Telehealth engagement and treatment strategies for adults living with alcohol use disorder: A sequential multiple assignment randomized trial protocol

Abstract

Background:

Treatment rates for alcohol use disorder (AUD) among adults in the United States (US) are low. Healthcare systems with electronic health records (EHRs) could identify those experiencing AUD and conduct outreach then treatment to reduce drinking and/or AUD symptoms. Here, we describe the protocol for a sequential, multiple assignment, randomized trial (SMART) of multiple technology-driven adaptive interventions (ADIs) to increase AUD care and improve outcomes among adults within a large healthcare system.

Methods:

We plan to enroll 400 adults in this two-stage SMART. Potential participants identified by alcohol indicators in the EHR are screened for eligibility criteria: current drinking, AUD, and lack of current AUD psychotherapy treatment. Participants are randomized to Stage 1 interventions: 1 brief motivational interviewing (MI)-focused phone session, or 4 weeks of interactive therapeutic messaging delivered via a patient-portal like system. Non-responders at a 4-week post-test are re-randomized to an enhanced dose of their Stage 1 intervention (i.e., additional phone session or 4 weeks of messaging) or to an 8-session MI-Cognitive Behavioral Therapy (MI-CBT) protocol via video/phone. Outcomes (e.g., drinking, treatment utilization) are measured at 4-, 8-, and 12-months post-enrollment.

Conclusions:

This innovative SMART evaluates critical technology-driven ADI strategies within an AUD engagement and care model. We will identify the most effective initial strategy (phone session vs. portal messaging) and, among non-responders, we will assess the most effective follow-up strategy (additional Stage 1 dose vs. MI-CBT). Outcomes will inform the utility of scalable ADIs to engage and improve outcomes for patients with untreated AUD.

Trial Registration:

ClinicalTrials.gov #NCT05594238. University of Michigan HUM00210546.

1.0. INTRODUCTION

In 2023, 10.9% of adults in the United States (US) had an alcohol use disorder¹ (AUD), a chronic medical condition that impairs health and quality of life. Only 7.8% of adults with AUD received treatment in the past year.¹ Given longstanding under-utilization of AUD treatment,² and recent increases excessive drinking- and AUD-attributable deaths^3,4, work is needed to close this treatment gap.

Several barriers contribute to these low treatment rates. The National Survey on Drug Use and Health (NSDUH) highlights types of reasons that people with substance use disorders (SUDs) do not seek care, including lack of readiness (e.g., not ready to stop/cut back or to start treatment) or willingness (e.g., believing the problem can be handled alone).¹ Per the trans-theoretical model of change, ambivalence about change (including fluctuating readiness/willingness) is a hallmark feature of AUD and other SUDs.⁵ Other common treatment barriers include ability to engage (e.g., cost/insurance coverage, transportation, need for childcare or flexible appointments, knowing where to get treatment) and stigma-related concerns (e.g., worried what people would think, privacy, fear of consequences [loss of job or children]), which can be greater for minoritized or marginalized populations.^1,6,7

Thus, current approaches to engaging people in AUD care and the types of care delivered appear insufficient. Strategies like screening, brief intervention, and referral to treatment (SBIRT), prove helpful for some. Yet, given the onus on patients to overcome barriers to follow-up on referrals on their own, SBIRT may be insufficient for those with fluctuating motivation and readiness. Collaborative care is also helpful to some⁸, allowing for care in familiar settings (e.g., primary care). Moreover, the rapid uptake of telehealth⁹ since the COVID-19 pandemic has created space for more accessible care that can overcome some key barriers (e.g., stigma concerns, transportation), though the sheer availability of telehealth did not result in a marked uptick in AUD treatment initiation.¹⁰

Solving the AUD care gap will likely take multiple, new strategies with an array of low-burden treatment options aligned with patients’ preferences, needs, and values (i.e., patient-centered care¹¹). Accordingly, as an alternative to requiring that patients seek out treatment, we previously called for “patient-first” care models that employ a proactive outreach approach to engagement, incorporating “screening and identification of untreated AUD, but offer[ing] direct, accessible patient-centered treatment via telehealth”¹² strategies consistent with the spectrum of abstinence to harm reduction. Critical is the need for outreach to non-treatment seeking patients to be rooted in non-stigmatizing strategies that facilitate problem recognition and build motivation to change.¹³

With diversity in AUD severity, patient barriers, and preferences for care, AUD engagement and care strategies should not be one-size-fits-all and studies testing active interventions vs. control groups to determine efficacy, in a broad sense, do not guide the field with regard to what works best for whom. Adaptive interventions (ADIs) are promising in this regard as they can be tailored to patients’ needs based on assessments, with intervention dose or type based on initial response.^14,15 Sequential, multiple assignment, randomized trials (SMARTs)¹⁶ help identify ADIs to allow for efficient resource allocation for diverse patient populations. In the present SMART, we are using patient-first outreach to test novel combinations of telehealth-based strategies employing Motivational Interviewing (MI) and Cognitive Behavioral Therapy (CBT) principles, two empirically-supported psychotherapies for AUD.

In this SMART, Stage 1 interventions designed to prompt reduced drinking and/or engagement in community-based AUD care include: 1) a telephone-delivered MI session (called T-Engage), or 2) 2-way therapeutic messaging using MI and CBT delivered via a patient portal-like system (P-Engage). Telephone-based alcohol interventions are acceptable and feasible^17–21, and could engage people who would not otherwise seek treatment.²² T-Engage is based on prior 1–2 session MI interventions.^23–25 Patient portals are acceptable, effective tools to prompt self-monitoring and communication for chronic diseases and are available in 90% of US health systems.^26–29 We have used portal-like systems in prior studies involving MI and CBT^30,31, but those efforts have not exercised portals’ full potential for AUD care. Non-responders to Stage 1 interventions are re-randomized to either an enhanced dose of their first condition or they step-up to telehealth-delivered MI-CBT.

Our purpose herein is to detail the SMART protocol as it relates to primary aims: 1) Compare ADIs that begin with telephone vs. portal therapy on alcohol outcomes and AUD treatment utilization, and 2) Among first stage non-responders, identify the most effective second stage strategy (enhanced dose vs. step-up) to improve alcohol outcomes and AUD treatment utilization.

2.0. METHODS

2.1. Setting and Design Overview

With IRB approval (HUM #00210546), we plan to enroll 400 patients (ages 18–70) from the Michigan Medicine (MM) healthcare system headquartered in Ann Arbor, Michigan. MM uses a comprehensive, integrated EHR which allows for researchers to identify lists of potential participants for recruitment based on EHR data (e.g., diagnoses, clinic visits). We outreach to potential participants remotely (e.g., letter, text, email, calls).

Our SMART design (Figure 1) includes two stages of randomized interventions post-baseline assessment. In Stage 1, participants are randomized to the single T-Engage session or the 4-week P-Engage condition. After 4 weeks, participants complete a post-test, with non-responders re-randomized to Stage 2 conditions (enhanced dose or step-up). For those initially assigned T-Engage, non-responders are randomized to a second T-engage session (enhanced T-Engage) or stepped-up to an 8-session telehealth-delivered MI-CBT protocol. For those initially assigned to P-Engage, non-responders are randomized to either 4 additional weeks of messaging (enhanced P-Engage) or stepped-up to the same MI-CBT protocol. Individuals who do not complete post-test are considered non-responders. Responders are not re-randomized and receive no additional interventions. All participants are assessed at 4-, 8-, and 12-month follow-ups.

Figure 1.

SMART study design

Planned primary outcomes are: AUD treatment initiation and engagement (based on the Healthcare Effectiveness Data and Information Set^32,33) at 4-months post-baseline and alcohol consumption (e.g., drinking days, heavy drinking days) assessed at 4-, 8-, and 12-months.

2.2. SMART inclusion/exclusion criteria

Via EHR abstraction, we obtain lists of patients at elevated likelihood of experiencing AUD meeting these criteria: age 18–70, had a MM health care visit in the past 2 years, and past-year alcohol-related diagnosis (abuse, dependence, use, etc.) or elevated Alcohol Use Disorders Identification Test-Consumption (AUDIT-C) score.³⁴ Patients with diagnoses reflecting impaired cognition to consent (e.g., Alzheimer’s disease, severe intellectual disabilities) are excluded from the list.

First, staff review each patient’s EHR to confirm age and identify clear exclusions (e.g., current end-of-life care, non-English speaking). Then staff invite potential participants to screen for trial eligibility (e.g., online self-administered survey or staff-administered phone survey). Trial inclusions based on EHR review and screening are: age 18–70 at EHR review, current AUD, past-month heavy alcohol use (>4 drinks/day for men, >3 for women), internet and phone access, and ability to consent. Trial exclusion criteria include: inability to communicate in English, circumstances precluding informed consent or active participation (e.g., cognitive conditions, end-of-life care), past 90-day psychotherapy addressing AUD/alcohol, prior complex alcohol withdrawal, and current cancer treatment excluding skin cancers (these patients have indicated they are too ill to participate; added criterion on 8/14/2024). For self-administered screened eligible participants, staff confirm eligibility during baseline assessments.

2.4. Recruitment, enrollment, consenting, randomizations, and assessments

Recruitment methods are similar to prior work^12,30,35 and recruitment is ongoing (initiated 7/25/2023). At first, we initiated recruitment by mailing a letter with a $2 pre-incentive, followed by additional contacts (e.g., phone, email, text messages). Given logistical challenges to obtaining and mailing this incentive, we removed this process on 10/3/2024, but continued all other forms of outreach and began compensating participants $10 (gift cards) for screening. Consent to screen is obtained verbally or online. Crisis supports and mental health/SUD treatment are provided to all screened (via web link to resources also shown in study emails and text messages). Eligible participants are invited to enroll in the SMART with informed consent for the trial occurring by phone or online.

Those enrolling complete a baseline assessment ($40 compensation), including survey measures (online self-administered or interview-administered via phone) and a semi-structured Timeline Follow Back (TLFB)-based interview of substance use and treatment history. When assessments conclude, participants are randomized to their Stage 1 condition balanced by sex (male/female) and AUD severity (severe vs. mild/moderate) with blocks of four, using a computer-based randomization program. Research staff are unaware of condition until assignment, when they then share the participant’s condition and next steps. For T-Engage, staff schedule the ~40–45-minute phone session or deliver it immediately, when possible. For P-Engage, staff describe the condition and register the participant in our electronic portal with a welcome message sent immediately, sharing information on how to login and interact with study interventionists.

At 4-weeks, participants are invited to complete a brief post-test assessment ($15 compensation) to determine response to Stage 1 interventions. Response is defined as initiation of AUD psychotherapy treatment since baseline or no past 2-week heavy alcohol use (>4 drinks for men, >3 for women). Those who do not complete the post-test within 7 days are considered non-responders. Randomized participants are notified of their Stage 2 condition (non-responders: enhanced T-Engage/P-Engage or new MI-CBT condition); the intervention ends for responders who are instructed to continue follow-ups. Assessments (surveys + TLFB-based interview) to evaluate outcomes occur at 4-, 8-, and 12-months post-baseline ($40, $45, $50 compensation, respectively); assessors are blind to condition assignment. Participants do not receive compensation for intervention engagement. All staff are trained in risk management procedures to address acute risks during study interactions (e.g., suicidality/homicidality), with clinical supervisors on-call.

2.5. Interventions

The manualized interventions primarily focus on alcohol; other substances used are addressed when applicable. All participants are provided community resources (e.g., mental health, SUD treatment, social services) at enrollment, during regular study communications, and at the end of interventions.

2.5.1. T-Engage:

T-Engage occurs over ~40–45 min via telephone; interventionists trained in MI strategies follow on-screen prompts to enhance session fidelity. Table 1 displays the general components of T-Engage strategies. Based on prior work^23–25, T-Engage as delivered in Stage 1 focuses on building problem recognition and enhancing motivation to change (e.g., reduce/quit drinking, seek community treatment). In Stage 2, non-responders who receive Enhanced T-Engage have a follow-up conversation to build motivation toward change and/or to problem solve barriers to change.

Table 1.

T-Engage Session Structure

Stage 1	Stage 2
Part 1. Introduction: Interventionist introduction • Outline of conversation • Confirm consent • Confirm location • Permission to audio record	Part 1. Introduction: Interventionist introduction • Outline of conversation • Confirm consent • Confirm location • Permission to audio record
Part 2. What’s important?: Build rapport • Elicit and affirm participant goals and strengths	Part 2. What’s new?: Build rapport • Elicit update on patient’s values/goals and how things have gone since first T-Engage session • Identify any items that stood out to participant from first session • Affirm goals/strengths
Part 3. Where does drinking fit?: Explore participant alcohol use (e.g., current drinking patterns, observations of changes to or impacts of drinking) • Begin to develop discrepancy	Part 3. What changes, if any, have you made specific to alcohol?: Recent impacts of alcohol use • Assess for changes and affirm steps and success or revisit prior reasons for change • Identify barriers to change/challenges encountered
Part 4. How does it affect you?: Explore concerns/consequences of drinking level • Decisional balance (pros of changing, cons of staying the same) • Provide AUD psychoeducation and explore participant response	Part 4. Moving forward: Assess interest in (further) change • Review how to obtain support (e.g., social support, resources, treatment including medications) for desired changes • Address logistical barriers and treatment seeking
Part 5. What are your choices?: Explore thoughts and goals on making changes to drinking (e.g., reduction, preventing consequences, treatment including medications) • Brainstorm strategies to meet goals (e.g., social supports, resources) • Discuss community treatment options, addressing ambivalence and encourage follow-up	Part 5. Wrap-up: Strategic summary • Share resources (e.g., support groups, treatment locators, medication assistance) • Affirm self-efficacy and plans • Thank participant and transition to research tasks (e.g., reminder of upcoming assessments, etc.)
Part 6. Wrap up and next steps: Share resources (e.g., support groups, treatment locators, medication assistance) • Affirm self-efficacy to enact any plans made • Thank participant and transition to research tasks (e.g., reminder of upcoming post-test survey, etc.)

2.5.2. P-Engage.

P-Engage is a 4-week messaging intervention delivered by interventionists using a system that mimics how health system patient portals function. Like prior work³⁰, rooted in MI, interventionists send messages at least twice weekly over 4 weeks (See Table 2) to engage patients in discussing their alcohol use and goals, providing psychoeducation or CBT strategies, treatment referral information, support, etc., as indicated. Message order and text is not standardized (except introductory/closing messages), with topics seeking to enhance problem recognition and facilitate motivation to change, similar to T-Engage (see Table 2).

Table 2.

P-Engage Messaging Intervention

P-Engage Message Topics
1. Start the conversation. Introduce the portal experience • Build portal engagement, trust/rapport • Identify key goals and strengths • Provide resource page (includes treatment resources such as therapy, medication, mutual help) and psychoeducation about withdrawal
2. Understanding alcohol. Elicit and explore role of alcohol in participant’s life • Review motives/triggers • Elicit consequences experienced and benefits of change • Develop discrepancy between current drinking and goals/values
3. Stress/mental health. Understand participant concerns, if any, about stress, mood, anxiety, mental health topics • Explore relationships of these with drinking • Brainstorm alternatives for managing stress/mental health; referrals as needed • Affirm coping skills
4. Alcohol and health. Elicit participant concerns on impact of alcohol use on their health • Reflect and explore health-related benefits of change • Affirm health goals • Develop discrepancy between current drinking and health goals • EPE for psychoeducation as relevant to participant’s health
5. Social support/social situations. Build awareness of their drinking in social situations • Enhance self-efficacy for navigating social situations if reducing drinking • Reinforce obtaining social support related to changes in drinking and/or seeking care for AUD (including therapy, mutual support, medications)
6. Readiness/Use of care. Elicit beliefs about treatment, discuss types of treatment • Reinforce personal benefits of treatment engagement • Barriers, planning, how to seek care (e.g., share resources for treatment locators, medication, mutual help groups, other mental health concerns)
7. Problem-solving. Build problem-solving skills related to alcohol, its impact, treatment, etc. • Evaluate pros/cons • Reflect and affirm personal strengths for managing challenges
8. Wrapping up. Strategic summary of portal-based interactions • Eliciting and planning next steps for goals • Share resources

All messages are tailored to the unique participant based on baseline measures (e.g., values, mental health symptoms) and any prior portal responses received, with attention to affirming participants’ strengths. For those randomized in Stage 2, enhanced P-Engage continues messaging over 4 more weeks, diving deeper into prior content, problem solving barriers, and/or providing support or other tailored therapeutic techniques. When participants reply to messages, the interventionists continue the conversation by using MI skills, exploring therapeutic content from the participant’s reply, or seamlessly transitioning to another alcohol-related topic.

2.5.3. Stage 2 MI-CBT:

Those randomized to “step up” are invited to complete an 8-session MI-CBT protocol via telehealth (video or phone, per patient preference), which is rooted in prior CBT treatments infused with MI skills.^36–39 Similar to prior work¹², the 8-session topics include: 1) Getting to Know You and CBT, 2) Cravings and Urges, 3) Understanding Triggers, 4) Exploring Thoughts, 5) Exploring Mood, 6) Decisions and Problem-Solving, 7) Finding Support and Establishing Boundaries, and 8) Progress Review and Future Planning. Each session includes agenda setting, review of the prior session’s assigned home practice and past-week drinking, reviewing the session topic and eliciting participant perspectives, and relevant handouts (mailed or sent electronically). Portions of sessions that are psycho-educational are delivered consistent with MI’s “Elicit-Provide-Elicit” approach, honoring participants’ expertise while addressing relevant knowledge gaps. Interventionists work with participants on individualized goals related to alcohol consumption (e.g., reduction, cessation, harm reduction) and do not require a specific consumption-related goal. Participants are only asked to engage in the conversations and consider the relevance of the CBT strategies as applied to their lives now, in the past, in the future, or in a hypothetical change scenario (e.g., “If you decided to quit in the future, how might you manage cravings/urges in different settings?,” “What supports would you lean on?”). Medications for AUD and mutual help groups (e.g., AA, SMART Recovery) are explored using the Elicit-Provide-Elicit framework and revisited during subsequent sessions when clinically relevant (e.g. reports of withdrawal, strong urges, difficulty reducing use, or desire for additional support). Interventionists monitor all conversations for change talk (even those focused on hypothetical scenarios) and utilize MI strategies to reinforce potential benefits of and steps toward change. MI-CBT sessions last ~40–50 minutes, occur approximately weekly, and are audio-recorded (with consent).

2.5.4. Interventionist training and fidelity monitoring.

Interventionists received training in the specific strategies (e.g., T-Engage, P-Engage, MI-CBT), with manuals to guide training and intervention delivery. Interventionists complete foundational MI training via online modules and live training. Clinical supervision occurs approximately weekly based on patient flow, with review of audio-recordings (e.g., for T-Engage and MI-CBT) and/or portal messaging (e.g., transcripts, co-drafting responses with a supervisor, etc.) for each patient. A master’s-level clinical psychologist with experience in substance use interventions using MI (including training from the MI Network of Trainers), CBT, telehealth and portal-based approaches provides supervision. Staff are required to meet benchmark fidelity thresholds and content adherence prior to seeing participants enrolled in the trial.

2.6. Anticipated Sample Characteristics

We project enrolling N=400 adults experiencing AUD ages 18–70 with demographics mirroring characteristics of patient population at the study site, with the exception that we plan to over-sample Black/African American individuals and enroll ~50% male/female sex. We estimate the sample will be ~65–70% White, ~25% Black and/or African American (inclusive of multi-racial Black or African American), with 5–10% from other backgrounds; about 5–10% are expected to be Hispanic/Latinx.

2.7. Eligibility and outcome measures

Outcome measures are repeated from screening/baseline across 4-, 8-, and 12-month assessments. Below, we focus on measures used to determine eligibility, intervention response at 4-weeks, and our primary outcomes (pre-registered at ClinicalTrials.gov).

2.7.1. Trial eligibility:

A screening questionnaire assesses demographics (e.g., age, race, sex), trial eligibility, and AUD severity. To reliably determine current (i.e., past-year) AUD symptoms and diagnosis, we use the Alcohol Symptom Checklist (ASC).⁴⁰ Selection of at least 2 AUD symptoms results in meeting diagnostic criteria for AUD; number of symptoms selected determines severity (2–3=mild, 4–5=moderate, 6+=severe). For past-month heavy drinking eligibility, we used an item based on the NSDUH.⁴¹ We also ask participants about reliable phone and internet access based on Pew Research Center items.⁴² The exclusion for past 90-day AUD psychotherapy is assessed via 2 NSDUH-based items reflecting inpatient and outpatient therapy.⁴³ Using items from our prior work¹², we query exclusions: 1) alcohol withdrawal risk based on clinical symptoms of severe withdrawal (e.g., delirium tremens, seizures, hallucinations), 2) current enrollment in another substance use study, and 3) current treatment for cancer other than skin cancer.

2.7.2. Primary outcomes.

Registered primary outcomes include AUD psychotherapy treatment initiation and engagement (≥2 psychotherapy attended; items based on screening questions above), which are assessed at the 4-month follow-up. Primary alcohol consumption outcomes over 4-, 8-, and 12-months are past 30-day drinking days and heavy drinking days measured via the reliable and valid 30-day TLFB.^44,45

2.7.3. Exploratory outcomes.

Planned exploratory outcomes include: 1) any AUD treatment (including medication or inpatient or residential, not limited to psychotherapy as in the primary outcome); 2) total AUD psychotherapy sessions attended; 3) quality of life; and 4) alcohol consequences.

2.8. Planned outcomes analysis

Analyses will be performed considering intent-to-treat. We will conduct 2-sided hypothesis tests (α=.05) for our Aim 1 and 2 analyses considering the primary outcome and focus on estimation and confidence intervals for secondary outcomes and exploratory aims.

2.8.1. Aim 1.

In the primary outcome analysis of treatment utilization measured at the 4-month follow-up, we will compare initial interventions using a logistic regression model controlling for stratification factors of sex and AUD severity. In the primary outcome analysis of alcohol consumption, we will examine the percent of drinking days and percent heavy drinking days in the past 30 days using generalized linear mixed effects regression of the follow-up outcomes adjusting for baseline values and stratification variables (and we will use a parallel approach for exploratory outcomes noted above). This ANCOVA strategy accounts for regression to the mean and provides unbiased intervention effect estimates when analyzing outcomes with baseline and follow-up data. We will assess the distribution of alcohol use to see if a binomial, Poisson, Negative Binomial, or beta-binomial is most appropriate⁴⁶. Random effects for the intercept and time with an unstructured within-person correlation structure for the residual errors will be used. We will employ diagnostics to determine suitability of more parsimonious (e.g., autoregressive) correlation structures.

2.8.2. Aim 2.

These analyses consider only those who do not respond to their initial intervention and compares arms A+F vs. B+E (Figure 1 and Table 3). Outcome analyses will be similar to Aim 1, but only for non-responders and considering the treatment group as the strategy of step-up (MI-CBT) vs. enhancement (e.g., additional T-Engage session or 4 more weeks of portal messaging).

Table 3.

Adaptive Interventions (ADIs)

ADI	Stage 1	Respond?	Stage 2	Sub-group
1	T-Engage	Yes No	None Enhanced T-Engage	C B
2	T-Engage	Yes No	None MI-CBT	C A
3	P-Engage	Yes No	None Enhanced P-Engage	D E
4	P-Engage	Yes No	None MI-CBT	D F

2.8.3. Exploratory ADI Analyses:

Exploratory analyses will estimate the embedded ADIs and investigate baseline and time-varying moderators to further optimize ADIs to inform future care delivery in health systems. We will estimate the treatment utilization and alcohol consumption outcomes for each of the 4 embedded ADIs using longitudinal data and weighted, generalized estimating equations.^47,48 Each ADI (subgroups A+C, B+C, D+E, D+F) will be estimated with 95% confidence interval and ranked in terms of effectiveness. Each participant is weighted inversely to the known randomization probabilities to correct for the bias by design. We will also explore baseline and time-varying moderators of the 4 embedded ADIs to develop more tailored ADIs using Q-learning or backward regression. These analyses will provide important information on whom the ADIs are most useful for.

2.8.1. Power analysis and sample size.

The total sample size of 400 is based on Aim 2 (thus, also sufficient for powering Aim 1 because of larger sample), to compare non-responders who are offered step-up to telehealth MI-CBT versus those receiving enhanced T-Engage and P-Engage (subgroups A+F vs. B+E, see Table 3), using alcohol consumption and treatment initiation/engagement as primary outcomes. We assumed 80–90% of individuals will be non-responders to Stage 1 interventions (i.e., at least 64 in each non-responder intervention group). We evaluated power across a range of assumptions, including a 20% attrition rate. Power calculations presented herein are conservative in adjusting for attrition and because they do not account for potential power advantages arising from strategies used to address missing data or repeated measures. For the outcome of treatment initiation, we assumed a rate of 13.7%, per prior work with non-treatment seeking AUD patients.⁸ In that case, even assuming 20% attrition, we would have 89% power to detect a treatment effect if the initiation rate in the groups that step up to MI-CBT are ≥30%, which is slightly lower than the 31.6% observed in prior work.⁸ We hypothesize even larger effects likely due to decreased barriers with virtual care. For alcohol consumption, we assumed 70% drinking days in the prior month (estimated from prior literature^49,50) and have 80% power to see a difference of at least 17% between groups. The same power and difference remain if the proportion of drinking days are lower, e.g. 50%. With the N ranging from 160–180 in each Stage 1 condition we will have 80% power to detect a difference in proportions for utilization of 13% (from 13.7% to 26.3% or higher) and difference in proportions of drinking days of 15% (from 70% to 55%).

3. SUMMARY

3.1. Overview of SMART

This SMART is examining virtual strategies to enhance treatment engagement and delivery, and improve alcohol outcomes, among non-treatment seeking AUD patients connected to a healthcare system. Adults are identified and screened for AUD and initially randomized to a phone-delivered session or 4-weeks of interactive messaging, with interventionists employing MI, CBT principles, and psychoeducation to promote reduced drinking and/or treatment engagement. After 4-weeks, treatment ends for those who initiated AUD psychotherapy or report no heavy alcohol use in the past two weeks, whereas non-responders are re-randomized to Stage 2 interventions: 1) enhanced dose of their first intervention (e.g., telephone session or 4 weeks of portal messaging), or 2) 8-sessions of telehealth-delivered MI-CBT. The primary outcomes are alcohol consumption and treatment utilization (initiation and engagement). Findings will address the needs of responders and non-responders and evaluate how intervention combinations may be used to engage and deliver care to patients with untreated AUD in healthcare systems.

3.2. Importance of ADIs to Engage Patients with AUD

Despite the high prevalence of AUD, there is a gap in care where patients suffering symptoms do not typically receive effective interventions. Outreach-based low-threshold virtual strategies may help close this gap by addressing barriers (e.g., stigma, access, availability, one-size-fits-all approaches). The integration of diverse strategies, ranging from a brief intervention style single session to asynchronous messaging over several weeks, to MI-CBT, in a SMART design allows for determining intervention sequences that best fit individual needs and preferences. If effective, there is a pathway for implementation given that most healthcare systems have EHRs⁵¹ (to be used for patient identification), and there are existing payment models, including: reimbursement for telehealth⁵² and emerging structures for cost-recovery for portal-based messaging.⁵³

3.3. Novel Features of the Current Study

Our SMART offers unique, low-burden Stage 1 interventions to facilitate drinking reduction and/or care engagement, with the potential for enhanced dose or stepping up to a higher level of telehealth care when indicated. Although clinicians routinely adjust the recommended dose of AUD interventions (e.g., intensive outpatient programming, varying session frequency), sequencing and dose of novel outreach-based remote strategies to engage patients or deliver care have not been examined in the large untreated AUD population. Our extension of the “patient-first” model¹², where outreach connects patients to convenient telephone or portal-based therapy, is innovative and provides an example of how existing infrastructures can be re-imagined for new models of care.

3.4. Limitations

Potential limitations of our approach exist. Although we are targeting enrollment of 50% women and 25% Black and/or African American individuals, the geographic region and reach of the single healthcare site limits broader representation. Additionally, those without specific technology access are excluded, though we find these exclusions were rare in pilot work. Other exclusions reflect higher risk individuals who could benefit from services, but would require more intensive clinical management, though generally our inclusion criteria are broad and pragmatic. Also, because we only recruit individuals who have EHR data reflecting alcohol-related diagnoses or positive AUDIT screenings, it is possible that sampled participants have a greater likelihood of prior treatment which could influence willingness to enroll in an AUD-focused trial and/or their understanding of or engagement with treatment. Sensitivity analyses of trial outcomes can probe this issue, though recent work has found treatment history was not a strong predictor of alcohol outcomes.⁵⁴ Finally, although the current study is limited to psychotherapies, future work could integrate pharmacological treatments given their efficacy.

4.0. Conclusions

This article outlines a SMART protocol testing the effectiveness of a diverse array of technology-based strategies to engage and deliver treatment to adults with AUD, with varying dose and intensity based on initial intervention response. This design advances a personalized approach to AUD care, with an emphasis on strategies that aim to be low burden and appealing to patients. These strategies have the potential to open the door to further engagement in care that can aid patients in managing a chronic, lifelong condition that, for many, is left untreated. This humanistic, patient-first approach has potential for high impact through utilizing scalable approaches that rely on common infrastructure and can be potentially translated to other SUD populations and related conditions, expanding reach and options for care.