The study team included clinicians (CAJ, MPD, and BKN), individuals with expertise in qualitative methods (SLK), implementation science (SLK), health services research (BKN, SLK, and JDP), and digital health technologies (DHT) (CAJ, BKN, MPD, and JDP). All members of the study team were employed by the University of Michigan (U-M), a large midwestern public university (CAJ, SLK, MPD, JDP, BKN) or a Veteran’s Health Administration hospital (SLK, JDP). Given the qualitative nature of this study and the aim of understanding participants’ subjective experiences, a constructivist research approach was used.

We used a descriptive qualitative study design to examine patient and clinician experiences with mHealth apps [46]. The study occurred at U-M from September 2023 to August 2024. The study adhered to the standards for reporting qualitative research reporting guidelines [47].

The interview guides were developed using constructs from the technology acceptance model (TAM) [50,51], the unified theory of acceptance and use of technology (UTAUT) [52,53], the theoretical domains framework [54], and the technology integration model (TIM) [55].

The patient interview guide included two domains: (1) technical and material factors and (2) social and personal factors (Multimedia Appendix 1). The clinician interview guide included three domains: (1) technical and material factors, (2) social and personal factors, and (3) policy and organizational factors (Multimedia Appendix 1). The technical and material factors domain included constructs such as perceived usefulness and perceived ease of use. Constructs in the social and personal factors domain included attitudes, self-efficacy, identity, and social influence. Finally, workflow, training, and education were the constructs included in the policy and organizational factors domain.

The patient guide included 19 questions (plus probing questions), and the clinician guide included 14 questions (plus probing questions) (supplement). Both interview guides were iteratively updated as new information was encountered during interviews, and the final guides were approved by members of the study team (CAJ, SLK, BKN, JDP, and MPD).

The interviews were conducted either virtually via the Zoom platform or by phone by a clinician (CAJ) and a nonclinician research associate (SY). Prior to interviews, clinicians viewed a brief presentation about the LowSalt4Life app to (1) facilitate a shared understanding of the general capabilities of mHealth apps, (2) evaluate their attitudes about the LowSalt4Life app, and (3) have a common reference app to allow comparison of clinician and patient attitudes about mHealth apps in general. Each interview was audio recorded and transcribed.

We conducted a deductive content analysis [56-58] using a rapid qualitative analysis approach to identify subcategories and categories within the domains of the theoretical frameworks. Rapid qualitative analysis is a particularly useful method when the data collection is targeted (eg, selection of interviewees) and processes (interview protocol) are highly structured [59-61]. The rapid qualitative analysis approach allowed us to maintain rigor while also working within the constraints of another study. Rapid qualitative analysis involves three analytic steps: (1) summarization of individual transcripts, (2) consolidation of transcript summaries by participant type (eg, patient or clinician), and (3) in-depth analysis [62].

Two study team members (CAJ, SLK) developed a summary template based on the domains in the interview guide, then independently reviewed, summarized, and discussed 3 interview transcripts to test and refine the template and summarization process. Thereafter, concurrent with ongoing data collection, each interview transcript was reviewed in detail, and transcript summaries that organized the data into the relevant domains were generated by the first author (CAJ). Data from the transcript summaries were transferred to patient or clinician data matrices to allow visualization of the data across domains, constructs, and participants. A detailed review of the data matrices by several study team members (CAJ, SLK, SY) led to the identification of patient and clinician subcategories within each domain. Through triangulation, the subcategories were then aligned to create broader categories by placing the patient and clinician data in a table that allowed comparison across domains and constructs (). CAJ and SLK met biweekly to discuss the data and resolve any discrepancies. The categories were finalized during meetings with members of the research team (CAJ, SLK, SY, JDP, and BKN).

The study protocol (HUM00238747) was approved by the U-M Institutional Review Board. All participants provided written informed consent. All the study data was deidentified and stored in a Health Insurance Portability and Accountability Act–compliant, U-M–approved electronic database to ensure privacy and confidentiality of participants. Patients received a US $25 gift card after completing the interview. Clinicians did not receive an incentive for participating in the study.

Prior to initiating the study, we planned to interview 20 clinicians and 20 patients. However, data saturation (no new analytical information identified by the research team) was reached after 20 clinician interviews and 28 patient interviews. Among the 20 clinicians interviewed, the mean number of years in practice was 17 (SD 11.6) years (Table 1) and most were general internists (n=12, 60%). The clinician interviews lasted between 22 to 49 minutes. The mean age of the 28 interviewed patients was 59 (SD 12.1) years, and 16 (57.1%) patients were female (Table 1). Two (7.1%) patients self-identified as Asian, 4 (14.2%) self-identified as Black/African American, 21 (75%) self-identified as White, 1 (3.6%) self-identified as unknown, and none self-identified as Hispanic. The patient interviews lasted between 21 and 63 minutes.

Patient and clinician attitudes about mHealth apps fit within 1 of 7 categories related to the 3 theory-based domains (Table 2). Within the technical and material factors domain, the categories identified were apps as tools to improve health by extending care, the role of apps in enhancing the patient–clinician relationship, and the need for simplicity and efficiency in app design. Categories in the social and personal factors domain included prior experience with mHealth apps, comfort with technology, and the influence of recommendations from trusted sources. Education and hands-on experience were the category identified in the policy and organizational factors domain. These categories are described in more detail below.

mHealth apps are an increasingly important part of the health care landscape and can enhance patients’ autonomy and engagement with their health care. Our findings show that overall, patients and clinicians have positive attitudes about mHealth apps, including their potential usefulness in extending care outside of traditional health care settings and enhancing the patient–clinician relationship. Patients and clinicians identified several common factors to facilitate the recommendation or adoption of mHealth apps. These factors included the importance of the ease of use of an app, receiving recommendations about mHealth apps from trusted sources, and access to education about or hands-on experience with mHealth apps. Prior experience with mHealth apps and comfort using technology were also described as facilitating factors by both patients and clinicians.

Other studies have indicated that both patients and clinicians have concerns about mHealth apps eroding the patient–clinician relationship and therapeutic alliances because of less face-to-face interaction [63]. This concern is well-founded as the introduction of technology can lead to a form of automation bias in which clinicians over-rely on mHealth apps and their outputs, or inappropriate delegation of tasks to mHealth apps [64]. Clinicians will need to exercise vigilance and make nuanced decisions about the appropriateness of mHealth apps for a given task to both ensure safety and that the patient–clinician relationship remains intact. Further, measures to avoid patient overreliance on mHealth apps will need to be in place because this could also negatively impact the patient–clinician relationship [65]. In a study by Preiser and colleagues [66], general practitioners in Germany noted that a symptom checker app could result in decreased unnecessary office visits and decreased clinician workload. However, the clinicians preferred that patients err on the side of caution and seek their help early to circumvent avoidable suffering. Ultimately, technology and safe, humane care should not be treated as mutually exclusive. Consistent with the TIM [55], this will require further exploration of how mHealth apps can extend patient and clinician capabilities, and both patients and clinicians acknowledging the strengths and limitations of technology and one another [67].

While mass media campaigns [68] and direct-to-consumer marketing [69] have proven effective at influencing patient health decisions, patients in our study noted that their attitudes about an app would be most influenced by trusted individuals (eg, friends, family, and health care providers). Therefore, clinicians must actively engage with patients and their family, friends, and caregivers to encourage the adoption and sustained use of safe and effective mHealth apps [70]. Further, studies have shown that patients’ adoption of health-promoting behaviors may be influenced by individuals in nontraditional health care settings. For example, in an RCT by Victor and colleagues [71], Black male patients with uncontrolled hypertension were assigned to either an intervention group in which barbers at Black-owned barbershops encouraged meetings in barbershops with specialty-trained pharmacists who prescribed antihypertensive medication, or to a control group in which the barbers encouraged lifestyle modifications and doctor appointments. While both groups experienced improvement in their blood pressure at 6 months, the improvement was much greater in the group that met with the pharmacists and were prescribed antihypertensives. This study highlights both the benefits of treating patients in their natural environments and the benefits of leveraging the influence of trusted members of the community to promote healthy behaviors. Partnerships such as this should be explored to promote the adoption and sustained use of mHealth apps that have proven safe and effective.

Clinicians in our study noted that trusted colleagues, specialty organizations, and evidence-based guidelines, each of which may be related to the subjective norms described by UTAUT, would strongly influence their decision to recommend an app to a patient. However, clinicians did not describe app certification or approval by a regulatory agency (eg, Food and Drug Administration [FDA]) as necessary for recommending an app to a patient. In other studies, clinicians have identified a lack of certification as a barrier to recommending mHealth apps. For instance, in a study by Schroeder et al [36], clinicians expressed that they view mHealth apps similarly to other medical products and that apps should be given an official certification to ensure more trust. This study was conducted in Germany, which in 2020 became the first country in the world to approve certain mHealth apps for prescription and coverage by health insurers [26]. Additionally, Germany maintains a directory of approved mHealth apps managed by the German Federal Institute for Drugs and Medical Devices known as Digitale Gesundheitsanwendungen. As of July 2024, there were 56 mHealth apps listed by the German Federal Institute for Drugs and Medical Devices, and over 274,377 prescriptions were recorded from 2020‐2023, with Zanadio (Sidekick Health Germany), a weight loss app, prescribed most frequently during that period [72]. Another study based in the United Kingdom showed that clinicians believed the most important factor encouraging mHealth app prescription was if it had a National Health Service stamp of approval [32]. The US FDA plays a role in the regulation of mHealth apps [43] but the vast majority of mHealth apps are not regulated by the FDA. While it may not be feasible for the FDA to rigorously evaluate every app, attention should be given to developing interventions that facilitate clinician confidence in the safety and efficacy of evidence-based mHealth apps as studies have shown that many apps in app stores with high ratings often have low usability or clinical usefulness [73].

Our study showed that patients and clinicians have differing beliefs about patients’ level of tech savviness. While most patients reported using mHealth apps, clinicians generally believed that older adults would not be interested, or they would be unable to effectively use mHealth apps [63]. A recent study showed that over 81% of older adults use DHTs, and over 40% reported any use of mHealth apps [74]. Our findings suggest that clinician bias could be a barrier to recommending mHealth apps to those who may benefit most (eg, older adults). During the COVID-19 pandemic, the use of DHTs significantly increased across all age groups. However, disparities widened with older adults significantly less likely to use DHTs compared to younger adults [75]. Still, older adults’ increased use of DHTs during the COVID-19 pandemic suggests that they are interested and capable of using mHealth apps, but prior to the pandemic, they were not using them for various reasons that should be explored. Moreover, views expressed by older adults in our study are consistent with this as a 72-year-old patient stated: “I am the type where if there is a need to use an app, especially for healthcare I would learn to use the app.” Older adults are likely to have a greater burden of chronic disease [11] and impairments (eg, physical and cognitive) that can limit their ability to use DHTs, and this should be considered when recommending mHealth apps to patients. The results from our study suggest that clinicians should take an individualized approach to recommending mHealth apps to older adults. Notably, age moderates the constructs in the UTAUT framework (eg, performance expectancy, effort expectancy, social influence, and facilitating conditions) that predict behavior intention to use technologies [52,53]. When studying older adults’ use of mHealth apps, using a framework that does not include age, but does include variables associated with the aging process (eg, memory or physical limitations) may add more nuance and avoid introducing age-related bias. Further, interventions are needed to both address clinician bias [76] and to encourage patients and their families or caregivers to advocate for themselves.

Clinicians also noted barriers to recommending mHealth apps that included getting patients oriented, time constraints, a need for support from staff, and a lack of optimized workflows to facilitate review and use of data generated by mHealth apps, which highlights the importance of the environment and situational context described in the TIM [55]. These concerns must be weighed against the benefits to patients. For instance, patient-generated data has been shown to increase patient health literacy and to help patients better engage with their health providers [77,78]. User-centered mHealth apps also have the potential to increase adherence by providing individualized, just-in-time reminders and advice during vulnerable times outside of clinics and hospitals as demonstrated in conditions such as human immunodeficiency virus and congestive heart failure [79-81]. Like other studies, clinicians in our study expressed interest in having data from mHealth apps integrated into electronic health records (EHR) for efficient visualization, analysis, and application. This type of integration would require policy and organization-level interventions that support the flow of new data streams, interoperability between systems, and application of data from mHealth apps to improve patient–clinician engagement and health outcomes. Furthermore, health care organizations and app developers would need to partner with EHR vendors to ensure that their products are interoperable, allowing seamless data exchange between the EHR and mHealth apps. Collaborations between patients, clinicians, health system leaders, and developers will be necessary to ensure that apps address relevant clinical problems, are easy to use, and simple to set up. Thus, user-centered design principles addressing the patient, caregiver, and clinician perspectives must be used to facilitate recommendation by clinicians and adoption by patients.

James et al [82] described the “4E’s (ease, evidence, experience and expectations)” that inform the mindlines influencing clinicians’ decisions to recommend mHealth apps. Mindlines are collectively reinforced, internalized tacit guidelines, which are informed by brief reading, but mainly (by) clinicians’ interactions with each other (as well as) opinion leaders, patients, and pharmaceutical representatives and by other sources of largely tacit knowledge that build on their early training and their own and their colleagues’ experiences [83]. Our study reinforces the potential influence of the 4E’s described by James and raises the question of whether the mindlines concept and the 4E’s informing them could be expanded to patients. However, it should be noted that a category related to expectations did not emerge in our study, which could be due to the theoretical frameworks used and the constructs used. Therefore, the role of patients’ and clinicians’ expectations around mHealth apps should be explored.

Strengths of this study include using an mHealth app that was being actively studied and deployed as a common frame of reference for discussions about mHealth apps with patients and clinicians. Additionally, we combined constructs from well-established technology adoption (eg, TAM and UTAUT), and implementation science (eg, theoretical domains framework and TIM) frameworks to account for limitations of each framework, which allowed us to identify facilitators and barriers that could inform development of interventions to increase recommendation and adoption of mHealth apps by patients and clinicians.

This study also had key limitations. First, it occurred at a single academic health center in a metropolitan area. There are known geographic differences in smartphone, internet, and DHT use, with lower rates of access among patients in rural areas. Second, our study only included patients enrolled in the LowSalt4Life-2 study, a study of a single mHealth app. Hence, the results may be affected by selection bias as participants may have had a favorable view of mHealth apps and interest as indicated by their participation in the study. Third, the qualitative design of our study does not allow us to infer that the attitudes expressed by participants in our study are shared by others. Fourth, because we used a deductive analytic approach that included domains from established frameworks, we may not have identified additional relevant or novel attitudes that may have been discovered with an inductive approach. Finally, we cannot conclude that addressing the identified patient and clinician attitudes will result in increased recommendation or adoption of mHealth apps. Our study did not measure intention to use, behavior intention, use behavior, or technology use, which are downstream constructs in the TAM, UTAUT, and TIM frameworks, respectively [50-53,55]. Thus, studies are needed to determine the impact of interventions that address the identified attitudes on patient adoption and clinician recommendation of mHealth apps.

In this qualitative study, patients and clinicians generally had a positive view of the potential benefits of mHealth apps. Both groups also identified several similar factors that facilitate the use of mHealth apps. However, overall, patients’ and clinicians’ views about older patients’ interest in and ability to use technology differed. As mHealth apps are developed and implemented, careful consideration must be given to both patients’ and clinicians’ attitudes and agreement and disagreement between these key stakeholders. This approach will allow the development of implementation interventions that facilitate recommendation of mHealth apps by clinicians and adoption of mHealth apps by patients.