Preventing Prescription Stimulant Diversion and Misuse via a Web-Based Intervention: A Randomized Controlled Trial

Abstract

Prescription stimulant diversion (i.e., giving, selling, or trading one’s medication) and non-medical prescription stimulant use (i.e., using in ways not prescribed) are common among undergraduates; however, few evidence-based interventions target these behaviors. This study evaluated the efficacy and feasibility of a 30-minute, interactive web-based intervention providing psychoeducation around diversion and non-medical use, practice refusing medication requests, and medication adherence strategies. Students (M_age=20.42 years; 74% female; 86% White) with current stimulant prescriptions from three US universities were randomized to the intervention (n=128) or attention-matched placebo (n=121) in a single-blind design, with 1- and 2-month boosters and 3- and 6-month follow-ups. Primary outcomes were diversion, non-medical use, and diversion intentions; secondary outcomes were perceived norms, perceived risk, self-efficacy to resist diversion and non-medical use, and prescriber communication. Contrary to pre-registered hypotheses, intervention participants did not report decreases in primary outcomes. There were small-to-medium effects on secondary outcomes of risk perceptions (d=0.39 [0.12–0.68]), perceived non-medical use norms (d=0.51 [0.24–0.76]), and self-efficacy to avoid non-medical use (d=0.47 [0.10–0.85]), but not on perceived diversion norms, self-efficacy to avoid diversion, and prescriber communication. Post-hoc analyses showed a 76% reduction in odds of any diversion (OR=0.24 [0.08–0.68]) and a 60% reduction of any non-medical use (OR=0.40 [0.21–0.77]) for intervention participants during the 6-month follow-up period. This intervention was acceptable and feasible to implement and evidenced some efficacy in modifying risk perceptions, self-efficacy, and perceived norms. Since diversion and misuse episodes were not reduced, future intervention refinements may tailor content to different levels of diversion and misuse risk. Registered in ClinicalTrials.gov on May 12, 2021: NCT04885166.

Attention-deficit/hyperactivity disorder (ADHD), a condition distinguished by inattentive, impulsive, and/or hyperactive symptoms (American Psychiatric Association, 2022), is associated with greater academic, interpersonal, and psychological challenges and increased risk of substance use during adolescence and young adulthood (Weyandt & DuPaul, 2013). Stimulant medications are the most common intervention for ADHD and effectively reduce symptoms (Coghill et al., 2017). Despite their efficacy, prescription stimulants pose risks, particularly in college settings, as these medications are the most frequently diverted (i.e., given away, sold, or traded) and misused class of medications on college campuses (Weyandt & DuPaul, 2013). These risks have been exacerbated by a rise in stimulant prescriptions in recent years, which could increase the availability of medication for misuse or diversion. Between 2003–2015, there was a 344% increase in stimulant prescriptions for females ages 15–44 (Anderson et al., 2018). More recently, between 2020–2021, there was a 15.1% increase in prescriptions for individuals 15–19 years and a 19.2% increase for 20- to 24-year-olds (Danielson et al., 2023). Further, recent research has shown that universities with higher rates of prescribed stimulant use evidenced higher rates of non-medical prescription stimulant use (Summit et al., 2024).

An estimated 8–17% of college students engage in non-medical prescription stimulant use, which involves using medication without a prescription or in ways not prescribed, predominantly for cognitive enhancement purposes (Benson et al., 2015; Wong et al., 2022). Non-medical use is associated with a host of negative short- and long-term outcomes, including other illicit drug use, lower grades, disrupted sleep, irritability, headaches, dizziness, negative mood, and substance use disorders (Faraone et al., 2020). Though non-prescribed students misuse stimulant medications, 32–57% of prescribed students (or students with an ADHD diagnosis) also reported non-medical use, which has been defined differently across studies, but typically involves one or more of the following: (1) taking a higher dose, (2) intentionally getting high, (3) intentionally using with alcohol or other drugs, (4) taking the medication for a reason other than prescribed, or (5) taking someone else’s medication (Francis et al., 2022; Gallucci et al., 2015; Holt et al., 2018; Sepulveda et al., 2011). Of note, in a recent review including both prescribed adolescents and adults, the prevalence estimate of past-year non-medical use was slightly lower (22.6%) (Forrest et al., 2025). Among students with prescriptions, using higher doses, more often, via different routes of administration (e.g., intranasal), and alongside other psychoactive substances, can potentiate the reinforcing properties of stimulants and increase the risk of addiction (Compton & Volkow, 2006). Taken together, these findings suggest that taking stimulants without a prescription or in ways they were not prescribed is common and may lead to negative consequences for individuals with and without ADHD.

One-third to over one-half of college students (36–58%) who are prescribed stimulants also report a lifetime history of prescription stimulant diversion (Gallucci et al., 2015; Holt et al., 2018; Schultz et al., 2017), although a recent review including both adolescents and adults reported a lower prevalence estimate (17.9%) (Forrest et al. 2025). The prevalence of diversion is consistent with the finding that most individuals who misuse stimulants obtain them from friends, peers, and acquaintances with prescriptions (Faraone et al., 2020). Even if one does not divert, being approached to divert one’s stimulant medication is common, reported by 58–76% of prescribed college students (Gallucci et al., 2015; Holt et al., 2018). Moreover, the frequency and intensity with which students are approached for their medications is a significant predictor of diversion (Holt et al., 2018), suggesting that helping students to turn down requests for their medication ultimately may reduce diversion and non-medical use.

The college setting poses unique challenges to adhering to one’s medication and resisting diversion. Students are seeking more autonomy, engaging in more risky behavior, experiencing less parental oversight, managing their medication independently, and navigating more variability in their day-to-day schedules (Schaefer et al., 2017). These conditions create “a perfect storm for non-adherence during the transition to college” (Schaefer et al., 2017, p. 707): students may use less frequently to avoid side effects, mix infrequent use with misuse, use to get high, and/or overuse during academically demanding times. Moreover, diversion and non-medical use among prescribed students tend to co-occur: diversion is more likely among students who misuse their medication (DeSantis et al., 2013; Sepulveda et al., 2011). Reducing diversion would decrease the availability of stimulants used for non-medical use. Deterring non-medical use among prescribed students may reduce psychological morbidity.

There are few available interventions that target non-medical use and diversion. Several publicly available curricula address diversion, such as Generation Rx’s “The Impact of Misusing Prescription Stimulants” toolkit; however, to our knowledge, none are specific to prescribed students, address nonadherence, or have been evaluated using a randomized controlled trial. A recent trial focused on the prevention of non-medical use in first-year students showed that, compared to a control condition, students who received the preventive intervention had lower rates of non-medical use three months later, with changes in expectations for academic enhancement partially mediating the intervention’s effects (Antshel et al., 2025). Because this intervention was universal, however, it did not specifically target prescribed students or address unique challenges faced by students with prescriptions (i.e., diversion and non-medical use). In a separate pre-post design trial, Molina and colleagues (2020) evaluated a primary care provider-delivered intervention targeting diversion for college-aged individuals with stimulant prescriptions. Although the intervention did not decrease diversion, participants reported being approached for their medication less often, lower intentions to divert, and less frequent disclosure of their prescription status, suggesting that risk factors for diversion are modifiable (Molina et al., 2020).

To address the lack of interventions for students with stimulant prescriptions, we used a randomized controlled trial to evaluate a novel web-based intervention focused on preventing and/or reducing diversion and non-medical use. We delivered the intervention on a web-based platform because web-based interventions have been deployed successfully to reduce alcohol use in college students (Prosser et al., 2018) and were comparable to brief in-person interventions (Carey et al., 2016). In addition, they are more accessible and practical to disseminate; they can be delivered in a more standardized fashion; they are more cost-effective because they require less staff training and effort (Ondersma et al., 2011); and they allow students to engage with the content privately and at their own pace (Carey et al., 2009). Our intervention engaged participants using a web-based experiential learning platform (Kognito Conversation Platform^™) developed and hosted by a health simulation company, Kognito¹. The platform utilizes virtual humans and interactive graphics to deliver content, allows users to participate in a conversation with responsive virtual humans for the purpose of skill building, and provides feedback from a virtual coach during the conversation and following the conversation, noting which responses were more/less efficacious (Albright et al., 2016). Research evaluating interventions utilizing this platform has shown that they led to significant change in health-related cognitions and behaviors among family members supporting military personnel and healthcare providers (Albright et al., 2012, 2018; Kuzma et al., 2018) and better preparation among college students, faculty, and staff to assist college students in distress (Rein et al., 2018; Smith-Millman et al., 2022).

Theoretical Background

Several theoretical and conceptual models informed the development of the intervention. Social learning theory suggests that nonmedical prescription drug use is more likely when one’s peers engage in this behavior, likely because the behavior is perceived as more normative and less deviant, and the benefits of the behavior outweigh the costs (Peralta & Steele, 2010). These elements also can explain the diversion of prescription stimulants. Diversion is more likely among students who report non-medical use in their peer group and who overestimate the prevalence of this behavior on campus (DeSantis et al., 2013). Since these students are more likely to be approached and to see their peers engaging in non-medical use, they may view this behavior (and, by association, diversion) as less deviant. Further, if one’s peers are engaging in non-medical use, the social benefits of diversion likely outweigh the costs, because the prescribed student can provide a desired substance to his/her peers.

The theory of planned behavior (Ajzen, 1991), which contends that behavioral intentions and outcomes are a product of attitudes, perceptions of how normative a behavior is, and self-efficacy over the behavior of interest, in addition to the theory of triadic influence (Bavarian et al., 2014) provide similar, but complementary frameworks for understanding diversion and medication misuse. Non-medical use is more common among college students who hold fewer concerns about the negative health effects of illicit prescription stimulant use, perceive non-medical use to be more socially acceptable, and endorse poorer self-efficacy to avoid stimulant use (Bavarian et al., 2014; Judson & Langdon, 2009). Similar mechanisms (i.e., perceived norms, risks, and self-efficacy to avoid diversion) likely underlie diversion decisions, which is consistent with a conceptual model of diversion risk described by Molina et al. (2021).

Together, these theories point to several modifiable cognitive and behavioral factors related to diversion and non-medical use. Hawkins et al. (1992) noted that enhancing social influence resistance strategies is one way to reduce the risk of substance misuse. Taught through instruction, modeling, and role-play, these strategies allow students to recognize and resist influences to engage in substance use and to anticipate challenges they may encounter when resisting (Botvin, 1986, as cited in Hawkins et al., 1992). One prior study examined college students’ perceived effectiveness of resistance strategies for stimulant medication requests with respect to how well they deterred future requests and preserved the requestor-requestee relationship. Findings showed that internal (e.g., “I don’t feel comfortable sharing my medication”) and external explanations (e.g., “If I run out early, my doctor won’t prescribe me more”) were most effective; direct refusals with validation (“I know it’s hard to get all of your work done, but I can’t give you any of my meds”) and alternatives (e.g., “I can’t share with you, but I’ll study with you in the library”) were moderately effective; and excuses (e.g., “I don’t have any pills on me”) were least effective (Holt et al., 2020). Since some students may not have been approached for their medication yet, enhancing self-efficacy to resist requests could inoculate students against diversion in the future, consistent with predictions from the theory of planned behavior. Changing students’ perceptions of the prevalence of diversion and non-medical use (i.e., descriptive norms) also may be an effective approach for preventing diversion. This intervention strategy was among the most efficacious in reducing problematic alcohol use in college settings (Reid & Carey, 2015). Finally, increasing student-prescriber communication frequency may deter both diversion and non-medical use since prescribers can inquire about side effects, nonadherence, and diversion pressure (DeSantis et al., 2013).

Aims and Hypotheses

Our primary aim was to evaluate the efficacy of a brief, web-based intervention focused on reducing non-medical use and diversion for students with stimulant prescriptions. We hypothesized that, compared to a placebo condition, participants receiving the active intervention would report fewer episodes of non-medical use and diversion and lower intentions to divert at the 3- or 6-month follow-up assessments. We also hypothesized that the intervention group would evidence more change on the secondary outcomes: increased self-efficacy to avoid diversion and misuse, perceived harm associated with diversion and misuse, prescriber communication, and lower normative perceptions of non-medical use and diversion. Our secondary aim was to assess the acceptability of the intervention by examining data on usability, usefulness, and booster session engagement.

Method

Participants

Between May 2021-November 2023, 235 college students were recruited from three US universities in Texas, Wyoming, and Connecticut. Mean age was 20.42 years (SD=1.80); there was a relatively even distribution of participants by undergraduate class year. Most participants identified as white (n=202; 86%) and reported female sex assigned at birth (n=174; 74%). Inclusion criteria were a current (last 90 days) prescription for a stimulant medication, part- or full-time enrollment at one of the three university sites, expected enrollment through the 6-month duration of the study, and age 17–25 years. Individuals 18–25 have the highest rates of non-medical use of all demographic groups (Schepis et al., 2022); given their university enrollment, 17-year-olds were included because they would likely turn 18 during study enrollment. There were no exclusion criteria.

Design & Procedure

To recruit, we circulated flyers, e-mails, and business cards and made announcements on campus, social media, in undergraduate courses, and via a research participant pool at the Wyoming site. To reduce the potential impact of demand characteristics, we advertised the purpose of the study as better understanding the experiences of undergraduate and graduate students who are prescribed stimulants; specifically, their psychological well-being, academic and social challenges they may face, and their experiences with substance use. Thus, prospective participants were not initially informed that they would be participating in an intervention study and were blind to the study conditions. Recruitment materials included a link to a one-minute online screening survey, which addressed the inclusion criteria, type of stimulant medication prescribed, and lifetime history of stimulant prescription diversion. We scheduled baseline sessions with students who met the criteria and consented. Participants were stratified by lifetime history of diversion and randomized to the active intervention or placebo by the principal investigator at each site. The randomization sequence was generated using the Research Randomizer website (https://www.randomizer.org).

Research team members conducted the 60- to 75-minute baseline session over videoconference. Team members answered questions about the study/consent/assessments, shared links to the assessments and to the web-based presentation, and monitored participant engagement remotely. Participants first completed the baseline measures. Active intervention participants then viewed an interactive web-based simulation; placebo participants viewed an interactive web-based tutorial focused on common psychological disorders and their treatments. Immediately post-intervention, participants completed a brief usability and usefulness survey and measures of perceived norms, perceived risk, self-efficacy, and intentions to divert (described below). The latter four constructs were not preregistered for analytic examination immediately post-intervention.

There were four follow-up contacts. Links to online booster sessions (described below) were distributed via e-mail at one and two months. Links to online follow-up assessments were distributed at 3 and 6 months, with e-mail and text reminders for non-responders. We chose a 6-month time frame for follow-up because it would encompass a midterm or final exam period, when peers may be most likely to request medication due to academic stress. At the end of the study, all participants were debriefed about the purpose of the study, their group assignment, and our reasons for employing blinding; placebo participants were provided access to the active intervention. Participants were compensated up to $100 for completing all aspects of the study. This study was approved by Trinity College’s IRB using a single-site IRB model and was monitored by an independent Data and Safety Monitoring Board (DSMB). The DSMB monitored outcomes (i.e., diversion and non-medical use) annually to ensure differences between the intervention and placebo groups were not unduly large (Cohen’s d=0.8). The trial was preregistered on clinicaltrials.gov [NCT04885166].

Active Intervention

The 25- to 30-minute web-based, active intervention (see Online Supplementary Appendix A) was divided into three modules, all delivered by a virtual human (with a voice actor) who introduced himself as a college student who takes stimulant medication for ADHD. Participants completed comprehension questions (with corrective feedback as needed) after each module.

In Module 1, the virtual human asked participants to indicate their prevalence estimates of diversion and non-medical use; participants then received corrective feedback. The module also addressed potential legal, health, and disciplinary risks associated with these behaviors and reasons it can be appealing to divert stimulant medication. Finally, the virtual human challenged several myths related to non-medical use, namely that among students without a prescription, non-medical use was not associated with significant improvements in cognitive functioning or academic performance, despite perceptions of enhanced functioning.

In Module 2, students learned effective strategies for resisting medication requests. The module encouraged students to utilize internal explanations, external explanations, alternatives, and direct refusals with validation to turn down requests for their medication. Excuses were discouraged given that they might encourage future requests. The module also discussed strategies (e.g., not offering one’s medication) to reduce the likelihood of being approached.

Participants practiced using the resistance strategies in an interactive conversation with a virtual human by “playing” a student with ADHD who has a prescription. A virtual peer asked for the participant’s medication to study for an upcoming exam. Participants chose from one of four responses differentiated by type of refusal strategy several times throughout the conversation; participants also could choose to divert. Although participants’ chosen strategies placed them on slightly different conversational “paths”, the virtual peer always requested the medication a second time and, in some cases, offered money. Throughout the conversation, participants could access a virtual coach, who provided feedback about the effectiveness of the resistance strategies utilized. Participants received feedback at the end of the module outlining which of their responses were more/less effective.

In Module 3, the virtual human discussed how lack of adherence to one’s stimulant prescription can be disadvantageous in the short- and long-term (e.g., trouble focusing, more difficulty graduating from college). Next, the module discussed reasons why prescribed students might misuse medication (e.g., academic, recreational), followed by strategies students could employ to avoid procrastination, distraction, and mixing their prescription with other substances. Finally, students were encouraged to communicate with their prescriber about the symptoms of the condition they are being treated for, the frequency with which they take their prescribed dose of medication, and any side effects.

Placebo

In the 25- to 30-minute web-based tutorial, students learned about the prevalence of psychological disorders in college students, their etiologies, psychiatric medications, and evidence-based therapies. ADHD and stimulant medications were addressed, but diversion and non-medical use were not. To mirror the interactive components of the active intervention, participants completed quizzes throughout the presentation that reinforced key ideas introduced in the presentation (e.g., “How common is major depression in college students?”).

Booster sessions

For both the intervention and placebo groups, key content from the web-based presentations was divided into two booster sessions, available as brief slide decks accessed via an e-mailed web link. Each booster session took approximately five minutes and contained five comprehension questions for which participants were compensated $1 for each correct response.

Measures

Demographics and ADHD Symptomatology

At baseline, we inquired about age, sex, gender identity, race/ethnicity, class year, and the age at which participants were first prescribed a stimulant. Participants reported their socioeconomic status (SES) using a nine-point scale adapted from Adler et al. (2000), with 1=highest SES, 9=lowest SES. Participants also rated the frequency (0=never, 4=very often) with which they experienced ADHD-related symptoms using an 18-item scale (Adler et al., 2006). Responses were averaged to create an overall ADHD symptom severity score (α = .84), with higher scores indicating greater symptomatology.

Primary Outcomes

Diversion.

At baseline, 3-months, and 6-months, participants reported how many times they (1) gave away or (2) sold/traded their stimulant medication in the past three months, with anchor dates provided to facilitate recall. The responses were combined to yield a composite score of diversion episodes.

Intentions to Divert

were assessed at baseline, immediately post intervention, 3- and 6-months with two questions from the Behavior, Expectancies, Attitudes and College Health Questionnaire (Bavarian et al., 2013) adapted to address diversion instead of non-medical use: “How likely is it that you will give away [or sell, trade] your stimulant medication in the next three months?”. These questions had a four-point response scale (1=very unlikely, 4=very likely).

Non-Medical Use.

Participants reported on non-medical use at baseline, 3- and 6-months. Specifically, they indicated the number of days in the last three months when they (a) used alternative routes of administration, (b) took more than their recommended dose, (c) took someone else’s stimulant medication, (d) took their stimulant with other drugs in order to experience intoxicating effects, and (e) intentionally got high on their stimulant medication (adapted from Sepulveda et al., 2011; Wilens et al., 2006). Although we included taking less of one’s prescribed medication as part of the non-medical use variable in our preregistered plan, we elected to exclude “taking less” to be consistent with how most other researchers have operationalized misuse of a prescribed stimulant (e.g., Bavarian et al., 2014; Darredeau et al., 2007; Rabiner et al., 2009).

Secondary Outcomes

Perceived Norms.

At baseline, immediately post-intervention, and at 3-month follow-up, participants used two 0 to 100 scales to indicate the percentage of students they believe engage in non-medical use or diversion (DeSantis et al., 2013).

Perceived Risk.

At baseline, immediately post-intervention, and at 3-month follow-up, we used a 5-item scale to assess perceived health, legal, and disciplinary risks associated with non-medical use and diversion (e.g., “How much do people risk harming themselves (physically or in other ways) if they take stimulants non-medically?” or “use their prescription in a way a prescriber did not intend?”) (adapted from Arria et al., 2008; Harris et al., 2015). Response choices ranged from 1=no risk to 4=great risk; reliabilities were ɑ=.69, .80, .82 (baseline, post, 3-months).

Self-Efficacy to Resist Diversion and Nonadherence.

At baseline, post-intervention, and at the 3- and 6-month follow-ups, we assessed self-efficacy with a modified version of the avoidance self-efficacy scale (Bavarian et al., 2013). We asked participants to “Please rate your confidence to (1) resist giving away your medication, (2) resist selling your medication, and (3) avoid using your stimulant medication in a way it was not prescribed.” Responses ranged from 1=not at all confident to 5=completely confident. We created an avoidance self-efficacy score for diversion by summing the first two questions (rs=.72, .63, .70, .46). We examined the question focused on efficacy to avoid nonadherence, or using one’s medication in a way it was not prescribed, separately.

Prescriber Communication.

Participants used Timeline (Wray et al., 2019) to report on prescriber communication. Timeline is a web-based application modeled on the Timeline Follow-Back (Sobell & Sobell, 1992), an assessment used to retrospectively assess substance use and other behaviors. Participants first labeled personally important events during the last 90 days on an electronic calendar to facilitate recall. Then, participants indicated at baseline, 3- and 6-months the number of days in the past 90 days they communicated with their prescriber regarding their stimulant prescription and any concerns they had regarding the dose, frequency of administration, and/or side effects. This measure was adapted from DeSantis et al.’s (2013) 3-item measure of patient-prescriber communication.

User Experience and Engagement

User Satisfaction

was assessed immediately following the intervention with 13 items from the Post-Study System Usability Questionnaire (Lewis, 1992, 2022). The questionnaire inquired about the usefulness, information quality, and interface quality of the web-based presentation using a seven-point scale, 1=strongly disagree to 7=strongly agree. Reliability was excellent (ɑ=.94).

Intervention Usability

was assessed immediately following the intervention with 15 items from the Health Information Technology Usability Evaluation Scale (Schnall et al., 2018; Yen et al., 2014). Items related to the perceived usefulness, user control, and impact of the web-based presentation. Although the same stems were used, the questions differed between the intervention and placebo groups to match the content of the presentations. Responses ranged from 1=strongly disagree to 5=strongly agree. Reliability was excellent (ɑ=.94).

Booster Engagement.

We assessed online booster session engagement on a 0–10 scale by summing the correct answers to the five comprehension questions from each online booster.

Fidelity and Supervision

To maximize standardization of study procedures across sites, research personnel (i.e., advanced undergraduates, graduate students, doctoral-level psychologists) used a detailed manual for recruiting, evaluating screening data, documenting participant data, conducting the baseline session, and distributing booster session and follow-up assessments. Doctoral-level psychologists trained and supervised research personnel and conducted annual check-ins at each site to train new research assistants and to reiterate study procedures with existing personnel.

Data Analysis

To estimate power, we used the effect size (d=.20) from a meta-analysis of 43 interventions evaluating longer-term effects (>6 weeks) on alcohol consumption for brief (~15–20 minutes) interactive interventions with college students compared to assessment-only controls (Carey et al., 2009). We estimated 80% power to detect a similar size effect using a repeated measures design with an autocorrelation of 0.6 for a total of 300 students (150 per group).

To compare the intervention to the placebo across assessment time points, we used generalized estimating equations (GEE). The GEE models used an unstructured covariance matrix and robust estimator. The models included site (school), group, time point, and the interaction between group and time. The placebo group served as the reference category for group and baseline was the reference category for time point. Thus, model parameter estimates represent the difference between groups in the change from baseline to a specific time point. The two self-efficacy variables were modeled as ordinal outcomes; the remaining count outcomes (e.g., times diverted, times misused) were modeled with a negative binomial distribution and log link. For measures of effect size, we report Cohen’s d, with 95% confidence intervals (CIs). Statistical significance was set at an alpha level of .05 and analyses were conducted in SPSS v29. To reduce the risk of type I errors while also maximizing power, we applied a false discovery rate correction to account for multiple tests of both the primary and secondary outcomes, respectively. We presented adjusted p-values in the text and both adjusted and unadjusted p-values in the tables (Benjamini & Hochberg, 1996). Instrumentation and analysis code are available upon reasonable request from the first author.

Results

We recruited and consented 249 students; however, only 235 completed the baseline session, which was fewer than the 300 participants we aimed to recruit. This number of participants reduced power to 68% to detect the effect size planned for. Table 1 displays the characteristics for the overall sample (N=235) at baseline and by treatment group. The majority of participants (n=215; 91%) reported being prescribed stimulant medication for ADHD; a smaller number (n=9; 4%) reported ADHD and a comorbid disorder (e.g., anxiety); narcolepsy (n=2; <1%); binge eating disorder (n=2; <1%); or other (n=7; 3%). Regarding study retention, booster completion at one month was comparable in both conditions (92%), but lower among placebo participants at two months (82%) compared to the intervention group (93%). There was a differential rate of dropout based on treatment condition: fewer placebo participants were retained at the 6-month follow-up (80%) compared to the intervention participants (90%) [χ²(1)=5.12, p=.024]. Figure 1 details participant flow.

Table 1.

Participant Characteristics at Baseline for Overall Sample and by Study Group

Characteristic	Placebo (n=113)	Intervention (n=122)	Overall (N=235)
Age (M, SD)	20.56 (1.81)	20.29 (1.79)	20.42 (1.80)
Female sex assigned at birth	76 (67%)	98 (80%)	174 (74%)
Gender identity
Woman	66 (58%)	82 (67%)	148 (63%)
Man	35 (31%)	20 (16%)	55 (23%)
Genderqueer	3 (3%)	8 (7%)	11 (5%)
Transgender man	1 (1%)	1 (1%)	2 (1%)
Transgender woman	1 (1%)	0 (0%)	1 (<1%)
Gender variant or nonconforming	5 (4%)	6 (5%)	11 (5%)
Other	2 (2%)	3 (2%)	5 (2%)
Decline to answer	0 (0%)	2 (2%)	2 (1%)
Race
White/non-Hispanic	98 (87%)	104 (85%)	202 (86%)
African American/Black	2 (2%)	1 (1%)	3 (1%)
American Indian/Alaskan Native	2 (2%)	0 (0%)	2 (1%)
Asian/Asian American	0 (0%)	5 (4%)	5 (2%)
Bi/Multiracial	8 (8%)	10 (8%)	18 (8%)
Other	3 (3%)	2 (2%)	5 (2%)
Hispanic/Latino	19 (17%)	26 (21%)	45 (19%)
Class year
Freshman	24 (21%)	31 (25%)	55 (23%)
Sophomore	28 (25%)	27 (22%)	55 (23%)
Junior	24 (21%)	35 (29%)	59 (25%)
Senior	35 (31%)	26 (21%)	61 (26%)
Graduate student	2 (2%)	3 (3%)	5 (2%)
Site
Texas	47 (42%)	55 (45%)	102 (43%)
Connecticut	40 (35%)	41 (34%)	81 (34%)
Wyoming	26 (23%)	26 (21%)	52 (22%)
Socioeconomic status (M, SD)	4.92 (1.76)	5.04 (1.75)	4.98 (1.75)
Age first prescribed stimulant (M, SD)	15.54 (5.05)	16.55 (4.28)	16.06 (4.69)
ADHD symptom severity (M, SD)	46.10 (10.40)	46.75 (9.05)	46.43 (9.71)

Note. After rounding percentages may not equal 100. Participant counts by site are slightly lower than those in the CONSORT diagram because they reflect only eligible participants who received the intervention, as opposed to the total number of participants who were randomized at each site.

Figure 1.

CONSORT Flow Diagram

Primary Outcomes

Prescription Stimulant Diversion

For number of times diverted, there was no significant change by group at 3 months (B=−0.35, p=.499, d=0.09) or 6 months (B=−0.33, p=.499 (.489), d=0.09). We performed sensitivity analyses adjusting for sex – it was not associated with diversion, nor did it change the interpretation of the findings (the effect of group remained nonsignificant).

Non-Medical Prescription Stimulant Use

For the model examining total number of non-medical use episodes, there was an excessive number of zeros coupled with several very high values; thus, the negative binomial model for count data did not fit well. We used the Tweedie model instead, which fit better, and converted the exponential coefficients into Cohen’s ds (Moshitch & Nelken, 2014). There was no significant change by group for number of non-medical use episodes at 3 (B=0.58, p=.489, d=−0.32) or 6 months (B=0.15, p=.845, d=−0.09) (Table 2). We performed sensitivity analyses adjusting for sex – it was not associated with non-medical use, nor did it change the interpretation of the findings (the effect of group remained nonsignificant).

Table 2.

Group Comparisons on Primary Outcomes of Diversion, Non-Medical Prescription Stimulant Use, and Intentions to Divert

Outcome	n	Placebo	Intervention	Effect Size	p	Adj-p
Number of Times Diverted
Baseline	235	0.45 ± 0.16	0.39 ± 0.10
3-Month	201	0.41 ± 0.17	0.25 ± 0.09	d= 0.09 (−0.19–0.37)	0.454	0.499
6-Month	200	0.30 ± 0.08	0.18 ± 0.08	d= 0.09 (−0.18–0.37)	0.444	0.499
Number of Non-Medical Use Episodes
Baseline	235	3.00 ± 10.55	1.89 ± 4.58
3-Month	201	1.80 ± 4.60	2.02 ± 12.30	d= −0.32 (−1.10–0.38)	0.372	0.489
6-Month	200	2.03 ± 7.26	1.49 ± 11.11	d= −0.09 (−0.94–0.77)	0.845	0.845
Intention to Divert
Baseline	235	2.46 ± 1.21	2.46 ± 1.05
Post-Intervention	233	2.41 ± 1.21	2.17 ± 0.57	d= 0.25 (0.09–0.51)	0.012	0.066
3-Month	201	2.57 ± 1.27	2.33 ± 0.91	d= 0.23 (−0.05–0.50)	0.102	0.187
6-Month	200	2.53 ± 0.97	2.29 ± 0.82	d= 0.19 (−0.10–0.46)	0.052	0.151

Note. Adj-p = Adjusted p-value using false discovery rate correction.

Intentions to Divert Stimulant Medication

There was no effect of the intervention on intentions immediately after the intervention (B=−0.10, p=.066, d=0.25), at 3 months (B=−0.10, p=.187, d=0.23), or 6 months (B=−0.10, p=.151, d=0.19) (Table 2).

Secondary Outcomes

Perceived Norms.

There were no group differences in perceived norms for diversion immediately after the intervention (B=1.23, p=.676, d=−0.14) or at 3 months (B=−0.24, p=.931, d=−0.03) (Table 3). In contrast, for perceived norms for non-medical use there was a medium effect of the intervention immediately after (B=−9.72, p=.001, d=0.51) but not at 3 months (B=−5.24, p=.207, d=0.25) (Table 3).

Table 3.

Group Comparisons on Secondary Outcomes of Perceived Norms, Perceived Risk, Self-Efficacy, and Prescriber Communication

Outcome	n	Placebo	Intervention	Effect Size	p	Adj-p
Perceived Norms Diversion
Baseline	234	32.7 ± 20.8	33.5 ± 20.0
Post-Intervention	235	27.1 ± 19.6	29.1 ± 9.2	d= −0.14 (−0.39–0.12)	0.588	0.676
3-Month	201	27.2 ± 15.2	27.6 ± 13.1	d= −0.03 (−0.31–0.25)	0.931	0.931
Perceived Norms Non-Medical Use
Baseline	234	39.7 ± 22.2	41.1 ± 23.6
Post-Intervention	235	32.6 ± 19.9	24.2 ± 11.6	d= 0.51 (0.24–0.76)	<.001	0.001
3-Month	201	33.8 ± 18.3	29.9 ± 16.5	d= 0.25 (−0.03–0.53)	0.074	0.207
Perceived Risk
Baseline	234	17.1 ± 2.3	16.6 ± 2.8
Post-Intervention	235	17.8 ± 2.5	18.6 ± 2.1	d= 0.37 (0.12–0.63)	<.001	<.001
3-Month	201	17.1 ± 2.9	18.0 ± 2.6	d= 0.39 (0.12–0.68)	<.001	<.001
Self-Efficacy to Avoid Giving Away & Selling
Baseline	235	9.40 ± 1.35	9.48 ± 1.14
Post-Intervention	235	9.78 ± 0.58	9.83 ± 0.60	d= 0.37 (−0.31–1.05)	0.288	0.474
3-Month	201	9.49 ± 0.99	9.65 ± 0.84	d= 0.14 (−0.41–0.68)	0.628	0.676
6-Month	200	9.46 ± 0.84	9.69 ± 0.78	d= 0.16 (−0.48–0.81)	0.611	0.676
Self-Efficacy to Avoid Nonadherence
Baseline	235	4.55 ± 0.91	4.45 ± 0.98
Post-Intervention	235	4.53 ± 0.95	4.56 ± 0.74	d= 0.13 (−0.21–0.44)	0.499	0.676
3-Month	201	4.59 ± 0.82	4.56 ± 0.86	d= 0.20 (−0.18–0.58)	0.305	0.474
6-Month	200	4.56 ± 0.75	4.63 ± 0.79	d= 0.47 (0.10–0.85)	0.013	0.046
Prescriber Communication
Baseline	234	1.02 ± 1.42	1.08 ± 1.31
3-Month	194	0.88 ± 1.23	0.64 ± 0.96	d= 0.22 (−0.07–0.50)	0.094	0.219
6-Month	196	0.77 ± 1.17	0.58 ± 1.02	d= 0.20 (−0.08–0.49)	0.186	0.372

Note. Adj-p = Adjusted p-value using false discovery rate correction.

Perceived Risk of Diversion and Non-Medical Use.

The intervention had a small to medium effect on perceptions of risk in the hypothesized direction immediately after the intervention (B=−0.65, p<.001, d=0.37) and at 3 months (B=−0.55, p<.001, d=0.39) (Table 3).

Self-efficacy to Resist Diversion and Nonadherence.

There was no significant change in self-efficacy to avoid diversion after the intervention (B=−0.68, p=.474, d=0.37), at 3 months (B=−0.25, p=.676, d=0.14), or at 6 months (B=0.30, p=.676, d=0.16). There was a medium effect on self-efficacy to avoiding nonadherence favoring the intervention, but only at 6 months (B=0.86, p=.046, d=0.47) (Table 3).

Prescriber Communication.

There was no significant change by group in days of prescriber communication at 3 months (B=−0.39, p=.219, d=0.22) or 6 months (B=−0.33, p=.372, d=0.20); effect sizes suggested a small effect in the opposite direction of what was hypothesized (Table 3).

User Experience and Engagement

Intervention group participants reported a mean user satisfaction rating of 6.47 (SD=0.60) on a 7-point scale, while the placebo reported a mean of 6.31 (SD=0.63) [t(233)=1.96, p=.052]. The placebo group reported a mean usability score of 4.21 (SD=0.52) on a 5-point scale, which was slightly higher than the intervention group (M=4.05, SD=0.70) [t(222.78)=2.05, p=.042]. Among participants who completed one or both booster sessions, accuracy in responding was significantly higher among intervention participants (M=9.46, SD=0.75) compared to placebo participants (M=7.83, SD=1.51; t(127.48) = −9.46, p<.001).

Post-Hoc Analyses

Given the relatively low rates of diversion and non-medical use in our study, we conducted additional post-hoc analyses that were not preregistered. Specifically, we examined (1) any diversion or non-medical use during the 6-month study period, (2) change in diversion or non-medical use status between baseline and 6 months (e.g., transition from diversion to non-diversion status), and (3) number needed to treat to prevent one instance of diversion or non-medical use. Binomial outcomes (e.g., any diversion, any non-medical use) were modeled with a binomial distribution and logit link. We report odds ratios (OR) with 95% CIs as a measure of effect size. To evaluate whether more intervention participants migrated from diversion to non-diversion status compared to the placebo group, we used the McNemar test.

Prescription Stimulant Diversion

The intervention group had 76% lower odds of any diversion during the 6-month study period [B=−1.43, p=.007, OR=0.24 (CIs: 0.08– 0.68)]. Specifically, 21% of the placebo group diverted at least once, compared to 12% of the intervention group. Regarding diversion group status, the McNemar change test showed significant change in the intervention group (p=.007), with 12% migrating from ‘diversion’ to ‘no diversion’ status compared to 4% of the placebo group. Only 2% of the intervention group migrated from ‘no diversion’ to ‘diversion’ status, compared to 6% of the placebo (see Supplemental Figure 1). The number needed to treat to prevent one instance of diversion was 20 and 19 individuals, respectively, for the 3- and 6-month time points.

Non-Medical Prescription Stimulant Use

The intervention group had 60% lower odds of any non-medical use during the study period [B=−0.91, p=.006, OR=0.40 (CIs: 0.21–0.77)]. Specifically, 44% of the placebo group misused at least once compared to 27% of the intervention group. Regarding non-medical use group status, the McNemar change test showed significant change in the intervention group (p<.001), with 30% migrating from ‘non-medical use’ to ‘no non-medical use’ status compared to 13% of the placebo group. Only 5% of the intervention group migrated from ‘no non-medical use’ to ‘non-medical use’ status, compared to 9% of the placebo group (see Supplemental Figure 1). The number needed to treat to prevent one instance of non-medical use was 9.7 and 5.7 individuals, respectively, for the 3- and 6-month time points.

Discussion

This randomized controlled trial is the first study to evaluate a brief, web-based intervention targeting both stimulant diversion and non-medical use in college students with stimulant prescriptions. Regarding our primary outcomes, intervention participants did not report a significant reduction in number of diversion or non-medical use episodes or intentions, contrary to our pre-specified hypotheses. Regarding secondary outcomes, the intervention showed a small-to-medium effect on perceived risks of diversion and non-medical use and a medium effect immediately post-intervention on perceived norms for non-medical use and on self-efficacy to avoid nonadherence 6 months after the intervention. Contrary to our pre-specified hypotheses, intervention participants did not evidence a reduction in perceived norms for diversion, nor did they report increased self-efficacy to avoid diversion or more frequent communication with a prescriber.

There are several explanations as to why our intervention may have had a limited effect on the primary and secondary outcomes. The lack of reduction in the number of diversion and misuse episodes may have been due, in part, to floor effects for diversion and misuse episodes and intentions to divert at baseline. Diversion, in particular, was less frequent among our participants than in prior cross-sectional studies (e.g., Gallucci et al., 2015; Holt et al., 2018), suggesting that a longitudinal RCT may attract students at lower risk for these behaviors. This explanation is corroborated by the relatively high self-efficacy ratings to avoid diversion among all participants. This ceiling effect also may account for the lack of change in self-efficacy to avoid diversion for intervention participants. Since participants did not know they were getting an intervention and the study did not target students who necessarily were concerned about and/or wanted to make a change to their behavior, we may have enrolled a less motivated and/or lower risk sample with less propensity to benefit from the intervention. Indeed, qualitative feedback from intervention participants after study completion suggested that those who did not have experience with diversion found the content to be less relevant. We also did not observe an effect of the intervention on frequency of prescriber communication. It is possible some students still rely on their parents/guardians to navigate the healthcare system and/or assist with refills. Qualitative feedback suggested that there was substantial variability in (1) access to one’s prescriber, (2) expectations for frequency of contact, and (3) student perceptions of when it was appropriate to reach out to a prescriber. For these reasons, it may have been difficult to change this behavior.

Notably, intervention effects on one of the key mechanisms to influence intentions and behavior in the Theory of Planned Behavior, perceived norms, either were not maintained over time (for non-medical use) or were not present (for diversion). Mean perceived norms for diversion at all time points were lower than prevalence estimates of diversion in the literature and the estimate included in the intervention materials, suggesting a possible floor effect for this variable. At the same time, the large standard deviations for perceived diversion norms suggested heterogeneity among participants. Given the influence of perceived prevalence of peer substance use on one’s own drinking behavior in other studies (Perkins, 2002), we see potential for refinement of the intervention to address perceived norms in a way that is more durable (in the case of non-medical use) and that is more responsive to differences in participants’ perceived norms at baseline.

Our finding that the intervention changed perceptions of risk associated with diversion and non-medical use, a secondary outcome, contrasted with Molina et al. (2020), where harm perceptions did not change following an intervention. We addressed a broad range of risks (i.e., health, legal, disciplinary) consistently throughout the intervention modules, which may account for this favorable outcome. That is, risks were introduced in Module 1, reiterated in Module 2, when the participant’s conversation partner noted specific health concerns related to sharing her medication, and revisited in Module 3, when the risks of misusing stimulant medication were reviewed.

Regarding our secondary aim to examine the feasibility and acceptability of the intervention, satisfaction ratings were high (i.e., between the top two ratings) among intervention participants, suggesting that the web-based program was feasible to navigate, and the information was logically organized. With respect to usefulness, the placebo group reported slightly higher ratings, which might be due to the breadth of topics related to mental health included in their presentation; yet the intervention group also rated the intervention to be useful (M=4.05 out of 5). Our finding that booster completion and accuracy rates were higher in the intervention group, however, as were retention rates, suggests that the intervention group might have found the information more personally relevant given that they demonstrated more consistent engagement over six months.

Although the pre-registered hypotheses for our primary outcomes were not supported, post-hoc analyses showed that participants had significantly lower odds of any diversion and non-medical use in the 6 months following the intervention. Most other brief interventions impacted intentions (e.g., LaBelle et al., 2020; Looby et al., 2013; Molina et al., 2020) but were unsuccessful in producing behavior change. A universal intervention by Antshel and colleagues (2025) resulted in lower rates of non-medical use compared to a control condition, though follow-up was brief (i.e., 3 months), diversion was not assessed, and participants with ADHD and/or a stimulant prescription were excluded from the study. Participants in our intervention condition were significantly less likely to begin diverting or engaging in non-medical use when they denied doing so at baseline compared to placebo participants; and they were less likely to continue diverting or misuse behavior during the follow-up if they had endorsed these behaviors at baseline. A relatively large number of students (20) would need to receive the intervention to prevent one case of diversion; however, one student may divert to several students, so the intervention may have positive downstream effects beyond one student. Taken together, while encouraging, these post-hoc analyses require replication and should be interpreted with caution given that they are a departure from the pre-specified primary outcomes.

Limitations and Future Directions

Our findings should be considered in the context of several limitations. First, we did not achieve our target sample size largely due to challenges with recruitment at the Wyoming site, where we only achieved 50% of our recruitment goal. Accordingly, our power to detect effects of the intervention may have been diminished. The reasons for these recruitment challenges were not clear, although this site was in a US region (West) where stimulant prescriptions are lowest compared to other US regions (Vaddadi et al., 2021). Also, recruitment occurred almost entirely during the public health emergency phase of COVID-19. Although our study was remote, there were fewer opportunities for in-person recruitment and a subset of students were dealing with COVID-19 infections, factors that might have hampered recruitment and deterred participation. Second, a national shortage in the availability of prescription stimulants, first recognized by the US Food and Drug Administration (FDA) in 2022 (FDA, 2023) and still active at the time of this publication, may have reduced the likelihood of diversion, as students may have been less likely to possess excess medication. The shortage also might have prompted less consistent adherence; specifically, students might have taken less than their prescribed dose if they were concerned about securing their next prescription. Alternatively, students might have been less likely to take more medication than prescribed if they anticipated that it might be difficult to fill their prescription. Future research should explore how students have navigated this prolonged shortage and what effects (if any) it has had on students’ symptoms and well-being, medication adherence and management, and willingness to divert.

Third, our findings may not generalize to all students with stimulant prescriptions. Our sample was disproportionately female and rates of diversion were relatively low in our sample. Future research testing the intervention should focus more explicitly on recruiting men, students at greater risk for diversion and non-medical use (e.g., students with a history of these behaviors and/or intentions to engage in them), and possibly on students transitioning to college. Participants at greater risk for diversion and non-medical use may have been less responsive to our recruitment methods; thus, we may observe stronger effects in the future if we target students with concerns about diversion or medication adherence.

Fourth, our assessments were self-reported and thus may have been subject to problems associated with retrospective recall. Linking participants to assessments via text message, instead of e-mail (as was done in the current study), and allowing participants to self-administer the intervention and assessments may improve participation, particularly among higher-risk participants who may not wish to meet with a research assistant, and minimize the time burden. Publicizing the study during routine health or mental health care visits also might help to reach higher risk participants. Finally, since the effects of our intervention on the primary outcomes of diversion and non-medical use only were detectable in post-hoc analyses, future research might consider focusing on a longer time frame for the assessment of primary outcomes and an examination of transitions from no divert to divert (and no misuse to misuse) status and vice versa. Also pertaining to our measures, the correlation for the self-efficacy to avoid diversion measure at 6 months was lower than the other time points. A more comprehensive measure of self-efficacy likely would be more reliable and may capture more variability among participants.

In the future, it may be beneficial to refine the intervention content so that it is better tailored to a participant’s risk of diversion and non-medical use at study entry. Regarding diversion, it was notable that base rates were relatively low; at the same time, post-hoc analyses showed that a significant number of intervention participants transitioned from diverter to non-diverter status during the study period (which was not true for the placebo group), suggesting that these subgroups have varied needs that may be best met with different content. For example, if a participant denies a history of diversion and reports no diversion intentions, a simulated conversation may be unnecessary. Instead, the intervention might prompt students to elaborate on their reasons for not diverting and to reinforce these reasons at future points in time. Regarding non-medical use, given that different types of medical misuse (e.g., taking more than prescribed, mixing with other substances, etc.) may result in different consequences, participants could receive information both about risks and on how/why to avoid specific types of misuse they endorsed. Communication with a prescriber still could be encouraged; however, the intervention may need to more explicitly address when, how, and why to connect with a prescriber and acknowledge that the timing and nature of communication may vary depending on the student-prescriber relationship. With respect to diversion and non-medical use, incorporating personalized normative feedback, whereby personalized information based on baseline behavior is provided to correct overestimated normative perceptions (or to reinforce underestimations), may bolster the intervention effects on perceived norms modification and subsequent intentions and behavior. Prevalence estimates of both diversion and non-medical use in college students should be monitored so that this content reflects the most recent data available. Overall, we see value in continuing to examine the efficacy of this intervention with adjustments to content; the sample(s) engaged; the timeframe of assessment; and how the intervention is presented at the outset.

It also may be beneficial to examine additional individual and/or sociocultural characteristics of participants. For example, given that recent research has shown that being prescribed stimulants after age 10 and for less than one year was associated with greater risk for non-medical use among adolescents (McCabe et al., 2024), future research could examine how age of initiation/duration of prescription stimulant treatment relate to diversion and non-medical use risk and intervention response. Relatedly, a more thorough assessment of students’ psychiatric conditions and prescriptions for other medications could elucidate whether students’ experience navigating multiple diagnoses and/or medications is associated with a differential response to the type of intervention tested in the present study. Indeed, a recent study showed that adults with ADHD and comorbid conditions were less likely to adhere to their stimulant prescription than those without comorbidities (Jeun et al., 2024). Finally, assessing sociocultural influences on diversion and non-medical use outlined in the theory of triadic influence (e.g., perceptions of campus drug culture, religiosity, exposure to prescription drug content in traditional and social media) (Bavarian et al., 2013, 2014) may elucidate where or for whom an intervention could be most impactful.

In conclusion, college students with stimulant prescriptions are at risk for diverting and misusing their medication, interrelated behaviors that pose risks to them and to their nonprescribed peers. The present study, a rigorous test of a new intervention, adds to the very limited literature on interventions for diversion and non-medical use. Due in part to floor effects and limited power, the intervention did not significantly reduce the number of diversion or non-medical use episodes, as originally hypothesized, but it did lead to declines in perceived norms for non-medical use, increased perceptions of risk, and self-efficacy to avoid medication nonadherence. Further, the intervention was feasible to implement and perceived to be useful. With the introduction of adaptive content, more targeted recruitment of a larger sample of students at risk for diversion and non-medical use, streamlining of the study assessments, and a single primary outcome whose timeframe better reflects the real-world frequency of diversion in particular, the intervention ultimately may prove to be both a cost-effective and scalable option for college personnel and prescribers of stimulant medication to address diversion and non-medical use.

Data Availability

Instrumentation and analysis code that support the findings of this study are available upon reasonable request from the corresponding author. The dataset is not publicly available to preserve the privacy of research participants. The intervention script/manual is available in the supplemental online materials.