, and complete follow-up surveys each spring semester thereafter. A total of 12,365 students were enrolled through this pipeline. We achieved consistent participation rates of 67–68% across all freshman classes, which compares quite favorably to cooperation rates of 22–40% reported in other web-based college surveys (Cook, Heath, & Thompson, 2000; Jans & Roman, 2007; National Survey of Student Engagement, 2011; Sax, Gilmartin, & Bryant, 2003). All data are entered in a registry, available broadly to all investigators with relevant research interests related to substance use and related behavioral health outcomes. Through the registry, participants can be selected for additional spin-off studies. Information on accessing S4S registry data is available on the website (spit4science.vcu.edu), and we encourage all investigators who have interest in working with this rich dataset to contact the corresponding author.

The large, diverse undergraduate population at VCU makes it an ideal environment to pilot the effectiveness of prevention intervention. The ability to successfully implement effective tailored interventions will depend on studying etiological pathways, prevention/intervention, and underlying mechanisms specific to individual groups. VCU has nearly 50% minority students, and 40% first generation college students. Sample demographics for the S4S participants do not differ significantly from the overall university study population: 17% identify as Asian; 17% identify as African American, 6% identify as Hispanic, 51% identify as Caucasian, 6% identify as more than 1 race, and 3% report other/unknown; 60% are female and 40% male. Analyses of the Spit for Science data formed the foundation for the development of our personalized feedback program, and the registry eventually will be used to recruit subjects for a randomized clinical trial to evaluate the effectiveness of personalized feedback programs as compared to standard BMIs.

Substance use in the S4S sample is considerable, and consistent with other national surveys of college students (Cho et al., 2015; Schulenberg et al., 2019; Substance Abuse and Mental Health Services Administration, 2016): 75.3% of incoming freshmen have tried alcohol on at least one occasion (and virtually all of those individuals report drinking on 5+ occasions). This proportion rises steadily across the college years. By spring of participants’ freshman year, 83% of participants report lifetime alcohol use, 87% by sophomore year, 92% by junior year and 94% by senior year. However, quantity and frequency of substance use show significant variability: among freshmen, 29% report drinking monthly or less, but 20% of drinkers report consuming alcohol more than 2x/week, and 5+ drinks/occasion; 53% of drinkers have blacked out; 30.2% of freshmen display no alcohol dependence symptoms, while 20.3% display 1 symptom, and 26%, 13%, and 10.2% meet criteria for mild, moderate, and severe alcohol dependence (AD), respectively. Marijuana use is also considerable and rises steadily with 45% of freshmen reporting lifetime cannabis use, 50% of sophomores, 56% of juniors and 59% of seniors.

Selection of underlying risk dimensions

We focused the initial personalized feedback on four domains. This choice was both practical and theoretical. For the initial development and testing phase, we wanted to get the program launched by building out a manageable number of domains for which we would provide feedback and resources. We also felt that a more limited number of domains would be more digestible and memorable for students. For these reasons, we settled on four initial domains, three of which reflect externalizing risk (Sensation-Seeking, Impulsivity, and Extraversion), and one of which reflects internalizing risk (Neuroticism). The disproportionate focus on externalizing risk factors is based on the literature suggesting that risky substance use in college students is more strongly influenced by externalizing characteristics than internalizing characteristics at this developmental stage (Dick et al., 2014). In the Spit for Science sample, Sensation-Seeking and other facets of impulsivity were measured using the UPPS (Lynam, Smith, Whiteside, & Cyders, 2006). This measure was developed via exploratory factor analyses of the Big Five factor model and additional commonly used measures of impulsivity. The resulting UPPS scale has four primary subscales which were administered in short form in S4S: urgency (alpha=.722, which was subsequently subdivided into positive and negative urgency; Cyders et al., 2007), (lack of) premeditation (alpha=.759), (lack of) perseverance (alpha=.683), and sensation seeking (alpha=.628). Each subscale consists of 3 items with 4 response categories ranging from “Disagree strongly” to “Agree strongly”. We collapsed three of the primary UPPS scales (urgency, lack of premeditation, lack of perseverance) into one Impulsivity domain, based on the decision that dividing impulsivity into many sub-facets would become too complicated and introduce confusion for the students. All of these impulsivity sub-facets (positive and negative urgency, lack of premeditation, lack of perseverance) loaded onto the same ANA factor, along with multiple other measures of impulsivity, with factor loadings >.5 on the impaired executive control ANA domain (Kwako et al., 2019). Conscientiousness was inversely correlated with this domain, and sensation-seeking did not load on the factor; accordingly, there is some theoretical rationale for this choice as well. We also note that we will assess students’ beliefs as to how well each of their assigned profiles “fits” them, so we will be able to empirically evaluate whether collapsing these impulsivity facets leads to reductions in perceived fit as compared to the other domains. We chose to provide feedback on Sensation-Seeking separately, due to the robust literature associating this facet of impulsivity with risky substance use (Earleywine & Finn, 1991; Quinn & Harden, 2013; Zuckerman, 2007). Sensation-seeking is also more strongly related to the incentive salience ANA domain. Extraversion (α = .84) was measured in S4S using the Big Five Inventory (BFI) (John & Srivastava, 1999). We chose to include this domain in our personalized risk feedback because data from S4S and other studies (Acton, 2003; Kendler, Myers, & Dick, 2015; Knyazev, 2004; R. E. Tarter, 1988) indicate that extraversion is associated with substance use in emerging adults. Neuroticism was also measured using the BFI (John & Srivastava, 1999) (α = .81), and reflects a disposition toward internalizing, also referred to as negative emotionality in the ANA.

Item Selection for Personalized Feedback Program (PFP).

Because we recognized that the PFP must be brief to be engaging to students, we conducted factor analyses with the items of each scale described above (UPPS scales, BFI Extraversion and Neuroticism scales), and selected the two item(s) with the highest factor loading for each subscale. For each selected item, we ran descriptive statistics with the corresponding scale scores in the Spit for Science sample, to ensure that the selected single response items adequately captured overall scores. All representative items selected for corresponding subscales had factor loadings >.75. We then computed descriptive statistics for alcohol and cannabis outcomes, including AD symptom counts, grams of ethanol, drinking endorsement rates, and heavy episodic drinking (HED) (4+ or 5+ drinks per sitting for females and males, respectively) to ensure that the reduced items also meaningfully predicted varying degrees of substance/alcohol (mis)use. For example, for the sensation seeking dimension of the UPPS, the item “I quite enjoy taking risks” had the highest factor loading (.94) on the sensation seeking factor. Responses on this item were positively associated with various alcohol and marijuana outcomes (Table 1).

Table 1.

Relationship between a sensation-seeking item selected for the PFP “I quite enjoy taking risks” with alcohol and cannabis outcomes

I quite enjoy taking risks	% Ever had a drink	Mean alcohol dependence symptom counts (SD)	Mean grams of ETOH (SD)	% Ever used Cannabis	% Used Cannabis 6+ times	Mean cannabis use disorder symptom counts (SD)	% Use cannabis at least 2–3 per week
1. Disagree strongly	76.80%	1.03(1.40)	104(390)	27.89%	16.50%	0.6 (1.53)	5.15%
2. Disagree a little	80.30%	1.16(1.43)	140(337)	35.95%	18.99%	0.57 (1.53)	5.90%
3. Agree a little	84.30%	1.40(1.52)	227(467)	47.15%	27.56%	0.93 (2.04)	9.97%
4. Agree Strongly	89.50%	1.83(1.70)	376(674)	60.99%	41.05%	1.52 (2.6)	17.64%

Once the subset of initial items had been selected, we worked with a programmer to create a personalized feedback platform, very similar to that used to provide personalized feedback surrounding alcohol use in current BMIs, except that the questions/feedback focus on the underlying risk domains and how these traits relate to substance use, rather than on current alcohol use patterns.

We recruited students from the psychology undergraduate research pool, comprised of students enrolled in psychology courses for which there are research participation requirements, for semi-structured focus groups to solicit feedback on the initial PFP prototype. Six focus groups were conducted on-line via Zoom, due to restrictions imposed by the COVID-19 pandemic, in groups of 2–4 students. Initial feedback from the 20 students who participated in the focus groups was overall very positive; participants thought that the program was engaging and graphically pleasing, liked that the content felt personalized to them based on the quiz and avatar builder, and expressed optimism at the opportunity to engage with the many resources offered. A majority of participants noted that they wished the quiz included more questions as well as a neutral response option.

Based on this feedback, we expanded the number of items included in the PFP from 10 to 20 by selecting additional items with the highest correlations with total scale scores for each domain. Two additional items were included for each of the extraversion, neuroticism, and sensation seeking scales; and one additional item was included for each of the four impulsivity sub-scales (negative and positive urgency, lack of premeditation, and lack of perseverance). Table 2 shows the final items selected for inclusion in the PFP. We used the response options “strongly disagree”, “disagree”, “neither agree nor disagree”, agree”, and “strongly agree” for all items. These response options are parallel to the BFI which includes the neutral response option (compared to the UPPS scale which excludes this option) in order to maintain continuity across response options for all items, and in response to the feedback from initial focus groups.

Table 2.

Final items selected for inclusion in the PFP

Dimension	Item
Sensation seeking	I quite enjoy taking risks.
	I would enjoy the sensation of skiing fast down a high mountain slope.
	I welcome new and exciting experiences and sensations, even if they are a little frightening and unconventional.
	I would like to learn to fly an airplane.
Extraversion	I see myself as someone who is outgoing, sociable.
	I am someone who tends to be quiet. (reverse-coded)
	I am someone who is talkative.
	I see myself as someone who generates a lot of enthusiasm.
Neuroticism	I see myself as someone who is relaxed, handles stress well. (reverse-coded)
	I worry a lot.
	I see myself as someone who is depressed, blue.
	I see myself as someone who is emotionally stable, not easily upset. (reverse-coded)
Impulsivity	When I am upset I often act without thinking.
	Sometimes when I feel bad, I can’t seem to stop what I am doing even though it is making me feel worse.
	I usually think carefully before doing anything. (reverse-coded)
	I like to stop and think things over before I do them. (reverse-coded)
	I finish what I start. (reverse-coded)
	Once I get going on something I hate to stop. (reverse-coded)
	I tend to act without thinking when I am really excited.
	I tend to lose control when I am in a great mood.

Results

Beta version of the Personalized Feedback Program (PFP)

Together with our colleagues at the university health and wellness center, we created text introducing the PFP, explaining that virtually all health behaviors are influenced by genetic predispositions and environments, and that by knowing about their own predispositions, students can use that information to understand their own risk and make better choices to impact their own health and happiness. Students answer the questions for the 20 selected items tapping into the four domains described above. After answering all items, and before receiving their feedback, students create personalized graphical avatars that then accompany the presentation of their risk profiles (Figure 1). These personalized avatars are an effort to have the material further resonate with the students and feel personalized. Hair styles and skin tones representing multiple racial/ethnic identities were included, and expanded after initial student focus group feedback, to facilitate personalization for diverse students. Students’ profiles can be shared via social media, which generates a link for individuals receiving the material to go through the program themselves and receive their own profile and associated feedback/resources. In this way, we aim for the program to have the potential for “organic spread”, meaning that students will willingly share the program and additional students can be exposed to the material.

Figure 1.

Example personalized graphical avatar accompanied by risk profile across the four domains

To generate the profiles, the PFP sums the students’ responses to the items for each domain and compares the students’ score to a pre-calculated cut-off score determined by the median score based on our administration of the survey previously in the undergraduate sample. Students are then classified in a binary fashion on each domain according to their response profile (Sensation Seeker/Playin’ it Safe; Extravert/Introvert; Spontaneous/Conscientious; Worrier/Cool as a Cucumber). We note that the labels do not directly correspond to the dimensions as labeled in the research literature, for example labeling high impulsivity as “Spontaneous”, and high neuroticism as “Worrier”. These changes were made in consultation with colleagues at our health and wellness center to use nomenclature that would not be stigmatizing to the students in order to conform to the motivational spirit of personalized feedback programs. Individuals who have a score equal to or greater than the median score are assigned the riskier categorization, and individuals who have a score below the median score are assigned to the less risky categorization.

We also provide dimensional feedback in addition to the binary assignments, by presenting a bar for each domain, anchored with the domain labels at each end (see Figure 2). The bar reflects which quartile the student’s score places them into, based on cut-offs from the undergraduate sample. In this way we aim to provide students with a rough indication of how high or low they are on each dimension in addition to their binary assignment.

Figure 2.

Example dimensional presentation of risk profiles for the four domains

The decision to assign binary labels in addition to the dimensional feedback was based on the extensive literature suggesting that individuals resonate with labels. This phenomenon is evident in everyday life from the prevalence of BuzzFeed personality quiz results shared on young adults’ Facebook feeds (Berberick & McAllister, 2016), to the popularity of astrological signs and horoscopes in newspapers and other media (Allum, 2010; Bader, Baker, & Mencken, 2017), to the widespread utilization of the Myers-Briggs Type Indicator (MBTI) in the workplace (Rideout & Richardson, 1989; Varvel, Adams, Pridie, & Ruiz Ulloa, 2004). The ubiquity of such label-producing systems may be a result of their ability to satisfy two basic human motivations: the need to belong to a group and the need to feel like a unique individual (Hornsey & Jetten, 2004). When analyzing the popularity of on-line identity quizzes, Berberick & McAllister (2016) postulated that these quizzes “appear to serve the needs of users through giving them shareable, simple, prepackaged content that allows them to easily represent their self-brand to friends...in a highly accessible, fun, and playful way.” We aimed to mimic the structure of such quizzes with our program by providing students with approachable personality types that empower them to both better understand their unique traits and share their types with friends to see what commonalities may exist.

After receiving their profile, students can access nonjudgmental feedback about each of their temperamental patterns, about how these patterns might put them at elevated risk for alcohol or other substance use/mental health challenges, and about available resources to support them. Information is provided for each of the four domains by clicking on the assigned domain label. Figure 3 provides sample feedback for the Sensation Seeker profile. Students also have the ability to explore the other personality types not assigned to them.

Figure 3.

Personalized feedback for the Sensation Seeker profile

Once students fully examine their unique profile, they are directed to a section that prompts them to explore their goals. Students are given the opportunity to articulate their personal short- and long-term goals, and asked to identify the behaviors, based on their unique personality profiles, that can either help or hinder progress toward those goals. Students are once again given the specific institutional resources which are available and asked to identify which resources they plan to use to help them reach their goals. This approach capitalizes on the evidence-based practice of goal exploration and setting as a facilitator of behavior change (Bailey, 2017; King, 2001), and uses goal and action planning constructs from BMIs which engage students in reinforcing commitment to change (Center for Substance Abuse Treatment, 1999). Students’ results are collated in “Your Personalized Strategy for Success”, a section that summarizes their personalized profile, goals, and university resources, that can be downloaded/emailed to themselves. The PFP takes approximately 15–20 minutes for students to complete (with total time obviously dependent on how much time they spend writing out goals and reviewing their results).

Finally, there is a “Choose Your Own Adventure” section that is designed to help students weigh the outcomes of their decisions and how their choices align with their internal values and stated goals. Fashioned to incorporate these basic principles of Motivational Interviewing (Miller & Rollnick, 2013), this section encourages students to reflect on how their day-to-day actions are connected to their long-term goals, and what discrepancies might arise between the two. Students are presented with various scenarios (e.g., a friend shows up at your door and wants you to go to a party; you are studying for an exam) and response choices (e.g., go to the party, stay home and study, go to the party for just a little bit). Each response leads to a different outcome, either moving them towards or away from their stated goals. These outcomes prompt students to reflect on the pros/cons of each decision and whether other choices would have been more in line with their goals. Graphics are personalized to the student’s avatar, and a path visually reinforces how different choices move them closer or further from their prepopulated personalized goals. This section is intended to empower students to connect their patterns of behavior or decision making with goal attainment, in the context of their natural personality styles.

We are currently in the process of launching the next phase of evaluating the PFP, an open trial which will involve administering the PFP to a large group of undergraduate students, with an accompanying survey to assess initial responses to the PFP, including future intentions regarding alcohol and drug use and campus resource use, as well as satisfaction with the PFP. A subset of the students will complete a 30 day follow-up to assess changes in substance use, mental health, and campus utilization, followed by a second round of focus groups. As is standard practice by our team, all planned analyses will be pre-registered in the Open Science Framework (OSF). Following the open trial, we plan to make a second round of revisions to the PFP, and launch a randomized clinical trial (RCT). The RCT will evaluate the effectiveness of the PFP as compared to a waitlist control and a computer-delivered intervention based on BMI content/principles. Additionally, we plan to test a combined condition in which students will receive the Personalized Feedback Program, in conjunction with the computer-delivered intervention. This trial will be registered through ClinicalTrials.gov and on OSF.

Discussion

In this paper, we describe the development of a new approach to substance use prevention among college students that shifts focus away from current substance use, attending instead to etiological predisposing risk factors. We view this approach as complementary to existing BMI strategies, which provide more detailed information about current substance use and associated risks and harm reduction strategies. In fact, we hypothesize that receiving information that one is at elevated risk (through the PFP) may enhance students’ interest in, and application of, the substance use material routinely included in BMIs. Further, by focusing on underlying risk factors, many of which impact multiple forms of substance use, we hypothesize that personalized risk assessments provide an approach for addressing alcohol and other drug use, which is increasingly common, particularly among young adults. An additional benefit of focusing on underlying risk factors is that risk information can be provided for students who have not yet initiated drinking, or for those who are early in their drinking careers.

The on-line nature of the personalized feedback platform will allow for easy adoption, dissemination, and modification. We anticipate that this program will be most useful as a universal or selective prevention program, and/or as an adjunct to in-person, clinician-assisted BMIs for indicated students. We hope that enhanced graphic design and gamification, as well as the personalization of feedback, designed to parallel popular social media “quizzes” will appeal to emerging adults. Although we are initially developing the platform for college students, upon demonstration of efficacy, modifications to make the PFP appropriate and available to other groups (younger students, non-college students) would be straightforward.

Our personalized feedback program aims to address a central challenge in the prevention field: how do you get large numbers of young people to willingly (and dare we add, enthusiastically) engage with prevention materials designed to provide helpful information for improving health outcomes. In our experience, completing alcohol programming is not something that most students willingly want to do. We found that, even when incentivized to complete a voluntary alcohol education module online, only a very small number of students (n<40 out of our incoming class of several thousand) chose to participate. Universities have various ways of addressing student’s reluctance, to include making training mandatory for course registration, or mandatory for students with conduct infractions as a condition of continued enrollment at the university. We propose another, more radical idea: that by making the material more personalized and engaging we may be able to get students to want to participate, and, beyond that, by making use of sharing and social media, students can also actively recruit other students into prevention programming. Human beings have great interest in learning about themselves, as evidenced by the tremendous popularity of mainstream personality tests, such as the Myers-Briggs, and more recently, countless social media quizzes that purport to do everything from tell you what Marvel superhero or dog breed you most closely resemble, to providing wine recommendations specific to your personality! There has been exponential growth in individuals’ participation in on-line direct to consumer genetic testing over the past several years (Janssens, 2019), with many of the most widely accessed “genetic risk scores” available through free on-line websites relating to substance use and mental health outcomes (Folkersen et al., 2019). Accordingly, we believe that there is tremendous potential for expanding prevention programming to incorporate the delivery of personalized risk information that extends beyond current substance use, and indexes other known risk dimensions that predict future problematic substance use.

In addition to assessing whether expanded personalized feedback is an effective means of preventing or reducing risky substance use, as this literature grows it will also be important to assess how personalized feedback programs impact behavior change. Despite the prevalence of brief alcohol interventions, the mechanisms by which these programs lead to behavior change remains largely unknown, with many theoretical frameworks specified, and measurement of intermediary mechanisms in college intervention studies often inconsistent or absent (Miller et al., 2015). The emerging literature on personalized prevention has also thus far largely failed to specify mechanisms of change (Conrod et al., 2011, 2013). We propose that personalized feedback programs may create behavior change via mechanisms described in the Health Belief Model, which posits that an individual’s understanding of their susceptibility to particular health problems influences one’s motivation for preventive behaviors (Rosenstock, 1974). We hypothesize that increased knowledge of one’s underlying risk for substance use problems will enhance motivation to engage in healthier behaviors and increase readiness to change behavior among those already engaging in risky drinking. Readiness to change is believed to be a central target for BMI effectiveness (Miller et al., 2015).

Another primary goal of our intervention is to increase participants’ awareness and use of campus supports and other resources to help them deal with their risks more effectively. To determine if this type of change has taken place, and to address the impact that previous and existing support services may have on outcome, we plan to obtain university records about service utilization, and to assess any community-based services students might have received that would not be in college records.

We have chosen to build out the initial version of our personalized feedback program to focus on a subset of personality dimensions (1) that have been robustly associated with substance use across countless studies with varying population-characteristics (Dennhardt & Murphy, 2013; Jessor, 1991; Kendler, Gardner, & Dick, 2011; Li et al., 2017; Shadur & Lejuez, 2015; Slutske et al., 2002), (2) that index underlying etiological risk and map onto the developmental genetics literature and underlying neurobiological mechanisms involved in substance use outcomes, and (3) that we believe will be of interest to students, toward our goal of creating natural engagement and spread. We are actively engaging students in the creation of the personalized feedback program, soliciting feedback via small focus groups of students recruited based upon differing personality profiles to ensure broad representation of feedback across the dimensions.

We note that personalized feedback could be expanded to incorporate many other dimensions of relevance. For example, personalized feedback modules could target neurobehavioral and neurobiological risk factors associated with externalizing and internalizing pathways, such as those described in Kwako et al. (2016; 2019). A well-established neurobehavioral risk factor is impulsive decision-making (Vassileva & Conrod, 2019). Studies reveal that simply providing personalized feedback on tasks of decision-making is associated with improved neurobehavioral performance (Brand, 2008; Brand, Laier, Pawlikowski, & Markowitsch, 2009). Recently, novel neurocognitive remediation strategies based on behavioral economic and neuro-economic approaches that target delay discounting, a key impulsivity dimension associated with constricted temporal horizons and excessive valuation of immediate rewards despite long-term negative consequences, have demonstrated promising results (Bickel, Johnson, Koffarnus, MacKillop, & Murphy, 2014; Bickel et al., 2016; Gray & MacKillop, 2015). Though no existing neurocognitive intervention programs to date incorporate personalized feedback, there is evidence that providing direct feedback to patients from their results on clinical neuropsychological testing improves quality of life and social adjustment (Rosado et al., 2018). Similarly, neurofeedback, one of the earliest personalized feedback interventions which uses real-time displays of brain activity (fMRI) to teach self-regulation of brain function (Watanabe, Sasaki, Shibata, & Kawato, 2017), holds promise for relapse prevention to control predisposing risk factors such as stress and craving (Pandria, Athanasiou, Konstantara, Karagianni, & Bamidis, 2020).

Mapping the externalizing and internalizing personality risk profiles onto integrative neuroscience-based RDoC frameworks (Kwako et al., 2016, 2019), promises to identify novel targets for prevention and intervention that may supplement existing programs and inform the development of new programs (Krystal & O’Malley, 2015; Litten et al., 2015; Sher, 2015). There is an unexplored therapeutic potential of integrating personality with neurocognitive and other intervention modules, based on individual risk factors, which may optimize the precision and efficacy of targeted intervention approaches for addictions. The development and evaluation of personalized feedback interventions will also set the stage for the incorporation of genetic information into personalized prevention intervention programming. There has been an exponential increase in the number of individuals participating in direct to consumer genetic testing (Regalado, 2019), and uploading their genetic data to publicly available websites to calculate genetic risk scores (Folkersen et al., 2019). In one of the most popular of these websites, substance use and mental health outcomes make up half of the top twelve most requested genetic risk scores (Folkersen et al., 2019), with alcohol use disorder being third. We have found that a high percentage of emerging adults (~80–90%) report being interested in receiving their genetic risk information for substance use and mental health disorders (Driver, Kuo, Dick, & Spit for Science Working Group, 2020). It has been suggested that this information can be used to motivate individuals to practice healthier lifestyle behaviors (McBride, Koehly, Sanderson, & Kaphingst, 2010); thus, providing genetic risk information alongside other personalized information may further prevent or reduce risky drinking behaviors. We believe the delivery of personalized genetic risk information will be increasingly commonplace in the future (Ashley, 2015), and our platform naturally lends itself to these future developments.

Although we believe personalized feedback programs have great potential, ultimately, the proof will be in the proverbial pudding. Through this paper, we hope to encourage more researchers to experiment with making pudding.

Public Health Significance:

College represents a unique opportunity to intervene and have positive life-course altering health benefits for a significant, and increasingly diverse portion of the population. Identifying innovative new programs to curtail risky substance use is a public health imperative. Integrating findings from basic genetic epidemiological research about pathways of risk represents one such novel approach.