This is an open-access article distributed under the terms of the Creative Commons Attribution License (https://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided the original work, first published in JMIR mHealth and uHealth, is properly cited. The complete bibliographic information, a link to the original publication on https://mhealth.jmir.org/, as well as this copyright and license information must be included.
Mobile health (mHealth) care apps are a promising technology to monitor and control health individually and cost-effectively with a technology that is widely used, affordable, and ubiquitous in many people’s lives. Download statistics show that
We aimed to compare the factors influencing the acceptance of lifestyle and therapy apps to better understand what drives and hinders the use of mHealth apps.
We applied the established unified theory of acceptance and use of technology 2 (UTAUT2) technology acceptance model to evaluate mHealth apps via an online questionnaire with 707 German participants. Moreover, trust and privacy concerns were added to the model and, in a between-subject study design, the influence of these predictors on behavioral intention to use apps was compared between lifestyle and therapy apps.
The results show that the model only weakly predicted the intention to use mHealth apps (
The results indicate that, rather than by utilitarian factors like usefulness, mHealth app acceptance is influenced by emotional factors like hedonic motivation and partly by habit, social influence, and trust. Overall, the findings give evidence that for the health care context, new and extended acceptance models need to be developed with an integration of user diversity, especially individuals’ prior experience with apps and mHealth.
Due to their affordability and ubiquity in people’s everyday lives [
mHealth apps encompass a variety of health-related services (eg, support of diagnostics and treatment), tracking of infection processes (eg, contact tracing during the COVID-19 pandemic), remote monitoring, and medicine intake reminders [
Users’ technology acceptance is one decisive factor for the adoption and widespread use of technologies, including mHealth apps, but it can also be a barrier if the diverse requirements of the potential users are not understood [
This research aims to improve the understanding of users’ decisions to use, or reject the use of, mHealth apps. To do so, we have built on and extended the established and widely used UTAUT2 acceptance model. As mHealth apps are essentially based on the collection and analysis—and often also transmission—of user data, privacy concerns can be a reason for rejection. Medical data are perceived as very sensitive and, thus, even more reluctantly disclosed [
Moreover, we tested for differences in acceptance patterns between two types of mHealth apps (research question 1): currently, therapy apps targeted at existing illnesses and ailments are far less often used than lifestyle apps that, for example, should improve fitness and prevent health problems [
Another important research duty regarding the acceptance of mHealth apps is the integration of effects of user diversity on technology acceptance. While demographic factors, such as age, have already been a focus of research (eg, Deng et al [
Proposed research model. H: hypothesis; RQ: research question; UTAUT2: unified theory of acceptance and use of technology 2.
Our research provides new insights into the individual and context-specific acceptance patterns for mHealth apps. This is important for revealing drivers to build acceptance as well as for identifying barriers that need to be reduced. User acceptance is one of the keys to a successful mHealth app rollout and to harnessing the full potential of mHealth for health care systems and for the improvement of quality of life and therapy for patients.
With the ever-increasing use of technology, the acceptance of new devices and software has advanced as an important focus of research. To understand the future use of new implementations, it is necessary to understand what factors influence human behavior. Based on psychological theories (eg, the theory of reasoned action [
The most recent extension in the line of technology acceptance models is the UTAUT2 [
Performance expectancy describes the perception that using the technology will provide benefits to the user and is, thus, tied to the perception of usefulness [
Based on this body of research, hypothesis 1 is as follows: Performance expectancy influences the intention to use mHealth apps.
Effort expectancy describes the expected ease of using the technology [
In this study, we, therefore, again examined this relationship using hypothesis 2: Effort expectancy influences the intention to use mHealth apps.
Social influence is “the degree to which an individual perceives that important others believe he or she should use the new system” [
Therefore, hypothesis 3 is as follows: Social influence affects the intention to use mHealth apps.
Facilitating conditions refer to the perceptions “of the resources and support available to perform a behavior” [
In this regard, we put forward hypothesis 4: Facilitating conditions influence the intention to use mHealth apps.
Hedonic motivation is the gratification counterpart to the utilitarian measure of usefulness represented by performance expectancy. Hedonic motivation refers to fun, pleasure, and enjoyment with the use of technology [
Considering these mixed results, we examined the following relationship as hypothesis 5: Hedonic motivation influences the intention to use mHealth apps.
Habit is operationalized within the UTAUT2 framework as a self-reported perception of a customary use of the respective technology [
Therefore, in this study, we only assessed habit for current users of the mHealth apps in question and proposed hypothesis 6: Habit influences the intention to use mHealth apps of current users of mHealth apps.
In the UTAUT2, a seventh predictor is the price value. Despite its significant effect on use intention as shown in some UTAUT2 studies [
As shown, a multitude of studies applied the UTAUT2 acceptance model in diverse application contexts. However, many researchers also extended and adapted the model to better fit the needs of the specific context of research [
One major barrier to the use of digital and connected technologies is privacy concerns [
Privacy can be defined as users’ rights to control the flow of personal information [
A large body of research shows that privacy concerns negatively influence users’ intention to provide information [
Based on these empirical results, we proposed hypothesis 7: Privacy concerns influence the intention to use mHealth apps.
Another important factor for the acceptance of information technologies and mHealth is trust [
In line with research on privacy concerns and acceptance, it could be shown that technology acceptance models should be extended by trust as an influencing factor, for example, regarding travel apps [
This leads to our last hypothesis, hypothesis 8: Trust in mHealth apps influences the intention to use mHealth apps.
The available spectrum of mHealth apps is very broad. Apps related to a healthy lifestyle (eg, fitness, diet, and sleep-monitoring apps)—further on called
Even though the market for lifestyle apps is twice as big as for health care apps and therapeutically orientated apps [
Using apps for a general healthy lifestyle may be influenced by different motives and barriers than using apps for therapy for existing illnesses. Initial empirical evidence shows that acceptance patterns differ between different contexts of digital health technologies [
For these reasons, we pose research question 1: How does the influence of the proposed factors on use intention differ between lifestyle and therapy apps?
People are diverse and so are their evaluation and acceptance of technologies. Besides the highly individual perceptions of, for example, performance, privacy, or influences through peers and habit, user acceptance varies depending on users’ characteristics. In general, current users of mHealth apps are rather young, female, and highly educated [
However, sociodemographic factors, specifically age, may just be carrier variables for the underlying reasons and user characteristics. The adequate know-how of handling mHealth apps, as well as the fit between needs and target groups, are factors that might have an impact on use intention [
Therefore, we pose research question 2: How do diverse user characteristics like sociodemographics, experience and literacy with mHealth apps, as well as personal dispositions relate to the acceptance of mHealth apps?
We used an online questionnaire with a between-subject design to assess the opinions and acceptance by the participants of either lifestyle or therapy apps and to evaluate our hypotheses and research questions.
In the introduction section of the questionnaire, a brief orientation on the topic of the study was given. The respondents were also reminded of their rights and informed on how the collected data would be dealt with. We encouraged them to answer freely, as there were neither “correct” nor “incorrect” answers, and we let them know that we were only interested in their perspective on this timely topic. Respondents were also informed that participation was voluntary and that they were free to quit at any time. Before starting the questionnaire, participants gave consent to collection of their data.
In the next part of the questionnaire—cataloguing the participants’ characteristics, attitudes, and experiences—their state of health was first assessed. Besides their subjective health status, questions regarding their experiences with different types of health problems were asked (ie, back and joint pain, headaches and migraine, cardiovascular diseases, allergies and food intolerances, metabolic illnesses, dementia, and a non-option). To measure their experiences with apps, the participants indicated their experience with and use of apps in general and with mHealth apps. For the use of digital health apps in particular, users require skills to search, select, appraise, and apply online health information. Therefore, digital health literacy was assessed in the next part of the questionnaire using the instrument by Van Der Vaart and Drossaert [
Constructs used in the questionnaire with their respective sources.
Constructs | Subconstructs | Source upon which the construct was based |
UTAUT2a constructs |
Performance expectancy Effort expectancy Social influence Facilitating conditions Hedonic motivation Habit (only answered by users) Behavioral intention (for users) Behavioral intention (for nonusers)b |
Venkatesh et al [ |
Perceived trust | N/Ac | Körber [ |
Information privacy concerns |
Perceived surveillance Perceived intrusion Secondary use of personal information |
Xu et al [ |
Digital health literacy |
Operational skills Navigation skills Information searching Evaluating reliability Determining relevance Adding self-generated content Protecting privacy |
Van Der Vaart and Drossaert [ |
Disposition to value privacy | N/A | Xu et al [ |
Propensity to trust (adapted to apps) | N/A | Körber [ |
aUTAUT2: unified theory of acceptance and use of technology 2.
bThis construct was adapted from Venkatesh et al [
cN/A: not applicable; the construct in this row did not have any subconstructs.
In the main part of the questionnaire, participants were randomly assigned to evaluate either lifestyle apps or therapy apps. The evaluation started with introducing the respective mHealth apps. We assessed performance expectancy, effort expectancy, social influence, facilitating conditions, hedonic motivation, habit, and behavioral intention using the items by Venkatesh et al [
To assess privacy perceptions, the Mobile Users’ Concerns for Information Privacy [
All items were assessed on 6-point symmetric Likert scales ranging from 1 (low agreement) to 6 (strong agreement). Items were randomized to prevent biases. The language of the questionnaire was German, as only German participants were recruited; therefore, items were translated into German. For trust, no validated German translation was available. Therefore, the items were forward-translated by a German native speaker and, to test the comprehensibility and correct translations, two authors translated these again back into English. The results were compared to the original items, and deviations were settled in a discussion. During the translation process, we also adapted the items to the mHealth context and discussed these adaptations.
To assure a realistic and empathetic evaluation of mHealth apps, the formulation of the items was, on the one hand, individually adapted to the participants already being users of such apps or nonusers. For example, one item for behavioral intention was modified in the following way: “I intend to continue using such a [medical or lifestyle] app” for users versus “I intend to use such a [medical or lifestyle] app in the future” for nonusers. Also, only current and prior users were asked to answer questions concerning habit. Moreover, as therapy apps are less widespread and, therefore, less familiar, the description of therapy apps was illustrated by an example of apps that matched the participants’ experienced health problems, as in the questionnaire section about state of health. For those participants who had not experienced any of these health problems before, a general description with several examples was used.
Before distributing the study, we pretested the questionnaire and participants reported back on comprehensibility issues. Only after those issues had been eliminated did we start the data acquisition.
Participants were recruited from a university seminar and its attendees’ social contacts. Participants accessed the questionnaire via a weblink that was given to them. The comparison between the app types used a between-subject design. Thus, each participant either answered the items regarding therapy apps or lifestyle apps. The app type to be evaluated was assigned randomly. The participants volunteered to take part in the study and were not rewarded for their efforts. Data were collected in May and June of 2019.
The recruitment method was chosen with the aim to reach mHealth users as well as nonusers of therapy and lifestyle apps. Additionally, participants of different age groups were recruited. However, in accordance with the technical requirements of mHealth use, only participants with access to the internet and digital devices were targeted. Today, young people, in particular, use mHealth apps [
The following sections will detail the analysis methods as well as regulations we applied to our data.
We checked reliability by using Cronbach α and applied a threshold of a>.70 for all scales not included in the structural model (ie, disposition to value privacy, propensity to trust apps, and digital health literacy). Additionally, as some of the translated German items were not validated previously, we conducted an exploratory factor analysis on the model constructs, confirming the validity of the items (
Our research model was tested using partial least squares (PLS) structural equation modeling (SEM). PLS is a component-based SEM method that is suitable for exploratively testing new models [
The software SmartPLS (version 3.3) was used for the SEM modeling [
As we tailored the questionnaire distinctly, using the targeted app examples, we had to ensure that this did not introduce a systematic error between different apps. We first checked whether the correlations between behavioral intention and the predictor variables differed significantly between the eight app examples used. No such differences were prevalent so that, in the final analysis, no differentiation was made between the participants evaluating therapy apps with different health or ailment foci. For all analyses, a significance level of 5% was set.
To describe how demographics and other user characteristics might be associated with our model variables, we used correlation analysis. To deal with suboptimal normality of our data, we used bias-corrected and accelerated bootstrapping [
Of 951 people who started the questionnaire, 799 completed it (84.0% completion rate). Further, 92 participants with a response time shorter than 50% of the median response time (<16 minutes, 17 seconds) were labeled as speeders and excluded. Finally, 707 participants were included in the analysis.
Access to the anonymized data set can be requested on the Open Science Framework repository [
The demographic characteristics of the sample are depicted in
Demographic characteristics of the sample comparing participants evaluating lifestyle apps and therapy apps (N=707).
Characteristic | Participants evaluating lifestyle apps (n=355) | Participants evaluating therapy apps (n=352) | |||
Age (years), mean (SD) | 36.4 (18.1) | 37.3 (16.8) | |||
|
|
|
|||
|
Women | 222 (62.5) | 206 (58.5) | ||
|
Men | 133 (37.5) | 146 (41.5) | ||
|
|
|
|||
|
No certificate | 6 (1.7) | 9 (2.6) | ||
|
Certificate of secondary education | 25 (7.0) | 25 (7.1) | ||
|
General certificate of secondary education | 59 (16.6) | 63 (17.9) | ||
|
General qualification for university entrance | 262 (73.8) | 255 (72.4) |
Most of the 707 participants used digital technologies: 695 participants (98.3%) owned a smartphone. Only 33 participants (4.7%) did not use apps regularly. Correspondingly, the participants’ self-rated app familiarity was quite high (mean 4.32, SD 1.36), as rated on a scale from 1 (very low agreement) to 6 (very high agreement). In contrast, the familiarity with health apps was lower (mean 3.44, SD 2.56). Of the 355 participants evaluating the lifestyle mHealth apps, 110 (31.0%) were current users and 82 (23.1%) had used a lifestyle app before. Only 18 (5.1%) of the 352 participants assigned to the therapy app evaluation group were current users, and 18 (5.1%) had used a therapy app before.
Disclosing information about health status was optional in order to not be too invasive regarding the participants’ privacy. Most of the 707 participants reported their health status as “good” (n=305, 43.1%), “very good” (n=216, 30.6%), or “excellent” (n=65, 9.2%). Out of 707 participants, 5 (0.7%) reported their health status as “very bad,” 25 (3.5%) reported it as “bad,” and 83 (11.7%) reported it as “rather bad.” Out of 707 participants, 8 (1.1%) chose not to answer. Out of 707 participants, 26.9% (n=190) lived with a chronic illness, 11.9% (n=84) depended on a medical assistive device, and 31.4% (n=222) needed regular checkups with their physician.
On average, the sample showed a neutral propensity to trust apps in general (mean 3.07, SD 0.79) and a slightly stronger than neutral disposition to value privacy (mean 3.88, SD 1.14). The mean digital health literacy was quite high (mean 4.53, SD 0.75).
To assess the quality of the measurement model, the guideline by Hair et al [
Evaluation of the validity included convergent validity (average variance extracted >0.5) and discriminant validity, using the Fornell-Larcker criterion. Mobile users’ information privacy concerns were modeled as higher-order models because the latent factor privacy concerns was based on three subdimensions. Therefore, validity criteria did not apply to the discriminant validity between the subdimensions themselves or between the subdimensions and the overall scale privacy concerns.
The resulting path coefficients of the model for both lifestyle and therapy apps are depicted in
The structural model with path coefficients juxtaposed for lifestyle and therapy apps (significance based on bootstrapping; lifestyle: n=355; therapy: n=352). PLS-SEM: partial least squares structural equation modeling; adj: adjusted.
The results revealed that only 19% of the variance in behavioral intention for both types of mHealth apps could be explained by the extended UTAUT2 model. The variables correspondingly showed only weak predictive relevance for behavioral intentions (Q=0.119 for both app types). Most hypothesized variables showed no significant relationship to behavioral intention. Regarding both types of apps, neither the UTAUT2 constructs of performance expectancy, effort expectancy, and facilitating conditions nor privacy concerns predicted acceptance. Hedonic motivation was the only included construct that had a significant impact on behavioral intention for both app types (lifestyle: 0.196,
The MGA confirmed these differences between the evaluation pattern for the app types. Significant differences were present regarding the relationships of habit (D=0.264
Bias-corrected and accelerated bootstrapped 95% CIs for the evaluation of lifestyle and therapy apps and significance of the difference in path coefficients between the two app types based on multigroup analysis (MGA).
Relationship | Lifestyle apps (n=355), 95% CI | Therapy apps (n=352), 95% CI | Significance of MGA, |
|
|
||||
|
Performance expectancy | –0.069 to 0.185 | –0.002 to 0.227 | .57 |
|
Effort expectancy | –0.100 to 0.141 | –0.152 to 0.040 | .41 |
|
Facilitating conditions | –0.125 to 0.123 | –0.078 to –0.093 | .98 |
|
Social influence | –0.197 to 0.013 | 0.089 to 0.275 | <.001 |
|
Habit | 0.141 to 0.381 | –0.126 to 0.106 | .002 |
|
Hedonic motivation | 0.061 to 0.328 | 0.214 to 0.470 | .12 |
|
Trust | –0.027 to 0.214 | 0.146 to 0.399 | .04 |
|
Privacy concerns | –0.160 to 0.049 | –0.041 to 0.110 | .19 |
|
||||
|
Perceived surveillance | 0.870 to 0.951 | 0.850 to 0.911 | .38 |
|
Perceived intrusion | 0.885 to 0.951 | 0.924 to 0.959 | .38 |
|
Secondary use | 0.907 to 0.945 | 0.894 to 0.939 | .51 |
The validated UTAUT2 model including the additional constructs of privacy concerns and trust showed only weak predictive relevance in explaining why people use mHealth apps and why not. The UTAUT2 postulates that age, gender, and experience moderate the relationships of the predictor variables with behavioral intention [
These moderators were not included in our model, as we focused on the direct relationships. Additionally, other human factors have been shown to influence the acceptance of digital technologies and mHealth. Therefore, in an exploratory attempt to decipher how user diversity influences mHealth acceptance and usage, we calculated correlations to get first hints as to what may influence behavioral intention. These results shall not represent a detailed analysis but should give first insights into the impact of selected user characteristics on the acceptance of mHealth apps.
Pearson correlation coefficients of user factors with behavioral intention to use mobile health apps with bias-corrected and accelerated 95% CIs (N=707).
User factor | Correlation with behavioral intention, |
95% CI | |
Age | –0.150 | –0.255 to –0.034 | .004 |
Gender | –0.075 | –0.152 to 0.004 | .048 |
Education level | 0.088 | 0.001 to 0.171 | .02 |
App familiarity | 0.142 | 0.054 to 0.240 | .007 |
Health app familiarity | 0.469 | 0.379 to 0.548 | <.001 |
Digital health literacy | 0.215 | 0.119 to 0.313 | <.001 |
Privacy disposition | –0.194 | –0.299 to –0.083 | <.001 |
Propensity to trust apps | 0.191 | 0.88 to 0.291 | <.001 |
The objective of this study was to increase the understanding of users’ acceptance and decision patterns to use mHealth apps and which factors impact acceptance for lifestyle apps compared to therapy apps. Therefore, we applied the established technology acceptance model UTAUT2 [
In this study, the UTAUT2 model with its extensions can only explain a small amount of variance in the intention to use mHealth apps (approximately 20%). From the original validated UTAUT2 model, only the constructs hedonic motivation, habit, and social influence partly predict the intention to use mHealth apps. The constructs effort expectancy and performance expectancy, which are similarly modeled as main aspects in other established acceptance models like the technology acceptance model [
Also, in contrast to previous studies (eg, Guo et al [
Our results further suggest that instead of the more “utilitarian” aspects of perceived usefulness and performance of mHealth apps, it is rather the “emotional” aspects, such as fun, prior experiences, and recommendations by peers, that are important for their use. This finding must be confirmed and further analyzed in future studies, but it indicates that, on the one hand, approaches that address user experience, such as gamification, are important for mHealth apps of both types as the hedonic motivation influenced use intention for both app types. On the other hand, personal and peer experiences are very influential, whereby a widespread use of mHealth apps becomes even more important.
Besides the general model, our results revealed differences in the importance of some predictors for lifestyle and therapy apps. In our sample and in general, lifestyle apps were far more frequently used than therapy apps. The categorization of mHealth apps into lifestyle and therapy apps is not disjunct, as some apps may have functions providing both. In our study, the introduction given to the participants clearly distinguished between apps used to improve fitness, nutrition, and similar for “lifestyle” and those apps providing support for dealing with a prevailing illness. However, for future research, a classification of mHealth apps that is commonly agreed upon is vital, as is the simplification of research on context differences.
Habit emerged as a significant acceptance factor only for the lifestyle apps, which may be explained by the more widespread use and larger proportion of users in the sample. However, as the sample of users for therapy apps was very small (ie, only 36 participants), these results have to be interpreted with caution. The behavioral intention to use therapy apps was, in contrast to lifestyle apps, also influenced by social influence and trust. In this medical context, the participants need more than fun to use the app and, rather, should search for more reliable and trustworthy apps. Similar results have been found by Schomakers et al [
All in all, the predictive relevance of the factors in the extended UTAUT2 model is rather weak. This confirms other authors’ opinions regarding health care technologies, in that the established models can only be cautiously applied and need further adaptations, or rather, new models for the special health care context are needed (eg, Ziefle and Wilkowska [
The results of our exploratory analysis of the relationship between different user factors and behavioral intention to use mHealth apps imply that user diversity is an important aspect that needs to be considered. In particular, experience showed a strong relationship with use intention in this preliminary analysis, particularly the experience with health apps, but also with apps in general. The same was true for digital health literacy. Further empirical research and analysis of user diversity, especially the importance of experience, is needed, but this first result hints at a developing acceptance. When more and more people use mHealth apps, including therapy apps, the increased familiarity combined with habit and social influence may increase acceptance within the population. On the other hand, it cannot be assumed that everybody has the experience and the skills to use mHealth apps. Digital health literacy has to be developed and, following the gray digital divide, older people in particular, who are also more prone to chronic conditions, need support in getting to know these digital helpers.
Despite the valuable insights into decision patterns regarding mHealth apps, this online questionnaire approach needs to be considered methodologically. Instead of actual adoption behavior,
Besides the young and healthy persons still improving their health via lifestyle apps, thereby preventing chronic diseases, very important target groups for mHealth are older people and people with health problems. Here, mHealth can unfold its potential in directly supporting therapy and monitoring diseases, thereby improving quality of care and relieving the health care systems in a short time. These user groups should be further researched as they are underrepresented within our sample. Also, the German nationality of the participants limits the implications from this research, as attitudes toward technologies are highly influenced by cultures (eg, Trepte et al [
As no validated translation of the UTAUT2 items to German was known to us when planning the study, and no validated adaptation to the health care context was yet available, the use of unvalidated translations and adaptations of the items might have lowered the validity of our results. We could statistically assure a good validity and reliability of our items; nevertheless, the use of validated scales is highly recommended for future research (for a German translation of UTAUT2 items, see Harborth and Pape [
In this study, an extended UTAUT2 technology acceptance model was used to predict behavioral intention to use mHealth apps. Only a few hypothesized predictors (ie, hedonic motivation, habit, and social influence) showed a significant relationship to use intention, and the model only explained a comparably small amount of variance (approximately 20%). These factors indicate that more emotional factors than utilitarian usefulness influence mHealth app acceptance, adding a piece to the understanding of the mHealth acceptance puzzle. Small differences in the decision patterns were prevalent between the acceptance of lifestyle apps (eg, for fitness, nutrition, and sleep) and therapy apps (eg, for the monitoring and treatment of back pain, migraine, and cardiovascular diseases). In future research, the results need to be replicated, as the generalizability from our rather young sample is limited. However, our results in combination with previous research indicate that the UTAUT2 model, which was developed for the acceptance and use of mobile internet technologies in general, is not very suitable to predict mHealth use. The health care context needs improved and adapted technology acceptance models, which must also include human factors, such as experience, to account for user diversity.
Questionnaire items, sources, and translations.
Exploratory factor analysis.
multigroup analysis
mobile health
partial least squares
structural equation modeling
unified theory of acceptance and use of technology 2
We would like to thank all participants for openly sharing their opinions.
None declared.