Determinants of Fitness App Usage and Moderating Impacts of Education-, Motivation-, and Gamification- related App Features on Physical Activity Intentions: A Cross-Sectional Survey

Background: Smartphone fitness apps have been considered promising tools for promoting physical activity and health. Yet, the user-perceived factors and app features that influence users’ adoption of these apps and subsequent associations with intentions to be physically active remain uncertain. Objective: Building upon the second version of the Unified Theory of Acceptance and Use of Technology (UTAUT2), this study aims to examine the influence of the seven UTAUT2 determinants as well as the moderating effects of different smartphone fitness app features (i.e., education-, motivation-, and gamification-related) on individuals’ usage intentions and their behavioral intentions to be physically active. Methods: Data from 839 U.S. residents who reported having used at least one smartphone fitness app were collected via an online survey. A confirmatory factor analysis was performed and path modeling was used to test the hypotheses and explore the influence of app features on the structural relationships. Results: The seven determinants explain 73% of the variance in behavioral intentions to use fitness apps. Performance expectancy (? = .42, P < .001), effort expectancy (? = .10, P = .001), facilitating conditions (? = .09, P = .002), hedonic motivation (? = .06, P = .03), price value (? = .11, P < .001), and habit (? = .35, P < .001) are positively related to behavioral intentions to use fitness apps. There is no significant relation between social influence and behavioral intentions to use fitness apps. The behavioral intentions to use fitness apps relate positively to intentions to be physically active (? = .12, P < .001; R2 = .02). Education-related app features moderate the effect of performance expectancy; motivation-related features moderate the effects of performance expectancy, facilitating conditions, and habit; and gamification-related features moderate the effect of hedonic motivation on usage intentions. Follow-up tests are employed to describe the nature of the interaction effects. Conclusions: The study identifies important drivers of the usage of fitness apps. Smartphone app features should be designed to increase the likelihood of app usage and hence physical activity by supporting users in achieving their goals and facilitating habit formation. Target-group specific preferences for education-, motivation-, and gamification-related app features should be taken into account. For example, since performance expectancy has a high predictive power of intended usage for consumers who appreciate motivation-related features, apps targeting these users should focus on goal achievement-related features (e.g., goal setting and monitoring). Future research might look into the mechanisms of these moderation effects, and their long-term influence on physical activity levels. (JMIR Preprints 26/11/2020:26063) DOI: https://doi.org/10.2196/preprints.26063 Preprint Settings 1) Would you like to publish your submitted manuscript as preprint? Please make my preprint PDF available to anyone at any time (recommended). Please make my preprint PDF available only to logged-in users; I understand that my title and abstract will remain visible to all users. https://preprints.jmir.org/preprint/26063 [unpublished, peer-reviewed preprint] JMIR Preprints Yang et al Only make the preprint title and abstract visible. No, I do not wish to publish my submitted manuscript as a preprint. 2) If accepted for publication in a JMIR journal, would you like the PDF to be visible to the public? Yes, please make my accepted manuscript PDF available to anyone at any time (Recommended). Yes, but please make my accepted manuscript PDF available only to logged-in users; I understand that the title and abstract will remain visible to all users (see Important note, above). I also understand that if I later pay to participate in <a href="https://jmir.zendesk.com/hc/en-us/articles/360008899632-What-is-the-PubMed-Now-ahead-of-print-option-when-I-pay-the-APF-" target="_blank">JMIR’s PubMed Now! service</a> service, my accepted manuscript PDF will automatically be made openly available. Yes, but only make the title and abstract visible (see Important note, above). I understand that if I later pay to participate in <a href="https://jmir.zendesk.com/hc/en-us/articles/360008899632-What-is-the-PubMed-Now-ahead-of-print-option-when-I-pay-the-APF-" target="_blank">JMIR’s PubMed Now! service</a> service, my accepted manuscript PDF will automatically be made openly available. https://preprints.jmir.org/preprint/26063 [unpublished, peer-reviewed preprint] JMIR Preprints Yang et al


Table of Contents
Only make the preprint title and abstract visible. No, I do not wish to publish my submitted manuscript as a preprint. 2) If accepted for publication in a JMIR journal, would you like the PDF to be visible to the public?
Yes, please make my accepted manuscript PDF available to anyone at any time (Recommended).

Yes, but please make my accepted manuscript PDF available only to logged-in users; I understand that the title and abstract will
Yes, but only make the title and abstract visible (see Important note, above). I understand that if I later pay to participate in <a href="http

Original Manuscript
ABSTRACT Background: Smartphone fitness apps are considered promising tools for promoting physical activity and health. However, it is still unclear which user-perceived factors and app features encourage users to download the apps with the intention of being physically active.
Objectives: Building upon the second version of the Unified Theory of Acceptance and Use of Technology (UTAUT2), this study aims to examine the association of the seven UTAUT2 determinants with, as well as the moderating effects of different smartphone fitness app features (i.e., education-, motivation-, and gamification-related) and individual differences (i.e., age, gender, and experience) on, individuals' app usage intentions and their behavioral intentions of being physically active.
Methods: Data from 839 U.S. residents who reported having used at least one smartphone fitness app were collected via an online survey. A confirmatory factor analysis was performed and path modeling was used to test the hypotheses and explore the influence of moderators on the structural relationships.

Results:
The determinants explain 76% of the variance in behavioral intentions of using fitness apps.
Habit (β = .42, P < .001), performance expectancy (β = .36, P < .001), facilitating conditions (β = .15, P < .001), price value (β = .13, P < .001), and effort expectancy (β = .09, P = .04) were positively related to behavioral intentions of using fitness apps, whereas social influence and hedonic motivation were non-significant predictors. Behavioral intentions of using fitness apps related positively to intentions of being physically active (β = .12, P < .001; R 2 = .02). Education-related app features moderated the association of performance expectancy and habit with app usage intentions; motivation-related features moderated the association of performance expectancy, facilitating conditions, and habit with usage intentions; and gamification-related features moderated the association of hedonic motivation with usage intentions. Age moderated the association of effort expectancy with usage intentions, and gender moderated the association of performance expectancy

INTRODUCTION
Today, there are 3.8 billion smartphone users worldwide [1], and about half of them consider their smartphones something "they could not live without" ( [2], p. 237). Numerous smartphone applications (apps) have been developed to allow users to go beyond basic voice calling and texting to social media, gaming, as well as managing one's health and fitness. In March 2021, 101,481 apps in the Google Play Store and 158,558 apps in the Apple App Store were available to users in the category of health and fitness [3,4]. These apps aim to promote physical activity and healthy lifestyles [5,6]. It is important to increase our understanding of what factors influence users' adoption of these apps, and subsequent associations with intentions of engaging in healthy behaviors -both from the perspective of public health and management (e.g., app providers), -because stakeholders in these domains are (or should be) interested in finding ways to promote healthy lifestyles via digitization in general and the use of mobile devices in particular.
The most widely used theoretical frameworks to explain users' adoption or use of technology are the Technology Acceptance Model [7] and the Unified Theory of Acceptance and Use of Technology (UTAUT) [8]. The two models focus on organizational contexts. In consumer settings, the second version of the UTAUT (i.e., UTAUT2) has been developed to explain individuals' acceptance of new technology [9]. Since the UTAUT2's first application (studying the acceptance of the mobile Internet), it has been used to explain smartphone app adoption and usage [10,11], among other applications. As regards previous empirical studies on mobile health and fitness apps, important gaps exist in the research. First, previous studies have left out essential determinants that the UTAUT2 incorporates (e.g., habit and hedonic motivation). Given the importance of habit [12] and hedonic motivation [13], the sole focus on the four determinants proposed by the UTAUT seems insufficient (e.g., [14,15]). Second, the relationship between intentions of using fitness apps and intentions of being physically active has not been explored. Assessing the downstream effects of intentions of using fitness apps is important, because downloaded but unused apps, or apps that are unable to motivate people to become or remain physically active, will have few health benefits [5,16]. Third, understanding of whether different fitness app features moderate the relationships of the UTAUT2 determinants and behavioral intentions of using the app is lacking. Previous research has categorized app features, such as education-related vs. motivation-related features [17], but has not considered their influence on structural relationships that aim to explain app usage intentions and physical activity intentions. Lastly, despite the fact that the moderating effects of individual difference variables (e.g., age, gender, and experience) have been theorized and empirically assessed [9], they have largely been neglected in prior research on mobile health and fitness apps [18][19][20][21]. Yet, their relevance was shown in a post hoc meta-analysis, for example, in which age was a significant moderator [22].
This study aims to partially fill these gaps and answer four research questions: (i) What are the relationships between the UTAUT2 determinants and individuals' behavioral intentions of using fitness apps? (ii) What is the downstream relationship between behavioral intentions of using fitness apps and intentions of being physically active? (iii) Do fitness app features moderate the relationships between the UTAUT2 determinants and intentions of using fitness apps? (iv) Are there individual differences regarding age, gender, and user experience in the relationships between the UTAUT2 determinants and intentions of using fitness apps?
To answer the research questions, we applied and extended the UTAUT2 model. A sample of 839 individuals was surveyed to test our hypotheses. Path modeling was used to test hypotheses.
Before we present the results of the study, in the following, we review the extant literature on determinants of fitness app usage, develop the hypotheses, and present the methodology of our approach.

Smartphone fitness apps
Along with the growing consensus on the health benefits of physical activity [23], a myriad of fitness wearables and smartphone fitness apps have been developed to quantify and promote physical activity. Fitness wearables are "devices that offer training plans, assist with activity tracking, and generally collect and process health-related data" ( [24], p. 1), while fitness apps refer to "the self-contained programs for smartphones designed for the purpose of getting fit" ( [25], p. 11).
The present study focuses on smartphone fitness apps.
Despite the potential of smartphone fitness apps to deliver cost-effective physical activity and health promotions, their effectiveness has been insufficiently established [5,16,26,27]. In particular, the effectiveness of the use of fitness apps or app-based interventions was found to be modest or short-lived [5,16]. In previous studies, only a limited number of factors that were considered by researchers were based on theories or behavior change techniques [16,26,27]. Furthermore, only a small number of fitness apps have undergone rigorous evidence-based evaluation in controlled trials [28]. There are some quality concerns in the reporting of the studies, too. For example, only a few studies reported whether fitness apps were based on human behavior change theories [28,29]. In the following, we outline the factors that might predict individuals' behavioral intentions of using fitness apps (and their downstream effects), building upon theories that have been identified as relevant in the information systems literature, particularly UTAUT2.

Determinants of the behavioral intentions of using fitness apps
Venkatesh et al. [8]  Hew et al. [20] applied the UTAUT2 to examine the factors that affect smartphone app adoption in general, considering the moderators of gender and education. They found that all but two factors (i.e., social influence and price value) were significant determinants, with habit exerting the strongest effect. Gender and education were non-significant moderators. Most important to the present research, previous studies used the UTAUT2 to investigate the determinants of behavioral intentions of using fitness-promoting smartwatches [18] and fitness apps [19,30]. However, none of them considered individual difference factors as moderators and none of them took into account the impact of app features on the proposed relationships.
Specifically, Beh et al. [18] found positive relationships between performance expectancy, effort expectancy, facilitating conditions, as well as hedonic motivation and behavioral intentions of using smartwatches for fitness and health monitoring purposes. The authors postulated that perceived vulnerability of developing chronic diseases and perceived severity of chronic diseases would moderate the effects, but found only weak support for their hypotheses. Dhiman et al. [19] found that effort expectancy, social influence, price value, and habit related positively to fitness app adoption intentions. They considered self-efficacy to be a predictor of effort expectancy and innovativeness as a predictor of habit. Both relationships were found to be significant. Yuan et al. [30] did not consider any mediators and found that performance expectancy, hedonic motivation, price value, and habit were predictors of behavioral intentions of continuously using health and fitness apps, but that effort expectancy, social influence, and facilitating conditions were non-significant predictors. There are important limitations to these studies. First, the downstream effects on intentions of being physically active were not assessed in any of the studies. The linkage of fitness app usage intentions and intentions of being physically active is important, because health benefits will only be realized if intended app usage motivates people to become or remain physically active. Second, none of the studies considered app features to be relevant moderators, despite the fact that previous research showed that app features such as gamification might moderate the effects of UTAUT2 determinants on app usage intentions [31], and despite the fact that the consideration of risk-perception factors (instead of app features) was largely unsuccessful [18]. Third, only one study assessed the moderating role of age, gender, and experience. The authors did not, however, include these variables in the model due to non-significant findings [30]. Thus, important similarities with, and differences to, the original UTAUT2 studies as regards the influence of age, gender, and experience remain largely unknown. The present study aims to partly fill those gaps.
Building upon UTAUT2, we first propose that the seven UTAUT2 determinants relate positively with individuals' intentions of using fitness apps. Second, we postulate positive downstream relationships with intentions of being physically active. Third, we pose a research question that considers three prominent app features (i.e., education-, motivation-, and gamificationrelated) as moderators of the relationships between the seven UTAUT2 determinants and behavioral intentions of using the app. Lastly, we explore the moderating effects of individual differences (i.e., age, gender, and experience) on the relationship between the seven UTAUT2 determinants and behavioral intentions of using the app. In the following, we derive our hypotheses.

Performance expectancy
Performance expectancy is defined as the "degree to which using a technology will provide benefits to consumers in performing certain activities" ( [9], p. 159). It was the strongest predictor of behavioral intentions in the original UTAUT study [8] and is a pivotal determinant of new technology acceptance in the areas of healthcare [32,33] and fitness wearables [21,34]. In the context of the present study, performance expectancy refers to the degree to which a user believes that using a particular fitness app would help improve their fitness. Previous studies showed a positive relationship between performance expectancy and intentions of using fitness apps [15,30].
Since the perception that fitness apps help people reach their fitness-related goals should be of high relevance to users, we propose that: H1: Performance expectancy relates positively with individuals' behavioral intentions of using fitness apps.

Effort expectancy
Effort expectancy refers to "the degree of ease associated with consumers' use of technology" ( [9], p. 159), similar to the perceived ease of use as described in the Technology Acceptance Model [7]. Within the current study, effort expectancy assesses the perceived ease of use of fitness apps. The easier the individuals believe the fitness apps are to use, the higher their intentions of using them.
Prior studies revealed a positive relationship between effort expectancy and behavioral intentions of using fitness apps [15,19] and fitness wearables [18,34]. Since people should be interested in intuitive and easy app usage, we expect that: H2: Effort expectancy relates positively with individuals' behavioral intentions of using fitness apps.

Social influence
Social influence is defined as "the extent to which consumers perceive that important others (e.g., friends, peers) believe they should use a particular technology" ( [9], p. 159). Social influence plays a particular role when users lack information about the usage [37]. In the context of fitness apps, previous studies revealed inconsistent results on the impact of social influence on behavioral intentions of using fitness apps. It was a positive predictor of Chinese university students' [15] and Indian users' usage intentions [19], while it did not predict the intentions of college-aged U.S. residents [30]. Given the positive effect of social influence postulated in the original UTAUT2 [9] and the importance of social support in being physically active [36,37], we assume that: H3: Social influence relates positively with individuals' behavioral intention of using fitness apps.

Facilitating conditions
Facilitating conditions refer to "consumers' perceptions of the resources and support available to perform a behavior" ( [9], p. 159). In the context of the present research, it reflects the support from resources (e.g., ubiquitous Internet connection for smartphones) and the required knowledge (e.g., experience of smartphone use) to be able to use fitness apps. The original UTAUT2 study [9] as well as studies considering the acceptance of general apps [20] and fitness wearables [18] showed that facilitating conditions increases acceptance. Thus, we postulate that: H4: Facilitating conditions relate positively with individuals' behavioral intentions of using fitness apps.

Price value
Price value is defined as "consumers' cognitive trade-off between the perceived benefits of a technology and the monetary cost of using it" ( [9], p. 161). Individuals expect a higher quality of services when they have to pay more for them [30,38]. In the fitness app context, providers offer three main patterns of pricing: free, paid, or freemium (i.e., free base app use, but additional features need to be paid for). Even if an app can be used for free, individuals might nevertheless consider other cost aspects, such as personal time costs or psychological costs. Previous studies have found a positive relationship between price value considerations and behavioral intentions of using the mobile Internet [9], healthcare wearables [39], and fitness apps [19,30]. Because high value for a given price can be assumed to be perceived positively by individuals, we propose that: H5: Price value relates positively with individuals' behavioral intentions of using fitness apps.

Hedonic motivation
Hedonic motivation refers to "the fun or pleasure derived from using a technology" ( [9], p. 161). If individuals' intrinsic motivation is high, they typically have high levels of hedonic motivation [40]. A meta-analysis revealed that 53 (equivalent to 58% of the included studies) UTAUT2-related empirical studies included hedonic motivation as a factor, whereas 43 out of these 53 studies found a positive relationship between hedonic motivation and behavioral intentions of using the technology [13]. Hedonic motivation has a positive effect on intentions of adopting healthcare wearables [18,21] and fitness apps [30]. Thus, we suggest that if a user has fun using a fitness app, they will be more likely to use it. H6 is stated as follows: H6: Hedonic motivation relates positively with individuals' behavioral intentions of using fitness apps.

Habit
Habit refers to "the extent to which people tend to perform behavior automatically", and was found to be a positive predictor of behavioral intentions of using the mobile Internet ( [9], p. 161).
About 35% of UTAUT2-related empirical studies utilized habit as a construct [12]. Most importantly, 15 out of the 18 studies revealed positive associations between habit and intentions [12]. In the context of this study, we consider habit to be an important predictor, because smartphones are a central means by which individuals can manage and facilitate their daily lives [2], and because individuals use their smartphone (and potentially fitness apps [19,30]) by habit. We thus propose that: H7: Habit relates positively with individuals' behavioral intentions of using fitness apps.

Downstream consequence of behavioral intentions of using fitness apps
Fitness apps aim to promote users' fitness levels. Since we can assume that people who download these apps are (at least partly) committed to reaching this goal, we postulate that higher intentions of using fitness apps relate positively with people's willingness to be physically active in the future. The claim can be substantiated by consistency theories, arguing that cognitive consistency fosters updates on the expectancy regarding an outcome or a state (here: to be physically active) [e.g., [41]]. To date, however, none of the UTAUT2-based studies have examined the relationship between usage intentions of new technology that aims to promote fitness (or health) and the downstream consequence on behavioral intentions of engaging in physical activity-related behaviors.
Two recent systematic reviews concluded that the effects of fitness apps on physical activity levels are present, but modest in magnitude [5,16]. Previously formed intentions at the individual level might be an explanatory variable for these effects. Thus, H9 is stated as follows: H8: Behavioral intentions of using fitness apps relate positively with behavioral intentions of being physically active.

Moderating effects of fitness app features
Smartphone apps have certain features, that is, the set of operational functions that an app is able to perform (e.g., gaming). The essence of fitness app features may be summarized within the socalled behavior change techniques (e.g., goal setting, monitoring, and acquisition of knowledge) [42]. Besides, various frameworks of features that are implemented in fitness apps have been proposed. For example, Mollee et al. [43] identified user input, textual/numerical overviews, social sharing, and general instructions as the most implemented features of fitness apps. Rabin & Bock [44] suggested that fitness tracking, the tracking of progress toward fitness goals, and the integration of features that increase enjoyment (e.g., music), are user-desired features. Other studies focused on the social features of fitness apps (e.g., sharing or comparing steps and receiving social support) [45], while a review concluded that the evidence of social app features to promote fitness was limited [36].
Conroy et al. [17] took an empirical approach to cluster fitness apps in terms of features and used cluster analysis to identify two broad categories: motivation-related and education-related features. Motivation-related app features emphasize the social-and self-regulation of fitness (e.g., tracking, feedback, social support, goal setting, and reward features). Education-related app features focus on fitness education (e.g., instructions, coaching, learning) [17]. These two clusters do not include gamification-related features, which have become relevant to help individuals improve their health and fitness [46]. Gamification-related features use game design elements to make the user experience playful and enjoyable [47,49]. In the current study, we thus consider gamification-related features besides motivation-and education-related features of fitness apps.
The literature on apps in general (without a focus on physical activity) has taken into account app features as moderators of the relationship between acceptance determinants and behavioral intentions of using apps [31,49]. Yet, it remains unclear whether the UTAUT2 determinants interact with fitness app features to explain behavioral intentions of using these apps. Such interaction effects might explain the rather modest effects found in systematic reviews on the effects of fitness apps on physical activity [5,16]. To explore this issue, we formulate the following research question: Do fitness app features moderate the relationships between the UTAUT2 determinants and behavioral intentions of using fitness apps?

Moderating effects of individual differences
The moderating effects of age, gender, and experience -so-called individual difference variables -on the relationships between UTAUT2 determinants and behavioral intentions have been proposed and empirically tested in the original UTAUT2 study [9]. In particular, it was theorized that age moderated the relationships between the seven UTAUT2 determinants and behavioral intentions such that the effects are stronger among young (vs. old) users for performance expectancy, effort expectancy, and hedonic motivation, but weaker for social influence, facilitating conditions, price value, and habit [8,9]. Gender was postulated to moderate the relationship between the seven UTAUT2 determinants and behavioral intentions such that the effects are stronger among females (vs. males) for effort expectancy, social influence, facilitating conditions, and price value, but weaker for performance expectancy, hedonic motivation, and habit [8,9]. Experience was postulated to moderate the relationships between five UTAUT2 determinants and behavioral intentions such that the effects are stronger among users in the early (vs. late) stage of experience for effort expectancy, social influence, facilitating conditions, and hedonic motivation, but weaker for habit [8,9]. Threeand four-way interactions of age, gender, and experience were included in the original UTAUT2 study, too [9]. Despite the fact that the original studies supported these proposed moderator relationships, previous studies on mobile health and fitness apps applying the UTAUT or UTAUT2 did not fully take them into account (e.g., [15, 18-21, 50, 51]). The moderators have been metaanalyzed and suggested as worth studying [22] or noted as future work [19]. To partly fill this research gap, we state the following research question: Are there individual differences in the relationships between the UTAUT2 determinants and intentions of using fitness apps?

Study Design and Procedure
This study applied a cross-sectional online survey design and the results were reported according to the CHERRIES checklist [52]. Using a convenience sampling technique, we recruited 867 Amazon Mechanical Turk (MTurk) workers in March 2020. This sample size was considered sufficient based on a rule-of-thumb [53], as well as similar studies on fitness app acceptance [19,30].
Participants were limited to healthy adults who were between 18 and 65 years old, owned a smartphone, and had downloaded at least one smartphone fitness app. Participants were also required to be able to read and understand English, and be located in the U.S. (i.e., U.S. residents).
Participants who met the eligibility criteria were invited to participate in the MTurk online survey,

Measures
The UTAUT2 items for the seven determinants and behavioral intentions of using apps were adapted to the context of the present study [9] (see Table 2). They were measured on a seven-point rating scale ranging from 1 (strongly disagree) to 7 (strongly agree). Behavioral intentions of being physically active were gauged by two separate measures. First, intentions were measured via an adaptation of the International Physical Activity Questionnaire Short Form (IPAQ-SF) [54], which covers a time span of four weeks into the future. The sum of the values (measured in MET-min/week) was calculated following established data processing guidelines [55]. Second, it was measured using a single question: "To what degree do you want to be physically active in the next four weeks?" (1 = Not at all, 7 = Very much) [56]. Individual difference variables of age and gender were self-reported. Experience was measured with a single item: "When did you download a fitness app for the first time? -() months ago", as done in the original UTAUT2 study [9].
Participants also rated the features of their most preferred app with importance ratings (1 = not important at all, 7 = extremely important). Importance ratings were used, because apps typically have multiple features and because the user's perspective of the features is important in the present study (e.g., [57]). The items for education-and motivation-related app features were formulated in agreement with previous cluster classifications [17] and substantive content of behavior change techniques [42]. Gamification-related app features were operationalized based on extant literature in gamification and fitness apps [47,48]. All three app features were measured via three items each.
Examples of items are: "How important to you are app features that motivate you to be physically active?", for motivation-related features; "How important to you are app features that educate yourself about how to exercise best?", for education-related features; and "How important to you are app features to enjoy yourself while exercising?", for gamification-related features (see Table 2).

Statistical Analyses
Normality was evaluated using multivariate skewness and kurtosis [58]. We conducted a confirmatory factor analysis to evaluate the internal reliability, convergent validity, and discriminant validity of the measurement model [59]. For internal reliability, we examined Cronbach's alpha (α > .70) and construct reliability (CR > .70). We employed the average variance extracted (AVE, > .50) and factor loadings for convergent validity [60]. Path modeling (maximum likelihood) was used to test the hypotheses. The variables were mean-centered prior to the analysis, and gender was coded as a dummy variable (0 = female, 1 = male). For significant interaction effects between the UTAUT2 determinants and app features, follow-up tests were performed to observe how the moderator changes the hypothesized relationships, as recommended by Dawson [62]. The data analyses were performed with R (RStudio, Boston, MA, USA), and the lavaan package [63]. The level of significance was set at P < .05 (twotailed). [Insert Table 1]  [Insert Table 2]

Structural Model and Hypotheses Testing
We used path modeling to test the hypotheses. The model was established by modeling the hypothesized paths among the UTAUT2 determinants, behavioral intentions of using fitness apps, intentions of being physically active, as well as the three app features (see Figure 1). supported, while H3 and H5 were not (Table 4, Figure 2).
[Insert Table 4] The testing of the interaction effects of app features and the seven UTAUT2 determinants was performed next. Education-related app features moderated the relationships between performance expectancy and behavioral intentions of using fitness apps (β = -.08, SE = .03, P = .014) as well as between habit and behavioral intentions of using fitness apps (β = .08, SE = .03, P = .009). The testing of the interaction effects of individual differences and the seven UTAUT2 determinants also revealed that age moderated the relationship between effort expectancy and behavioral intentions of using fitness apps (β = -.11, SE = .04, P = .008). Gender moderated the relationships between performance expectancy and behavioral intentions of using fitness apps (β = .13, SE = .06, P = .03) as well as habit and behavioral intentions (β = -.12, SE = .05, P = .02).

Motivation
Experience was a non-significant moderator. Additionally, the joint moderating tests (three-and fourway effects) taking into account individual differences revealed a significant three-way interaction for age, gender, and hedonic motivation (β = -.14, SE = .06, P = .02); a significant three-way interaction for age, experience, and effort expectancy (β = .09, SE = .03, P = .007); and a significant three-way interaction of age, experience, and habit on behavioral intentions of using fitness apps (β = -.12, SE = .04, P = .004). There were no significant four-way interaction effects.
Subsequently, we conducted follow-up tests to describe how the moderators change the relationships (Table 5), considering low (-1 SD) and high (+1 SD of the mean) values of the moderators. First, when education-related features were rated important, the relationship between performance expectancy and usage intentions was weaker compared to when this feature was rated unimportant. Second, when education-related features were rated important, the relationship between habit and usage intentions was stronger compared to when these features were rated unimportant.
Third, when motivation-related features were rated important, the relationship between performance expectancy and usage intentions was stronger, the relationship between facilitating conditions and usage intentions became non-significant, and the relationship between habit and usage intentions was weaker compared to when these features were rated unimportant. Fourth, when gamification-related features were rated important, the relationship between hedonic motivation and usage intentions was stronger but still non-significant compared to when this feature was rated unimportant. Furthermore, the relationship between effort expectancy and usage intentions was positive for younger, but nonsignificant for older users. Lastly, the relationship between performance expectancy and usage intentions was stronger among males, while the relationship between habit and usage intentions was stronger among females.

DISCUSSION
The purpose of the study was to examine the influence of the UTAUT2 determinants, as well as the moderating effects of different smartphone fitness app features (i.e., education-, motivation-, and gamification-related) and individual differences (i.e., age, gender, and experience) on individuals' app usage intentions and their behavioral intentions of being physically active. The results showed that habit and performance expectancy were the two strongest predictors of individuals' intentions of using fitness apps. The effects of performance expectancy were greater when motivation-related features were rated important and when education-related features were rated less important, as well as for males; and the effects of habit were greater when educationrelated features were rated important and when motivation-related features were rated less important, as well for females. Age moderated the relationship between effort expectancy and app usage intentions. Individuals' intentions of using fitness apps predicted their intentions of being physically active, using two different means of measuring future physical activity.

Theoretical Contribution
We contribute to the mobile health and physical activity literature in several ways. Answering the first research question -What are the relationships between the UTAUT2 determinants and intentions of using smartphone fitness apps? -, we found positive relationships between habit, performance expectancy, facilitating conditions, price value, and effort expectancy and behavioral intentions of using fitness apps. Habit and performance expectancy were found to be the most important predictors of intentions of using fitness apps, consistent with prior studies (e.g., habit: [19,20,30]; performance expectancy: [15,30,50]). Positive relationships have also been identified for effort expectancy [18][19][20], facilitating conditions [18,20,21], and price value [19,21,30].
Social influence was a non-significant predictor of intentions [18,20,30]. Interestingly, the latter finding is not due to the high domain-specific experience of users (given the non-significant interaction effect of social influence and experience), who might have relied less on peer opinions for their evaluations and intentions than low-experience users. Furthermore, in contrast to the original UTAUT2 study [9] and prior studies [18,20,21,30], but in agreement with Dhiman et al. [19], we found a non-significant relationship between hedonic motivation and app usage intentions. This may be explained by the fitness app users' high demands on app usage to achieve their physical activity goals, compared to the fun or pleasure derived from the apps. Still, focusing solely on the four determinants proposed by the first version of UTAUT, as realized by [14,15,34], may be insufficient. Habit, in particular, is the strongest determinant that was linked to intentions of using fitness apps in the present study.
Answering the second research question -What is the downstream relationship between behavioral intentions of using fitness apps and intentions of being physically active? -, we contribute to UTAUT2-based research by showing that app usage intentions have important downstream consequences. In particular, individuals have greater intentions of being physically active when they have higher intentions of using the fitness apps. Assessing the downstream effect of intentions of using fitness apps is important, because downloaded but unused apps or apps unable to motivate people to become or remain physically active will have little health effects [5,16]. The positive relationship between fitness app usage intentions and physical activity intentions indicates that app usage might indeed motivate people to become or remain active. The findings thus contribute to previous research into whether, and when, mobile health and fitness apps may help individuals become physically active [65,66]. However, it should be noted that individuals' intentions of being physically active are affected by numerous correlates and determinants (e.g., self-efficacy, sociodemographic variables, sport club membership, among others) [67], and the intention-behavior gap is considerable [68]. Thus, adding these factors and incorporating the measurement of actual physical activity may be warranted in the future.
Answering the third research question -Do fitness app features moderate the relationships between the UTAUT2 determinants and intentions of using fitness apps? -, the present study contributes to previous research that categorized app features [17], yet ignored their influence on the structural relationships proposed by the UTAUT2. Based on our exploratory analysis, we found six relevant interaction effects. One of the most intuitive findings was that, when motivation-related features were rated important, the relationship between performance expectancy and intentions was strong. Research into goal achievement [69,70] might explain the interaction effect: individuals who are interested in improving their physical activity levels, or keeping them at certain levels, might use the app exactly for this purpose. Among the three features, motivational elements aim most directly to help users stick to their goals and plans [71]; since there is goal congruence, the effect is strong [72]. When motivation-related features were rated important, the relationship between facilitating conditions and usage intentions was not significant. This makes sense, because people who lack resources and capacities are more dependent on help from others compared to people who do have these resources and capacities, particularly when motivation features are not considered to be crucial (i.e., motivation might 'not be the problem'). Also, when motivation-related features were important, the relationship between habit and intentions was weaker compared to when this feature was unimportant. This finding might indicate that when habits have been formed, features that motivate individuals to be active (e.g., reminders) become less important to these app users [73].
The present study also found that performance expectancy had a greater effect on usage intentions when education-related features were rated unimportant. In this case, individuals might be less interested in being educated -an aspect that might make them feel distracted from achieving their goals. Further, the effect of habit on usage intentions was stronger when education-related features were rated important. This may be explained by the fact that individuals' habits are formed best when they are exposed to education-related cues when using an app (e.g., how and when to exercise best) [74]. Regarding the interaction of hedonic motivation and gamification-related features, no final conclusions can be drawn. While research into intrinsic motivation [75] and flow [76] may lead us to propose that intrinsic motivation, as a principal source of enjoyment, may be enhanced by the gamification app features (e.g., apps using incommensurate gamification elements [likes]) [77], the follow-up tests did not reach significant levels in the present study.
Answering the fourth research question -Are there individual differences regarding age, gender, and user experience between the relationships of the UTAUT2 determinants and intentions of using fitness apps? -the present study found partly significant, partly non-significant moderating effects of age, gender, and experience. First, the relationship between effort expectancy and app usage intentions was stronger among younger individuals, which agrees with the original UTAUT2 study [8,9] and a meta-analysis (i.e., age group of 25 to 30-year-olds) [22]. Second, the relationship between performance expectancy and usage intentions was stronger among males, which is consistent with the original UTAUT2 study. In contrast, the relationship between habit and usage intentions was stronger among females [9]. Thus, females were not more sensitive to new cues, which might have weakened the effect of habit on behavioral intentions. In the context of fitness apps, females may indeed be prone to cues that help them form health-related habits, because they are interested in health-and body appearance-related topics. Lastly, in the present study, experience was a non-significant moderator as regards the interaction effects of the UTAUT2 determinants on app usage intentions. Thus, differences in experiences between users might be less relevant today -a time in which smartphone users can easily add and delete new apps and in which users are technology-savvy.

Managerial Implications
This study provides implications for smartphone app designers and managers. First, they can be advised to focus on habit formation and performance (e.g., goal-setting) when designing fitness apps and tailoring them to potential users. Meeting users' expectations concerning facilitating conditions, price value, and effort expectancy will also increase the likelihood of the app being accepted. Second, practitioners should highlight certain app features depending on user preferences.
For example, motivation-related features are important drivers of app usage intentions for those target group users who value performance (education-related features might be less relevant here); habit formation and facilitating conditions are less important to these individuals. Third, health professionals should consider age and gender differences among users with regard to the effects of effort expectancy (age) as well as performance expectancy and habit (gender). Lastly, practitioners may also be advised to monitor whether app usage intentions have a positive correlation to intentions of, or even actual, physical activities so that immediate action can be taken when users lose track of their original goals (having already downloaded the app).

Limitations and Outlook
This study has some limitations. First, the generalizability of our findings is limited. We used a non-representative sample of U.S. residents who owned a smartphone and have used fitness apps before. Future studies may consider rather inexperienced people with fitness apps to reveal the influence of UTAUT2 determinants on usage intentions at the early-or pre-adoption stage. Second, given the present research design, we did not consider one specific fitness app, but participants stated their preferred app, and rated the features of this app. By these means, we considered a variety of apps (which might be beneficial for external validity, given the myriad of apps on the market [3,4]).
Researchers might collaborate with certain providers and use real-world app data, and objectively measure actual physical activity to validate our findings. Third, we relied on self-reported physical activity intentions using a single-measure and the IPAQ-SF. For the latter, overreporting is common (e.g., around 84% [78]). Lastly, future research could look into the mechanisms of moderation effects on individuals' behavioral intentions of using apps, incorporate app features into mobile health interventions accordingly, and evaluate their long-term influence on physical activity levels.    Notes. *** P < .001; ** P < .01; * P < .05. β: Unstandardized path coefficient; SE: Standard Error.   Note. In agreement with the original UTAUT2 study [9], experience was postulated to not moderate the relationships between performance expectancy as well as price value and behavioral intentions of using fitness apps.
40 Figure 2. Path modeling results on the relationship between the UTAUT2 determinants and behavioral intentions of using fitness apps as well as the downstream effects on intentions of being physically active.
Note. *** P < .001; ** P < .01; * P < .05. The figure shows the main effects of the seven UTAUT2 determinants. See Table 4 for interaction effects with app features, age, gender, and experience.

Figures
Path modeling results on the relationship between the UTAUT2 determinants and behavioral intentions of using fitness apps as well as the downstream effects on intentions of being physically active. Note. *** P < .001; ** P < .01; * P < .05. The figure shows the main effects of the seven UTAUT2 determinants. See Table 4 for interaction effects with app features, age, gender, and experience.