Improving healthy lifestyle behavior is an effective strategy to decrease mortality and increase health-related quality of life [1,2]. Current digital health technologies provide meaningful contributions to the design of healthy lifestyle interventions and the dissemination of such interventions [3]. A combination of self-tracking, goal-setting, and feedback in automated interventions has been indicated by many to be an effective approach for increasing healthy lifestyle behavior [3-5]. Self-tracking is “the practice of systematically recording information about one’s diet, health, or activities, typically by means of a mobile phone, so as to discover behavioral patterns that may then be adjusted to help improve one’s physical or mental well-being” [6]. Components that might be important for self-tracking are the self-tracking device, validity, the effort required of the participant to perform self-tracking, and the presentation of summary data to the user [7].

Goal-setting and feedback are components that can be provided via so-called persuasive eCoaching. This new term is a contraction of the terms “persuasive technology” and “eCoaching.” We refer to persuasive eCoaching as the use of technology during coaching to motivate and stimulate (groups of) people to change attitudes, behaviors, and rituals [8]. Oinas-Kukkonen and Harjumaa’s persuasive system design (PSD) model [9] describes such persuasive technologies that are expected to positively influence health behavior change. This PSD model builds upon earlier research by Fogg [10] and divides the persuasive components into 4 main categories: primary task support, dialogue support, system credibility support, and social support. These categories contain additional components such as personalization and reminders. To make the PSD model more complete for persuasive eCoaching, some coaching components that can be provided via technology can be added, namely educational coaching, goal-setting, and feedback.

The integration of self-tracking and persuasive eCoaching in fully automated healthy lifestyle interventions creates new opportunities for healthy lifestyle management. First, self-tracking devices enable the objective tracking of lifestyle behavior such as physical activity, heart rate, or sleep. This objective measurement of one’s lifestyle pattern can be more reliable than people’s own estimations based on their memory and biological sensing of their lifestyle patterns [11-13]. More reliable measurements could become an essential component in lifestyle behavior change, enabling a greater awareness of people’s current lifestyles [14]. Second, data from wearable devices can generate automated, personally relevant feedback 24/7. Previous research suggests that this just-in-time tailored feedback contributes to the sustainable use of the intervention [8,15,16]. Third, more and more people own devices that are suitable for eHealth interventions [3]. Even among ethnic minorities and the elderly, the use of mobile phones and computers is rising [17-19]. This suggests a certain scalability for such interventions and maybe even cost-effectiveness due to the fact that no human effort is required to carry them out.

Besides these opportunities, applying the combination of self-tracking and persuasive eCoaching in automatic eHealth interventions also gives rise to a few challenges. These challenges concern privacy issues, trust, and ethics due to personally sensitive data being obtained and stored [20-22]. Concerning ethics, suggestions based on self-tracking data that are invalidated or unsupervised might end up being incorrect or even harmful [21]. In addition, individuals need to be able to understand and interpret the self-tracking data [23].

Despite the challenges, the combination of self-tracking and persuasive eCoaching to promote a healthier lifestyle is promising [3-5], and consequently, interventions employing this combination are becoming more common [5]. To our knowledge, no literature review has been conducted to identify the key components of such interventions. Knowledge about these key components can serve as input for future development of healthy lifestyle interventions that combine self-tracking and persuasive eCoaching, which in turn might increase the effect on health outcomes, usability, and adherence. Usability and adherence are important effect measures of eHealth interventions as they are prerequisites for the intervention to positively influence health or health behavior. In addition, it is worthwhile to identify the specific way a key component should be designed to create positive effects on health outcomes, usability, and adherence. “Effect on health outcomes” here means the effects of the lifestyle intervention on both changes in healthy lifestyle behavior (eg, an increase in physical activity) as well as changes in health status (eg, improved blood levels or weight loss). “Usability” here means the user’s satisfaction with the technology and its ease of use [24]. “Adherence” here means the extent to which the technology is used as intended [15].

Key components of interest are self-tracking components (eg, type of device and presentation of summary data to the user), persuasive eCoaching components (eg, elements of the PSD model such as personalization and suggestion), and other intervention components (eg, the underlying behavior change theory and cocreation with end users). This review addresses the following research questions: (1) What are key components for the effectiveness on health outcomes, usability, and/or adherence of automated healthy lifestyle interventions combining self-tracking and persuasive eCoaching? and (2) In which way should key components be designed to contribute to effectiveness on health outcomes, usability, and/or adherence?

This review is part of an overarching project for the development of a workplace stress management intervention that combines self-tracking and persuasive eCoaching. To ensure systematic and holistic development and implementation of the eHealth intervention, the Center for eHealth Research (CeHRes) roadmap is adhered to throughout this project [8]. This evidence-based roadmap aims to improve the uptake and impact of eHealth technologies and is based on a participatory development approach, persuasive design techniques, and business modeling. The first step consists of contextual inquiry. This step aims to identify key components from the literature and from users and other stakeholders who will affect—or will be affected by—the intervention.

As technology continues to evolve rapidly, this particular scoping review methodology was chosen for this review study because it allowed us to obtain a quick overview of the current literature on the topic. The fact that this field is rapidly evolving is illustrated by the development of Fitbit self-tracking devices. Ever since the first Fitbit tracker was released at the end of 2009, 13 more Fitbit trackers have been released [25].

Another reason to conduct a scoping review on this topic is that a scoping review is not limited to randomized controlled trials (RCTs) [26,27]. To properly identify the scope of this topic, studies evaluating the effect on health outcomes, usability, and adherence are required. Studies regarding the latter two will primarily be qualitative studies [24].

Arksey and O’Malley’s scoping review methodology [26] was applied. This methodology comprises the following steps: (1) identifying the research question; (2) identifying relevant studies; (3) study selection; (4) charting the data; (5) collating, summarizing, and reporting the results; and (6) consultation. A number of additional recommendations by Levac et al [27] were followed, namely: providing a clear purpose for the scoping review, review of full-text articles by 2 independent reviewers to decide on their inclusion, collectively developing the data-charting form with the research team, continually extracting data and updating the data-charting form, inclusion of the consultation step (an optional step according to Arksey and O’Malley [26]), and providing a clear consultation purpose.

A systematic literature search was performed in PubMed, EMBASE, PsycINFO, and Scopus covering the period from January 1, 2013 to January 31, 2016. PubMed and EMBASE were chosen for their wide coverage of scientific journals, whereas PsycINFO was chosen for its specific relevance to this review’s subject. Scopus was searched because of its multidisciplinary scope, which allows for identification of articles outside the medical field, such as in engineering. We decided to include no publication from before 2013 as technologies described in publications before 2013 seem less comparable with technologies described in newer publications. To illustrate, publications containing the search terms “Fitbit” and “smartwatch” increased from a negligible number before 2013 to hundreds from 2013 onward, reflecting the rise of personal monitoring devices [28]. These personal monitoring devices represent newer self-tracking technologies that simplify the collection and combining of personal data and enable more personalized healthy lifestyle interventions [29]. Including older publications, in which technological advances are not displayed, might lead to less relevant findings [30].

This study’s search strategy was created in collaboration with a University of Twente librarian, based on 3 main components: (1) self-tracking, (2) persuasive eCoaching, and (3) healthy lifestyle interventions. Related search keywords were identified using MeSH and EMTREE terms, PubReMiner, synonyms, keywords from relevant articles, and self-determined search terms (see Multimedia Appendix 1).

Our aim was to include articles that described fully automated interventions. However, we found that many articles involved a fully automated intervention in addition to human coaching, which we call blended coaching. As the scoping review methodology allows for post hoc decisions [27], we then decided to also include blended coaching interventions because we expected to find relevant results in these studies. Other inclusion criteria were that the articles had to be written in English or Dutch and had to be journal articles. Excluded publications included reviews, study protocols, study populations outside the age range of 18-66 years, publications lacking empirical data, and paper-based or personally reported tracking. This age range is in line with the target group of our overarching research project that focuses on the working population. In a lot of European countries, the retirement age is gradually increasing toward 67 years [31].

The results of the search query were uploaded into the EndNote X7 reference manager (Thomson Reuters, Philadelphia, PA, USA) and independently assessed by two reviewers to decide on their inclusion based on title, abstract, and full-text (the review team was comprised of AL and HO for selection based on title and abstract, and AL, HO, LP, and MG for selection based on full-text, with AL reviewing all full-text articles). Differences were fully discussed until consensus was reached.

In addition to the electronic database search, manual searching was performed in JMIR mHealth and uHealth for issues dated from January 1, 2013 to January 31, 2016. In addition, a check was performed on the bibliographic reference lists of publications that remained after full-text selection of the search query or manual searching and did not describe interventions involving blended coaching, to identify any additional eligible publications.

The electronic database search and manual searching resulted in 394 publications and 59 publications, respectively, 98 of which were duplicates. After the final full-text selection, 27 publications remained [32-58]. The check of the reference lists resulted in 5 additional publications [59-63] (see the flowchart in Figure 1).

A data-charting form was created by the research team that included the following: study characteristics (eg, title, participants, outcomes of interest, and effectiveness), intervention characteristics (eg, short description of the intervention, self-tracking components, and persuasive eCoaching components), and advantages and limitations of the intervention and research according to the authors or reviewers (see Table 1). Next, the data-charting form was improved by several iterations between researchers and 2 consensus meetings of the whole research team.

Using the framework by Morrison et al [64], the interventions were divided into 3 categories in terms of their effect on health outcomes: high effective, low effective, and ineffective (see Table 2). For example, if the intervention group showed statistically significant improvements for steps per day and body mass index but not for blood pressure as a result of within-group analyses and was more effective than the comparison intervention group as a result of between-group analysis, the intervention was categorized as “high effective.” No distinction has been made between the different outcomes, as long as they were related to health (eg, healthy lifestyle behavior or health status). Study quality was assessed by evaluating the rigor of the study designs based on the established hierarchy of study designs [65].

Data on usability was extracted from included publications if participants in the study expressed a preference for a component or a specific component’s design that increased their level of satisfaction regarding the technology or its ease of use.

Data on adherence was extracted if the studies described the way in which participants should use the intervention and presented results on the participants’ adherence to that intended use. Originally, we intended to divide the studies into incremental categories of adherence. Unfortunately, the data extracted from the studies did not allow us to do so. Additionally, data on adherence was extracted if participants expressed an expectation that a specific component could increase their adherence in using the technology.

Persuasive eCoaching components were extracted from included publications using the PSD model [9]. Solely, persuasive components were coded when they were executed by the technology and not by human effort, which is in line with the use of the PSD model as described in the review by Kelders et al [15]. As data extraction progressed, we decided to include 3 persuasive eCoaching components in the data-charting form: educational coaching, goal-setting, and feedback. The reason for this decision was that these components were often described in the intervention’s design and comprised coaching strategies that could be delivered via technology. In addition, the specific reasons for allowing social support in the design were often omitted. Consequently, we could not link design elements to specific social support components in the PSD model and therefore created the component social support in general (see Multimedia Appendix 2 for an overview).

The self-tracking components and other intervention components were identified using qualitative analysis of the data from publications. Components were added or changed continuously as the qualitative data analysis progressed.

The consistency of the data-charting form was discussed by two reviewers (AL and HO), who focused on data extraction performed by one reviewer (AL) on 4 articles with various study designs (3 studies evaluating the effect on health outcomes and 1 study on usability) [32,35,50,53]. After their discussion, it appeared that persuasive eCoaching components and the advantages and limitations of the research and interventions were more prone to reviewer subjectivism than other components. Therefore, we decided to extract the data from 3 articles [42,46,49] gathered by two reviewers (AL and HO) independently of these components to increase consensus with regards to the data interpretation.

All relevant data was coded using the data-charting form in ATLAS.ti version 7.5 (Scientific Software Development GmbH, Berlin), a qualitative software package. In addition, short summaries were obtained from the data-charting forms to provide quick overviews. Qualitative analysis was used due to our interest in how and why components were applied and to observe patterns in the application of the components and their contribution to the effect on health outcomes and usability [66]. Following Arksey and O’Malley [26] and Levac et al [27], descriptive numerical summaries and thematic analyses were used for data analysis, resulting in an approach that is akin to a “narrative review” [26]. First, a descriptive numerical summary was used to create a numerical overview of specific self-tracking components, persuasive eCoaching components, and other components in the interventions categorized by their effect on health outcomes. Components were identified as key components if at least 50% of the interventions that showed effectiveness on health outcomes (high and low effective interventions) included the component. This 50% rule was applied to all persuasive eCoaching components, with exception of feedback, the self-tracking component validity, and the other intervention components, theory applied, results from other research applied, cocreation, design testing, integration of self-tracking and persuasive eCoaching, and blended coaching. Other components included in the data extraction were of a descriptive nature and could therefore not be treated as dichotomous components for which percentages could be calculated.

Second, thematic analysis was applied to obtain more insight into the various components’ specific designs and if these specific designs relate to the effectiveness of the interventions on health outcomes, usability, and/or adherence. When patterns were observed linking components and effectiveness, these components were then identified as key components. Additionally, thematic analysis directed the process of creating the data-charting form.

The aim of this consultation was to give meaning to and assess the applicability of the results by obtaining insight from other perspectives, beyond the research team’s own perspectives [26,27]. The consultation was carried out during the 11th International Conference on Persuasive Technology. The preliminary results of this scoping review [67] were presented and input was requested from experts in several fields during the workshop on Behavior Change Support Systems (BCSS 2016): Epic for Change, the Pillars for Persuasive Technology for Smart Societies. This consultation adjusted the scope to the way in which components are designed, to get a clear idea of how and why specific components do or do not contribute to effects on health outcomes, usability, and/or adherence.

An overview of the key components, categorized by effect measures (health outcomes, usability, and adherence), can be found in Multimedia Appendix 4. This table also provides an overview of which studies the key components are based on. Results regarding components have not been presented if too little data was present (results regarding that component from only one study) or no clear pattern could be observed between the component and the effectiveness (on health outcomes, usability, and/or adherence). Moreover, key components were not separately described for interventions utilizing blended coaching or automatic coaching, as too little data was present or no differences were observed for the key components between the 2 types of coaching in terms of their effects. The same holds for key components from studies describing results regarding usability based on expectations, and studies based on experiences. The key components presented below are divided according to the 3 effect measures: health outcomes, usability, and adherence.

This scoping review aimed to identify key components of self-tracking and persuasive eCoaching in automated healthy lifestyle interventions that contribute to the effectiveness on health outcomes, usability, and adherence.

First, the key components identified in this scoping review are identified as separate components. However, components might interact with each other and could lead to different outcomes than those anticipated based on knowledge of single components. One review on e-mental health interventions indicated that some combinations of persuasive technology components do indeed differ in terms of their synergy [90]. Further research is recommended to identify the most effective combination and dosage of the key components in healthy lifestyle interventions combining self-tracking and persuasive eCoaching. To date, most studies evaluating eHealth designs apply the traditional RCT method [91]. However, RCTs are often too time-consuming to keep up with the speed of technological developments and can explain little about separate elements and their contribution to effectiveness [91-93]. Riley and Rivera [94], Hekler et al [93], and Pham et al [91] advocated for new strategies to identify and design effective intervention components—“opening the black box.” Technology can provide a meaningful contribution to such discoveries, for example, by means of the “Model Predictive Control” [94]. This strategy changes the intensity and combination of intervention components on a daily basis by using the monitoring data provided by participants’ responses and other contextual factors.

Another strategy is suggested by Sieverink et al [94]. This strategy does not only attempt to open the black box, it also contributes to more insight into adherence. Sieverink et al provided preliminary results for the development of a log data protocol for eHealth technologies to identify their adherence level and effect on health outcomes [94]. They suggest collecting log data on the usage and intensity of usage for specific intervention components to be able to draw conclusions regarding adherence and linking such log data with effects on health outcomes to be able to draw conclusions regarding adherence to specific components and their effects on health outcomes.

Future research and design should focus specifically on the reminder design, social support, and the observation of patterns through data mining, as different designs seemed to influence effect on health outcomes and/or usability. It would be interesting for future research to test variations on components’ designs and their effects on health outcomes and usability.

Another recommendation is the application of personalization to account for the variation in preferences between groups of participants and even within groups of participants. Besides the fact that it is a practical way to account for the existing differences between users, participants also considered personalization to be useful. In addition to the application of personalization in the design, the observed differences both among and within groups also suggest that a needs assessment is required before and during the design phase of an intervention using self-tracking and persuasive eCoaching. Although cocreation is often mentioned as an important aspect in design models [95,96], only a few publications cited in this review described anything regarding cocreation between intervention developers and the target group. The importance of cocreation has been indicated as important to increase satisfaction with the design by other reviews [67,69]. Similar to the component of cocreation, few publications described anything about the theory underpinning the design of the intervention. The main message of other reviews in the field of eHealth is that the design of current eHealth interventions is often not based on existing theory [67,69]. The use of theory has been recognized as resulting in higher effectiveness [97].

One of the strengths of this research is that by applying qualitative research methods, an attempt has been made to not only describe which components might contribute to effectiveness but also which specific component design is most effective in terms of health outcomes, usability, and adherence. This is important because applying components from one theory in different interventions can result in various designs, whereas applying components from different theories can result in interventions with quite similar designs [93]. Another strength is the fact that we used both data from RCT studies and studies in real-life settings, providing a more realistic overview of the opportunities and challenges for interventions that combine self-tracking and persuasive eCoaching in practice than we would have been able to by relying only on results from RCTs.

One limitation of this scoping review is that potential biases might have influenced the results. First, publication bias could be present, indicated by the absence of negative effects reported and the higher number of high effective interventions included in this scoping review versus low effective or ineffective interventions. With results from only 3 ineffective studies, we could not come to conclusions about any one key component being more often applied in effective interventions compared to ineffective interventions. Due to this limitation, we introduced the 50% rule to identify key components. It should be mentioned that the most commonly applied components in interventions were, therefore, more likely to be identified as key components.

Second, we observed that interventions described in publications from 2013 already differ to some extent from interventions described in publications from 2016. The importance of certain identified key components for effectiveness might increase or decrease due to new technological developments. The main differences were the more frequent use of accelerometers and mobile sensors for self-tracking in newer publications and the delivery of the intervention via mobile phone in newer publications in comparison with computer in older publications. This trend is likely to continue [73,98]. An example of a key component that may become more important is the ability to enable or disable observation of trends and patterns. The use of mobile phone sensors enables collection of a wide spectrum of personal data. As indicated by this scoping review’s results, not every user is open to intensive data mining. As another example, the importance of applying the proper frequency and timing of reminders may increase . Mobile phone interventions can use a broader set of tools to send reminders than computer-based interventions. In addition, reminders cannot be ignored as easily due to visual or audible alerts [73].

Third, we did not make a distinction between health outcomes. It could be the case that interventions targeting a more intermediate health outcome (eg, an effect on physical activity instead of an effect on blood pressure) were more easily identified as high effective studies. Fourth, the extraction of data concerning persuasive eCoaching components is somewhat subjective, which was observed by the two researchers during data extraction comparison. Finally, we did not code intervention components from the actual interventions because technology is a rapidly evolving field of research and this would have taken a significant amount of time. Choices such as this one are characteristic of the scoping review methodology [26]. These limitations limited us in making a definite list of key components. However, we attempted to provide a first impression of key components in this relatively new field of research.

To our knowledge, this scoping review provides a first overview of key components and effects on health outcomes, usability, and adherence. The following key components and their specific design both seem to influence health outcomes and usability in a positive way: reduction by setting short term goals to eventually reach long-term goals, personalization of goals, praise, reminders to input self-tracking data into the technology, use of validity-tested devices, integration of self-tracking and persuasive eCoaching, and provision of face-to-face instruction during implementation. In addition, health outcomes or usability were not affected when more effort was requested from participants to input data into the technology. Unfortunately, we were limited in our ability to identify key components for adherence. Still, one key component identified for both usability and adherence is the provision of personalized content. Identification of key components for adherence is highly important because adherence is a prerequisite for interventions to be effective. This scoping review provides a first overview, and future research is needed to confirm the key components identified for effect on health outcomes and usability, identify key components for adherence, and study whether the key components represent an effective combination of components.