Integrating Theory with Education about Genetic Risk for Alcohol Use Disorder: The Effects of a Brief Online Educational Tool on Elements of the Health Belief Model

Precision medicine research efforts have advanced our understanding of the role of genetic factors in the development of complex health outcomes, such as cancer [1], heart disease [2], and psychiatric conditions [3]. Using newly developed methods, such as the calculation of polygenic risk scores, an individual’s genetic information can be used to provide insight into their underlying genetic predisposition to developing different health outcomes [4]. However, the utility of genetic risk information relies on the assumption that individuals will use the information to change their behavior in healthy ways in order to reduce their risk of developing health problems. Researchers have begun to investigate the impact of receiving genetic information for health conditions on a number of different health-related behaviors, such as smoking cessation, diet, physical activity, and sun protective behaviors [5, 10]. Despite the enthusiasm surrounding precision medicine, systematic reviews show that there is no consistent or strong evidence that receiving genetic risk information alone has an impact on preventative behaviors [5, 10].

There are two primary reasons why the provision of complex genetic information may not be associated with risk-reducing behavior change. First, genetic information for complex outcomes such as substance use and psychiatric conditions is provided as a polygenic risk score, rather than a diagnostic test result, which is what might be more typically imagined as the outcome of a genetic test. Polygenic risk scores are also quite different from genetic feedback based on single genetic variants that may relay information about increased risk for developing alcohol-related cancers or decreased risk for alcohol dependence [11, 13]. Polygenic risk scores capture an individual’s genetic risk by summing information about risk-enhancing variants detected across the genome [14] and, rather than providing definitive information about whether or not someone will develop a particular condition, instead provide an estimate of probability of developing a condition. Not only is probabilistic information difficult for people to grasp in general, polygenic risk scores are typically presented as relative risks – which are especially challenging [15]. There is additional complexity as a result of current limited predictive ability and limited utility in non-European populations [14]. Research has demonstrated that, in fact, there is considerable misunderstanding of complex genetic concepts [16, 18]. Another central concern is that most studies assessing behavior change in response to receiving genetic feedback were not guided by theory [6, 8, 10, 19]. Most of the interventions were not designed to target or address previously established determinants of behavior change including self-efficacy, response efficacy, or motivation [6, 8, 20]. Thus, the impact and utility of providing genetic risk information to individuals could be improved by developing personalized feedback programs and educational interventions that are informed by behavior change theories.

The Health Belief Model is one common conceptual framework used to explain behavior, changes in behavior, and maintenance of behaviors in relation to health-related outcomes [21]. The central tenet of this theoretical model postulates that perceived susceptibility to a condition, perceived severity of a condition, perceived benefits of an action, perceived barriers to completing an action, and perceived self-efficacy of completing an action are key determinants of motivation for health behavior [21]. Demographic and personal characteristics, such as age, education, and race/ethnicity, may modify these health perceptions [21].

Empirical work surrounding health behavior and behavior change theories show that interventions designed to incorporate and target key elements of the Health Belief Model are effective in promoting behaviors for positive outcomes [21]. For example, it was found that among tailored interventions, those based on components of the Health Belief Model and those that included a physician recommendation had a stronger effect on likelihood of getting a mammogram as compared to those that were tailored based on demographic characteristics or used motivational interviewing [22]. Additionally, a meta-analysis of 19 studies utilizing different theory-driven interventions demonstrated that the three interventions guided by the Health Belief Model significantly improved oral health [23].

The Health Belief Model has also been used to study alcohol and drug use behavior, with a few studies using the theory to design substance use and addiction-related educational interventions in order to modify substance use behaviors [24]. An intensive educational prevention for drug use based on the Health Belief Model was shown to significantly alter key dimensions of the Health Belief Model, including motivation and intentions to engage in preventative behaviors, in the intervention condition. The prevention program was designed for high school students and consisted of four 75-min sessions [25]. Male elementary school students in Iran had significant changes in beliefs related to drug dependency and preventative behaviors after receiving an educational intervention that consisted of two 1.5-h sessions with both lecture and group discussion components. There were significant correlations between Health Belief Model components, awareness, and preventative behaviors [26]. Finally, Younis and Naji [27] (2021) demonstrated a complex interrelatedness between components of the Health Belief Model, with significant correlations between perceived severity of substance use and perceived benefits of preventing addiction, between perceived benefits and perceived barriers, and between perceived benefits and intentions.

We designed a brief, online educational intervention intended to accompany the return of genetic risk information for alcohol use disorder (AUD) with the intention of achieving two main goals: increasing understanding of polygenic risk scores for AUD and promoting risk-reducing behavior change. To assess the efficacy of the educational intervention, we conducted a randomized controlled trial (RCT) that consisted of three conditions: a control condition in which no educational information was provided, an intervention condition in which participants received online educational information about AUD, and an intervention condition in which participants received online educational information about polygenic risk scores and AUD.

Figure 1 illustrates our conceptual model for the RCT and demonstrates how the provision of genetic information, along with accompanying educational information, may promote behavior change, as conceptualized using the Health Belief Model. The Health Belief Model suggests that individual factors, such as demographic characteristics and knowledge, influence individual health beliefs. These health beliefs include perceived risk of developing an outcome, perceived severity of having an outcome, perceived benefits of risk-reducing behaviors, and perceived barriers to risk-reducing behaviors, which influence the likelihood of an individual engaging in risk-reducing behaviors. As demonstrated by the green boxes in Figure 1, we expect that providing educational information about AUD and polygenic risk scores will impact one’s knowledge and, therefore, alter beliefs regarding perceived severity of having AUD, perceived benefits of risk-reducing behaviors, and perceived barriers to risk-reducing behaviors. The present study aimed to test the hypothesis that providing participants with educational information about AUD and polygenic risk scores will impact these elements of the Health Belief Model. As demonstrated by the blue boxes in Figure 1, we expect that providing personalized genetic risk information for AUD will impact one’s knowledge and alter perceived risk of developing AUD. Changes in these different health beliefs are expected to influence the likelihood of engaging in risk-reducing behaviors.

The main goal of the present study was to assess whether the educational information about polygenic risk scores and AUD altered elements of the Health Belief Model that are known to be associated with motivations and intentions to change behavior. We assessed the association between beliefs regarding AUD and demographic characteristics, evaluated the impact of the educational intervention on components of the Health Belief Model, and assessed whether the effect of the intervention was moderated by demographic characteristics. These analyses focused strictly on examining the effectiveness of the educational intervention. The RCT was conducted in a sample of college students because college students are entering a critical period for the development of AUD [28] and college students engage in high rates of risky drinking behaviors [29]. They are also a population with high use and acceptability of new technologies and online resources [30] and, due to their relative youth, will likely be impacted by the incorporation of genetic information into health provision. The present study can provide foundational knowledge about how to effectively provide education alongside the return of genetic feedback in hopes of promoting positive behavioral changes.

The data used in the present study were collected as part of a RCT that evaluated the impact of educational information on understanding of polygenic risk scores for AUD [15]. The RCT was conducted online with college students recruited from a diverse, urban university and consisted of three conditions: a control condition in which participants received no educational information (control condition), an intervention condition in which participants received general educational information about AUD (AUD Edu condition), and an intervention condition in which participants received educational information about AUD and polygenic risk scores (PRS Edu + AUD Edu condition). The educational information about AUD included a definition and information about the prevalence of AUD, consequences, and risk factors. In addition, simple graphics were used to communicate ways to reduce risk of developing alcohol problems, such as measuring drinks, finding alternative activities, and avoiding places that trigger drinking. The content was developed based on educational information available through public websites, including the National Institute of Alcohol Abuse and Alcoholism, Mayo Clinic, and the National Survey on Drug Use and Health. The educational information about polygenic risk scores provided a general explanation of genetic variation, risk variants, and the creation of polygenic risk scores and provided specific information regarding the application of polygenic risk in the context of AUD. The web-based educational tool used to communicate the educational content in the present study was adapted from the polygenic risk score dynamic explainer created by a team of geneticists, clinicians, software developers, and data visualization experts as a collaboration between the Broad Institute’s Cardiovascular Disease Initiative, IBM Research, and Massachusetts General Hospital [31]. Additional details about the design of the educational information are presented in Driver et al. 2022.

Participants responded to a series of survey items after completing the educational intervention specific to their condition. Because participants in the control condition did not receive educational information, these participants only completed the study survey. The survey was designed to assess different beliefs regarding AUD that correspond to elements of the Health Belief Model. The beliefs included perceived chance of developing AUD, worry about developing AUD, perceived susceptibility of developing alcohol problems, perceived severity of having alcohol problems, perceived benefits of taking actions to reduce risk of developing alcohol problems, and perceived barriers to taking risk-reducing actions. The analytic sample consisted of 325 participants (70.4% female; 43.6% white; mean age = 18.9), with 109 participants in the control condition, 105 participants in the AUD Edu condition, and 111 participants in the PRS Edu + AUD Edu condition. All procedures were approved by the University’s Institutional Review Board. The study details are further outlined in Driver et al. 2022.

Perceived chance of developing AUD was measured using an item adapted from Lipkus et al. [32] (2015). The original item was developed to assess the chance of developing a nicotine addiction as a consequence of cigarette smoking. The adapted item used in this study was “What do you think is your chance of developing alcohol use disorder?” Response options included “no chance,” “very unlikely,” “unlikely,” “moderately likely,” “likely,” “very likely,” and “certain to happen.” Chance of developing AUD was coded as a semicontinuous variable (range 0–6), with higher scores reflecting a greater perceived chance of developing AUD.

Worry about developing AUD was measured using an adapted version of the Cancer Worry Scale [33]. Eight items from the adapted version of the Cancer Worry Scale were scored from 1 to 4 and summed to create an overall score with a range of 8–32. Example items included “How often have you thought about your chances of getting alcohol use disorder?” with response options of “never,” “rarely,” “sometimes,” and “frequently” and “How concerned are you about the possibility of getting alcohol use disorder 1 day?” with response options of “not at all concerned,” “not very concerned,” “somewhat concerned,” and “very concerned.”

Perceived susceptibility of developing alcohol problems was measured using adapted items from the Health Beliefs about Mental Illness Instrument [34]. Five items were scored from 1 (strongly disagree) to 5 (strongly agree) and summed to create an overall score with a range of 5–25. Example items included “There is a good possibility that I will develop alcohol problems in the next 10 years” and “I feel I will develop alcohol problems in the future.”

Perceived severity of having alcohol problems was measured using adapted items from the Health Beliefs about Mental Illness Instrument [34]. Seven items were scored from 1 (strongly disagree) to 5 (strongly agree) and summed to create an overall score with a range of 7–35. Example items included “The thought of having alcohol problems scares me” and “Alcohol problems would threaten relationships with my family or friends.”

Perceived benefits of risk-reducing behavior were measured using adapted items from the Health Beliefs about Mental Illness Instrument [34]. Four items were rated from 1 (strongly disagree) to 5 (strongly agree) and summed to create an overall score with a range of 4–20. Example items included “Taking action to reduce my risk for developing alcohol problems would make me feel better about myself” and “A burden would be lifted off me if I were to take action to reduce my risk for developing alcohol problems.”

Perceived barriers to risk-reducing behavior were measured using adapted items from the Health Beliefs about Mental Illness Instrument [34]. Five items from the instrument were rated from 1 (strongly disagree) to 5 (strongly agree) and summed to create an overall score with a range of 5–25. Example items included “Taking action to reduce my risk for developing alcohol problems is embarrassing” and “Health professionals would not understand someone like me if I went to them because of my alcohol problems.” Table 1 displays information about measures used to assess beliefs regarding AUD, including the number of items that were included to assess each belief, the potential range of scores for each belief, and Cronbach’s alpha to measure internal consistency of the items used to assess each belief.

Participants were asked to indicate their sex assigned at birth. Sex was coded as male or female.

43.6% (n = 140) of the sample self-identified as White, 24.3% (n = 78) of the sample self-identified as Black/African American, 18.7% (n = 60) of the sample self-identified as Asian, 5.9% (n = 19) of the sample self-identified as Hispanic/Latino, 0.3% (n = 1) of the sample self-identified as American Indian/Alaska Native, 0.3% (n = 1) of the sample self-identified as Native Hawaiian/Pacific Islander, and 6.9% (n = 22) of the sample self-identified as more than one race. Because of the small sample sizes across several racial/ethnic backgrounds, race/ethnicity was coded as a binary variable to increase power to detect differences in beliefs regarding AUD. Individuals who self-identified as White were categorized as White, and individuals who self-identified as Asian, Black/African American, Hispanic/Latino, American Indian/Alaska Native, Native Hawaiian/Pacific Islander, or more than one race were categorized as non-White.

Drinking status was measured using the frequency item (“How often do you have a drink containing alcohol?”) from the Alcohol Use Disorder Identification Test (AUDIT) [35]. Because this was an early college sample and variation in frequency of use was limited, a binary variable to indicate drinking status was used in the analyses. Participants who responded “never” to the alcohol frequency item were categorized as nondrinkers to indicate that they had not previously used alcohol, and participants who responded at least “monthly or less” were categorized as drinkers to indicate that they had previously used alcohol.

Participants answered separate questions about alcohol problems for four types of biological relatives: mother, father, aunts/uncles/grandparents, and siblings [36]. An example question was: “Do you think your biological mother has ever had a drinking problem? (by drinking problem, we mean that her drinking caused problems at home, at work, with her health, or with the police, or that she received alcohol treatment).” The questions were repeated for each type of relative. Response options for each question were “yes,” “no,” and “I don’t know.” Family history items related to alcohol problems were combined into an overall binary variable to indicate whether or not the participant had any first- or second-degree relatives with a history of alcohol problems. 50.5% of the full sample reported having a family history of alcohol problems.

Correlations between beliefs regarding AUD were calculated. Independent sample t tests were used to assess whether there were mean differences in each belief regarding AUD (i.e., perceived chance of developing AUD, worry about developing AUD, perceived susceptibility of developing alcohol problems, perceived severity of having alcohol problems, perceived benefits of taking actions to reduce risk for developing alcohol problems, and perceived barriers to taking risk-reducing actions) across four demographic characteristics: sex, race/ethnicity, drinking status, and family history of alcohol problems. Follow-up linear regressions were used to assess whether the relationship between sex and beliefs regarding AUD, as well as race/ethnicity and beliefs regarding AUD, were consistent while controlling for frequency of alcohol use. ANOVA methods were used to assess mean differences in beliefs regarding AUD between the control condition, AUD Edu condition, and PRS Edu + AUD Edu condition. Post hoc tests were conducted to examine where the differences occurred between the three conditions. Two-way ANOVAs were used to assess whether the effect of the intervention condition was moderated by demographic characteristics. All analyses were conducted using R 4.0.4 software [37]. The analytic plan and hypotheses were preregistered through the Open Science Framework (osf.io/efh6j).

Table 2 displays correlations between beliefs regarding AUD. Chance of developing AUD, worry about developing AUD, and perceived susceptibility were moderately correlated (r > 0.60). This suggests these different constructs may be measuring similar beliefs related to one’s perceived absolute risk of developing AUD. Some of the beliefs were weakly associated; for example, the correlation between perceived severity of having alcohol problems and perceived barriers to risk-reducing actions was 0.13. Other beliefs were not significantly associated; for example, there was no relationship between perceived benefits and perceived susceptibility (r = 0.02). This suggests that some elements of the Health Belief Model are distinct beliefs.

In general, demographic characteristics were significantly associated with perceived chance of developing AUD, worry about developing AUD, perceived susceptibility of developing alcohol problems, and perceived severity of having alcohol problems, and were not associated with perceived benefits and barriers of risk-reducing actions. Table 3 displays results from the t tests used to compare mean differences in beliefs regarding AUD across demographic characteristics.

Individuals who self-identified as White compared to non-White had a higher perceived chance of developing AUD, worry about developing AUD, perceived susceptibility of developing alcohol problems, and perceived severity of having alcohol problems. This same pattern of results was observed for drinkers compared to nondrinkers, individuals with a family history of alcohol problems compared to those without a family history of alcohol problems, and females compared to males.

We ran a series of supplementary linear regression analyses to further examine whether sex and racial/ethnic differences in beliefs regarding AUD held after accounting for frequency of alcohol use, given prior literature of sex and racial/ethnic differences in rates of drinking [38, 40]. Results indicated that sex and racial/ethnic differences in beliefs regarding AUD remained consistent when controlling for frequency of alcohol use (Table 4).

Table 5 displays the means and standard deviations for each belief related to AUD across the three conditions as well as the results from ANOVAs that assessed whether there were mean differences in the beliefs across the conditions. There were no significant mean-level differences in worry about developing AUD, perceived susceptibility and severity of developing alcohol problems, or perceived benefits and barriers of risk-reducing actions across the three conditions. However, there was a significant difference in perceived chance of developing AUD between the three conditions. Post hoc analyses revealed that individuals in the PRS Edu + AUD Edu condition reported higher perceived chance of developing AUD as compared to individuals in the control condition (adj. p < 0.01), but there was no difference between the control condition and AUD Edu condition (adj. p = 0.07) or between the AUD Edu condition and PRS Edu + AUD Edu condition (adj. p = 0.70). In addition, there were no significant interactions between the intervention condition and demographic characteristics for any of the six beliefs regarding AUD (Table 6).

This study evaluated how a brief, online educational tool intended to increase understanding of polygenic risk scores for AUD and promote risk-reducing behavior change impacted beliefs regarding AUD. The findings from the present study do not support the hypothesis that providing participants with educational information about AUD alters worry about developing AUD, perceived susceptibility of developing alcohol problems, perceived severity of having alcohol problems, perceived benefits of risk-reducing actions, and perceived barriers to risk-reducing actions.

Overall, data from this study indicated that providing educational information about AUD and polygenic risk scores did not change beliefs regarding AUD, with the exception of perceived chance of developing AUD. Education about the role of genetic variants in conjunction with information about AUD appeared to increase one’s perception about their perceived chance of developing AUD. This could suggest that young adults underestimate their risk of developing AUD and education about the role of genetic factors in the development of AUD leads to changes in risk perception that may more accurately reflect their risk. This is contrary to the field of breast cancer, in which women generally overestimate their risk for breast cancer [41, 42], and education about risk through genetic counseling decreases their worry and risk perception [42, 43].

Results from this study showed there were key demographic differences in beliefs regarding AUD, specifically regarding perceived chance of developing AUD, worry about developing AUD, perceived susceptibility of developing alcohol problems, and perceived severity of having alcohol problems. Generally, as compared to those with lower risk (i.e., individuals who do not drink and those without a family history of alcohol problems), individuals with more risk (i.e., current drinkers and those with a family history of alcohol problems) believed that they were more at risk for developing alcohol problems and that alcohol problems would lead to more severe consequences in their lives. Interestingly, females indicated a greater perceived risk of developing AUD and were more worried about developing AUD, even though females tend to drink less than males [39, 40]. However, the differences in beliefs between males and females, and individuals who self-identified as White and individuals who did not self-identify as White, remained significant even after controlling for frequency of alcohol use. Interactions between the intervention condition and demographic characteristics were not significant, suggesting that the educational information was equally ineffective across the different demographic groups.

It is important to note that the educational information about AUD and polygenic risk scores was designed with the intention of providing participants with personalized genetic risk information for AUD. We have previously demonstrated that the intervention increases understanding of polygenic risk scores [15], and the receipt of hypothetical polygenic risk scores for AUD indicating elevated risk is associated with intentions to seek out additional information, talk to a healthcare provider about risk, and reduce drinking behavior [44]. This suggests that providing personalized risk information may be more effective in altering beliefs regarding AUD than providing brief online educational information alone.

Additionally, previous studies with more intensive drug- and alcohol-related educational interventions have been shown to significantly impact beliefs related to the Health Belief Model [25, 27]. These interventions included face-to-face components and interventions with multiple sessions or longer sessions. Genetic counseling may serve as a unique opportunity to educate individuals about their unique and personal risk for developing substance use disorders. Utilizing genetic counselors to return polygenic risk scores for AUD may be crucial in order to promote behavior change. In the process of genetic counseling, there is two-way communication between the genetic counselor and the individual which can lead to discussion about ways to reduce risk and ways to overcome barriers that the individual foresees. Genetic counseling may influence beliefs specifically related to perceived barriers and benefits of risk-reducing actions, which in turn could increase motivation to change behavior. Aspects of genetic counseling, such as providing an assessment of risk and using what is known about the etiology of the condition to help people identify strategies they can use to achieve their own desired health outcomes, overlap with empirically supported alcohol interventions [45].

Several limitations should be considered in the interpretation of these findings. First, we only assessed beliefs regarding AUD after the intervention, so analyses capitalized on between-group differences rather than individual changes in beliefs. Second, this study was intentionally conducted in a college sample due to elevated risk for developing alcohol problems in this population; however, these findings may not be generalizable to populations of more diverse ages and educational backgrounds. Third, because of small sample sizes across several racial/ethnic backgrounds, we coded race/ethnicity as a binary variable to increase power to detect differences in beliefs regarding AUD. This limited our ability to assess differences between each of the racial/ethnic groups. Fourth, the educational tool did not include assessment or feedback in terms of the individual’s current drinking behaviors, which is a component of current alcohol interventions targeted toward college students that are primarily based on a brief motivational intervention framework [45]. Future studies can be expanded to include the provision of personalized feedback on both genetic and behavioral factors.

In conclusion, results from this study demonstrate the need to better design and refine the educational information intended to accompany the return of genetic feedback for AUD in order to promote risk-reducing behaviors. Providing more personalized educational information and personalized genetic risk information may be more effective in altering beliefs regarding AUD in a way that increases motivation for beneficial changes in behavior. Future studies should assess how interventions can better capitalize on components of behavior change theories when providing personalized genetic risk information.

We would like to thank the Pattern Team at the Broad Institute for their collaboration. We would also like to thank the research participants who completed this study. JA offers gratitude to the Coast Salish Peoples, including the xʷməθkwəy̓əm (Musqueam), Skwxwú7mesh (Squamish), and Səl̓ílwətaʔ/Selilwitulh (Tsleil-Waututh) Nations, on whose traditional, unceded, and ancestral territory they have the privilege of working. They acknowledge that their efforts in reconciliation and repatriation need to go far beyond land acknowledgments, and we hold ourselves publicly accountable to learning and doing as much as we can to support efforts in returning the land to its rightful custodians.

This study protocol was reviewed and approved by the Virginia Commonwealth University Institutional Review Board, approval number HM20023103. Written informed consent was obtained from participants to participate in the study through REDCap.

The authors have no conflicts of interest to declare.

No funding sources were utilized in this research study.

Morgan N. Driver: conceptualization, methodology, formal analysis, investigation, data curation, writing – original draft, visualization, and project administration. Sally I-Chun Kuo: conceptualization, methodology, writing – review and editing, and supervision. Jehannine Austin: methodology and writing – review and editing. Danielle M. Dick: conceptualization, methodology, resources, writing – review and editing, funding acquisition, and supervision. All authors contributed to and have approved the final manuscript.

The data that support the findings of this study are not publicly available due to the privacy of research participants but are available from the corresponding author (DMD) upon reasonable request.