Taxonomies and models are useful tools for defining eHealth intervention content and delivery, enabling the comparison and analysis of research across studies and disciplines [1]. The Behavior Change Technique Taxonomy version 1 (BCTTv1) was developed to decrease ambiguity in defining specific characteristics inherent in health interventions [2], but it was developed outside the context of digital technology. In contrast, the Persuasive System Design Model (PSDM) was developed to define and evaluate the persuasive content in software solutions but did not have a specific focus on health [3]. Both the BCTTv1 and PSDM have been used to define eHealth interventions in the literature [4-9], with some researchers combining the taxonomies to include both the health and persuasive technology components [10-12]. However, there have been concerns regarding the ability of the taxonomies to capture the variability and dynamic nature of health interventions [13-15]. In addition, it is unclear how well the BCTTv1 and PSDM accurately define eHealth and whether they should be used alone or in combination. Therefore, a scoping review methodology was used to develop a relevant data set to explore how the BCTTv1 and PSDM capture the content and delivery of eHealth interventions. The scope was concerned with parent-focused eHealth as part of a program of studies investigating the use of technology to support parents with therapy home programs for children with special health care needs.

The BCTTv1 taxonomy defines 93 distinct behavior change techniques (BCTs) categorized into 16 behavioral determinants. The behavioral determinants include categories common to many behavior change theories, such as goals and planning, social support, and shaping knowledge [2,16]. The BCT represents the smallest component of a behavior change theory that could bring about a change in behavior under the right circumstances. As such, BCTs are described as the “active ingredients” of an intervention [17].

Persuasive technology [18] refers to the use of technology, such as computers, mobile devices, and the internet, to influence and change people’s attitudes or behaviors. Although technology is often described as a service delivery vehicle in eHealth, digital applications that are responsive to a user’s input can mimic social drivers of behavior and are more persuasive than static communication media such as print [19,20]. Persuasive technology ranges from simple nudges and reminders to more complex interactive systems that use personalized feedback and rewards to motivate and engage users. On the basis of the earlier work by Fogg [18] describing persuasive technology, Oinas-Kukkonen and Harjumaa [3] proposed the PSDM grouping of 28 persuasive system design (PSD) elements into 4 design principles. A fifth design principle was later added to capture the coaching elements found in many eHealth interventions [6]. Despite being called a model, the PSDM does not define the relationships between the design elements and principles [21] and, thus, can be considered a taxonomy akin to the BCTTv1 [12].

The BCTTv1 and PSDM have a different focus; however, there is considerable overlap in the categories and principles of these 2 taxonomies. Therefore, some systematic reviewers have merged labels within the 2 taxonomies ad hoc [9-12]. However, labels within the 2 taxonomies may describe different characteristics of the same eHealth intervention. For example, Self-Monitoring from the BCTTv1 is a BCT that includes activities such as using a diary or checklist to self-monitor behavior. This requires some discipline and commitment to keep track of what is being monitored. In persuasive technology, Self-Monitoring defines an automated feature of technology that requires little effort from the user but potentially requires the user to give up a level of privacy for the convenience of having the technology do the monitoring for them. Self-Monitoring as captured by these 2 taxonomies clearly represents different behaviors and experiences for the user and, therefore, potentially different outcomes.

Combining multiple taxonomies to define an intervention presents a challenge in terms of efficiency and usability. Although merging taxonomies based on similar labels or constructs may improve efficiency, it could also result in the omission of important ingredients or features. Therefore, this review explores the similarities and differences between the BCTTv1 and PSDM to gain a deeper understanding of their usability individually and in combination to define the content and delivery of digital interventions in parent-focused eHealth.

Scoping reviews are an alternative to systematic reviews when there is a need to clarify characteristics related to a concept in the literature [22,23]. As scoping reviews are not intended to inform practice or answer effectiveness questions, they do not include a critical appraisal of the literature and cannot be used to determine intervention effectiveness [24,25]. Scoping reviews require a systematic review approach and provide a methodologically rigorous structure; however, they are intended to be iterative. The search terms and questions are refined as the researchers become immersed in the available literature [24,25]. The PRISMA-ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses extension for Scoping Reviews) guidelines [26] accommodate changes in scoping review protocols over time and guided this scoping review report [27], along with the original framework provided by Arksey and O’Malley [24].

Parents of children with special health care needs were the population of interest in this scoping review. Children with special health care needs are defined as children with “chronic physical, developmental, behavioural, or emotional conditions who also require health and related services of a type or amount beyond that required of children generally” [28].

Many parent-focused eHealth interventions use computer-mediated interactions as a primary feature, involving mostly synchronous videoconferencing between the parent and the therapist [29]. However, persuasive technology was a key interest, and therefore, the eHealth interventions needed to predominantly include human-computer interaction to be included in this scoping review. In other words, the parents needed to interact directly with a digital application that was responsive to their input. In addition, the continuous evolution of technology highlights the importance of focusing on underlying principles that define the core functionality of eHealth interventions to facilitate research and future development [30-32]. Therefore, the context of this scoping review included all parent-focused eHealth interventions regardless of the underlying theories, application, platform, or device being used.

The search was pragmatically limited to articles published after 2008, when both the BCTTv1 and PSDM were first published [2,3], and the subsequent 10-year period in accordance with recommendations for limiting the number of articles in a scoping review [33]. This provided a comprehensive yet manageable database to examine the relationship, similarities, and differences between the BCTTv1 and PSDM, providing insights into how the active principles captured by these taxonomies could be applied in future studies.

The objective of this scoping review was to develop novel insights into how the BCTTv1 and PSDM, 2 increasingly popular taxonomies, identify active ingredients and features of eHealth interventions to inform future research and design.

This review answered two questions: (1) What are the active ingredients and persuasive technology features commonly found in parent-focused eHealth interventions for children with special health care needs? and (2) How do the active ingredients and features described in parent-focused eHealth literature overlap and interact with respect to the BCTTv1 and PSDM taxonomies?

A protocol for this scoping review was published elsewhere [34]. Consistent with the iterative and responsive nature of scoping reviews [33], three variations from the protocol are noted. First, the term “ingredients and features” replaced the term “principles” to decrease ambiguity in terminology. Second, the term persuasive eCoaching did not replace PSDM as originally suggested. Further immersion in the data and related literature found that the elements suggested by Lentferink et al [6] did not reflect an eCoaching process.

The third variation related to the illustration within the protocol depicting the relationship between the PSDM and BCTTv1. This figure has been updated (Figure 1) in light of scoping review findings. The labels within the figure were changed to more accurately represent the components of the behavior change models related to the 2 taxonomies (previously labeled Behavior Targets) and behavior outcomes (previously labeled Behavior). The additional PSD category and elements suggested to represent coaching elements of eHealth [6] were also included as PSD principles (increased from 4 to 5) and elements (increased from 28 to 32). The updated figure illustrates how the BCTTv1 and PSDM relate and link to behavioral outcomes through their related models: the Capability, Opportunity, and Motivation–Behavior (COM-B) model and the Fogg Behavior Model (FBM), respectively. Each model is made up of behavior components linked to BCTs and PSD elements that represent the behavioral determinants and design principle categories in the taxonomies. The 9 intervention functions linking BCTs to behavioral components in the BCTTv1 have also been included.

The search strategy was informed by guidelines for scoping reviews [24,26,35] and included a systematic search of MEDLINE (Ovid), Embase (Ovid), PsycINFO (Ovid), Scopus, Web of Science, CINAHL (EBSCOhost), and ERIC (Ovid) to identify potentially relevant articles. Keywords related to parent-focused eHealth were used, where the parent was the intended target of the intervention and their child’s health or behavior was the intended outcome. To be included, interventions required human-computer interaction. For the purpose of this review, human-computer interaction was conceptualized as user-generated content and an element of automation to support the parents’ behavior change. Interventions that were predominantly a replacement for face-to-face interventions or coaching (ie, computer-mediated synchronous communication), were not interactive (eg, static text, such as a digital version of an information pamphlet), or were solely intended as an education platform were excluded for this reason. In addition, as the aim of the larger study was to support parents, interventions that predominantly required the child rather than the parent to interact with the technology were also excluded. A detailed list of the inclusion and exclusion criteria can be found in Table 1.

Keywords were developed and refined through an exploratory search of the literature using Scopus. Keywords, Medical Subject Heading terms, and Excerpta Medica Tree terms were tested in Scopus and then extrapolated to match the criteria of the remaining databases in consultation with a medical librarian. Keywords (detailed in Multimedia Appendix 1) were used to conduct an electronic search of MEDLINE (Ovid), Embase (Ovid), PsycINFO (Ovid), Scopus, Web of Science, CINAHL (EBSCOhost), and ERIC (Ovid). The search yielded 9405 articles published between 2009 and 2018.

After duplicate removal, 4843 abstracts were screened stepwise at the title (n=3228, 66.65% remaining), abstract (n=136, 2.81% remaining), and full text (n=102, 2.11% remaining) levels. After duplicates were removed, titles and abstracts were imported into EndNote X7 (Clarivate Analytics). Publication titles were used to remove studies that were clearly not related to the scoping review purpose, were a summary of academic proceedings, or did not contain details of specific eHealth interventions.

The reasons for exclusion were recorded (Figure 1). As the unit of data analysis was the eHealth intervention, articles that did not have basic information describing the intervention were excluded unless there were other publications about the same eHealth intervention with sufficient details. This included pamphlets and other gray literature found through reference mining of the full-text articles. An additional 8 articles were included, bringing the final number of articles to 42.

Abstracts were screened independently by 2 study authors. All the retained articles were retrieved as full-text articles and reviewed independently by 2 authors. Throughout the process, disagreements were resolved through discussion between the 2 reviewers and, at times, a third author.

Publications describing the same intervention in a different country or population were collapsed into a single unit of analysis, as were publications reporting on different aspects of the same study. This resulted in 23 digital health interventions to analyze. Figure 2 illustrates the selection process using the PRISMA-ScR flow diagram [26].

Data from multiple publications on the same study or intervention were combined, as well as information about the intervention available on the web, to extract the most comprehensive details of the eHealth intervention. Data were charted in table form for content related to year and country of publication, health conditions covered, theories mentioned, and age range of the children targeted by the interventions. The type of technology hardware used (eg, tablets and smartphones) was also recorded. Any references to the BCTTv1, the PSDM, or their related models and theories were noted.

A qualitative process of deductive content analysis [37] of the extracted data occurred in iterative cycles of coding, refining codes and recoding to define the content, patterns, and relationships of each intervention’s active ingredients and features. Articles were coded in NVivo (version 12; QSR International) with matrix-coding queries used to look for relationships and patterns. The BCTTv1 and PSDM taxonomies were used as the codebook for this analysis. Therefore, tests of codebook reliability and validity were not considered necessary as codes were not added or removed. Definitions were clarified as needed throughout the coding process. Coders had varied previous experience and training. EJHS had previous experience applying the BCTTv1 in several projects [38-40]. MS had no previous coding experience but undertook web-based training in BCT coding [41] before starting this process. Where publications provided their own list of BCTs, these supplemented the retrospective codes identified but did not replace them.

Initially, 4 articles were read and coded independently by MS and JHS. The codes were discussed to reach an agreement on the definition and parameters of each code. As the definitions in the PSDM are not robust, the persuasive technology text [18] and course notes from the behavior design guide workshop [42] were used to supplement the understanding of PSDM coding use. Credibility was supported via both coders coding all articles independently and then discussing codes at weekly meetings where further refinements of code definitions were made and decision rules were documented. External advice from academic colleagues with expertise in the BCTTv1 and PSDM was sought once for each taxonomy when an agreement could not be reached between the 2 primary coders. A log of decisions that contributed to the final coding scheme was kept as an audit trail. Once all articles were coded and no further updates were made to the coding manual, all articles were uploaded into NVivo, reread, and recoded based on the final coding parameters.

Coding rules were expanded beyond the BCTTv1 guidelines when there was a lack of clarity regarding which of the 2 related codes to use. For example, if it was difficult to determine whether self-monitoring was directed at the behavior or the outcome, it was coded as both Self-monitoring of behavior and Self-monitoring of outcome(s). Knowledge and education were a large part of most interventions but were often described with insufficient detail for BCT coding requirements. Text related to education, instruction, or the provision of knowledge was not coded unless it clearly adhered to a BCTTv1 definition given that research suggests that education and knowledge on their own do not change behavior [29-31]. This contrasts with the approach taken by other authors, who coded any reference to education as Information About Health Consequences and Instruction on How to Perform the Behavior [28]. All text was coded in this way before checking the supplementary material where BCTs were detailed by the publication authors. Where supplementary material was provided listing the BCTs, this was reviewed after coding, and any additional codes were added.

NVivo queries were used to generate 3 matrix tables representing how the BCTs and PSD elements were represented in the data. The first table showed the number of times a BCT was coded with a PSD element (BCT-PSD pairs). The other 2 tables showed the number of times a BCT was coded with a different BCT (BCT-BCT pairs) or a PSD element was coded with a different PSD element (PSD-PSD pairs) in the same intervention. These matrix tables were exported to Microsoft Excel (Microsoft Corp) and enabled us to explore the relationship between BCTs and PSD elements as they were represented in the data.

By definition, persuasive technology requires human-computer interaction where the technology can form a “relationship” with the user. In this scoping review, technology was considered an active participant in the intervention, taking part in a “relationship” with the parent. In Figure 3, the 2 arrows in the rectangle represent this relationship in a parent-focused eHealth intervention, which was the focus of this review.

This scoping review mapped 23 parent-focused eHealth interventions included in 42 articles published between 2009 and 2018. The most common study type in the included articles was randomized controlled trials (15/42, 36%). The next most common study types were cohort (5/42, 12%) and qualitative (4/42, 10%) studies. The publications also included mixed methods studies (2/42, 5%), co-design (2/42, 5%), case series studies (2/42, 5%), case-control studies (2/42, 5%), action research (1/42, 2%), and a randomized controlled trial protocol (1/42, 2%). The remainder of the published materials (8/42, 19%) included reports, pamphlets, and supplementary material such as tables and images of the digital intervention that were included if they detailed the intervention content for coding purposes (Table 2).

The authors of the articles were from the United States (13/42, 31%); Australia (10/42, 24%); Europe (4/42, 10%); the United Kingdom (3/42, 7%); China, New Zealand, and Sweden (2/42, 5% per country); and Belgium, Canada, Israel, Italy, Japan, and Poland (1/42, 2% per country). The parent-focused eHealth interventions were delivered to parents with children ranging in age from 1 to 18 (mean 7, SD 5) years (Table 3).

The range of chronic childhood health and behavioral concerns is listed in Table 4 and covers medical, behavioral, and developmental issues. Children with cerebral palsy and other childhood physical disabilities were not represented.

This scoping review mapped the theories, models, and taxonomies described in the included publications. Just over half (13/23, 57%) of the interventions mentioned a behavior change theory. Social-cognitive theory was the most frequently mentioned (6/23, 26%), followed by self-determination theory (3/23, 13%). Only 9% (2/23) of the interventions referenced PSD (CLKonline and Strongest Families FASD), and only 4% (1/23) of the interventions (Health Heroes) referred to the Behavior Change Wheel [78] as the overarching theory for the COM-B model and the BCTTv1. This was also the only publication that included supplementary information listing the BCTs in the intervention. No articles referenced the FBM.

Some interventions were established treatment approaches adapted for a web-based medium, for example, BraveOnline [48] and TPOL [70]. Other interventions were novel approaches designed specifically for the eHealth intervention, such as Sensory Treat [68] and Health Heroes [60]. Most interventions (19/23, 83%) used a web-based platform, with 47% (9/19) of these also including a mobile phone or tablet. The remaining 17% (4/23) of the interventions were stand-alone mobile phone or tablet apps. The most popular features included in the interventions were interactive questionnaires and charts (17/23, 74%), followed by instructional videos (14/23, 61%).

Consistent with a focus on parents, all the included eHealth interventions (23/23, 100%) had the parent as the primary user, with the child (either the child’s behavior, health, or both) as the outcome target. The parent interacted directly with the digital platform, which could be linked to an intervention team, a preprogrammed intervention package (eg, a digital app), or a combination (Figure 3). Thus, the active ingredients and intervention features (ie, BCTs and PSD elements) coded in this review were those directed toward the parent, whose behavior change affected their child’s health and behavioral outcomes. However, multiple relationships could exist in any one intervention such as targeting the child or other family members as intervention users. In addition, intervention descriptions in the publication text were often focused on the parent-child interaction rather than describing the human-computer interaction between the digital intervention and the parent.

The following extract from the scoping review data set illustrates how this attention to who was being targeted influenced the coding process. In an eHealth intervention supporting parents in implementing activity schedules for children with autism spectrum disorder, the following text was coded:

If the child were the target, coding would have captured techniques and elements such as Prompts and Cues, Restructuring the Environment, and Information and Problem Solving. However, as the directional relationship was from the digital intervention to the parent, this extract was coded as the BCT Instructions on How to Perform a Behaviour and the PSD element Educational Coaching.

Next, we examined the overlap or interaction between BCTs and PSD elements using the BCT-PSD matrix table. An example of this table is illustrated in Table 7, which is a snapshot of the full BCT-PSD matrix table. For example, the BCTs goal setting, action planning, and problem-solving were frequently paired with PSD elements reduction, suggestions, and educational coaching. However, we observed that the more times each ingredient was coded, the greater the “scatter” of the BCT-PSD pairings. For example, the BCT Instructions on How to Perform a Behaviour was coded with a PSD element 142 times, representing 20 different PSD elements. Similarly, the PSD element Tailoring was paired with a BCT 200 times during the coding process, and these pairings included 33 different BCTs.

Table 7

Behavior change technique (BCT)–persuasive system design (PSD) element matrix table snapshot^a.

This scoping review only included English-language publications from before 2018. The COVID-19 pandemic has resulted in rapid shifts in web-based therapy provision that may not be described adequately in this review. However, capturing a decade of work before the COVID-19 pandemic also provides a useful baseline for future comparisons of change. In addition, the objective of this scoping review was to explore the utility of 2 taxonomies in defining the active ingredients and features of an intervention. The BCTTv1 and PSDM have not been updated since 2017. The data from the included articles provided sufficient information to ascertain the value of using both taxonomies in their entirety rather than merging or condensing them, as has occurred in other publications. Retrospective coding of publications based on taxonomy definitions was noted to be prone to errors, as has been reported elsewhere [14,89,90]. Similar errors may have occurred during our coding process. Explanations as to how we interpreted definitions have been included in the Methods section to increase transparency and decrease ambiguity where possible.

A total of 3 other scoping reviews related to the scope of this review have been published previously [91-93]. These publications did not address the objectives of this scoping review to map the active ingredients and intervention features in parent-focused eHealth and explore the use of both taxonomies. Greffin and Barros [91] investigated health-related parenting but did not address active ingredients and intervention features of behavior change and persuasive technology. Bradshaw et al [92] reviewed behavior change in parent-focused interventions but did not limit their review to eHealth and did not review persuasive technology. Asbjørnsen et al [93] looked at both behavior change and persuasive technology; however, the focus was on weight loss maintenance in adults. In addition, the authors used the 16 BCTTv1 categories rather than coding all 93 BCTs.

The 2 taxonomies representing behavior change and persuasive technology were found to code different constructs, discouraging merging or reduction. Merging taxonomies may fail to capture the persuasive ingredients inherent in technology, missing the unique role of technology in creating bonding relationships with the end user. Using both taxonomies to code interventions in this scoping review was useful in providing a comprehensive description of the active ingredients and features of an eHealth intervention. However, it is important to note that potential coding errors may render the accuracy of the description uncertain. Coding errors may result from limited space for comprehensive reporting of the active ingredients and intervention features or from ambiguity in the publications’ description of ingredients, features, behavior targets, and outcomes.

Commonly found BCTs and PSD elements did not necessarily represent the best BCTs or PSD elements to use, with some potentially impactful codes not being used at all in the parent-focused eHealth data set. Despite the time-consuming and cumbersome task of capturing all BCTs and PSD elements, identifying each of these constructs enables a much richer description of the ingredients within the interventions as well as providing opportunities to consider innovative approaches in future designs. This scoping review highlighted the need for supplementary material detailing the active ingredients and intervention features in publications and the benefit of using 2 taxonomies to address both behavior change and persuasive technology that together influence the experience and behaviors of the people using the technology.