Although mHealth is wide reaching, the uptake and availability of certain technologies are not equal across all populations. Therefore, researchers must consider the contextual factors of the target population before deciding which mHealth technologies will be used or developed to promote feasible implementation [28]. The target population should encompass all potential stakeholders (ie, those using the technology both directly or indirectly) and contextual factors derived from the physical environment or clinical workflow in which the technology is to be implemented [16]. For example, if a researcher develops an mHealth intervention that is inaccessible to the target population in the real-world context but then provides the technology to research participants to prove its efficacy, this mHealth intervention is only effective within the sphere of academia. Such research findings may still have implications in policy (eg, informing health care coverage to include novel technologies) but are unlikely to be translated quickly into practice to improve public health [17].

As SSBC begins to scale up across Canada, mHealth technologies provide an opportunity to improve long-term behavior change adherence while providing participants with continued support from program providers in an accessible manner. Although much of the mHealth research and development has focused on smartphones and associated apps, such technologies are less likely to reach rural Canadian populations (ie, Canadians at increased risk for developing T2D) [30]. However, cellular phone ownership and SMS text messaging rates are increasing, with 99.7% of Canadians covered by mobile phone networks [30]. To date, SSBC has been implemented within Kelowna, a metropolitan area. However, as SSBC expands to more rural communities, mHealth prompts sent via SMS text messaging have the capacity to reach and engage with the largest number of potential SSBC participants compared with the mHealth prompts sent via push notifications (through an mHealth platform) in the laboratory-based trial.

Before investing resources in developing an mHealth intervention, researchers should first evaluate the intervention’s potential to influence behaviors within a given context [16]. Common pitfalls in formative development include spending too much or too little time on this phase [16]. The use of existing data, conducting research using methods that prioritize speed and simplicity (eg, rapid prototyping and innovation sprints), conducting or consulting reviews of the literature, and conducting qualitative research with the target end users can aid in the evaluation of the potential utility of a given technology for the population of interest [29]. Furthermore, agile innovation suggests that creating a clear termination and evaluation plan a priori may avoid continuing to invest resources in an intervention that does not suit the intended users or target context [16].

To identify the potential utility of mHealth prompts sent via SMS text messaging as SSBC expands to more rural areas, exploratory analyses were conducted to identify how long participants self-reported that they exercised on their respective mHealth platforms during the laboratory-based trial (when they received mHealth prompts sent via push notifications) and community-based trial (when they received no additional BCTs to increase engagement) [31]. We found that 83% of participants in the laboratory-based trial (ie, those who received prompts) self-reported exercise on their mHealth platform for an average of 82 days, whereas only 34% of participants in the community-based trial (those who did not receive prompts) self-reported exercise on their mHealth platform for an average of 43 days.

Following this, a more in-depth analysis of the laboratory-based trial was conducted to determine the acute impact a prompt may have on self-monitoring (ie, any day in which they logged yes, no, or day off) and self-reported exercise (ie, only the days in which they logged that they completed exercise) in the week following compared with the week preceding a prompt [32]. Self-monitoring and self-reported exercise data from the mHealth platform were averaged over 1, 3, 5, and 7 days before and after a prompt for the first and second half of the 12-month follow-up phase and were compared using t tests. The impact of the prompt was strongest in the first half of the year, with no significant differences found in the second half of the year. In the first half of the year, self-monitoring significantly increased in the 3 days following a prompt (P<.001; d=0.60), and self-reported exercise significantly increased in the 3 (P<.001; d=0.37), 5 (P=.04; d=0.14), and 7 days (P=.02; d=0.15) following a prompt.

Together, these secondary analyses provide preliminary evidence that mHealth prompts may be useful within the SSBC context, and more comprehensive development of an mHealth-prompting (to be sent in the future via SMS text messaging to improve scalability) intervention is warranted.

Once the potential utility of an mHealth-prompting intervention was identified, a scoping review was conducted to identify how existing DPPs develop mHealth-prompting interventions [33]. The results from this review highlight that mHealth prompts are typically delivered via SMS text messaging, followed by push notifications, and that they are generally well-received by participants. However, both the development of prompt content (ie, what BCTs are used and how theory influenced content) and delivery of prompts (eg, frequency, timing, and duration of prompting interventions) are often underreported, highlighting the need for structured development and rigorous evaluation of mHealth-prompting interventions before deployment.

In addition to conducting our own review, previous reviews assessing SMS text messaging interventions were also consulted. To date, many reviews have assessed the impact of SMS text messaging on health behavior change interventions [34-41], and meta-analyses have consistently shown that SMS text messaging interventions have a significant effect on hemoglobin A_1c among individuals with diabetes [39,40,42], weight loss [36,37], and physical activity [38,43]. It is widely accepted that behavior change interventions should be developed using theory, past evidence, and formative research [20,44] and that sufficient intervention detail be reported to allow for replication [45,46]. However, many SMS text messaging interventions lack rigorous development and thorough reporting, thereby limiting their utility in future intervention development and implementation [47].

In addition to how SMS text messaging interventions are developed, the description of theoretical mechanisms within message content and reporting of delivery characteristics (eg, timing and frequency) are largely unspecified in the literature [48]. With respect to SMS text messaging delivery, the evaluation of weight management SMS text messaging interventions by Skinner et al [36] included subgroup analyses examining intervention duration (<6 months vs 6 months vs 12 months), message frequency (daily vs weekly or biweekly vs personalized), and 1- versus 2-way messaging and found no significant subgroup differences. The authors noted that poor intervention descriptions within publications may have affected their ability to accurately code aspects of intervention delivery. Although a trend was noted that less frequent messaging (weekly or biweekly) was associated with greater reductions in weight (mean −2.88 kg, 95% CI −4.56 to −1.21 kg) than for daily messages (mean −1.56 kg, 95% CI −2.26 to −0.86 kg), the authors concluded that the mechanisms of action by which SMS text messaging programs lead to these effects remain largely unclear, and further investigation into message delivery and content features is warranted. This is mirrored by other reviews that note the heterogeneity in message timing, frequency, and duration within the existing literature and call for future research to determine the optimal message dose for health behavior change [35,49].

The selection of intervention components should be performed early in the development process to identify different component levels. Component levels include the different doses of an intervention component one wishes to test (eg, high or low dose or turning a component on or off). By selecting components early, adequate time may be invested in preparatory development work to select appropriate component levels and consider how they will be tested. The detection of key intervention components should be based on theory, evidence, and end users [28,50,51].

Steps 1 and 2 may help to clarify intervention components that are either (1) based on existing evidence or (2) currently under researched and may be studied to move the field forward. In this way, evidence-based intervention components may be incorporated into the intervention package, and where evidence for a potential intervention component does not yet exist, researchers are able to further study the mHealth intervention component of interest (eg, if there is no existing evidence on how the frequency of intervention contact may influence a target population, but previous theory or evidence suggests that it should influence their behaviors, inclusion of that delivery component can allow a new line of evidence to be built for the target population).

The integration of previous research evidence and theory can provide insight into what intervention components have been previously identified as having the potential to be effective; however, unless these studies were conducted with the population of interest, they are unlikely to provide guidance on which intervention components are most important or can be best implemented within a given context [28]. To ground mHealth development within the implementation context, the Person-Based Approach suggests engaging in in-depth qualitative research [28]. Understanding user perspectives and accommodating their priorities within the candidate intervention components can aid in maximizing the acceptability and uptake of interventions when they are at the implementation stage [28].

Based on qualitative work profiling SSBC participants, participants wanted continued support from their coach and additional information regarding prediabetes after the completion of the training phase, and they found mHealth technologies to be facilitators for maintaining their long-term behavior change [26,27]. From the results of the scoping review, the intervention components to be tested in the SSBC MOST optimization trial included SMS text messaging content and delivery components (ie, timing and frequency of prompt delivery) [33]. By understanding what message content SSBC participants would like to see (and the underlying theoretical mechanisms) and by optimizing SMS text messaging delivery to best facilitate behavior change, this research may reduce intervention costs and promote user engagement, ultimately reducing T2D risk.

The MOST recommends that interventions be developed and optimized to meet resource-related implementation considerations (eg, constraints on personnel time, costs, or complexity of the selected technology). Typically, interventions are developed without considering affordability or the ability to be implemented as designed, which results in considerable research funds and time devoted to establishing efficacious interventions that are not practical in the real-world context [52]. By accounting for implementation constraints at the onset, the MOST aims to achieve intervention EASE by balancing the effectiveness of an intervention against relevant implementation constraints on affordability (eg, can be developed and delivered within certain budgetary constraints), scalability (eg, can be implemented with high fidelity), and efficiency (eg, contains only active intervention components).

By considering potential end users when designing mHealth content, researchers can tailor development and delivery based on the intended implementation context, thereby improving translation into practice [10]. Specifically, these constraints can help balance the intervention EASE while developing a specific optimization objective. Intervention optimization objectives can be broad (eg, achieving the best outcomes for the lowest price) or specific (eg, improving physical activity by a minimum of 15 minutes per day while keeping the overall intervention cost <US $300), and the level of specificity will depend on the constraints outlined at the onset and the requirements and resources of potential end users.

When developing the SMS text messaging intervention, constraints were placed on the messages themselves, and the platform that was chosen to deliver the messages. The optimization objective was defined as identifying the SMS text messaging intervention that increases physical activity adherence most during the SSBC follow-up phase, given the identified constraints.

To improve the implementation of evidence-based mHealth research in real-world contexts, it has been suggested that intervention content should be developed (1) using a dynamic and iterative process including end users, (2) based on existing research evidence and theory, and (3) with the implementation context in mind (eg, clinical workflows) [28,53]. The integration of specific theoretical frameworks such as the Behaviour Change Wheel (BCW) has been suggested to systematically identify candidate intervention components and develop theory-based mHealth content that is underpinned by specific mechanisms of action (note that the use of such theoretical frameworks can also be helpful when developing a conceptual model) [21]. The BCW is a synthesis of 33 behavior change theories that collate the barriers and facilitators needed to change a target behavior.

The development and refinement of SMS text messaging content for SSBC were split into 3 parts: identification of BCTs, message development, and message evaluation and refinement.

Where previous experimental evidence exists, as identified in the review of existing literature (step 2), it should be used to inform intervention component levels, warranting further testing. If no experimental evidence exists for different component levels or the objective is to further the science of mHealth development and evaluation, different components or component levels may be chosen. At this stage, intervention developers should begin to think about the study design for future intervention optimization trials; these can include factorial, sequential multiple assignment randomized trials (SMART) and microrandomized trials to name a few. The design of the optimization trial is driven by the research questions and available resources. When conducting a factorial experiment, the most efficient way (and the one recommended by the MOST) is to stick with 2 levels for each component (eg, on or off or high or low dose). Once the MOST optimization trial design has been decided, researchers can consider the candidate intervention component levels to be tested. Component levels can be as broad or as specific as needed and often depend on the nature of the intervention and research questions.

Creating a conceptual model to link intervention components to their theoretical mediators and expected outcomes may improve the ability of an intervention package to be effectively adapted to different contexts or settings [60]. Developing a conceptual model within the preparation phase of the MOST can be completed using a number of potential theories. As pinpointing a specific theory that fits one’s needs can be overwhelming, the use of a theoretical framework such as the BCW can aid mHealth researchers in developing a conceptual framework based on theoretical constructs [21].

Based on the identified intervention components of prompt content and message delivery (timing and frequency), a conceptual model was created (Figure 3) to outline how these components are anticipated to influence target behaviors (dietary and physical activity behavior change adherence). Although each individual message is made up of different BCTs targeting different barriers and facilitators, the content generally aims to increase an individual’s motivation and self-efficacy to adhere to the diet- and exercise-related behavior changes made during the SSBC training phase. From our secondary analyses, it appears that the prompting dose may influence a participant’s behaviors by encouraging them to continue to self-monitor their behaviors.

Before the optimization phase, Collins et al [20] suggest conducting a feasibility study to establish whether each candidate component and component level can be feasibly implemented as intended. Feasibility studies are a common step in the process of developing and translating social science and public health interventions. Conducting a feasibility study can provide necessary information on participant recruitment, intervention acceptability, and the feasibility of administering an intervention as intended. This information may be used to aid in the decision to (or not to) conduct a fully powered optimization and efficacy trial. Furthermore, this may ensure that resources are not wasted in conducting an optimization trial if the intervention itself cannot be delivered as intended [61]. The primary aim of a feasibility study should be to inform the feasibility or acceptability of an intervention and identify modifications that need to be included within the intervention before a large-scale trial. As feasibility studies are not fully powered trials, conducting inferential statistical tests is discouraged, as P values rely on sample size [61]. If a researcher needs to make decisions about the inclusion or exclusion of an intervention component or component level, they should identify a priori how they intend to use the feasibility study data to inform this decision making. It is recommended that participant acceptability and preferences be used instead of P values in addition to effect sizes (which are less reliant on sample size).

To test the feasibility and acceptability of the message delivery platform before optimization, a feasibility study is currently being conducted [59]. In addition to general intervention feasibility (ie, can it be delivered as intended), this study aims to assess participant preferences regarding message timing and frequency to further refine intervention levels. The feasibility study will result in a final set of mHealth-prompting delivery characteristics for further testing in the second phase of the MOST, using a factorial experiment.