The availability of digital behavioral health interventions has surged in recent years. However, when deployed in research trials and disseminated in real-world practice, the uptake of and engagement with many digital interventions are often lower than desired [1-3]. In turn, this results in the unrealized potential of both immediate and sustained health outcome improvements among the intended users [3]. Although numerous factors may contribute to low uptake, one of the most important is a suboptimal fit between the characteristics of the technology and the needs, skills, and context of the user. To adequately address these complex interrelationships, the development of digital behavioral health interventions should include a theory-driven, iterative, and human-centered design (HCD) process.

Importantly, there are a number of models and frameworks available to guide the digital intervention development processes [4-9]. The Model for Internet Interventions [4] is a key example that has been used as a basis for the development of many digital health programs [10-13]. This model can help researchers distinguish and operationalize various components of digital behavioral health interventions and identify the relationships among the components. More specifically, this model posits that to explain behavior change across digital behavioral health interventions, it is necessary to consider design-related components, areas, and elements, including user characteristics, environment, intervention content, level of intervention support, and targeted outcomes [4]. Furthermore, the Model for Internet Interventions highlights 8 main areas that comprise the digital health application used to deliver the intervention (ie, appearance, behavioral prescriptions, burdens, content, delivery, message, participation, and assessment). To improve the likelihood of digital health applications meeting the needs and requirements of users, HCD and instructional design processes should be applied when considering and manipulating these 8 areas.

Although this and other models and frameworks provide helpful principles and guidelines [4-9], they are not intended to be a step-by-step prescription for digital intervention development. A scoping review of 160 papers regarding research activities for participatory eHealth development processes identified a variety of methods and products [14]. However, there was no evidence of an optimal single-step approach for developing digital behavioral health intervention applications [14]. The findings from this narrative review highlighted the importance of researchers and developers selecting the most appropriate objectives for and methods for developing the context and user characteristics of their digital behavioral health interventions.

Despite the flexibility in methodological processes, the application of HCD and instructional design processes in the development of digital behavioral health interventions is consistently recommended [15]. The International Organization for Standardization defines HCD as “an approach to systems designs and development that aims to make interactive systems more usable by focusing on the use of the system and applying human factors and usability knowledge and techniques” [16]. Interdependent design activities include understanding and specifying the context of use, specifying the user requirements, producing design solutions, and evaluating the design. In addition, the International Organization for Standardization posits that six requirements must be met for an HCD process: (1) the design is based upon an explicit understanding of users, tasks, and environments; (2) users are involved throughout design and development; (3) the design is driven and refined by user-centered evaluation; (4) the process is iterative; (5) the design addresses the entire user experience; and (6) the design team includes multidisciplinary skills and perspectives [16].

Complementary to HCD concepts, instructional design processes involve designers executing cycles of continuous formative evaluation to ensure that the intervention meets the users’ needs, prior knowledge, and experience [15]. These instructional design processes should involve setting measurable learning objectives or performance requirements, assessing the users’ achievement of the targeted outcomes, and revising the program components until the desired outcomes are achieved [15,17-20]. Ultimately, the ability of digital behavioral health interventions to achieve sustained engagement and desired behavioral outcomes is enhanced when using scientific frameworks along with proven and context-specific HCD and instructional design approaches during the development process.

Although HCD processes are important across different contexts, the need for them is magnified in rural regions, including Appalachia. Historically, extending digital interventions into the Appalachian region has been hindered by digital divide [21,22]. However, similar to other rural American communities, Appalachia is making great strides in narrowing the digital divide and closing the rural-urban gap in home broadband internet connection and smartphone ownership [23-26]. However, little is known about how rural Appalachian adults engage with digital behavioral health interventions. As such, human-centered and instructional design processes are especially important for interventions targeting this and similarly underserved regions where access to digital behavioral health interventions has been limited.

In addition to the geographic and digital divide context of Appalachia, the behavioral context of sugar-sweetened beverage (SSB; eg, soda or pop, sweet tea, sports and energy drinks, and fruit drinks) consumption is noteworthy. SSBs are the single largest source of calories in the US diet and account for approximately 7% of the total energy intake of the US adults [27]. In Appalachia, SSB intake is disproportionately higher and accounts for an average of 14% of the total energy intake—twice as high as national estimates [28]. Consistent with both national and Appalachia-specific data showing that SSB is the largest contributor to added sugar intake [27-29], excessive SSB consumption is undeniably pervasive in Appalachia. However, there are only a few known SSB consumption–specific behavioral interventions that have used digital technologies or applied HCD processes [30-33].

The primary objective of this study was to describe a flexible, step-by-step approach to and an HCD process for developing digital behavioral health interventions by illustrating iSIPsmarter as an example. iSIPsmarter is a technology-based behavior and health literacy intervention aimed at improving SSB consumption behaviors among Appalachian adults. The secondary objective is to illustrate the key components and characteristics of iSIPsmarter that resulted from the HCD process. Although iSIPsmarter’s content is specific to improving SSB consumption behaviors among Appalachian adults, the intent is that the HCD process will have wide applications in the development of digital behavioral health interventions across multiple geographic and behavioral contexts.

The development process for the digital behavioral health intervention iSIPsmarter was guided by the Model for Internet Interventions [4] and by best practices in HCD [16] and instructional design processes [17-20], which is being evaluated in a randomized controlled trial (Trial Registration: NCT05030753). Figure 1 illustrates the conceptual integration of this model and these processes, along with key definitions. Specifically, the development process of iSIPsmarter included 13 semistructured interviews across three nonsequential iterative phases: (1) contextual, (2) prototype testing, and (3) end user testing phases.

This formative, flexible, and step-by-step approach allowed the advisory team participants to provide insights at all phases of the development process and interact with multiple prototype drafts developed by the content development team (CDT).

The following sections describe the considerations for the CDT, advisory team participants, adaptation context, iSIPsmarter intervention, and adaptation process. Subsequently, the 3 phases of data collection and the data analysis strategy are described.

This program evaluation project involved activities that do not represent human participant research and, therefore, did not require submission to the institutional review board (Human Research Protection Program Standard Operating Procedures [34]).

The development of digital behavioral health interventions should be conducted by a multidisciplinary CDT [15,16,35]. This includes subject matter experts, behavioral psychologists, web designers, instructional designers, and developers and programmers.

The multidisciplinary CDT involved in the development of iSIPsmarter included 4 doctoral-level researchers: 2 (50%) SSB content experts who had led prior SSB behavioral intervention trials in the targeted Appalachian region, 1 (25%) expert on digital behavioral health technology interventions, and 1 (25%) instructional design expert. This team was further supported by research staff experienced in nutrition, public health, and digital health development and implementation, including those with expertise in the design and creation of robust user interface and user experience. For the duration of this HCD process, the team met weekly to collaboratively draft advisory team interview guides with clear objectives, develop prototypes, and respond to feedback from the participants. Likewise, the CDT also wrote content for the iSIPsmarter cores, developed media-rich interactions and videos, drafted personas and vignettes, and developed an integrated internet-based platform.

When developing digital behavioral health interventions, the selection of advisory team members is an important decision. Ideally, the participants should represent the intended end user of the intervention in terms of demographic and cultural characteristics, health literacy and digital literacy skills, geographic location, and patterns of the targeted behaviors [16].

The advisory team members involved in the development and design of iSIPsmarter were a convenience sample of prior SIPsmartER participants who resided in the Appalachia region. The advisory team members also had to have internet access and the ability to review the iSIPsmarter program materials on the web. These members were diverse in terms of age (21 to 60 years), gender, and socioeconomic status. Of the 15 prior SIPsmartER participants invited, 7 (47%) joined the advisory team. These participants were involved throughout the intervention design and development processes, with the first interview conducted in February 2019 and the final interview in December 2021. In total, the advisory team members were compensated with up to US $200 in electronic gift cards for their participation.

iSIPsmarter is a technology-based behavior and health literacy intervention aimed at improving SSB behaviors among Appalachia adults [36]. iSIPsmarter was adapted from the evidence-based intervention SIPsmartER [37]. SIPsmartER is a 6-month intervention that includes 3 in-person small-group classes, 1 live teach-back call, and 11 interactive voice response (IVR) calls; the intervention also comprises a 12-month maintenance phase that includes monthly IVR calls (refer to the database Evidence-Based Cancer Control Programs for details [38]). The classes cover key behavioral content and action planning, whereas the IVR calls engage users in reporting SSB intake and personal action planning. Several formative research projects among Appalachian adults initially guided the development of SIPsmartER [39-41]. The content and strategies of SIPsmartER are guided by the theory of planned behavior and health literacy, numeracy, and media literacy concepts [42-46]. In brief, SIPsmartER was found to be effective at reducing and maintaining SSB consumption behaviors through a full-scale randomized controlled trial and a pilot dissemination and implementation trial conducted in collaboration with the Virginia Department of Health districts in Appalachia, Virginia [29,37,47-58]. Although SIPsmartER had been highly successful in reducing SSB consumption, transitioning from its original structure to a fully digital modality presented an opportunity to focus on optimizing scalability and reach.

At the onset of developing iSIPsmarter, the CDT had ample experience understanding and intervening on SSB consumption in the targeted region and had over a decade’s worth of rich qualitative, observational, and experimental data from the previous SIPsmartER trials. As such, content and behavior change techniques related to SSB consumption were relatively well established when embarking on the digital intervention development. However, little was known about the opportunities, barriers, and access to technology in the targeted population for intervention delivery or data collection purposes. Similarly, weight self-monitoring and the use of cellular-enabled scales to encourage in-home weighing had not been previously explored in the targeted population and region. Therefore, the intention of the iSIPsmarter adaptation process was largely focused on changing the mode of delivery and adding key content around weight while maintaining other core components and the cultural relevance for the intended Appalachia target audience.

In its final form, iSIPsmarter comprises 6 internet-delivered interactive cores, which are metered out sequentially over time, with a new core becoming available 1 week after completion of the previous core; an integrated SMS text message strategy to engage users in tracking and reporting SSB consumption behaviors; and an electronic cellular-enabled scale for in-home weighing [36]. At the end of cores 2 to 6, the users evaluate their SSB and weight diary data and set a personalized SSB consumption and weight action plan for the upcoming week. After the sixth core is completed, there is a recurring maintenance core where users can continue behavioral tracking and personalized action planning. To assist with mastering key content and behavioral strategies, the cores also include PDF resources that users can view or print. iSIPsmarter is highly interactive and contains a media-rich format of text, audio, graphics, animation, interaction, and video. Vignettes are woven throughout the intervention, which include stories (based on the experiences of past SIPsmartER participants) to model and describe situations related to setting goals, behavior changes, resolving barriers, and encountering slips. Finally, iSIPsmarter includes a stepped care engagement strategy (ie, human-supported text messages, followed by phone calls if needed) to provide encouragement, technical assistance, and strategies to promote core completion. The adaptation processes, HCD approach, and findings leading to the final content and structure of iSIPsmarter are further detailed in the subsequent sections.

The Model for Internet Interventions highlights 8 areas to consider when developing digital behavioral health interventions (Figure 1) [4]. These areas are defined and illustrated within the adaptation context for iSIPsmarter.

The Model for Internet Interventions (Figure 1) also illustrates other design-related components and elements (eg, support, user characteristics, environment, and digital health intervention use) [4]. Many of these components were considered in the iSIPsmarter development process. As an example, the model describes supports (eg, email, phone, and face-to-face) that can directly impact users’ adherence to digital health applications. During the development and design of iSIPsmarter, interviews were conducted with the advisory team members to assess their understanding and perceptions of program reminders and gather inputs on the timing and content of stepped care messages.

After each round of interviews, the researchers who conducted the interviews reviewed the audio transcripts and created interview summary documents that summarized each participant’s response to each question. Following each interview series, these summary documents with screenshots were shared with the CDT to incorporate feedback into the ongoing intervention content and programmatic development.

The analysis and data interpretation process involved several steps [61,62]. First, the 2 researchers who conducted the interviews reviewed the summary documents and independently identified key takeaway statements for each of the 13 interview rounds. Then, they met to build consensus and finalize the interview-level takeaway statements. Second, 2 researchers independently examined and deductively summarized the interview-level takeaway statements as higher-level findings and incorporated them into all 3 HCD phases, as aligned with the Model for Internet Interventions. During this second step, they also used CDT meeting minutes and artifacts from the development process (eg, prototype versions and drafts of core content) to inform higher-level findings. Again, the 2 researchers met to build consensus on the overarching phase-level findings. Finally, to illustrate the phase-level findings, key quotes from the advisory team members were extracted from the transcripts.

Interviews 1, 3, 11, and 12 focused on the contextual and technology-related aspects of intervention delivery and data collection. The following key findings emerged:

These findings are illustrated by several key quotes from the advisory team members. One of the members expressed that they face difficulty in responding to SMS text messages owing to a lack of signal but that they would be able to email:

In reference to monitoring weight on the iSIPsmarter dashboard, one of the members said the following:

The prototype testing phase allowed the CDT to iteratively address and respond to the user feedback. Interviews 2 and 8 focused on the dashboard and diaries; interviews 5, 6, and 7 were dedicated to the action planning concepts; and interviews 4, 5, and 6 focused on the personas. The final dashboard illustrated in Figure 2 contains a few examples of key intervention components, and a sample action plan is shown in Multimedia Appendix 2. For the persona-focused interview, 9 persona journey maps were shared with the advisory team participants, one of which is illustrated in Figure 3. On the basis of the participants’ insights into and reflections on the journey maps, the 9 initial personas were narrowed down to 6 and then further refined.

The main phase 2 findings are summarized as follows:

Highlighting different patterns in SSB consumption behaviors among the vignettes was identified as a helpful messaging approach to improve the personalization and tailoring of content.

The following key advisory team member quotes support the findings:

In reference to the ability to navigate the action planning process, one of the members said the following:

In reference to the personas being realistic and relatable, the following was said:

Interviews 9, 10, and 13 focused on the end user testing. Table 1 highlights the final iSIPsmarter overview that was evaluated in this phase, including the asset summary, the user objectives, and a description of interactions or videos. The cores and interactions were developed and programmed in a manner that allows the users to go back and review the content and repeat the interactions to master skill development. Related to action planning, the underlying programmatic structure in cores 3 to 6 and the maintenance core are identical. However, the personalized feedback loops change based on the user’s diary data, progress toward their prior goals, and their new goals. The final SSB message bank includes 14 barriers and 80 strategies, and the weight-related message bank includes 13 barriers and 115 strategies. Therefore, each time a user completes an action plan, the personalized feedback has the potential to look very different (Multimedia Appendix 2).

On the basis of phase 2 findings, the CDT fully developed 6 vignettes and mapped key content through each of the iSIPsmarter cores and several interactions. Table 2 illustrates a final vignette, along with several examples of how the vignette is integrated across the cores.

The following phase 3 findings emerged when the participants were allowed to access the programmed intervention in their own environment:

A few key quotes from the advisory team member have been illustrated to support these findings. In reference to the interaction aimed at teaching and reinforcing the amount of sugar in drinks, one of the members said the following:

In reference to the personalized feedback, the following was expressed:

In reference to the overall iSIPsmarter user experience, the following was said:

This paper presents a flexible, step-by-step approach to and an HCD process for developing digital behavioral health interventions using iSIPsmarter as an illustrative example. By applying the Model for Internet Interventions [4] as well as best practices in HCD [16] and instructional design processes [17-20], we have been able to clarify assumptions and address key areas of the iSIPsmarter intervention that were hypothesized to support participants’ engagement and promote behavior change. As evidenced by the rich feedback received from the advisory team members, the human-centered methodology was instrumental in our development process. Likewise, the value of our robust process is exemplified by our resulting user-informed, high-quality products (eg, iSIPsmarter core components, vignettes, dashboard, and personalized action plan). Our approach can be interpreted within the context of the findings of recent narrative scoping review by Kip et al [14], in which a variety of methods (eg, interviews, focus groups, questionnaires, card sorting, and usability testing) and products (eg, prototypes, personas, and behavior change strategies) used for participatory eHealth development processes have been highlighted. Similar to the conclusions of Kip et al [14], our approach illustrates the importance of researchers and developers using the most appropriate methods to match their objectives and user characteristic context.

In terms of the iSIPsmarter development process, some of the biggest challenges faced when transitioning from the original group class structure to the digital structure include (1) replicating engagement provided by participant-to-participant and participant-to-instructor communication and relationships; (2) balancing the demands and cognitive load of the behavior change content (especially planning for the potential low health literacy skills of our targeted users) and trying to mimic visual, experiential, and hands-on class-based activities; (3) building flexibility to account for varying levels of internet and SMS text message accessibility among users; (4) creating a seamless user experience by integrating internet, SMS text messages, and sensor information; (5) operationalizing all aspects of the intervention content for digital delivery; and (6) automating all elements while ensuring that all permutations were considered. Similar challenges have been highlighted in a few other digital behavioral health intervention adaptation papers [14]. However, these issues are often overlooked and underreported in the literature, which may partly explain the low uptake of digital behavioral health interventions as well as the suboptimal fit between the characteristics of the technology and user needs. Simply converting evidence-based content traditionally delivered face-to-face to a web-based or digital format, without adequate attention to these and other challenges, is an insufficient approach to engaging the intended users or promoting improved health outcomes. Researchers and developers should rely on the established models and frameworks, such as the Model for Internet Interventions [4-9], to help anticipate and guide key decisions when embarking on the development of digital behavioral health interventions.

Although several behavioral interventions targeting SSB consumption reduction among adults have been developed and evaluated [30], only a few have used scalable digital approaches. Moreover, only one other known adult-focused digital intervention targeting SSB consumption reduction has applied user-centered methodologies in the formative phases of intervention development [31]. Similar to our study, the study by Tonkin et al [31] focused on disadvantaged and nonurban adults, and its findings revealed the importance of understanding the available technology and patterns of its use as well as participants’ preference for stories, role modeling, and gamification, which foster engagement with the intervention. In addition, digital weight self-monitoring has become a cornerstone of many weight-related behavioral interventions, and greater adherence to self-monitoring is associated with better outcomes [63-65]. Unfortunately, consistency and disengagement in digital weight self-monitoring are known to be problematic [66]. Although several studies have investigated experiences of self-monitoring at the conclusion of interventions [67,68], there is a dearth of published studies that have applied HCD processes in formative intervention development stages to understand and build-in behavioral strategies to address potential personal and environmental barriers to digital self-monitoring [69].

On the basis of our experiences with iSIPsmarter, we offer 6 broad considerations for other teams developing or adapting digital behavioral health interventions (Textbox 1).

In terms of study implications, the efficacy of iSIPsmarter in reducing SSB consumption in rural Appalachian adults is currently being evaluated in a randomized controlled trial that includes a 2-group design (iSIPsmarter vs static Participant Education website) with 4 assessment points (Clinical Trial Registry: NCT05030753) [36]. When efficacy and other summative data are available, they will provide additional insights to inform the potential value of applying an HCD process to build iSIPsmarter as well as identify future areas of study.

Overall, 3 key limitations of this study should be considered. First, generalizability may be limited by the relatively small sample of advisory team members, the targeted rural Appalachia region, and SSB-specific content. Second, qualitative interviews can result in socially desirable responses. However, efforts were made by our trained interviewers to minimize this potential limitation by probing and clarifying the participant responses. This approach resulted in the participants providing thoughtful and constructive critiques of iSIPsmarter. Third, although we applied the Model for Internet Interventions, we limited the focus and scope of our manuscript to highlight the 8 main areas that comprise digital health applications [4]. Future efforts could focus more broadly on describing the other design-related components and elements that guided the iSIPsmarter adaptation process. Despite these potential limitations, we hope that the illustrated processes, scientific frameworks, and context-specific instructional design methodology will have wide applications in the development of digital behavioral health interventions across multiple geographic and behavioral contexts.

Our process emphasizes the value of researchers and developers applying the existing models and frameworks as well as best practices in HCD and instructional design processes in digital intervention development processes. Together, the complementary skills of the CDT and advisory team members were invaluable in the iSIPsmarter adaptation process. The importance of the contextual and iterative prototype testing phases was largely reinforced by the overwhelming positive feedback received in the user testing phase. By illustrating iSIPsmarter content, we have highlighted the user-informed, high-quality products that resulted from our robust HCD process.