With advances in health care assisting more people to live longer, the number of people aged over 60 years is projected to reach nearly 2.1 billion, representing 22% of the world’s total population [1]. In China, the proportion of the population aged 60 years and above will increase from 12.4% (168 million) in 2010 to 28% (402 million) in 2040 [2]. This demographic change further intensifies the conflict between the medical care demands of elders and the limited medical care resources. Indeed, mobile health (mHealth) has the potential to enable the elderly to experience longer and healthier lives by transforming health care services and clinical interventions for elderly users [3]. mHealth services are believed to be beneficial for elderly people because they can bring about multiple positive outcomes, such as health care costs savings, individually tailored health information and services, and a more effective health service process [4,5]. Therefore, mHealth is regarded as an important approach of extending the traditional (ie, offline) health services and satisfying the medical demands of an increasingly aging population.

However, older people are often found to be less innovative toward mHealth services [6]. There is limited usage of mHealth services to manage chronic diseases among the elderly [7]. In China, the development of mHealth services continues to remain at the infancy stage, and the adoption rate of such services among older people remains low [8]. The low adoption rate of mHealth services by the elderly has prompted the Chinese government to reconsider their developing strategies, to encompass the rapidly growing number of elderly users to alleviate the pressures of the aging population. In addition, although the elderly could obtain great potential benefits by using health information technology (HIT), previous research has not examined the relationship between HIT use among elderly users from their use of health services [9,10].

Due to their unique physical and psychological characteristics, elderly users may need to expend much more effort and time familiarizing themselves with information technology (IT) than the younger user groups [11]. In addition, declining physiological conditions [12] and limited social resources [13] are hurdles for elderly individuals in making health-related decisions independently. To serve the elderly population through mHealth services, service providers need to understand the behavior of this special group and the antecedents that influence their acceptance and usage of mHealth.

Building user trust is the key to promoting the adoption of mHealth services by elderly users. Trust could alleviate uncertainties and risks when users encounter new IT or services (eg, purchasing books on the Web and mobile payment services) [14,15]. In this context, trust transfer is a means of building customer trust in an unknown target through a trusted party [16]. Customer trust can be developed through trust transference from offline health care services, which are trusted by the elderly, to mHealth services with which they are not familiar. Although trust plays an important role in mHealth adoption, very few previous studies have investigated the development of trust in mHealth services [17,18]. Furthermore, as mHealth services are developed based on the traditional service in a new context, that is, a mobile context, users’ perceptions on mHealth may be influenced by the offline traditional services, such as trust. However, research on the trust transfer from an offline context to a mobile channel is also underexplored in both the information system (IS) and health care fields. In addition, although considerable research on trust transfer has been conducted in the IS domain, inadequate research has been undertaken on trust transfer in mHealth research.

Our study aims to develop a trust transfer model to investigate the adoption of mHealth services by the elderly and to address the following research question: To what extent is trust transfer a means of establishing the initial trust of elderly users in their adoption of mHealth services? In this study, based on the trust transfer theory, we incorporate both declining physiological conditions and support from hospitals into our research model to explore their moderating effects on the trust transfer process. To validate the research model and proposed hypotheses, a survey comprising 395 elderly users in China was conducted to analyze the research model.

All the measures in the survey were adapted or adopted from existing prevalidated instruments. The measures of trust in offline health services and trust in mHealth services were adapted from the study by Gefen et al [25]. The intention to use the mHealth service was measured using the items from the study by Deng et al [8]. The measures of declining physiological conditions were adapted from the study by Phang et al [12]. The measures of support from hospitals were adapted from the measures of vendor support in the study by Thong et al [48]. In fact, Thong et al indicated that vendor support results in a higher level of IS effectiveness. In the context of our study, the hospital of our study launched an mHealth service to extend the traditional offline health services. The role of the hospital is very similar to the role of the vendor. Therefore, we modified the measurement items of vendor support according to our research context.

To improve the content validity of the questionnaire, we revised the questions and deleted similarly phrased items based on feedback from 2 mHealth researchers and a pretest with 30 elderly users. All items included in the survey were measured on a 7-point Likert scale ranging from 1 to 7, with “1” representing strongly disagree and “7” representing strongly agree. The details of the constructs and measurements are presented in Multimedia Appendix 1.

We conducted our survey at Aerospace Center hospital in Beijing, the capital of China. At the end of 2016, this hospital had launched a health management platform aiming to provide reliable daily health care services for patients through mobile services. Therefore, this target hospital is an appropriate site for data collection. The mHealth services provided by this platform included a routine appointment out-patient clinic, follow-up visits, medicine reminders, medical records, and real-time positioning. Permission was obtained, and proper arrangements were made by the management board of the hospital for successful data collection. We followed 3 screening criteria to select the target participants: (1) patients coming to a physical examination center for medical examination, from which we could assume that they cared about their health; (2) patients who did not have a serious disease; and (3) patients aged 60 years or above. As these elderly participants could find it difficult to complete questionnaires unaided because of literacy problems, we accordingly employed 6 postgraduate students to assist the participants in completing the questionnaires. Furthermore, considering that most of the participants might be unfamiliar with mHealth services, a copy of an instruction manual of mHealth services was shown to the participants before they completed the questionnaire.

In fact, 500 copies of the survey questionnaire were distributed, and 395 usable copies were obtained, accounting for a response rate of 79.0% (395/500). We excluded 105 participants because, owing to several reasons, they were not our target subjects. First, these participants were illiterate. Second, their declining abilities regarding reading, writing, and speaking hindered their use of mobile services. Third, these participants did not own mobile phones, which could support the use of mHealth services. Accordingly, we selected the remaining respondents as our target subjects to use in the data analysis. Among these participants, approximately 49.1% (194/395) were males and 50.9% (201/395) were females. In fact, 316 out of 395 participants (80.0%) were aged from 60 years to 70 years. In addition, 75.9% (300/395) of the participants had attended high school. All respondents were given ¥10 (US $1.5) supermarket coupons for their participation.

Partial least squares (PLS) was used to test the research model because of the several advantages of this technique. First, PLS can predict all loadings and weights of indicators and causal relationships among constructs in multistage models [49,50]. Second, compared with covariance-based (CB) structural equation modeling (SEM), PLS is the most suitable technique for models with formative constructs and is appropriate for relatively small samples [51], which is the case in our study. In addition, PLS provides a good approximation of CB-SEM in terms of final estimates [50,51]. Given these considerations, we adopted the PLS to analyze our research model.

The data analysis was conducted in 2 stages. In the first stage, the measurement model (ie, based on the reliability, validity, common method bias, and multicollinearity of the constructs) was assessed to ensure its appropriateness. In the second stage, the structural model was examined and the stated hypotheses were tested [52].

The internal reliability, convergent validity, and discriminant validity were examined to assess the measurement model [53]. The reliability of the constructs was tested using Cronbach alpha, composite reliability (CR), and average variance extracted (AVE) [54]. In our study, the threshold values of the CR and AVE were .70 and .50, respectively, which is consistent with the work by Chin [55]. The value of Cronbach alpha, which was greater than the threshold .70, indicated adequate construct reliability [56]. CRs for these constructs ranged from .906 to .957, and the AVE varied from .765 to .881. All the values of Cronbach alpha and CR were above the threshold values, indicating good construct reliability [54]. All the item loadings of each construct were significant and above the suggested cut-off value (.700), indicating convergent validity [55]. As shown in Table 1, the loadings of all items were much greater than the cross-loadings on other constructs, and the correlations of any 2 constructs were significantly smaller than the square root of the AVE of each construct, further indicating acceptable discriminant validity. In addition, we tested the potential issue of multicollinearity through the use of the variance inflation factor (VIF). As a rule of thumb, the threshold value of the VIF is less than or equal to 10, indicating no presence of multicollinearity. The results indicate that all VIFs are less than 7, thereby suggesting that there is no multicollinearity among or between the independent variables.

Common method bias could be a potential concern as the data were collected from a self-reported survey [57]. We first tested for common method bias according to Harman single-factor test [58]. According to the results, we found that the factors accounted for 81.5% of the variance and the first factor only explained 25.1% of the variance, thus indicating that common method bias was not likely to have been an issue.

The results of the structural model are recorded in Figure 2. They indicate that trust in mHealth services (beta=.556; t₃₉₄=11.174; P<.001) had significant effects on the intention to use mHealth services. Hence, hypothesis 1 is supported. The results demonstrate that trust in offline health services has a significant effect on trust in mHealth services (beta=.583; t₃₉₄=14.528; P<.001), thus supporting hypothesis 2.

The moderating effects of declining physiological conditions and support from physicians were further tested. Declining physiological conditions were perceived to have a positive moderating effect on the relationship between trust in offline health services and trust in mHealth services (beta=.140; t₃₉₄=2.723; P=.003), thus lending support to hypothesis 3.

Support from hospitals was seen to negatively moderate the relationship between trust in mHealth services and intention to use mHealth services (beta=−.084; t₃₉₄=1.725; P=.04). Hence, hypothesis 4 is not supported.

In addition, the mediating effect of trust in mHealth services in the proposed model was tested according to the procedures provided by Baron and Kenny [59]. First, the relationship between trust in offline health services and intention to use mHealth services was tested (beta=.532; t₃₉₄=10.252; P<.001). Then, we tested the relationship between trust in mHealth services and trust in offline health services (beta=.582; t₃₉₄=13.191; P<.001). Finally, We tested the relationship between trust in offline health services and intention to use mHealth services (beta=.202; t₃₉₄=3.420; P<.001), and the relationship between trust in mHealth services and intention to use mHealth services (beta=.566; t₃₉₄=11.988; P<.001). Therefore, the relationship between trust in offline health services and intention to use mHealth services is partially mediated by trust in mHealth services. The results of each hypothesis are summarized in Table 2.

This study yields several important findings. First, trust in mHealth services positively affects elderly users’ use intentions. This highlights the proposition that trust in mHealth services is of importance in predicting the adoption of mHealth services. Second, trust in offline health services has significant effects on trust in mHealth services, which indicates that trust in offline health services can be transferred to the mobile environment. This suggests that mHealth service providers may be able to swiftly build elderly users’ trust in mHealth services through leveraging their existing trust in offline health services, thus leading to a higher adoption rate of mHealth services. Third, declining physiological conditions are seen to strengthen the relationship between trust in offline health services and trust in mHealth services. As physiological capabilities decline with age, elderly users are less capable of using new IT apps. As elderly users with declining physiological conditions need to exert more effort to evaluate the competence and ability of mHealth services, they rather tend to rely more on their prior experience and knowledge of offline health services to cultivate trust in mHealth services. Fourth, support from hospitals is seen to weaken the association between trust in mHealth services and the intention to use mHealth services. Contrary to our hypotheses, support from hospitals decreases patients’ trust in mHealth services because they depend more on physicians’ support instead of expending extra effort on evaluating the competence and ability of mHealth services. There is possibly a mismatch between their perceptions and their physicians’ advice.

This study can enrich and advance our theoretical understanding in several ways. First, it extended the trust transfer theory to the offline context in transition to mHealth services. Previous studies of the trust transfer theory have mainly focused on the perspectives of offline to online and online to mobile channels [16,20,22,60]. On our part, we explored the trust transfer process in the context of the new mHealth services from a cross-environment perspective. As the offline to mobile service transition has become a trend for most offline services, our research into this phenomenon provides a valuable reference. Our study can be seen as an attempt to fill this research gap by providing a cornerstone for further theoretical development.

Second, to comprehensively understand the IT acceptance behavior of the elderly, this research introduced a variable from gerontology—declining physiological conditions. Previous IS research on the IT acceptance behavior of a specific group is rare, except for the studies by Phang et al [12] and Deng et al [8]. On our part, the moderating effect of declining physiological conditions was tested in the trust transfer from offline to mobile channels. Indeed, exploring the moderating role of declining physiological conditions on trust transfer will facilitate the understanding of which conditions are more effective and thus further extend our knowledge of the trust transfer theory.

Third, support from hospitals was introduced in the context of research on mHealth services. Social support is positively associated with health behavior in the health care literature [61-63]. However, studies on the role of social support in moderating the associations between these 2 trust elements and intention to use have been limited. We suggest that future studies should take social support into account in an attempt to better understand how trust elements influence elderly users’ attitude changes and health behaviors.

This study also reflects several practical implications. First, the important role of the relationship between offline health services and mHealth services suggests that offline trust can be used as an enabler that encourages a provider of mHealth services to expand from the offline to mobile channels. Accordingly, mHealth service providers are encouraged to increase their cooperation with hospitals that are trusted by patients when marketing their mHealth services to elderly users. This may lead to higher adoption rates than simply promoting mHealth services in isolation.

Second, hospitals should undertake responsibility for ensuring that the elderly use mHealth services because they possess limited health literacy and limited experience with mHealth services. However, hospitals need to provide support to elderly users in a proper way to avoid them being overly dependent on hospitals. This is because elderly users who rely too much on support from hospitals are less likely to develop trust in mHealth services, which can weaken their intention to use mHealth services.

Third, mHealth service providers need to understand the different behaviors among different groups of users of mHealth services. Service providers are urged to employ tailored strategies to promote their mHealth services to elderly users when they develop marketing campaigns targeting the building of trust in a mobile environment.

As with all empirical research, this study has its limitations. First, the study did not include users of all age groups. Elderly users were taken as the sample in our study because this specific group accounts for a large portion of all users of mHealth services. Our results need to be interpreted with caution for applications in other population and age groups. Second, this study was conducted in China, and the results may be applicable only in cultural contexts similar to those of the Chinese mainland. We suggest a similar study in a Western context for comparing the results across different cultures. Third, although the explanatory power of the model is acceptable (48.3% for intention to use), we still advocate the potential to enhance our explanatory power through taking additional factors into consideration, in future research.

In conclusion, we reiterate that mHealth services are regarded as an essential means to alleviate the conflicts between the medical demands of an increasingly aging population and limited medical care resources. However, the adoption rate of mHealth services among elderly users still remains low. Our research draws upon the trust transfer theory and builds a comprehensive framework by integrating declining physiological conditions and support from hospitals to investigate the initial trust-building mechanism of mHealth services. The results indicate that trust in offline health services has a significant effect on trust in mHealth services, thereby leading to an intention to use mHealth services. Declining physiological conditions is seen to positively moderate the association between trust in offline health services and trust in mHealth services; however, contrary to our hypotheses, support from hospitals weakens the association between trust in mHealth services and the intention to use mHealth services. The results provide a good explanatory power to predict the mHealth use intentions of elderly users. These findings have advanced the trust transfer theory and enriched the literature on mHealth services. Accordingly, mHealth services practitioners can better understand how to leverage the benefits of trust transfer and the characteristics of elderly users to promote their mHealth services.