A nonsystematic literature search was performed to identify original research papers that examined the selected persuasive strategies in the context of PA mHealth interventions. A nonsystematic search was deemed appropriate because the objective of this study is to gain insights into the various operationalizations of strategies and their influence on the effectiveness of the strategy and not to provide a complete overview of current literature or a list of effective implementations and characteristics of the strategies. The PRISMA ScR (Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews) criteria were used to guide the reporting of the methods and findings (Multimedia Appendix 1) [28]. The protocol for this review was not registered.

Most papers were identified by checking the references of (recent) reviews in the same field and in the author’s personal libraries. As the research team is multidisciplinary, papers from the fields of behavior change, computer science, and gamification were included. When interesting papers were identified (either by previous reviews or in the library of authors), extensive snowball searches were performed (ie, the references of all interesting papers were checked for other relevant papers). We extended the snowball search with additional quick searches in Google Scholar when the initial search resulted in a small number of studies for a specific operationalization of a persuasive strategy (eg, sharing information on social media). The search included terms that refer to mHealth (eg, mobile devices, PA apps), PA behavior (eg, exercise, walking, running), the persuasive strategy (eg, sharing, social media, Facebook, Twitter), and adults. These additional searches increased the number of identified papers to a limited extent.

Studies were included if they (1) were (partly) delivered by a mobile system (eg, pedometer, SMS, mobile app), (2) reported PA outcomes (self-reported or objectively measured), (3) used an experimental trial (eg, randomized controlled trial, factorial design, pretest-posttest design), (4) examined at least one of the selected persuasive strategies, and (5) described this strategy in sufficient detail. Finally, (6) studies were only included when they specifically compared the effect of different designs or implementations of persuasive strategies. Therefore, papers were excluded if they only examined the effectiveness of the intervention as a whole (with multiple strategies). Studies were also excluded if they targeted children or individuals with a chronic disease (eg, patients with cardiovascular diseases, mental disorders, etc), as these target groups may have different needs [26]. There were no restrictions on publication year, sample size, or study duration.

To systematically analyze the included studies, a data charting list was developed in multiple review rounds with input from all coauthors (consistent with the guidelines for writing a scoping review [29]). The final data chart comprised 3 sections, namely (1) study characteristics, (2) technical implementations and design characteristics of persuasive strategies, and (3) study results (Multimedia Appendix 2). The study characteristics included information on the methodology of the study and other factors that can influence study outcomes (ie, characteristics of the participants and contextual factors [30]).

The technical implementations and design characteristics section was inspired by the Behavioral Intervention Technology (BIT) model [20]. The BIT model is grounded in 3 well-respected design models [11,31-33] and includes principles from both behavioral theories and technological features. The BIT delivery elements (subdivided into delivery systems and elements) and BIT workflow informed the technical implementation category (Multimedia Appendix 2). The BIT characteristics of the elements informed the design characteristic category. Most design characteristics were inferred from previous reviews [22-25] and theories of behavior change [33-35], as the BIT framework does not provide a detailed list of design characteristics. Furthermore, during the data extraction period, the chart was updated when new implementations or design characteristics were identified.

The final data chart section covered the study results. We examined whether there was a higher amount of PA compared with a control group (without the strategy) and compared with another intervention arm (with a different operationalization of the strategy, see Multimedia Appendix 2). The PA outcome measurement used was the step count of the participants, unless the paper did not measure this. In that case, the main PA outcome measurement was used instead, as described in this study. The outcomes were classified as positive when the PA outcome measurement of the study was significantly more effective, neutral when no effect was found, and negative when the implementation resulted in worse PA outcomes.

The general study characteristics were extracted for the study as a whole. The type of persuasive strategy and its design and technical implementation were extracted separately for each intervention arm because this differed between arms in the same study. One researcher (KS) performed the data extraction of all the included papers. To ensure that the data extraction was performed correctly for all persuasive strategies (in line with Levac et al [29]), the second reviewer (SW) performed data extraction of at least 20% of each of the persuasive strategies (22.6% on average, SD 0.9%). The interrater reliability was high (93.5% agreement).

The search yielded 29 original research papers (85 intervention arms) [36-64]. An overview of the study characteristics can be found in Multimedia Appendix 3 [36-64]. The results of the individual persuasive strategies are presented below. For each strategy, first, a short description or definition is provided. Second, a summary of the identified implementations and designs in the intervention arms is provided (a complete overview can be found in Multimedia Appendix 3). Third, an overview is given of which implementations have proven to be an effective addition in an intervention. Finally, the findings of studies that compared designs and implementations of strategies are provided.

Monitoring involves keeping track of your behavior or behavior outcomes [9]. Traditionally, users had to actively track their own PA behaviors (eg, by means of questionnaires and self-logging); however, nowadays, PA tracking can also be performed passively by using mobile devices, without posing a burden to the user. Almost all of the included studies used monitoring of behavior (28 studies and 82 intervention arms), apart from [54].

Goal setting is a strategy in which the individual either sets a goal or gets a goal assigned. CALO-RE distinguishes between unspecific behavioral goals, behavioral outcome goals, and action plans [9] (also referred to as implementation intentions [65]). It is a commonly used strategy [18], as reflected by the presence of the strategy in most of the included papers (n=26). Not all studies described goal setting in sufficient detail [41,42,61], and these were, therefore, excluded from this part of the analysis. The results of the remaining 23 studies (66 intervention arms) are listed below.

Reminders are cues, prompts, or triggers that push the user to perform a certain behavior, such as a notification or an email. Only reminders that were used to remind the user to engage in PA were considered. Reminders intended to remind the user to set a goal or to wear the device were not taken into account [40,50,59]. Reminders are used relatively frequently, almost half of the included studies incorporated this strategy (n=13, implemented in 34 intervention arms).

Rewards are reinforcers of behavior that can be given for attempts to reach a goal and for reaching the goal [9]. In line with the CALO-RE taxonomy [9], only rewards that incentivize performing PA behaviors were considered and not rewards that incentivize study participation. A total of 12 studies (29 different intervention arms) met these criteria.

Sharing is a social strategy in which users can actively offer and receive social support from others. It can be provided digitally (eg, sharing on Facebook) or in a real-life setting (eg, group meetings [48]). In total, 9 studies (15 intervention arms) were identified that included this strategy in their intervention.

A second social strategy is facilitating social comparison [9], which includes competition, collaboration, and social norm information. In total, 9 of the included studies enabled social comparisons (17 intervention arms and 11 did not include social comparison).

The main objectives of this study were to examine (1) the variation in the technical implementations and design characteristics of common persuasive strategies and (2) if previous mHealth studies examined the influence of the different operationalizations of these persuasive strategies on the effectiveness of the intervention to persuade the user to engage in PA. Similar to previous work on feedback [22], we found that the technical implementations and design characteristics of the examined persuasive strategies vary greatly between studies. For instance, 29 different goal types were identified in the 29 included studies. The goals differed in terms of behavior type (eg, step count, time walked) and difficulty (eg, 10,000 steps per day, 7000 steps per day).

Only a few of the identified implementations and designs were thoroughly examined for effectiveness, especially regarding the technical implementation. Interestingly, the studies that examined this showed that some implementations and designs are more effective in increasing PA compared with others. By performing an in-depth analysis of these studies, this review provides important insights into which implementation types and designs are promising for PA mHealth interventions, as listed below. Furthermore, this study highlights that more research on effective technical implementations and design characteristics of persuasive strategies is essential.

Monitoring was not only the most frequently used strategy in the selected studies but also the least investigated one regarding implementation types and designs. Only one study compared technical implementations, which found no significant difference regarding intervention effectiveness between automatic tracking alone, or both automatic tracking and (active) self-monitoring [41]. Thus, although self-monitoring is arguably more demanding for the user, these results suggest that this additional effort does not influence the effectiveness of the intervention. A possible explanation is that individuals perceive wearing a second device as inconvenient or not pretty (eg, when wearing a dress) [40], which counterbalances the positive effects of the ease of use of automatic tracking. More research is needed to better understand how monitoring can be implemented to successfully support PA behavior.

The results suggest that the operationalization of goal setting influences the effectiveness of the intervention to increase PA behavior. In line with the Goal Setting Theory [34], this review demonstrates that goals that are tailored to the physical capabilities of the user and/or the context of the user are likely more effective in increasing PA than generic goals [36,37,50,55,62]. Next, challenging (but doable) goals appear to be more effective than easy goals. However, there is no consensus on how challenging the goal should be. Is a small increase of baseline steps a difficult enough goal (eg, 20% increase [50]) or is a large increase required (eg, a 100% increase [37])?

In contrast to the Goal Setting Theory [34] and other theories, such as the Self-Determination Theory (SDT) [67], positing that self-set goals are likely to be more effective, the results of this review indicate that adding self-set goals to an intervention does not increase effectiveness [58] and that goals set by the system might be more promising than self-set goals in increasing PA behavior [68]. However, people will likely choose to set their own goal when offered the choice [68]. Therefore, there might be a discrepancy between what the user wants and what is most effective. Whether a combination of both self-set and assigned goals is effective should be determined in future studies. Future efforts are also needed to explore the technical implementations of gamification, as none of the selected studies examined this.

One of the included studies in this study demonstrated that a promising approach to implement reminders is the use of glanceable displays, which is a constant but gentle reminder that resides at the background of the device [41]. However, as most implementation types of reminders were not compared with each other, it is unknown if certain designs of glanceable displays are more attractive and effective than others (eg, flowers or a robot army [41]) or if other types of reminders (such as text messages) are equally effective as glanceable displays.

With respect to the design characteristics of reminder text messages, tailoring the timing of the messages might improve the intervention [4,33,69]. One study examined this and found that optimally timed messages were perceived as more useful than randomly timed messages, although it did not result in more PA [42]. Next, regarding the content of the reminders, the results of this study suggest that generic reminders do not result in more PA compared with receiving no messages [61]. People perceive these messages as impersonal, boring [61] and stop reading them [70,71]. Tailored message content is likely more appropriate [23], as demonstrated in similar research domains [72]. However, none of the papers selected for this study examined this. Taken together, future studies should be conducted to better understand the effectiveness of tailoring (of both the content and the timing) of messages to increase PA behavior.

This study suggests that the design and implementation of rewards are important factors that motivate individuals to engage in PA. For instance, the results suggest that rewards are effective if they are cumulative (eg, with enough points, the participant receives a badge). Furthermore, receiving an immediate reward (eg, immediately after attaining the goal) is likely effective, whereas receiving delayed rewards seems less effective. This is in line with previous research and behavioral economics theory [35], which explain that immediate rewards are perceived as more valuable than rewards given in the future [73,74].

Although several studies have examined effective reward types (eg, comparing monetary rewards with charity donations), multiple important questions remain unanswered. For instance, it is neither known whether the value of the financial reward influences its effectiveness, nor is it known if certain types of virtual rewards (eg, badges, points, levels) are more effective than others. The use of rewards is sometimes criticized, as they mainly promote the extrinsic motivation to perform PA [75,76] and can inhibit the intrinsic motivation associated with lasting behavior change. Consequently, users likely stop performing PA after the reward is removed, as demonstrated in several of the selected studies of this study [44,51,52]. Therefore, it is important to gain a better understanding of how to design rewards that foster feelings of intrinsic motivation [77].

Sharing and giving social support are argued to foster feelings of relatedness to others, which is an important determinant of intrinsic motivation to perform PA according to the SDT [67,75]. However, of the 4 studies that examined the effect of adding sharing to the intervention, only 1 study found that having access to a sharing function (with Facebook) significantly increased the step count [56]. A possible explanation for the ineffectiveness of the other 3 implementations of sharing is that they used a delivery system that is not well integrated in people’s lives and are, therefore, not frequently used [40,49,60]. However, as no study has directly compared the effectiveness of different delivery systems, future research should examine whether the delivery system of social sharing indeed influences the effectiveness of the intervention. Other technical implementation or design characteristics also require future investigation, such as with whom the data are shared or the type of data that are shared (such as tips from peers or the performed PA behavior).

In contrast to previous work on PA (mHealth) interventions [7,14], we found that most interventions that included social comparisons did not result in significantly more PA compared with the same intervention without social comparison. There may be various reasons for this, such as the relatively small sample sizes used in the studies [38,47] because the strategy was not optimally operationalized or it could be that the strategy itself is not effective. As only 2 studies compared different design characteristics and no study examined the technical implementation of social comparisons, knowledge on effective operationalizations remains to be limited. However, it does seem that collaboration enhances participation in PA above competition (in line with previous research [38,78]). Notably, it might not be competition itself but the overemphasis on winning that can be counterproductive [78]. In general, it is thought that competition can increase PA when everybody has a reasonable chance to win [79], as also indicated in the exit interviews of Chen and Pu [38].

The important strengths of this study are the in-depth systematic analysis of multiple persuasive strategies by using a framework and the exploration of both the technical implementation and design characteristics of the interventions. By doing so, we went beyond examining which persuasive strategies are effective by also examining how these strategies can be effectively implemented and designed. Furthermore, by only including studies that evaluated the isolated effect of the selected persuasive strategies, an overview of implementations and designs could be provided that have been shown to be effective.

However, this study has some limitations. First, inherent to performing scoping reviews, we did not consider the quality of the included studies, which could have biased the results. In addition, as no systematic search was performed, important literature could have been overlooked. To minimize this risk, we performed cross-reference checks of previous (recent) reviews and snowball searches. In light of these limitations, the results of this study should be treated with caution. Second, no corrections for other factors that can influence the effectiveness of the intervention, such as the study duration [36], the geographical location of the study [30], or user friendliness of the intervention [27]. To illustrate, it is likely that the same intervention is not equally effective in different contexts (such as in different countries or seasons). Although we recognize the importance of these factors, it was beyond the scope of this paper to examine this. A third limitation is that the influence of other persuasive strategies of the intervention was not examined. For instance, it is possible that the investigated strategy did not increase PA behavior, even though its operationalization was good, as the other included persuasive strategies of the intervention were not set up properly.

Furthermore, the data extraction was limited to the amount of details in which the researchers described their intervention. For instance, most papers did not describe how they developed the reminder system (eg, whether they used a pool of messages or framed the messages positively or negatively). Therefore, we call on future studies to report the implementations and designs in more detail. For instance, authors can include screenshots of the apps or websites [43], videos of the app, and/or a user manual. To document mHealth interventions, researchers can use the framework provided in Multimedia Appendix 3 and, for instance, the work of Schembre et al [22], Elaheebocus et al [25], and/or Hoffman et al [80].

We realize that the number of implementation types is so diverse that it would cost considerable time and resources to investigate the individual design characteristics and technical implementations of the strategies. A promising approach to reduce this burden on researchers is to build a database in which mHealth PA interventions are described at a granular level, covering both the implementation characteristics of the included persuasive strategies and the study characteristics (as possible confounders). Advanced statistical testing (eg, Meta-Cart analyses [81]) and machine learning techniques [82] allow identifying which implementation characteristics seem effective for which target group.

Mobile exercise interventions have the potential to increase PA behaviors of individuals [1,6,7]. However, there is a limited understanding of how to effectively develop exercise interventions and its components (ie, their persuasive strategies). To increase this understanding, it is important to examine the operationalization of persuasive strategies and to evaluate its impact on the effectiveness of the intervention. The results of this study highlight the great variation in which monitoring, goal setting, reminders, rewards, sharing, and social comparison are being operationalized. Moreover, the findings of this study suggest that how a conceptual persuasive strategy is being translated into a practical delivery form can influence the effectiveness of the PA intervention. Thus, the operationalization of strategies in mHealth interventions should be critically considered when developing such interventions and before drawing conclusions on the effectiveness of the strategy as a whole. To advance the research field, future research should go beyond evaluating which persuasive strategies are effective by also examining how these strategies can be effectively implemented and designed.