Principal Findings
In this review, we found that the sources of human support could be categorized into the following: (1) peers and peer specialists, (2) health experts and practitioners, (3) trained coaches, and (4) members of the research team. Most individuals providing human support are paraprofessionals with no formal credentials nor professional training related to the intervention [79]. The advantage of using health experts and practitioners is that they are readily available since they are already embedded in the clinical setting. This can allow for a more seamless and streamlined implementation process. However, several system-level barriers must be considered. For example, providers may have concerns regarding the time and effort it takes to integrate interventions into their existing workflow [80], as their time and effort invested may not be reimbursed or compensated [81-84]. The use of paraprofessionals (eg, trained coaches) specific to the interventions could bridge this gap, as existing evidence suggests that interventions delivered by paraprofessionals are scalable, feasible, acceptable, and effective [85,86]. In addition, pairing paraprofessionals with DHIs can enhance treatment fidelity by minimizing human error or therapist drift, which is common in manualized treatment protocols [87]. Adequate training for paraprofessionals is key to ensuring fidelity and scalability. Rosenberg et al [79] offered an in-depth description of their coach training program based on learning theories and competency-based supervision.
One of the common sources of human support found in this review is peers and peer specialists. Peer-to-peer support could promote DHI engagement because their relationship is characterized by reciprocal accountability, wherein peers motivate each other to work toward their individual goals [19]. On the other hand, peer coaching models (eg, peer specialists) are frequently used to enhance intervention accessibility, engagement, and scalability by providing human support from coaches who share similar identities and lived experiences [88]. One common use of peer specialists can be found in recovery support for substance use [89], in which peers apply personal experience and knowledge to support individuals who are earlier in their recovery process [90,91]. Aligned with our review, Blonigen et al [44] used peer specialists to facilitate veteran participants’ engagement with a DHI for unhealthy alcohol use. Shared lived experiences give peer specialists credibility and offer participants a sense of belonging and hope for the future [19].
Although some studies have demonstrated the positive effects of these peer models [92-95], empirical support remains mixed [96-99]. These inconsistencies may be due, in part, to the lack of structured interventions, formalized training, and quality assurance mechanisms for peer specialists and other paraprofessional human supports [97,100]. A critical gap in the current literature is the lack of detailed descriptions of training protocols, core competencies, and supervision processes, which makes it difficult to assess fidelity, replicate successful interventions, or scale promising approaches. Developing and reporting standardized training protocols for human support providers in DHIs are essential to ensure consistency, maintain intervention quality, and enhance participant outcomes. This is especially important for peer specialists and other paraprofessional human support, who are not typically professionally trained for these specific roles and may require additional guidance. Such protocols should include clearly defined roles and responsibilities, communication strategies grounded in behavior change theory, ethical and boundary training, cultural competence, and ongoing supervision or fidelity monitoring.
Our findings suggest that the frequency and duration of human support interactions are closely tied to the mode of communication, with important implications for engagement and adherence in DHIs. This relationship aligns with the communication bandwidth principle proposed in SAM [31], which suggests that communication media differ in their ability to convey social presence, thus influencing the quality of the interaction and the strength of accountability. For example, phone and video calls—used in 64% of the interventions—typically involved less frequent but longer sessions, allowing for richer interpersonal exchanges that include verbal and, in the case of video, visual cues. These higher-bandwidth modes facilitate stronger perceptions of bond, legitimacy, and trust—core constructs of SAM—by supporting nuanced communication, emotional expression, and real-time responsiveness.
In contrast, text-based communication (also used in 64% of interventions) enabled more frequent but shorter and often asynchronous interactions. Although this lower-bandwidth mode may offer greater flexibility and accessibility, particularly for asynchronous support, it strips away nonverbal cues and could potentially diminish social presence. The findings imply that, although both modes have utility, higher-bandwidth communication may be more effective for establishing initial trust and accountability. In contrast, text-based support may be a more suitable supplementary strategy for maintaining ongoing engagement. Future DHI design should consider tailoring support frequency and duration based on the mode of communication and the complexity of the behavior change goals, striking a balance between efficiency and the need for relational depth.
The majority of studies (97%) used human support to promote accountability to improve engagement with DHIs, including monitoring users’ usage, sending reminders, checking in, setting goals, and discussing performance [101,102]. Each of these strategies may be positioned relative to each other on a spectrum based on the degree to which an approach is “active.” For example, using human support to monitor user performance alone is a passive approach as it does not necessarily promote engagement with DHIs. Performance monitoring is often coupled with more direct check-ins and reminders that allow participants to infer the social presence of human support. To improve engagement, human support can also set forth expectations by setting goals with users; for example, human support can help users set and review personal goals, as well as create plans for continued DHI use. When setting goals, the targets of the expectations can vary. In this review, most human support focused on process accountability, with a primary goal of increasing engagement with the DHI. On the other hand, Cheng et al [73] used automated support, quizzes, and email notifications to maintain accountability. Although the notifications were effective reminders to complete the weekly sessions and diary entries, quizzes prompted participants to pay more attention to the information as they promoted outcome accountability.
Although accountability was used across most studies, we did not see any use of instruments to measure accountability. Therefore, it is unknown whether the steps taken to ensure accountability had any effects, positive or negative, on accountability. Qualitative studies by Sayegh et al [65] and Wilhelmsen et al [60] interviewed participants about their experiences with the intervention. In each of these studies, participants described their experience with their human support, and their encounters resulted in different levels of accountability that promoted their use of DHIs. Parent participants in the study by Whiteside et al [61] also described observing that the interventions gave a sense of responsibility for their children’s symptom management. Meyerhoff et al [103] saw a need for supportive accountability measures for coached digital interventions and developed the Supportive Accountability Inventory (SAI). The SAI is a 6-item instrument assessing users’ perceptions of (1) whether they believe the coach pays attention to their use of DHIs and progress and (2) if their coach maintains expectations of user engagement with the DHIs and their effort toward treatment-related goals. Future studies should utilize validated measures such as the SAI [103] to assess the mechanisms by which accountability is achieved (eg, monitoring and expectation).
Additionally, many studies did not explicitly describe how bonds and legitimacy were used to cultivate accountability. Articles frequently used keywords such as connection, warmth, trust, and friendly, which may have been used to signal or allude to bonds (eg, [38,50,67]. However, without clear definitions or theoretical grounding, it becomes difficult to determine whether these references reflect bonds—which emphasize mutual liking, emotional connection, and respect (eg, therapeutic alliance) [104]—or relational legitimacy, which centers on perceptions of trustworthiness, benevolence, and expertise [77] and does not necessarily involve interpersonal liking [31]. One possible reason for this gap is that qualities associated with bonds or relational legitimacy are often assumed to be inherent in any supportive or therapeutic relationship. As such, these relational dynamics may be implicitly embedded in human-supported DHIs but are rarely made explicit in intervention design or measured in outcome evaluations.
To address this gap, future research should more deliberately integrate and measure these constructs using validated instruments. For example, the Working Alliance Inventory [105] could capture the strength of the bond, while the Credibility/Expectancy Questionnaire [106] may assess relational legitimacy and perceived coach expertise. Intervention protocols could also include structured relational strategies such as rapport-building scripts, credibility-enhancing coach introductions, and personalized messaging that reinforces benevolence and competence. Last, qualitative evaluations are essential for understanding how participants perceive relational legitimacy, which, in turn, influences their behaviors. Although many interventions aim to build legitimacy, few assess how it is experienced. Interviews and open-ended feedback can capture these relational nuances, offering critical insight into how legitimacy fosters engagement and adherence in DHIs.
In this scoping review, we found certain studies used human support as the main deliverers of the intervention content, corresponding to Types 1 and 2 in the taxonomy in Pineda et al [39]. As aforementioned, SAM was proposed to suggest the specific use of coaches to support user engagement and adherence to DHIs [31]. This aligns with Type 3 within the taxonomy. The tension among Types 1, 2, and 3 within this scoping review indicates variations in the application of SAM to DHIs. These studies did not appear to utilize SAM in the way that it was originally intended. As such, it is necessary to clearly define intervention goals and subsequent roles of coaches when considering the use of human support in DHIs.
We found that only 6 (17%) of the 36 studies focused on a youth population, yet youth are increasingly intertwined with technology, and many seek and share health advice through social media [107,108]. One study [61] targeted children with anxiety, but the intervention itself was designed for both parents and their children, some as young as 8 years old. In our review, we found that the use of peers (eg, [46-48,50]) is a common strategy that aligns with the social affordances described by Wong et al [109]. Since youth, adolescents, and young adults are particularly interested in and influenced by their peers, future studies may consider leveraging peer influence in accordance with SAM to enhance DHI engagement and adherence among younger populations. For example, adolescents and young adults may be motivated to adopt DHIs that incorporate social components, as these may foster social support and a sense of belonging [110,111].