Participants were recruited from ads placed online (eg, Craigslist, Reddit) and in a local newspaper containing a brief description of the study. Interested participants were phone screened for eligibility on alcohol use severity, frequency of binge drinking episodes, suicidal ideation, emotion dysregulation, mental health treatment history, and demographic characteristics. Inclusion criteria included: (1) being 18 years or older, (2) endorsing suicidal ideation in the past month (defined as endorsing a wish to die at least once in the past 30 days), (3) engaging in at least two episodes of heavy drinking in the past month, (4) high emotion dysregulation (defined as scoring one standard deviation above the normative mean on the 16-item Difficulties in Emotion Regulation Scale [32]), and (5) residing in the United States. Exclusion criteria included: (1) currently receiving psychotherapy; (2) diagnosis of bipolar I, schizophrenia, or schizoaffective disorder; (3) non-English speaking; (4) unable to read or write; and (5) having no access to a computer connected to the Internet. Of 398 individuals who expressed initial interest, 60 of 91 eligible individuals returned the informed consent forms and were randomized into the immediate-treatment (n=31) and waitlist (n=29) groups. One person randomized to the immediate-treatment group did not complete any of the assessments, leaving 59 people included in the analyses. The phone screening process helped us prevent participants from enrolling with multiple identities.

The study has been approved by the Human Subjects Division of the Institutional Review Board at the University of Washington (IRB# 50295). Through the initial phone screening, all interested individuals were briefed on the aims of the study and their rights as participants. Participants were informed that their answers to the questions were voluntary, and that they could withdraw from the study at any time. Informed consent forms (see Multimedia Appendix 1) containing details about the study, procedures, and possible risks and benefits involved were provided to all eligible participants. Prior to randomization, participants were required to indicate on the consent forms their preferred mode of communication for receiving study materials, and to return their signed consent forms to the study site. Referral resources for treatment and crisis services were sent by email to interested callers who were deemed ineligible for the study. All participants consented to have their nonidentifiable data published. Access to the dataset is restricted to the principal investigator and authorized research assistants only. Complete data from this study can be obtained by contacting the first author once data analysis has been completed. iDBT-ST is not currently available online.

This study focuses on sensitive topics such as alcohol use and suicidal behavior, and the phone screen and outcome assessments contain questions that could potentially be distressing to participants. As a safety measure, participants were encouraged to consult with research staff whenever they felt distressed during screening and later participation. Suicide risk was also monitored and addressed in ways that are discussed later in this paper.

Semistructured phone screening interviews were conducted with individuals who initiated contact via phone or email and remained interested after receiving information about the study. Phone screens were either prescheduled or conducted during the first call of contact, depending on caller preference, and were administered by trained research assistants. Eligible individuals received informed consent forms to sign and return, and were randomized once their signed forms were received at the study site.

Following randomization into the immediate-treatment and waitlist groups, participants in both conditions received a baseline assessment, followed by identical monthly assessments in the ensuing four months. Primary monthly outcomes included suicide ideation (Scale for Suicidal Ideation [33]), severity and frequency of alcohol consumption (Alcohol Use Disorders Identification Test; AUDIT [34], and Self-Administered Web-Based Timeline Followback [35,36]), and emotion dysregulation (Difficulties in Emotion Regulation Scale [37]). Secondary outcome measures included the Reasons for Living Inventory [38], Short Inventory of (Alcohol) Problems (SIP) [39], Patient Health Questionnaire [40], and Dialectical Behavior Therapy - Ways of Coping Checklist [41]. All participants also received brief online weekly assessments on suicide ideation and alcohol consumption. Suicide risk was monitored via weekly assessments.

Prior to treatment, participants also responded to their expectancy of outcome along with the credibility of the intervention (Credibility/Expectancies Questionnaire [42]). At the end of the treatment period, participants reported on their treatment satisfaction (Client Satisfaction Inventory-Short Form [43]) and therapeutic alliance (California Psychotherapeutic Alliance Scale [44]). Participants actively receiving treatment also received brief daily surveys about their urges to quit treatment, use of DBT skills, and engagement level.

All assessments were administered through an online survey platform (Qualtrics [45]) made accessible to participants via Uniform Resource Locator links in scripted emails sent at scheduled timepoints. As compensation for their participation, participants received up to US $120 in Amazon gift cards upon completing each monthly survey (baseline=$15, 4-week=$20, 8-week=$25, 12-week=$30, and 16-week=$30).

Eligible and consenting participants were randomized into the immediate-treatment (n=31) and waitlist groups (n=29). To control for potential covariates that may spuriously affect analytic results, an open source desktop application called MinimPy [46] was used to match participants for equal distribution between conditions on three variables. The three variables were: sex assigned at birth (Demographic Data Schedule), severity of suicide ideation (Suicide Behavior Questionnaire-Revised) and degree of disordered drinking (AUDIT).

Suicide risk was carefully monitored in the current study. Participants were assessed at the start and end of the recruitment phone screen, as well as throughout the course of their participation. High risk of suicide during phone-screening was indicated by endorsing 4 or more points on a 7-point Likert scale, or an increase of 3 or more points from the start of screening on urge to die by suicide. Throughout the 16-week trial period, weekly questionnaires assessing suicide risk were emailed to all participants. These emails included prompts on: (1) frequency of suicidal urges in the past week, (2) intensity of suicidal urges in the past week, (3) seriousness of acting out on suicidal urges in the past week, and (4) current suicidal urge, all using 5-point Likert scales. Participants endorsing a rating of 3 or higher on any item, or a 2-point increase from the previous week, would be called by phone and assessed for suicide risk. Additionally, the number for the National Suicide Prevention Lifeline was situated next to prompts on suicidality received by all participants.

The principal investigator was on call to intervene by phone at any time. Evidence-based procedures based on the University of Washington Risk Assessment Protocol (UWRAP) [47] were implemented by the principal investigator whenever further assessments of risk were warranted. The UWRAP guides the assessor through a suicide risk assessment procedure that evaluates the presence of a plan, access to means, likelihood of following through with a plan, being interrupted by others, et cetera. If a participant was determined to be at high risk based on this interview and the clinical judgment of the principal investigator, a safety plan was to be developed with the cooperation of the participant and the principal investigator. If a participant was unavailable by phone, a nonjudgmental caring email (see Multimedia Appendix 2) was sent, which expressed concern and encouraged participants to reach out to a suicide hotline or textline provided in the email.

One primary aim of iDBT-ST is to replace maladaptive behavioral strategies with alternative strategies for coping and emotion regulation. The current intervention comprises the skill-training component alone, which is administered in a computerized format. The intervention was modified from an online DBT skills training intervention for emotion dysregulation [48].

Participants randomized into the iDBT-ST group were immediately enrolled into the online treatment program, and received weekly sessions over a period of 8 weeks. After completing the eighth session, participants in active treatment were transitioned into a follow-up phase that entailed monthly and weekly assessments without treatment. Participants randomized into the waitlist group received repeated monthly assessments as well as weekly surveys for 8 weeks prior to receiving treatment.

The iDBT-ST intervention was administered through an online electronic learning server (Articulate Online [49]) which consists of 8 sessions (Table 1) accessible through an access portal individualized for each participant, with one new session activated each week. The sessions begin with a weekly assessment, followed by (1) a review of the last session at weeks 2-8, (2) an introduction of the current session, (3) skills teaching, and (4) summary and consolidation. The skills were taught via videos and graphics describing the skills and examples, along with interactive activities. The intervention completion rate of each participant was also monitored through the server. If consented by participants, reminders to complete sessions were sent via their preferred method before the next session became available. Individualized assignments of skill practice were provided following the completion of each session. See Figure 1, Figure 2, and Figure 3 for screenshots of the intervention.

Sessions 3 and 4 were designed to target problematic drinking, and included the new skill Dialectical Drinking. The skill of dialectical drinking describes the abstinence violation effect (see Figure 4), goal setting, pros and cons of maintaining or changing their drinking behavior, as well as an assessment of motivation and confidence to follow-through on their drinking goal. Dialectical drinking incorporates aspects of motivational interviewing [50] and harm reduction [51] as ways to nonjudgmentally reinforce a change toward goal-oriented drinking behavior. Problematic drinking is also targeted through the DBT skills of Community Reinforcement, whereby participants identify people, activities, and places that can reinforce their drinking goal, as well as Burning Bridges, whereby participants eliminate people, activities, and places that would reinforce their problematic drinking. In burning bridges, participants learned a variety of methods to manage alcohol cravings [15,16]. See Table 2 for common alcohol reduction strategies and their DBT counterparts.

All information provided by participants will be strictly confidential. Answers to phone screens and online assessments that will be kept indefinitely for research purposes were coded in a nonidentifying way, and stored in password-protected files on an encrypted server accessible only to the principle investigator and trained research assistants. Confidential records linking participant names with their respective research identification number will be deleted no later than September 30^th, 2018. All identifying information from ineligible individuals was deleted immediately after phone screening. Additionally, the study site is a secure research clinic equipped with noise-cancelling devices to prevent conversations with participants from being overheard. Voice messages left for participants were devoid of details about the study and other sensitive information that may compromise confidentiality. Any paper documents that contained confidential Patient Health Information are kept in a locked area in the study site where access is restricted, and will be shredded before disposal. These documents may be examined in strictly confidential checks conducted by government or university staff for legal and safety insurance.

Participants were informed of one scenario in which confidentially may be breached. That is, if a participant reveals an imminent high risk of danger to their own or someone else’s life, research staff may address that risk to police or emergency resources.

A power analysis was not conducted because the primary purpose of the data analyses is to describe the effect of iDBT-ST protocol treatment over time using repeated observation data in a preliminary manner. The results will be used to help determine if further research on the treatment program for suicidal heavy drinkers is necessary in larger samples.

Intervention effects will be evaluated by examining change outcomes as a function of treatment condition using both hierarchical linear modeling (HLM) [52] for outcomes with a normal distribution and generalized estimating equations (GEEs) [53,54] for count outcomes. Both HLM and GEEs accommodate observations that are missing at random, thus increasing the power of analyses. Analyses will be conducted using an intent-to-treat framework, in which all randomized and available participants will be included in the analyses. To examine changes over time for the entire sample, analyses will be conducted among all participants during all assessment waves (baseline to 16-week). A set of analyses will also be conducted using the first three assessment waves (baseline to 8-week) comparing immediate-treatment to no treatment, and at each condition’s relative pretreatment, midtreatment (4-week for iDBT-ST; 12-week for waitlist), and posttreatment (8-week for iDBT-ST; 16-week for waitlist) assessment phases.

Given the pilot nature of the study, all outcomes will be evaluated in terms of effect sizes as well as clinically significant change using Jacobson and Truax’s specifications [55]. Treatment effect sizes will be analyzed using Cohen’s d and Relative risk ratios for continuous and count outcomes, respectively. Cohen’s d will calculate within-condition effect sizes across all assessment waves. Descriptive analyses will also be conducted to examine acceptability.