In 2017, only approximately 37 million patients were admitted to US hospitals [1]. As the number of individuals needing hospitalization has increased in recent years [2], hospitals have mainly limited their beds to acute care and left the rest of treatments in the form of outpatient settings to the patients’ loved ones [3,4]. Unpaid family or informal caregivers are, therefore, the backbone of the long-term care system and offer the majority of long-term care that patients require in the United States [5]. In addition, informal caregiving is essential for sustaining adults with disabilities or chronic health conditions [6].

Caregiving responsibilities primarily include providing informational and emotional support, dealing with financial concerns, and managing medical care [4]. These responsibilities also consist of more basic support, such as assistance in eating, bathing, grocery shopping, and meal preparation [7]. Many informal caregivers experience positive feelings such as satisfaction, a sense of gratification for giving back to those who cared for them, and improved family relationships [8,9]. However, the nature and amount of such responsibilities may change caregivers’ lives; modify their emotions; and make them feel frustrated, fearful, and nervous, all resulting in a decline in their quality of life [10].

Estimates of the number of informal caregivers in the United States vary widely. According to the American Association of Retired Persons, there are currently more than 40 million unpaid caregivers in the United States [11], about 7 potential informal caregivers per adult [12]. Although studies indicate that the number of informal caregivers will continue to rise, it is expected that the number of individuals who need care will be far from the number of those who offer that care in the near future (about 4 potential informal caregivers per adult) [12,13]. There are 2 main reasons for this result. First, the number of individuals who will be over 65 years old by 2050 is expected to be 2 times more than that number in 2010; the majority of those individuals will be dealing with various chronic conditions and a decline in quality of life [14,15]. Second, the traditional American family structure has experienced fundamental changes in recent years. Family sizes continue to decrease as a result of higher rates of those who have never married or are divorced or are affected by infertility and childlessness. In addition, there is a higher chance that women, who are the backbone of informal caregiving, are in the workforce [16,17].

This shrinkage in the number of potential informal caregivers per adult increases the amount of responsibility and inflicts more physical and mental stress among the existing caregivers [14]. The ubiquity of mobile technology and its applications has the potential to reduce such stress. In general, mobile apps are reasonably priced and user friendly and offer an information repository collected from various sources [18-20]. Caregiving-related apps are specifically designed to provide users with a platform to gain appropriate and trusted information, manage medication taking, improve communication with care providers and support groups, connect with counterparts, reserve transportation, and manage the health condition of care receivers in an organized manner [21].

Although there are hardly any studies in the literature that investigate the role of caregiving-related apps in reducing caregiving-related stress [22], it is very likely that such apps significantly lessen the stress caregivers face by providing a convenient platform to receive informational and emotional support [14]. Despite the considerable role of caregiving apps in reducing stress and improving the overall quality of life among caregivers and although more than 57% of American caregivers have a smartphone, only 40% of them use a caregiving-related app [23]. This raises the concern of finding solutions to increase caregivers’ access to and effective use of such beneficial resources.

Some studies highlight the roles of caregivers’ digital literacy and sociodemographic factors on their natural propensity to use various internet-based tools and services for caregiving purposes in general [24,25]. However, the current understanding of caregivers’ intentions to use related mobile apps for their responsibilities is limited, and we could not find any published studies that directly investigated the influential factors.

As such, our objective is to provide insights into this issue. To explore the important factors that affect their intentions to integrate mobile apps into their routine caregiving responsibilities, we designed a cross-sectional study using the protection motivation theory (PMT) perspective [26]. PMT is among the most influential explanatory theories in the literature to predict an individual’s intention to adopt recommended actions (adoption of related mobile apps in this study) [27]. It has also been empirically applied to both technological and nontechnological solutions [28].

PMT suggests 2 consecutive appraisals that explain the process whereby individuals adopt a recommended action: threat and coping [29]. If an individual is exposed to a stressful or fearful situation, a personal perception of threat arises. If the threat is perceived to be appropriate and possibly harmful, coping appraisal will occur [28].

On the basis of the threatening event, individuals first assess the level of danger in 2 aspects: (1) perceived threat severity, which is the individual’s assessment of the seriousness of the threat, and (2) perceived threat vulnerability, which is the individual’s assessment of the likelihood of coming across the threat personally. Once the threat has been assessed, individuals assess their ability to cope with the threat and form their perceptions of the efficacy of the recommended action in 2 aspects: (1) self-efficacy, which is the individuals’ assessment of their capability to perform the recommended action, and (2) response efficacy, which is the individual’s assessment of the efficacy and benefits of the recommended action [26,29,30]. This efficacy assessment affects individuals’ intentions to adopt the recommended action. Figure 1 illustrates the PMT model.

To test the proposed model, we recruited participants through Amazon’s Mechanical Turk, which is a web-based crowdsourcing market for registered users to participate in various tasks and receive a predetermined amount of money upon successful completion [41]. Studies in the domain of information systems have confirmed the validity and reliability of the results [42,43]. Purposive sampling was used to target participants who met the following inclusion criteria [44]: (1) US residents, (2) owning and using a smartphone, (3) informal caregivers (individuals who give care to a friend or family member without payment) who provided at least 8 hours of care per week in the past year, and (4) currently not using any mobile app for caregiving purposes. We obtained approval from the Institutional Review Board for the explained approach.

We adapted measures from a set of empirically validated studies in the literature and used multi-item measures to enhance the validity and reliability of the measurement. All the main constructs were reflective. For the constructs drawn from PMT, measurement items were adapted from Witte [45] and modified to fit the context of our study. Under threat appraisal, 5 items (mean 3.76, SD 0.74) and 4 items (mean 4.03, SD 0.69) were used to assess perceived vulnerability (eg, “My care-receiver is at risk for getting health threats”) and perceived severity (eg, “I believe that threats to my care receiver's health are severe”), respectively. For coping appraisal, 6 items (mean 3.37, SD 0.86) assessed response efficacy (eg, “Using mobile apps is effective in monitoring my care-receiver’s health condition remotely”), and 4 items (mean 3.79, SD 0.78) were used to measure self-efficacy (eg, “I feel confident using mobile health applications for my caregiving activities”). Intention to use items was adapted from a widely used scale in the literature [46] (3 items; mean 2.98, SD 0.93) and changed appropriately to fit our study (eg, “I plan to use mobile apps to manage my care-receiver’s health status in the next 3 months”). Items for perceived self-autonomy were adapted from the study by Deci et al [47] (5 items; mean 3.89, SD 0.83) and altered to suit our context (eg, “I have some choice in what I want to do in my caregiving activities”). For all the aforementioned constructs, participants indicated their level of agreement with each item using a 5-point Likert-type scale ranging from strongly disagree (1) to strongly agree (5).

Questions were randomly shown to the participants to minimize the order-effect bias. Two experts reviewed the initial questionnaire to ensure face validity. We revised the survey based on their comments and feedback. We also conducted a pilot study among master’s students at a large university in the southwestern region of the United States and made appropriate changes to the survey based on the results. Previous literature, pilot tests, and numerous series of pretests allowed us to confirm the content validity of our instrument and measures [48]. The final measurement items are listed in Table 2. The Checklist for Reporting Results of Internet E-Survey is available in Multimedia Appendix 1 [49].

A total of 249 valid responses were collected. The average time that our responders provided care for their current care receivers and for various individuals in general was 50 and 83 months, respectively. A total of 32.5% (81/249) of the care receivers were financially dependent on their caregivers. Tables 3 and 4 summarize the key demographic variables of the respondents and their care receivers.

A partial least squares approach was used to test the proposed model and associated hypotheses. We used the SmartPLS software package (version 3.2.6, SmartPLS GmbH) to analyze the data [50]. To estimate the path coefficient weights and their significance, we used bootstrapping procedures with 5000 resamples.

To assess the measurement model, we conducted a confirmatory factor analysis. The results are summarized in Table 5. All of the latent constructs are modelled to be reflective. For each latent construct, the path loadings, t statistics, and SE were calculated. At α=.05, all the measures’ path loadings are significant with a value of more than 0.7 [51], indicating that more than 50% of the variance is shared between each construct’s items [52].

To examine convergent validity, both composite reliability and average variance extracted (AVE) were obtained [53]. All constructs met the acceptable score of 0.5 for AVE, confirming their reliability [54]. Moreover, the composite reliability values were between 0.874 and 0.927.

The square root of the AVE was also used to verify discriminant validity [54]. The results are presented in Table 6. For each latent construct, the square root of the AVE was higher than all its cross-correlations. Moreover, higher values of interconstruct correlations confirmed the greater variance among each construct’s specific measures compared with other measures [55]. All things considered, we can confirm that all the indicators were valid through their constructs.

As data for all the variables were collected in a single survey, common method variance (CMV) could have an excessive effect on the results [56]. Harman’s single factor test was conducted [57] to find out the extent of this effect. According to this approach, with a factor analysis, if only a single factor arises, or a factor explains the majority of variance among all the measures, we can conclude that CMV is a significant issue in the sample [56]. After conducting an exploratory factor analysis among all the items, 6 factors were obtained that explained more than 70% of the variance. All the extracted factors had an eigenvalue >1, and the highest factor accounted for only 33% of the variance. This indicates that CMV is not a serious concern in this data set.

The primary goal of this study is to investigate the factors that affect caregivers’ intentions to adopt caregiving-related mobile apps based on a model that contextualizes PMT. The results indicate support for the proposed model with decent explanatory power.

As the analyses illustrate, caregivers’ capabilities and skills to use mobile apps and the app’s effectiveness in responding to caregivers’ needs can predict their willingness to adopt related apps. In addition, our results indicate that the degree of control of caregivers over their responsibilities and the decisions they make for their care receivers can also increase the likelihood of adopting such mobile apps. Interestingly, the app’s effectiveness in responding to caregivers’ needs had the strongest effect on their intention to adopt such apps.

We also found that as the care receiver’s health status is perceived to be more severe, caregivers will count more on their capability to use a caregiving-related app. In addition, as the threat of facing unexpected health changes becomes more likely, caregivers will perceive the mobile app as a more efficient tool to help them with their responsibilities. These findings are consistent with previous studies on PMT [28].

However, we did not find enough evidence to support the effect of the care receiver’s severity of health status on the caregiver’s perception of the mobile app’s effectiveness. In addition, the results indicate that the effect of the care receiver’s vulnerability to unexpected health changes on the caregiver’s perception of their self-efficacy in using the mobile app is not significant. Although these results are not congruent with PMT, there are various other studies that confirm that individuals, in most cases, believe that threats either happen only to others or influence other individuals more than themselves [59]. This belief helps them maintain a sense of invulnerability and explains why we did not find evidence to support 2 of the hypotheses.

As mentioned earlier, caregiving apps play a considerable role in reducing stress and improving the overall quality of life of informal caregivers. Therefore, it is important to devote time, money, and effort to develop and promote caregiving-related apps to enhance both caregivers’ and care receivers’ well-being. In this regard, the results of this study provide several practical implications for developers, health care practitioners, and policy makers.

This study has some limitations. First, the survey was limited to US caregivers, raising the external validity issue. Future research is required to generalize the results and offer an inclusive perspective of caregivers’ adoption of related mobile apps in the context of other countries. Second, the collected information and derived results were based on self-reported data. This can increase the likelihood of some biases such as social desirability. Further studies may use a more inclusive data collection technique to measure adoption more rigorously. Finally, there is still some unexplained variance in the intention to adopt mobile apps. Although including more constructs may have improved the explanatory factor of the proposed model, we intended to maintain a parsimonious extension of PMT [28]. Additional studies are required to investigate other effective factors.

Grounded on PMT, this study investigates the factors that affect the intentions of informal caregivers to adopt related mobile apps for their routine caregiving responsibilities. The results of the survey of 249 US-based informal caregivers indicated that caregivers’ degree of control over their responsibilities and choices, their perception of the app’s effectiveness in responding to their needs, and their capability to use mobile apps play a positive role in their willingness to adopt caregiving-related apps. Furthermore, care receivers’ vulnerability to unpredicted health changes and the severity of their health status indirectly affect their caregivers’ intentions to adopt such mobile apps. These findings offer important contributions to the field and have significant implications for developers, health care practitioners, and policy makers.