Adaptive interventions for alcohol misuse and violent behaviors among adolescents and emerging adults in the emergency department: A sequential multiple assignment randomized controlled trial protocol

Abstract

Alcohol use and violent behaviors among youth are associated with morbidity and mortality. An emergency department (ED) visit provides an opportunity to initiate prevention efforts. Despite promising findings from our single session SafERteens brief intervention (BI), impact is limited by modest effect sizes, with data lacking on optimal boosters to enhance effects. This paper describes the protocol for a sequential, multiple assignment, randomized trial (SMART). Adolescents and emerging adults (ages 14–20) in the ED screening positive for alcohol use and violent behaviors (physical aggression) were randomly assigned to: 1) SafERteens BI + Text Messaging (TM), or 2) SafERteens BI + remote Health Coach (HC). Participants completed weekly surveys over 8 weeks after the ED visit to tailor intervention content and measure mechanisms of change. At one-month, intervention response/non-response is determined (e.g., binge drinking or violent behaviors). Responders are re-randomized to continued intervention condition (e.g., maintenance) or minimized condition (e.g., stepped down). Non-responders are re-randomized to continued condition (e.g., maintenance), or intensified condition (e.g., stepped up). Outcomes were measured at 4 and 8 months, including primary outcomes of alcohol consumption and violence, with secondary outcomes of alcohol consequences and violence consequences. Although the original goal was to enroll 700 participants, COVID-19 impacts on research diminished recruitment in this trial (enrolled n = 400). Nonetheless, the proposed SMART is highly innovative by blending real-time assessment methodologies with adaptive intervention delivery among teens with comorbid alcohol misuse and violent behaviors. Findings will inform the content and timing booster interventions to alter risk behavior trajectories.

1. Introduction

Although the initiation of alcohol use occurs during adolescence, alcohol misuse increases into emerging adulthood. In 2020, rates of past month binge drinking were: 4.1% (ages 14–17), 31.4% (ages 18–25), 22.9% (ages 26 and older) [1]. Alcohol use is associated with the leading causes of death for youth, including motor vehicle crash, overdose, suicide, and homicide [2,3]. In addition, in 2019, 21.9% of high school students reported a recent physical fight, 8.2% experienced dating violence, or 13.2% carried a weapon in the past year [4]. Violence is particularly concerning given health (e.g., injury, death) and social (e.g., imprisonment) consequences, which disproportionally impacts males who are Black or African American [5,6].

The relationship between alcohol use and violence can be explained by theories of problem behavior clustering [7], as well as the acute effects of alcohol intoxication [8–10], with additional systematic stressors for youth residing in economically disadvantaged communities [11]. Research supports the association among drinking and violence across the socio-ecological spectrum (e.g., individual, relationship, community) [12–16]. The emergency department (ED) setting provides an opportunity to initiate individual-focused prevention interventions for adolescents and emerging adults. The American Academy of Pediatrics and American College of Emergency Physicians, as well as government agencies (i.e., National Institute on Alcohol Abuse and Alcoholism, Substance Abuse and Mental Health Services Administration), endorse alcohol screening, brief intervention, and referral to treatment (SBIRT) approaches [17–19]. Further, universal screening of youth in EDs located in economically disadvantaged communities finds elevated rates of violence [20]. As noted by the Center for Disease Control’s technical package for youth violence prevention [21], a brief intervention focusing on violence prevention (i.e., SafERteens) in this setting is also promising [22]. This intervention reduces severe peer violence, dating violence, and/or alcohol consequences among youth [23–27]. The SafERteens session incorporates motivational interviewing (MI) using the “why” and “how” model of change [28–30] focusing on reducing risks and increasing promotive factors based on resiliency theory [31], enhancing coping with negative affect, self-efficacy and skills for avoiding alcohol/violence, prosocial relationships, and use of community services.

Despite positive results of SafERteens, the effect sizes from this single session intervention were modest. Adaptive interventions provide a potential solution to optimizing health outcomes, as they are tailored to needs over time, adjusting the dosage and/or type of intervention content offered based on intervention response [32,33], preserving the most intensive resources for those with the greatest needs. Sequential, multiple assignment, randomized trial (SMART) designs provide an innovative methodology to develop adaptive interventions [32,33]. Although SMARTs are becoming more common in the health literature [34,35], and substance use specifically [36–39], no SMARTs have tested early interventions for alcohol or violence among adolescents and emerging adults.

We are currently conducting a SMART delivering SafERteens to adolescents and emerging adults in the ED, who screen positive on a survey for recent binge drinking and violent behaviors, followed by various combinations of mobile health (mhealth) boosters, namely text messaging (TM) or health coach (HC) sessions. Prior research examining TM among adolescents and emerging adults, including a meta-analysis [40] and studies conducted in the ED [41–43], shows efficacy in reducing substance use. For example, prior ED based studies used TMs to conduct surveys about drinking, which were followed by automated, tailored text messages focusing on alcohol risk reduction. Also, our prior work shows automated TM focused more broadly on risk and promotive factors (e.g., goals, strengths, reasons to avoid fighting), were promising for youth violence prevention [44,45]. Specifically, our initial study included three TMs per day, including a daily survey message followed by two intervention focused TMs [44]; however, based on youth feedback, we reduced burden in a subsequent study to eliminate the survey message [45]. Next, another promising booster approach that we are testing is remote, telephone-based therapeutic support with a health coach, which was feasible and acceptable in a pilot study, with 50% of youth attending at least 5 of 8 sessions over 8 weeks [46].

Our SMART study is testing the efficacy of individualized sequences of interventions based on individual response [47]. Aim 1 focuses on examining the relative efficacy of initiating treatment with SafERteens, combined with either TM or remote HC, on primary (i.e., alcohol consumption, violence behaviors), secondary (i.e., alcohol consequences, violence consequences), and other outcomes (i.e., other drug use, victimization). Aim 2 focuses on determining which intervention strategy is necessary to enhance outcomes among responders to the initial treatment (continued intervention or stepped down) and among non-responders to the initial treatment (continued intervention or stepped up). Finally, exploratory aims are to compare the pre-specified 8 embedded interventions on outcomes. This paper describes the study protocol.

2. Methods

2.1. Design overview

The original goal was to enroll 700 adolescents and emerging adults (ages 14–20; herein termed youth) in the SMART. Youth screening positive for binge drinking and physical aggression were eligible for the SMART. Following completion of the baseline assessment and SafERteens session (in-person or via telephone), participants were randomly assigned to booster conditions: Text Messaging (TM) or remote Health Coach telephone call (HC). After receiving the SafERteens intervention, participants completed weekly surveys over 8 weeks to tailor intervention content, measure mechanisms of change, and determine response to interventions. At 1 month, participants were classified as responders or non-responders to the intervention. We defined an intervention responder as not reporting any binge drinking or aggression in weeks 3 and 4 (although they could in weeks 1 or 2 to allow time for the intervention to take effect); intervention non-responders were defined as those reporting these behaviors in weeks 3 or 4. Importantly, we defined those with missing survey data in weeks 3 or 4 as non-responders for conservative purposes. Responders in each arm were re-randomized to continued condition or minimized condition (e.g., stepped down); specifically, youth ‘responding’ to the TM or HC who are randomized to the stepped down condition received a resource brochure. In a parallel manner, non-responders were re-randomized to continued condition or intensified condition (e.g., stepped up); specifically, youth ‘not-responding’ to the TM who are randomized to the stepped-up condition receive HC calls, whereas youth ‘not-responding’ to the HC, who are randomized to the stepped up condition, received the intensified HC+ condition (e.g., HC’s send supportive text messages between calls). Outcomes were measured at 4 and 8 months (see Fig. 1).

Fig. 1.

Study SMART design.

2.2. Study setting, Institutional Board Approval & Trial Registration

Study sites included emergency departments in economically disadvantaged communities, originally planned for Hurley Medical Center (HMC) in Flint, Michigan, and expanded to Ascension St. John Hospital in Detroit, MI and Covenant Hospital in Saginaw, MI. The study protocol was approved by the University of Michigan Institutional Review Board for the use of Human Subjects in research (ClinicalTrials.gov NCT03344666, University of Michigan # HUM00109156), as well as local IRB boards at the individual study ED sites. A Certificate of Confidentiality was obtained for the study from the National Institute of Health.

2.3. SMART eligibility and exclusions

Patients (ages 14–20) presenting to the study EDs for any reason were identified via the electronic medical record for research assistants to approach (in-person or remotely, see Procedures) for screening to determine study eligibility. Patients were excluded if: 1) they could not provide informed consent/assent due to mental incompetence, incarceration, or unstable medical condition; 2) their parent/guardian was not available to provide consent (ages 14–17); or 3) they presented for acute suicidal ideation or attempt, sexual assault, or child abuse (see Fig. 1). Based on the screening survey, eligible participants for the SMART self-reported: a) past 4-month binge drinking based on the AUDIT-C [48] item, tailored by sex and age based on a blending of NIAAA recommendations [49] and consistency for clinical implementation (females: 3 or more ages 14–17, 4 or more for ages 18–20; males: 4 or more for ages 14–17, 5 or more for ages 18–20); and, b) past 4 month physical aggression based on prior work [23] combining 10 items modified from the Conflict Tactics Scale (CTS; i.e., slapped, pushed or shoved, slammed against wall, punched or hit with something that could hurt, beat someone up, kicked, pulled a knife on, pulled a gun on, used a knife on, used a gun on) and 5 items from Add health [50,51] (i.e., serious physical fighting, alcohol before fight, group fighting, hurt someone badly enough to need bandages or care).

2.4. Procedures for enrollment, consent, eligibility and randomization

Consent was obtained for those ages 18–20, and parental consent (and child assent) was obtained for those ages 14–17. Remuneration was: screening ($1.00 gift from a dollar store if in-person), $40 for the baseline, $10 for each weekly assessment (weeks 1–4), $15 for each weekly assessment (weeks 5–8; increased mid-way through the trial from $10), $40 for the 4-month assessment, and $40 for the 8-month assessment. The 4-month screening window was selected so that participants would be recently involved with binge drinking and/or physical aggression, and to reflect parallel time frames as the follow-up assessments (at 4-months and 8-months).

After completing the baseline assessment, all participants received the SafERteens intervention (in-person or remotely), which was followed by the initial computerized random assignment (blocks of 10) to TM or HC, stratified by sex and age group (14–17;18–20). The condition began the following week on Monday; weekly surveys were completed on Sunday (before 8 am Monday) to allow for tailoring of the TMs. The second randomization (stratified by sex and age group) occurred after the week 4 survey (based on responder classification described above), with the assigned condition beginning on the following Monday.

2.5. Intervention conditions

2.5.1. SafERteens BI

The SafERteens intervention [22,23] was based on motivational interviewing (MI) [30] and was facilitated by a tablet computer, which presents tailored screens to prompt content for the therapists and standardize delivery, using the “why” change and “how” change framework [28,29], collaboratively exploring behaviors using open-ended questions, reflections, and 0–10 rulers. Tailored computerized prompts included: session introduction, goals and strengths, experiences with fighting, experiences with alcohol and drugs, reasons to stay away from fighting, reasons to stay away from alcohol and drugs, and scenarios to identify tools to avoid fighting (i.e., peer violence/retaliation, dating violence) and to avoid risky alcohol and other drug use (i.e., coping with negative affect, social activities). The session ended with a session summary, discussion eliciting one next step to avoid drinking and fighting, and review of a resource brochure, whichcontained tools to reduce drinking and fighting as well as community resources (e.g., suicide hotlines, domestic violence shelters, healthcare, mental health/substance use treatment, financial/food/clothing assistance, employment, mentors, leisure activities). The brochure was also provided to youth (e.g., paper, emailed, texted) at baseline and follow-ups.

2.5.2. Text messages (TMs)

In prior work, we developed a library of TM [44,45] boosters delivered in the 2 months following the SafERteens intervention, which were expanded in this study to create a library of tailored, twice daily messages (based on weekly surveys). For example, alcohol messages were tailored based on reasons for drinking or not drinking that week, but did not reveal their specific drinking behaviors to preserve confidentiality; similarly, violence messages were tailored based on reasons for fighting or avoiding fighting that week. Tailoring defaulted to the prior week’s data for those with missing weekly survey data. Participants chose the time for TM delivery within a window of time (10 am-3 pm; 4 pm–9 pm). The morning messages focused on up-stream factors associated with alcohol use and/or violence, to enhance novelty and breadth of factors covers, including negative affect, impulsivity, sensation seeking or excitement, social support, hope, and leisure activities. The afternoon messages primarily focused on reasons and strategies to avoid/reduce involvement with alcohol and violence; however, one day focused on cannabis use. These ~1200 TMs were rated and edited by a group of paid youth advisors in Flint (with diversity by sex and race, including Black or African American individuals) to create the final message bank (see Table 1). The word “MCoach” preceded all TMs, based on youth feedback regarding confusion when receiving a TM from a 5 digit number they didn’t recognize; participants could to text “STOP” to end the TMs. Also, participants could “pull” TMs by texting MCOACH to obtain affirmations and tips related to prevention of violence and alcohol misuse. Finally, one day per week youth created an ultra-tailored message to be sent to themselves via a weekly survey question asking, “Type a message to remind yourself of your strengths or positive qualities that will help you be the person you want to be”. Notably, participants were not required to respond to the TMs.

Table 1.

Overview of text message content.

Day of Week	AM	PM
Sunday	Reminder: weekly survey	Reminder, Thank you message, & Affirmation
Monday	Affirmation: Negative Affect	Why/How: Marijuana motivational statement or Tool
Tuesday	Affirmation: Impulsivity	Why: Violence Motivational statement
Wednesday	Affirmation: Excitement	Why: Alcohol Motivational statement
Thursday	Affirmation: Social Support	Youth Created Message
Friday	Affirmation: Hope	How: Violence Tool
Saturday	Affirmation: Activities	How: Alcohol Tool

2.5.3. Remote health coach call (HC)

Our HC condition was informed by prior work [46], including qualitative coding of SafERteens sessions as well as exploratory analyses regarding risk and protective factors associated with response. The HC integrated alcohol and violence content within a framework of a healthy lifestyle for reaching goals, using a MI framework of “why” and “how” change, including open-ended questions, reflections, and summaries to elicit commitment talk to avoid/reduce alcohol and violence (and other drug use). At any time if youth indicated acute risk (e.g., intent to harm self or others, victimization), the HC collaboratively creates a safety plan based on our IRB approved protocol. The HC call used a 4-step framework: 1) Check-in: review weekly assessment; 2) Why: elicit and provide reflections and summaries about progress; 3) How: elicit and provide strategies for cutting down or avoiding consequences, if not now in the future; and 4) Closing: summarize the call. To enhance fidelity, HCs and participants collaboratively chose what they want to discuss across broad topics: coping, social support, community connections and leisure activities (Table 2). Participants in the HC condition, who were assessed as non-responders at 1 month and assigned to step up to HC+, received twice weekly individualized, supportive TMs from the HC sent from a study cell phone, in addition to their weekly calls, and could text the HC on a study cell phone with concerns arising in between calls. Regarding fidelity, after initial MI training, HCs were monitored via weekly individual supervision by a master’s level psychologist who was a member of the MI network of trainers.

Table 2.

Topics discussed in Health Coach Telehealth Calls.

Week	Topic
Coping	• Elicit coping motives/triggers for drinking and fighting. • Enhance self-efficacy to manage stressors, negative mood, and anger by affirming strengths and strategies, such as mindfulness.
Leisure activities	• Elicit social motives/triggers for drinking and fighting. • Elicit free-time activities that promote goals and to avoid drinking and fighting. • Elicit strategies for avoiding people and places involving drinking and fighting.
Social Support	• Elicit potential future consequences of drinking and fighting in relation to goals. • Collaboratively identify adult mentors and social support to reach goals.
Resources	• Elicit motives for use in relation to mental and physical health. • Collaboratively identify community resources to address motives, including adult mentors and leisure activities.

2.5.4. Summary of intervention conditions over time

Our study included 8 intervention strategies embedded in the SMART design, where each intervention strategy included an initial intervention, a second intervention for those determined to be responders, and a second intervention for those determined to be non-responders. Six of the eight intervention strategies were adaptive, such that the second intervention differs by response to the first intervention (e.g., ramp up or ramp down), and two were static, such that the second intervention is the same (e.g., continue with the intervention) regardless of response to initial intervention (see Table 3).

Table 3.

Overview of embedded interventions.

Embedded Intervention Strategy #	Adaptive or Static	First Intervention	Second Intervention for Responders	Second Intervention for Non-Responders
1	Static	BI +4 weeks Text Messages	4 weeks Text Messages	4 weeks Text Messages
2	Adaptive	BI +4 weeks Text Messages	4 weeks Text Messages	4 weeks Health Coach Call
3	Adaptive	BI +4 weeks Text Messages	Resource Brochure	4 weeks Text Messages
4	Adaptive	BI +4 weeks Text Messages	Resource Brochure	4 weeks Health Coach Calls
5	Static	BI +4 weeks HC	4 weeks Health Coach Calls	4 weeks Health Coach Calls
6	Adaptive	BI +4 weeks HC	4 weeks Health Coach Calls	4 weeks Health Coach Calls + Personalized Messages
7	Adaptive	BI +4 weeks HC	Resource Brochure	4 weeks Health Coach Calls
8	Adaptive	BI +4 weeks HC	Resource Brochure	4 weeks Health Coach Calls + Personalized Messages

2.6. Measures

2.6.1. Primary and secondary alcohol outcomes

Our primary alcohol consumption outcome variable is assessed using the past 30-day Timeline Follow Back (TLFB) interview [52,53]. A secondary outcome is alcohol-related consequences, measured with a modified Brief Young Adult Alcohol Consequences Questionnaire (B-YAACQ) [54], which includes 24 specific alcohol-related problems (e.g., blackouts, hangovers) assessed over the past four months (responses: ‘0 = None’ to ‘3=More than 5 times’). Specifically, we removed two questions from the B-YAACQ which are not frequently endorsed (“My physical appearance has been harmed by my drinking’ and ‘I have felt like I needed a drink after I’d gotten up (that is, before breakfast)”) and added two questions from the Original YAACQ (‘I have damaged or lost property after drinking’ and ‘I have gotten into physical fights because of drinking’).

2.6.2. Primary and secondary violence outcomes

Our primary violence outcome is physical aggression towards friends, neighbors, relatives, acquaintances, dating partners, and strangers, which is measured in the past 30-days using the modified TLFB interview (see prior work) [55]. A secondary outcome is violence consequences, assessed with 13 items expanded from prior work (i.e., hurt/injured, family/friends hurt injured, arrested/run from police, watch my back more/on guard, afraid someone get back at, guilty or down, trouble at work/school, hurt someone else, constant desire to fight, family/friends told me to stop fighting, arguments with family/friends, trouble getting along/lost friendship, tried but unable to stop fighting) [23].

2.6.3. Other exploratory outcomes

Other drug use is assessed with 7 items for frequency of use of cannabis, illicit drugs and misuse of prescription drugs (opioids, sedatives, stimulants) based on a modified ASSIST-Lite [56]. Victimization is assessed using the TLFB for past 30-day victimization as done in prior work [55].

2.7. Planned analyses

Using an “intent-to-treat” approach, all randomized participants will be included in analyses even if a youth does not engage in interventions. Missing data will be for a SMART as described by Shortreed et al. (2014), with sensitivity analyses conducted with and without imputation [57]. Aim 1 is to compare the efficacy of adaptive interventions that begin with BI+TM vs. BI+HC on reducing alcohol misuse and violent behaviors among youth in the ED. Hypotheses are that participants assigned to the HC will have significantly less primary (alcohol consumption; violent behaviors), secondary (alcohol consequences, violence consequences) and other outcomes (other drug use, victimization) than those assigned to the TM. Analyses will contrast interventions that begin with BI+TM vs. BI+HC on primary and secondary outcomes (see Fig. 1). The primary outcomes are assessed at baseline, 4 and 8 months. Using intent-to-treat analyses, generalized linear mixed models (GLMM³⁰³), also known as random effects or growth curve models), will be used to analyze the longitudinal data for each outcome. Each analysis will fit a GLMM with fixed effects time (categorical), group, and a group-by-time interaction terms, where group is an indicator of phase-one treatment (BI+TM vs BI+HC); we will adjust for the stratification variables of age and sex. The GLMMs include a random intercept and an unstructured within-person correlation structure for the residual errors. The aim of this analysis is whether there is a phase-one treatment difference using a likelihood ratio test for the addition of the treatment and treatment by time terms to the model excluding these. If the treatment by time interaction is not significant, the effect of treatment will be estimated from the simpler model with only treatment and time point (categorical) included in the model.

Aim 2 is to identify the most efficacious second-stage strategy based on response to optimize participant outcomes. Hypotheses are a) Among non-responders, youth who are “stepped up” will have better primary (alcohol consumption; violent behaviors), secondary (alcohol consequences, violence consequences) and other outcomes (other drug use, victimization) outcomes than those who continue with the initial intervention. b) Among responders, youth who are “stepped down” to control will have worse outcomes than those who continue with the initial intervention. This analysis compares cells for responders (is it better to continue with the intervention or step down the intervention) and for non-responders (is it better to continue or step up). Both primary and secondary longitudinal outcomes will be examined using GLMMs as described above, but (a) including only the subset of responders (or non-responders) to stage 1 treatment, (b) defining group as the second stage strategy (for responders to continue or step down, and for non-responders to continue or step-up initial intervention), and (c) using longitudinal outcomes at 4 and 8 months.

3. Results

3.1. Enrollment summary

Recruitment occurred 5/2018–10/2022. Despite significant challenges in research conducted in ED settings during the pandemic, we successfully enrolled 400 youth in the SMART. In total, 3865 participants completed the screening survey, with 496 (12.8%) screening positive for eligibility; 400 (80.7%) were enrolled in the SMART, received the SafERteens intervention session, and were randomized to the HC or TM arm. Follow-up assessments are ongoing. The average age of the sample was 18.4 (standard deviation = 1.5). Participants were 78.7% female sex and 21.3% male sex; gender identity was 77.3% female, 21.5% male, 0.8% genderqueer, 0% transgender male, 0% transgender female, and 0.5% other. Regarding race, participants were 39.5% White, 45.3% Black or African American, 0% Asian, 1.5% Native American/Alaskan Native, 0.3% Hawaiian/Pacific Islander, and 11.8% multi-racial, and 1.8% unknown; 8.8% indicated Latinx ethnicity.

3.2. Power analyses for enrolled sample size

The planned sample size for this study is based on Primary Aim 1, using alcohol use and physical violence as independent primary outcomes, to compare groups that begin with BI+HC vs. BI+TM. We assume Poisson distributions for the primary outcomes, with each outcome tested at a two-sided α = 0.05. If we consider the enrolled sample size of 400, for Aim 1, we can detect a rate ratio of 0.88 for alcohol use and 0.78 for violence. Given prior work, we anticipate that 50% of teens will respond to one month of BI+HC or BI+TM, so that each of the 8 subgroups is expected to include 50 participants. Then, Aim 2 includes comparisons within response status of those who continue initial intervention, or step up (for non-responders) or step-down (for responders); these are anticipated comparisons of 100 vs. 100. In aim 2, we will have 80% power to detect rate ratios of 0.83 for alcohol use and 0.70 for violence within each responder group. These rate ratios represent small to medium effect sizes [58]. Moreover, our SMART explores the effects of the adaptive interventions, each of which includes 2 arms (an initial treatment, a follow up treatment for responders, and a follow up treatment for non-responders). These adaptive interventions motivated the SMART. We will estimate the effect of each of the 8 embedded adaptive interventions (shown in Table 3) and will provide 95% confidence intervals. From ranking these adaptive interventions, we can take the best performing one to power a subsequent confirmatory trial comparing the best adaptive intervention to standard of care or static interventions. Finally, comparison of the 8 embedded interventions will be exploratory as originally planned, albeit with a smaller sample size.

4. Conclusions

4.1. Overview of study

This comparative efficacy SMART study is testing the efficacy of various combinations of booster interventions following receipt of the SafERteens intervention among adolescents and emerging adults in the ED reporting recent binge drinking and physical aggression. Aim 1 focuses on which is the best first stage intervention post-ED visit, TMs or HC calls, and Aim 2 focuses on the best intervention strategy for responders and non-responders. We selected a 1-month period for first stage intervention and a 1-month period for the second stage intervention to be clinically consistent with the early intervention focus of the study and based on the frequency of our variable used to determine response, namely binge drinking and physical aggression, which typically occur weekly or less among these youth. Exploratory aims are to compare the embedded 8 interventions (i.e., combinations of first phase intervention, plus second phase intervention for responders and second phase intervention for non-responders) on outcomes to inform future interventions. Challenges to study recruitment included the initial impact of the Flint Water Crisis, which delayed recruitment start-up and reduced youth patient volumes. Following this crisis, the impacts of COVID-19 on ED research, which were particularly salient in economically disadvantaged communities, presented additional challenges, including requiring pauses in recruitment for extended periods, fewer patients able to be approached (e.g., unable to approach those responding to “yes” to COVID screening questions during medical triage and/or in droplet precautions), and difficulties in hiring research staff to work in-person in the ED during a pandemic.

4.2. Importance of adaptive interventions

The proposed study is a major leap forward from single session brief interventions, and from studies which have used traditional designs to compare static intervention boosters. In contrast, this SMART is exploring the efficacy of individualized sequences of interventions for alcohol misuse and violent behaviors based on individual response, to maximize health outcomes among adolescents and emerging adults. Importantly, interventions for these youth should consider addressing concomitant risk factors (e.g., cannabis use, mental health) when tailoring intervention content.

4.3. Novel aspects of the design

The proposed study offers several innovations, with future translation in mind. For example, testing adaptive interventions using a SMART design will allow for identification of the most promising sequence(s) of interventions using sophisticated analyses. Next, weekly assessments will provide rich data, which is currently lacking, to explore mechanisms of change; this data also is used to tailor TMs for optimal efficacy. Comparing starting with TM versus with HC calls has significant potential for public health impact, mimicking real life clinical practice, with parsimony of resources based on need and scalability for clinical integration.

The timing and dose of assessments and intervention delivery in this SMART was carefully considered based on the study focus on youth with multiple risk behaviors associated with injury (binge drinking and violent behaviors), with anticipated variability in the severity of these behaviors among ages 14–20. Given the expected frequency of these behaviors, we chose weekly assessments over 8 weeks, whereas mHealth studies examining youth with more frequent risk behaviors may benefit from more frequent assessments (e.g., daily, multiple times per day). In parallel, we chose an 8-week intervention period based on pragmatic considerations to allow sufficient periods for the initial intervention (1 month) prior to determination of response/non-response, and for the second intervention (1 month), as well as to balance intervention dose with participant burden and future scalability. We chose weekly health coach sessions to increase the likelihood that at least 4 sessions would be completed over the 8 weeks (e.g., 2 sessions per month) given competing demands faced by youth and our pilot data [46]. Next, we selected twice daily TMs which did not require a response to cover earlier and later in the day and to reduce burden based on prior work [45]. Future studies are needed to determine optimal dose and timing of mhealth interventions.

4.4. Limitations

Given that prior work demonstrates the efficacy of the SafERteens BI vs. control, we elected to forgo a control group and compare the relative efficacy of adding TM and HC to this intervention. We are not collecting objective measures of self-report of alcohol consumption and violent behaviors, since collaterals may not know about involvement with these behaviors, biological measures for alcohol are not useful for prevention focused work, and criminal justice data are limited because youth violence may not involve the police. Instead, we will enhance reliability and validity of self-reported risk behaviors among youth by using valid assessments and ensuring confidentiality [59,60]. Due to COVID-19, we did not reach the planned sample size. Although we maintain power for a meaningful effect in aim 1 related to initial randomization, findings for aim 2 (e.g., re-randomization based on response) may be more exploratory. As planned, additional exploratory analyses include comparisons of the 8 embedded interventions, which will inform future adaptive interventions.

5. Conclusions

Future papers will examine the efficacy of ED-initiated interventions involving mhealth boosters on reducing involvement with drinking and fighting among adolescents and emerging adults. Findings from the SMART, which will identify optimal initial interventions (TM or HC) post-ED visit, and subsequent interventions based on response and non-response, will provide clues to inform future adaptive interventions.

Data availability

No data was used for the research described in the article.