Just-in-Time Support for Smoking, Vaping, and Eating Urges During High-Stress Moments Among Gender and Sexual Minority Adults in Dhaka

Overview

Gender and sexual minority adults in Bangladesh experience high levels of stigma-related stress. Acute stress can trigger strong, affect-driven urges to smoke, vape, or engage in stress-attributed eating. These urges can rise and fall within minutes to hours, operating outside the reach of conventional clinic-based interventions delivered days or weeks later. This pilot micro-randomized trial (MRT) evaluates whether delivering a brief, culturally adapted support message via smartphone during moments of high self-reported stress reduces the intensity of the urge for the participant's pre-specified target behaviour, relative to no message. Over a 24-day period, 115 participants will be scheduled for up to 5 ecological momentary assessments (EMA) per day. Whenever a participant reports an acute stress level of 5 or higher (0 to 10 scale) and satisfies pre-specified privacy, safety, behavioural, and spacing criteria, the app randomizes with probability 0.5 to deliver either a brief support message or no message. By contributing multiple randomized decision points over time, each participant serves as their own micro-control. The primary outcome is the intensity of the urge for the participant's pre-specified target behaviour, measured 45 to 60 minutes after randomization. An exploratory urge assessment is also collected 10 minutes after randomization; comparing the two time points will help establish whether the support message's effect is immediate, delayed, or both, and will inform the choice of follow-up window for a future, fully powered trial. Prior to trial launch, all EMA items, message content, notification logic, and digital safety protocols will be developed and tested with community members, reviewed by a Community Advisory Board, and locked. As a pilot optimization trial, its objectives are to estimate the short-term causal effect of the prompt, to identify the contexts in which that effect is larger or smaller, and to establish whether momentary assessment and just-in-time intervention research can be delivered safely, privately, and acceptably in this high-stigma setting. The study does not evaluate long-term behavioural maintenance and does not claim to address the underlying structural stigma and discrimination that produce minority stress.

Registered Study and Formative Phase: The registered interventional study is a micro-randomized trial (MRT) involving 115 participants over a 24-day assessment period. It is preceded by a formative cultural adaptation and technical refinement phase conducted to develop, optimize, and lock the momentary assessment items, support message library, notification wording, and digital safety features. This preliminary phase comprises cognitive walkthroughs and think-aloud interviews with approximately 10 community members, a short structured momentary assessment run to characterize the naturalistic distribution of momentary stress ratings, and a technology validation period. The formative phase involves no randomization and no assignment to any intervention. It is not part of the registered interventional cohort, is not included in the registered enrollment figure, and does not determine the study start date. Micro-Randomization Framework: Randomization occurs repeatedly within each participant rather than once between participants. Each participant is randomized at every decision point during the 24-day period that satisfies all predefined eligibility criteria, thereby contributing both intervention and control observations over time. The primary unit of randomization and proximal observation is the eligible decision point; repeated observations are nested within participants. Decision Point Definition: Randomization occurs at available high-stress decision points, defined as a completed momentary assessment at which all five of the following conditions are satisfied simultaneously: the assessment is completed within the allowed 30-minute response window; the participant reports a current stress level of 5 or higher on a 0 to 10 scale; the participant confirms that it is currently safe and private enough to receive a message; the participant has not already reported performing their designated primary target behavior since the preceding completed assessment; and no randomization has occurred in the preceding 90 minutes. Moments that fail any criterion are logged by the platform for system audit and feasibility monitoring but are not randomized. Randomization Procedure: At each available decision point, the digital platform randomizes prompt assignment independently, with a fixed probability of p = 0.50, to either delivery of a supportive coping prompt or no prompt (active assessment control). The randomization outcome and timestamp are securely logged. Assessment Schedule and Conflict Resolution: Participants receive five scheduled baseline momentary assessments per day, delivered at randomized times within five pre-specified daily time blocks, with one assessment per block across the 24-day period. Following every randomization, in both the prompt and no-prompt conditions, participants complete a brief micro-check assessment at 10 minutes, with a 5-minute response window, and a main proximal follow-up assessment at 45 minutes, with a 15-minute response window. To prevent assessment overload and survey collision, if a regularly scheduled baseline assessment falls within 60 minutes after a randomization, that scheduled assessment is automatically suppressed by the platform and recorded as protocol-suppressed rather than missing. Target Behaviour and Multi-Behavioral Logic: Participants who qualify on both smoking/vaping and stress-attributed eating designate a single primary target behavior at baseline. This designation is locked before the assessment period begins and remains fixed throughout the study to maintain primary outcome integrity. Urge intensity for both behaviors is measured at every momentary assessment to allow descriptive, exploratory characterization of urge patterns across behaviors, consistent with the pre-specified exploratory outcome measures; however, only the locked target behavior provides the primary confirmatory outcome. Urge scores for the two behaviors are analyzed independently and are never averaged or combined. Sample Size and Power Assumptions: The target enrolment is 115 participants. Under the planning assumptions regarding attrition, momentary assessment completion, high-stress trigger frequency, and availability, approximately 1,225 randomized decision points are expected. Detailed power assumptions and sensitivity analyses will be prespecified in the statistical analysis plan. Statistical Analysis Plan: The primary analysis estimates the proximal causal effect of prompt delivery on behaviour-specific urge intensity using weighted and centered least squares with robust standard errors. Binary secondary outcomes are analysed using the estimator for causal excursion effects with binary proximal outcomes. Secondary behavioural outcomes are powered for estimation rather than for confirmatory testing and are reported with confidence intervals to inform the design of a future trial. Interpretation, Estimands, and Scope: The control condition is "no message" within the context of active momentary monitoring, rather than a neutral placebo message. The estimated causal effect therefore reflects the incremental contribution of receiving a supportive coping prompt relative to active self-monitoring alone, and does not separate the effect of the prompt's content from the effect of being notified at all. Participants are not blinded to condition, and outcomes are self-reported. The 10-minute micro-check is delivered identically across both conditions, holding proximal self-monitoring and appraisal constant across arms. These design features are inherent to the JITAI framework and are stated as part of the estimand in all reports of findings. Safety, Privacy, and Advisory Governance: Privacy is treated as an essential component of participant safety. Notification wording is neutral and non-disclosing, and intervention content is displayed only after the participant opens the study application. Participants may decline, skip, or temporarily pause assessments at any time. Direct identifiers are stored separately from momentary assessment data, and access is restricted to authorized study personnel. A prespecified distress-response and referral protocol operates throughout the trial. A Community Advisory Board reviews the recruitment procedures, assessment items, support-message content, notification wording, digital safety protections, and distress-response procedures before the trial materials are finalized and locked.