We conducted 2 rounds of a modified Delphi exercise to establish expert consensus on the most important BCTs for inclusion in a digital intervention targeting co-occurring tobacco and alcohol use. We followed the Conducting and Reporting Delphi Studies (CREDES) guidelines [31] to guide the process.

We selected a modified Delphi methodology to enable structured, iterative input from a panel of experts while allowing asynchronous participation. In this study, we modified the classical Delphi approach [32] by presenting a predefined list of BCTs identified in the rapid review, rather than starting with open-ended questions. This modification streamlined the process, increased efficiency, and allowed participants to focus their expertise on evaluating and refining a curated set of concepts [33].

The research team developed a questionnaire based on findings from our rapid systematic review [34], which identified 27 commonly used BCTs in interventions targeting co-occurring alcohol and tobacco use. From these, 16 BCTs were prioritized for expert evaluation based on the research team’s assessment of their relevance and feasibility within the context of a 1-time, direct-to-patient digital intervention delivered through the STOP patient portal. BCTs were excluded if they required clinician involvement, medications, or repeated engagement that could not be accommodated within the digital portal. For example, pharmacological support (BCT 11.1; BCIO:007144) was not included because it typically involves medications that require a prescription, and their use must be carefully assessed by a health care professional. A 1-time digital intervention cannot provide the medical assessment, diagnosis, or prescription needed to determine if pharmacological support is appropriate for an individual, nor can it ensure proper supervision during use. Moreover, the STOP program already provides pharmacological support for smoking cessation through nicotine replacement therapy, which is not a prescription medication.

To facilitate more precise feedback, 3 BCTs—action planning (BCT 1.4; BCIO:007010), problem solving (BCT 1.2; BCIO:007008), and information about antecedents (BCT 4.2; BCIO:007052)—were disaggregated into their core components to better reflect how each could be operationalized within this format. Specifically, “problem solving” was divided into 3 items (identifying barriers, craving management strategies, and recognizing high-risk situations), action planning into 2 items (developing individualized change plans and reduction strategies), and information about antecedents into 2 items (providing general information about antecedents and explaining the link between smoking and drinking). This resulted in 20 items for expert rating. Each BCT was presented with a brief definition and an illustrative example drawn from the reviewed studies to support informed appraisal. Panelists were asked to rate each BCT on the acceptability, practicability, effectiveness, affordability, safety, and equity (APEASE) criteria [35]. The APEASE framework provides a structured approach to evaluating intervention components in terms of their real-world feasibility, impact, and alignment with public health values. Ratings took the form of an ordinal scale, with options ranging from “strongly agree” to “strongly disagree,” as well as an “uncertain” option. Additionally, participants were encouraged to provide free-text comments for each intervention strategy and to suggest additional alcohol intervention strategies that the STOP patient portal should include, allowing reconsideration of any potentially relevant BCTs during the consensus process.

In round 2, participants were invited to rerate BCTs that had not achieved consensus in the first round. They were also shown anonymized summaries of group-level ratings and qualitative feedback from round 1 to facilitate reflection on, and reconsideration of, their initial responses. Both rounds were delivered using Research Electronic Data Capture (REDCap; Vanderbilt University), a secure web-based platform for managing research data.

Panelists were eligible to participate if they had expertise in behavior change, tobacco and alcohol treatment, and digital interventions for substance use, as determined by the research team based on academic publications, clinical experience, or relevant program involvement. Eligible individuals included researchers, clinicians, and implementation specialists with relevant academic or professional experience.

We used a combination of convenience and snowball sampling techniques [36] to recruit a diverse panel. Members of the investigator team leveraged their professional networks and distributed a virtual recruitment flyer via email to 34 experts. In addition, personalized invitations were sent to 14 authors of the studies included in the rapid review, given their familiarity with the literature and relevant intervention components. Although there is no universally agreed-upon ideal sample size for Delphi studies, a minimum of 8 panelists is generally considered sufficient to support consensus building [37,38]. To ensure diverse perspectives and account for potential attrition, we aimed to recruit up to 15 panelists. All panelists took part in the Delphi study voluntarily, and their responses were collected anonymously.

The first survey was administered between August and September 2024, followed by a second round from November 2024 to January 2025. Panelists who agreed to participate were provided with a briefing note (Multimedia Appendix 1) and a link to the REDCap survey. The briefing note included (1) information about the sociodemographic characteristics of the STOP patient population, (2) details of the clinical treatment participants receive for smoking cessation, and (3) a list of BCTs identified by the rapid review as effective in addressing alcohol and tobacco use.

Once panelists opened the survey, they were directed to a consent form. After providing consent, they completed a sociodemographic questionnaire with items about their professional background; place of work; gender identity; and research experience in alcohol, tobacco, and behavioral science. They were then redirected to the first-round survey described previously.

After the first round, the research team sent the initial honorarium, analyzed the data, and created a summary of the survey responses that included data from the ordinal scale and the open-ended text fields. The team shared this summary with each participant when sending the second questionnaire, in which participants rerated the items and responded to existing comments if desired. The second-round survey included comments and points raised by panelists for each BCT, providing participants with the rationale underlying other panelists’ responses in the first-round survey. After completing the second round, the team sent panelists a thank-you email alongside their remaining honorarium.

Survey responses were included in the analysis if participants completed at least 1 survey question. Free-text responses from each round were collated and analyzed qualitatively using a thematic approach [39]. Two researchers independently reviewed all comments to identify recurring patterns and concepts and then met to discuss their findings. Through discussion, themes were developed that captured the key considerations and rationale provided by panelists regarding each BCT. These comments offered valuable context and helped clarify the rationale behind panelists’ agreement or disagreement with the BCTs. Several panelists contributed written comments across BCTs. Selected comments are presented in the Results section to highlight key themes and areas of concern.

Following each round, responses were summarized using measures of central tendency (median), distribution (range), and the proportion of ratings for each score assigned to each BCT. Although no universally accepted threshold exists for defining consensus, prior studies commonly apply a benchmark of approximately 70% to 80% agreement among participants [40].

For this study, a consensus threshold was defined a priori as follows:

This consensus approach has been successfully used in prior evaluations [33,41]. A qualitative analysis of free-text comments was conducted using a thematic approach to identify key patterns and insights [39]. Items that reached consensus as appropriate were included in the round 2 questionnaire with the points raised from the previous round and were rerated on the effectiveness criterion only. Items that reached consensus as inappropriate were not included in the round 2 questionnaire. For items that achieved partial consensus, the research team reviewed accompanying free-text comments and, based on this feedback, proposed revisions to improve their clarity and framing. These revised items constituted the round 2 questionnaire, along with a summary of points raised by panelists for each BCT that provided the rationale of other panelists when they answered the first survey.

This study was approved by the Centre for Addiction and Mental Health’s Quality Project Ethics Review (QPER-100). All participants were provided with a study briefing note and a digital consent form prior to participation. Informed consent was obtained electronically before completing the survey. To recognize the substantial time and expertise contributed by senior experts, panelists received a CAD $300 (CAD $1=US $0.71 as of Jun 18, 2026) honorarium. This was distributed in 2 installments: CAD $100 upon completion of round 1 and CAD $200 following completion of round 2. Depending on the participant’s preference, compensation was provided via an electronic gift card or electronic funds transfer.

Of the 48 experts invited to participate in the study, 29% (n=14) agreed to participate. Twelve (86%) experts completed the first round of the Delphi questionnaire, and 100% (n=14) of panelists participated in the second round. All 14 experts completed the sociodemographic survey. Overall, 14% (n=2) experts (panelists 6 and 11) did not complete the first round but participated in the second round, accounting for the discrepancy in participation between rounds. Invited experts who did not participate represented similar areas of expertise, including behavior change, tobacco and alcohol treatment, and digital interventions for substance use. All panelists were researchers with expertise in substance use, behavior change, and digital interventions. On average, participants reported 22.1 (SD 10) years of research experience related to alcohol, 17.0 (SD 11.2) years related to tobacco, and 23.2 (SD 10.2) years related to behavioral science. Sociodemographic characteristics of the panel are presented in Table 1.

Out of the 20 BCTs included in the first round, the panel reached consensus that 4 BCTs were appropriate for inclusion in a direct-to-patient digital intervention: goal setting, action planning (individualized change plan), action planning (reduction strategies), and feedback on behavior.

These techniques were considered practical, motivational, and feasible to implement digitally without real-time human support. However, 7 panelists offered several considerations regarding the application of these techniques. For example, in relation to goal setting, similar to comments made by other panelists, 1 panelist noted the following:

Overall, 3 BCTs reached consensus as inappropriate for inclusion in a direct-to-patient digital intervention: social support, social comparison, and comparative imagining of future outcomes. Although 7 panelists recognized their potential value in the free-text comments, they raised concerns about their feasibility, the need for tailoring, or the cognitive burden associated with their inclusion in a BI. For example, for comparative imagining of future outcomes, 1 panelist commented the following:

These BCTs were also among those with the lowest equity ratings (details in Multimedia Appendix 2).

Partial consensus was achieved for 13 BCTs: problem solving (identify barriers), problem solving (craving strategies), problem solving (identify situations), information about health consequences, information about antecedents, information about antecedents (link between alcohol and smoking), pros and cons, self-monitoring, behavioral substitution, information about social and environmental consequences, behavioral practice or rehearsal, credible source, and nonspecific reward.

Panelists also emphasized that these techniques could inadvertently exacerbate inequities if delivered without appropriate tailoring, particularly in cases where delivery depends on self-guided engagement. With respect to problem solving (identify barriers), 1 participant cautioned as follows:

Two general key considerations were identified in the panelists’ free-text responses. First, panelists emphasized the importance of including only BCTs that could be implemented without clinician oversight, highlighting the need for feasibility in the context of a 1-time, self-guided digital intervention. Second, panelists stressed the importance of prioritization, raising concerns about the cognitive and logistical burden associated with delivering a large number of components simultaneously. One participant commented on problem solving (craving strategies):

A summary of the round 1 results is presented in Table 2.

Out of the 14 BCTs included in the second round, 2 BCTs, reattribution and pros and cons, reached consensus as appropriate. Reattribution in this digital health context is defined as providing knowledge about what constitutes a standard drink and the safe drinking guidelines established by the World Health Organization to prompt users to reevaluate whether their drinking is “normal” or low-risk, thereby shifting their self-perception and perceived control over their behavior. This distinguishes it from purely informational techniques, which transfer knowledge without necessarily challenging causal beliefs or self-attribution, and from feedback-based techniques, which reflect observed behavior back to the user. Similar to comments shared by other panelists, 1 panelist pointed out how these techniques were consistent with guideline recommendations:

Five BCTS were deemed inappropriate: problem solving (identify situations), behavioral substitution, information about social and environmental consequences, behavioral practice or rehearsal, and nonspecific reward. Consistent with the concerns raised by other panelists, 1 panelist (panelist 10) commented that problem solving (identifying situations) “...sounds pretty involved and might be too intensive for the typical case.”

For behavioral substitution, 1 panelist (panelist 2) noted that suggested alternatives “would need some thought—also probably less equitable than others as these really depend on resources of the individual,” and recommended that “any suggestions offered should include options that are free or low cost.”

For nonspecific reward, 1 panelist (panelist 7) observed that “not all patients will have the same resources in terms of being able to reward themselves, potentially requiring money or time.”

Another panelist (panelist 5) cautioned that “rewards would need to be more powerful than drinking—might be difficult for some patients to conceptualise a type of reward that would work here.”

Another panelist commented regarding information about social and environmental consequences:

In total, 7 BCTs reached partial consensus: problem solving (identify barriers), problem solving (craving strategies), information about health consequences, information about antecedents, information about antecedents (link between alcohol and smoking), self-monitoring, and credible source. Similar to what other panelists had commented, 1 participant cautioned and agreed with fellow panelists’ previous points for information about antecedents and problem solving (identify barriers):

Panelists highlighted that certain BCTs will only be effective if they are continued over time. As 1 participant mentioned regarding problem solving (craving strategies):

Another point panelists emphasized was the importance of providing health information to participants. One participant commented the following:

The consensus ratings for the BCTs from the second survey are shown in Table 3, and the medians and consensus ratings for the BCTs across both rounds are shown in Multimedia Appendix 2. In total, based on the 2 rounds, 6 BCTs reached consensus as appropriate for inclusion in a patient portal targeting individuals with concurrent tobacco and alcohol use. An overview of the Delphi study process is shown in Figure 1.

This modified Delphi study aimed to identify BCTs that experts deemed appropriate for use in a 1-time, direct-to-patient digital intervention targeting dual use of alcohol and tobacco, with a specific focus on middle-aged adults with low income and education levels, those with a high prevalence of mental and physical health conditions, and those living in rural areas. Given these characteristics, the intervention must be simple, self-directed, and scalable, with minimal reliance on human facilitation or repeated engagement.

After 2 rounds, 6 BCTs achieved consensus as appropriate. Of these, 3 BCTs—goal setting (BCT 1.1; BCIO:007002) and the 2 variants of action planning (BCT 1.4; BCIO:007010)—fall within the Goals and Planning domain of the Behaviour Change Technique Taxonomy (version 1) [30]. The remaining techniques are distributed across other domains: Feedback and Monitoring (feedback on behavior: BCT 2.2; BCIO:007023), Shaping Knowledge (reattribution: BCT 4.3; BCIO:007053), and Comparison of Outcomes (pros and cons: BCT 9.2; BCIO:007069). This pattern reflects a preference for intervention strategies that emphasize goal-directed behavior, self-regulation, and intentional planning. These strategies enable individuals to translate intentions into action by setting specific goals, making concrete plans, and monitoring progress, which is particularly relevant in a direct-to-patient digital context where clinician involvement is minimal at the point of delivery, although patients may later have an opportunity to discuss their responses and plans with a health care professionals during follow-up care.

These findings are consistent with existing literature on digital behavior change interventions for substance use [42,43]. Goal setting and action planning are foundational self-regulation strategies and are among the most commonly used and effective BCTs for both smoking and alcohol reduction [44,45]. For example, in digital alcohol reduction interventions such as the Drink Less app [46], which includes modules for goal setting, self-monitoring, and pros and cons evaluation, users who engaged with goal-oriented content demonstrated more sustained reductions in alcohol use [47]. Although Drink Less involves repeated user engagement, the use of similar BCTs in a 1-time intervention may still serve as a low-touch prompt for behavior change, especially when grounded in motivational content and presented in a user-centered format. Digital interventions for smoking cessation, such as the Smoke Free app, also use self-monitoring, personalized feedback, and goal setting BCTs and have shown promise in promoting quit attempts and sustained abstinence [48]. Similarly, in a metaregression of physical activity and healthy eating interventions [49], the use of action planning and goal setting was significantly associated with greater behavior change. Although these findings were derived from a different behavioral domain (physical activity and healthy eating), they reinforce the theoretical value of these strategies across contexts. Reattribution and pros and cons target mechanisms such as causal attribution (BCIO:006032) and self-efficacy beliefs related to behavior (BCIO:006154). Although less frequently implemented in digital interventions, their selection reflects expert recognition of their motivational potential for the intended population.

The BCTs reattribution and pros and cons also achieved consensus for inclusion. These techniques align with theoretical frameworks such as the health belief model and the transtheoretical model, which emphasize cognitive restructuring and decisional balance as precursors to behavior change. Reattribution helps individuals reinterpret the causes of their behavior, potentially reducing fatalism and increasing perceived control, which are important constructs in substance use contexts [50,51]. Similarly, pros and cons is a well-established decisional balance technique that helps users reflect on the positive and negative consequences of behavior change. It engages mechanisms of evaluative belief (BCIO:006148) and readiness to change [22]. These strategies have been used successfully in apps such as Drink Less and the UK National Health Service Smokefree app, which includes a decisional balance tool for evaluating the advantages of quitting smoking [52,53]. Although reattribution is less commonly used in interventions [54], it was considered appropriate in this case, given the target population’s need for motivational support and increased self-efficacy. Research suggests that individuals who use both alcohol and tobacco are more likely to experience motivational barriers and fatalistic beliefs, which can make behavior change more difficult [55]. This highlights the importance of addressing these psychological factors through BCTs such as reattribution.

The expert panel also reached consensus on 8 BCTs as inappropriate in this context: social support (BCT 3.1; BCIO:007028), social comparison (BCT 6.2; BCIO:007073), comparative imagining of future outcomes (BCT 9.3; BCIO:007070), problem solving (identify situations: BCT 1.2; BCIO:007008), behavioral substitution (BCT 8.2; BCIO:007095), information about social and environmental consequences (BCT 5.3; BCIO:007064), behavioral practice rehearsal (BCT 8.1; BCIO:007094), and nonspecific reward (BCT 10.3; BCIO:007252). Experts expressed that these BCTs were less feasible due to their reliance on social interaction, personalization, or repeated engagement, which are incompatible with a 1-time, automated intervention. It is important to note that these techniques may still be valuable in other formats, such as clinician-delivered interventions or multisession digital programs, and their exclusion here reflects contextual limitations rather than their intrinsic effectiveness. However, given the constraints of a 1-time digital intervention, especially for populations with limited time, goal setting, action planning, and feedback on behavior emerged as the highest priority BCTs. These techniques were consistently rated as both effective and feasible, offering the greatest potential for impact within a brief, self-guided format. If intervention time or space is restricted, prioritizing these core self-regulation strategies is recommended.

Equity considerations appeared to further differentiate BCTs that failed to reach consensus from those that were ultimately included. Specifically, BCTs that did not reach consensus were more frequently characterized by panelists as requiring greater cognitive effort, sustained engagement, social resources, or material supports—factors that may disproportionately disadvantage individuals with lower socioeconomic status or limited digital access. This pattern is consistent with existing literature demonstrating that digital behavior change interventions can inadvertently widen health inequities when intervention components are more accessible or actionable for individuals with greater resources, higher health literacy, or stronger digital skills [56].

A key strength of this study is the diversity of the expert panel in terms of disciplinary background (behavioral science and substance use) and geographic representation, including participants from multiple countries. This broad perspective enhances the relevance of the findings to global digital health contexts and supports the design of 1-time digital interventions that address both substances. At the same time, the panel was composed exclusively of researchers. Although research expertise provides a valuable and rigorous foundation for early-stage intervention design, the absence of clinician and patient perspectives may have skewed BCT selection toward techniques that are theoretically robust but not necessarily optimal in practice. Given that the intervention is intended for direct use by patients enrolled in the STOP program, many of whom have low income and education levels, a high burden of mental and physical health conditions, and limited experience with digital health tools, patient perspectives on engagement, comprehensibility, and feasibility will be essential for further refinement.

The use of the APEASE criteria similarly reflects both a strength and a limitation. By requiring panelists to systematically evaluate affordability, practicability, effectiveness, acceptability, safety, and equity, the framework ensured that BCT selection was not driven by effectiveness evidence alone, representing a meaningful departure from conventional approaches to intervention design. A limitation of this study is that the findings are based on expert consensus rather than empirical evaluation of intervention effectiveness. Although the Delphi method is well suited to identifying BCTs perceived as appropriate and feasible, it does not provide evidence of their actual impact on behavior change outcomes. Therefore, the BCTs identified in this study should be interpreted as promising candidates for inclusion rather than as components proven to be effective in practice. The explicit inclusion of equity is particularly notable, as it highlights concerns about differential accessibility that are often absent from BCT selection processes. However, prior research has shown that digital health interventions can inadvertently contribute to intervention-generated inequalities when access, engagement, or benefit are patterned by socioeconomic status, digital literacy, or resource availability [57], and the equity assessments in this study were based on expert judgment rather than direct evidence from underserved groups. As such, the equity-related findings should be interpreted as hypothesis generating rather than definitive and require validation through empirical and participatory research. A qualitative study is planned to explore patients’ and health care professionals’ perspectives on the intervention’s design, with findings intended to inform future iterations and ensure that expert consensus is translated into content that is genuinely accessible and acceptable to the intended population.

This modified Delphi study identified a core set of BCTs (centered on goal setting, planning, feedback, and decisional balance) that experts judged to be appropriate for a 1-time, direct-to-patient digital intervention targeting dual alcohol and tobacco use. These findings provide preliminary guidance for the development of brief digital tools aimed at promoting self-regulation and behavior change in real-world, resource-constrained settings. Although expert consensus provides a useful foundation, future research should integrate patient perspectives and usability testing to ensure that these techniques resonate with intended users and support meaningful behavior change.