Our program implemented 2 complementary, parallel-arm, and randomized clinical trials with decentralized administration, summarized here and described in further detail elsewhere [18,22,23]. One trial (ClinicalTrials.gov ID NCT04076020), conducted in individuals with a rural Pennsylvania residence as determined using a definition of rural status as designated by the United States Census Bureau, aimed to recruit geographically remote individuals with AF. The second trial (ClinicalTrials.gov ID NCT04075994) focused on recruitment of individuals residing in metropolitan communities with a focus on economically depressed regions of Pittsburgh, Pennsylvania. Both trials prioritized recruitment from populations that have historically had limited representation in clinical trials for AF. Eligibility for participation in either trial included a diagnosis of AF, as confirmed by the electronic health record, and the prescription of oral anticoagulation for the purpose of thromboembolic stroke prevention in the setting of having AF. The rural and metropolitan trials aimed to recruit 264 and 240 participants, respectively, given differences in design and complementary study objectives.

The architecture of this program consisted of entirely remote recruitment, engagement, assessment, and retention. In effect, this process resulted in the absence of in-person contact between participants and study team members. Recruitment for both studies occurred using multiple approaches. Foremost, eligible individuals received an introductory letter cosigned by their physician provider, such as a physician or nurse practitioner to introduce the research study, accompanied by a brochure, contact information, and a stamped postcard to decline participation. Individuals who did not return the postcard within 2‐4 weeks received a telephone call as described by the letter. Participants also self-referred, having learned about the study from their physicians or material placed in clinic settings. Those interested in participating underwent telephone-based screening to verify appropriateness for the trial and review of the inclusion and exclusion criteria summarized in, Textbox 1. If eligible, potential participants scheduled a baseline interview and were then mailed the informed consent and materials for the baseline visit.

The study team obtained telephone-based informed consent at the start of the baseline visit. The informed consent outlined the study schedule, participant burden and compensation, access to the participant’s electronic health record, privacy and security protections in place for participant data in the electronic data management system, and provided contact information for the principal investigator of the study. In addition to providing telephone consent, participants returned a signed copy of the informed consent using a preaddressed, stamped envelope that accompanied study materials.

Consenting participants were randomized to intervention or control arms using electronic software [24,25]. Participants were provided with copies of assessments reproduced in 12- to 14-point font to enhance readability and facilitate their administration by telephone with trained assessors. All study materials were provided at a sixth grade reading level to ensure accessibility to participants. Table 1 lists the assessments conducted by study visit for both trials.

Consented participants received study smartphones accompanied by training on their use and summary guides on smartphone and app operation developed for this study specific to the intervention and control arms. Materials provided to participants in the rural study and randomized to the intervention are provided as an Appendix in Multimedia Appendix 1. Training on smartphone and app use followed a standardized curriculum and concluded when participants reached capacity to operate the phone at a level appropriate for study participation. Smartphone ownership or digital literacy was consequently not required for trial participation. Smartphones were programmed to enable features such as phone calls and texting without the the capacity for more advanced functions such as downloading apps. Participants randomized to the intervention arms of the trials had the relational agent preinstalled on the smartphone and were provided detailed instructions regarding its use as part of smartphone instruction.

Relational agents are human-computer interventions adapted for multiple settings to facilitate education, problem-solving, and behavioral approaches with patients. Previous work has demonstrated that such agents are accessible to individuals with limited health and digital literacy with varied medical conditions [37-40]. Here, we designed the agent to provide education about AF; problem-solving regarding intentional and unintentional nonadherence to medications; preparation for the medical encounter; and address symptoms common in AF (eg, irregular or rapid heart rates, dyspnea, and chest discomfort). Relational agent content was modified for use in a rural or metropolitan setting. Figure 1 presents the 2 relational agents used in the trials. Intervention participants also received the AliveCor Kardia Mobile device [41] for heart rate and rhythm monitoring, as prompted by the relational agent, to reinforce self-care and the correlation of symptoms with heart rate and rhythm assessments.

Participants randomized to the trials’ control arms had the WebMD (WebMD LLC) app preinstalled on the study-provided smartphones. Research assistants encouraged participants to use this application to learn more about AF, its management, and the tracking of medications and symptoms. Participants in the control arms of both trials received an informational session from study team members providing a brief overview of AF and complications such as stroke, heart functioning, and signs of an impending stroke derived from American Heart Association educational materials. To further distinguish the 2 trials, control participants in the metropolitan trial also received the AliveCor Kardia Mobile device with instruction on its use, and guidance to use it as for heart rate and rhythm monitoring. In both studies, heart rate and rhythm obtained using this device was monitored, categorized, and recorded by the study team.

The trials shared the primary outcome to improve adherence to oral anticoagulation for thromboembolic stroke prevention in individuals with AF, as measured by pharmacy claims data using the proportion of days covered [36] (a validated measure to quantify medication adherence using pharmacy claims) and by participant self-report [35]. The secondary outcomes of the trials were (1) health-related quality of life [42], measured using the disease-specific AF Effect on Quality of life [33] and general patient-reported outcomes measurement information system [32] measures; and (2) health care utilization, as measured by days of hospitalization and emergency room visits.

Both intervention and control applications were used for 4 months. Participants were sent a box with a prepaid label for returning the study phone and were informed that their second study payment was tied to a smartphone return. They were allowed to keep the Kardia Mobile device and received instruction from the study team on how to connect the device to their personal smartphone with the caveat that results would no longer be monitored by the study. Participants underwent repeat telephone assessments at 4-, 8-, and 12-months with simultaneous review of the electronic health record for hospitalization events. To assist with interviewer-administered phone-based instrument completion, participants were again mailed the packets of questionnaires summarized in Table 1.

Given the absence of direct, personal contact, the study developed remote strategies for participant engagement. Participants received regular newsletters also written at 6th-grade reading level for the duration of the study that provided additional education about the studies and updates. In addition, the team mailed birthday cards to participants yearly throughout the duration of the study. Finally, study participants were offered the opportunity to participate in qualitative assessments to further share their experience of AF [43,44]. These sessions were conducted by experienced qualitative researchers using remote video conferencing software.

The trials described here were registered in clinicaltrials.gov with registration numbers NCT04076020 and NCT04075994 and were approved by the University of Pittsburgh institutional review board. All research participants provided informed consent that allows for secondary analyses without additional consent. This manuscript used solely deidentified data. Participants received compensation up to US $150 for participation across the 4 study visits.

The decentralized research program described here demonstrates the successful enrollment of rural and metropolitan individuals with AF in clinical trials using a mobile health intervention. Our approach for remote engagement yielded geographic and racial diversity in study participants that exceeds many AF trials. Both trials fulfilled enrollment goals with participants maintaining excellent fidelity to the relational agent intervention and to the AliveCor Kardia Mobile device. Our studies consequently affirm the feasibility of conducting remote, decentralized trials with mobile health interventions, and affirm the demonstrated capacity of decentralized trials to enhance participant diversity [45,46].

AF is a complex syndrome treated with long-term, possibly lifelong, oral anticoagulation for stroke prevention, and managed with subspecialty care and procedures according to professional society guidelines [42]. Our program aimed to address the prominent challenges of health literacy, medication adherence, quality of life, and increased health care utilization that individuals with AF may experience. Our program was informed further by the consistent literature demonstrating the prominent associations of social and structural factors with care processes in AF and its related outcomes [47,48]. Community-based studies, registries, and health services analyses have identified that individuals of non-White race, lower educational attainment, lower income, and residence in neighborhoods with greater social deprivation experience more clinical adversity and limited access to AF-related care than their counterparts [15,16,49-52]. However, social and structural factors are not regularly captured in the conduct of research related to AF, potentially perpetuating disparities by precluding assessment of generalizability to individuals and populations that experience greater social disadvantage. By enrolling individuals with limited education and social resources, our program sought to enhance the access and generalizability of our research program.

Our approach further eliminated geographic and financial barriers to participation. As the intervention was delivered via smartphone, we provided smartphones to participants for the study period along with standardized instruction for their use, thereby eliminating access to contemporary technology as a barrier to trial participation. Our provision of study materials at a sixth-grade reading level and verbal administration of surveys by staff further diminished health literacy as an obstacle to engagement. Telephone-based visits likewise reduced geographic distance and travel as obstacles to participation. A further iteration of our programmatic design may include video-based visits and additional phenotypic characterization of study participants. Finally, as the literature documents both the increased adversity in women with AF accompanied by diminished participation in clinical trials [53-55], we emphasized the recruitment of women, achieving >60% (319/513) enrollment of women in both trials.

Remote trials have promise to implement novel digital interventions, such as the relational agent used here. Advantages include the provision of patient-centered education, relevant monitoring, availability, and increased attention and feedback to promote self-care. A meta-analysis determined that digital interventions have the potential to increase medication adherence–the primary outcome of our studies–by 10% (95% CI 1.00‐1.22) [56].

Concerns for the implementation of digital trials include technological literacy, access to services, and sustainability. Implementation of digital interventions necessitates attention to digital literacy, addressed here by the provision of standardized education and staff support regarding smartphone and device use. Increased dependence on technologies has necessitated the use of digital devices for communication and health maintenance but challenges approximately 20% of individuals reported limited digital literacy in one convenience sample [57]. In addition, the provision of digital technologies requires infrastructure for their effective use. Persistent disparities in Broadband access and coverage present an additional obstacle to the effective implementation of trials using digital technologies. Our study, conducted in metropolitan and rural Pennsylvania, benefited from most participants having adequate cellular coverage and access to the relational agent not being dependent on connectivity. Despite providing access to smartphones, several participants experienced challenges during the trial such as software updates and complications during use as is typical for mobile health trials.

The provision of smartphones to participants eliminated technology access as a barrier to participation. However, we recognize that such an approach would be challenging to sustain beyond the duration of the clinical trials described here. More sustained deployment of mobile health intervention requires assessment of (1) a budget impact, to appreciate the long-term costs (including technology infrastructure and maintenance) and savings of such an intervention and (2) further assessment of the facilitators and barriers that inform the implementation process of the intervention. The next steps of our program include evaluating the implementation process and ascertaining its cost-effectiveness.

We recognize several strengths of our programs. We conducted 2 decentralized trials that used a digital health intervention, exceeding recruitment goals in rural and metropolitan settings. Intervention participants demonstrated excellent fidelity with use of the relational agent. Our program also has important, noteworthy limitations that we consider foremost as pertinent to generalizability. First, we recognize rurality as highly heterogeneous and expect that our cohort of rural individuals is not representative of those in other rural contexts. Second, the rural trial was primarily White race, reflecting the region’s demographic composition, but again limiting the generalizability of our findings to more racially and ethnically diverse populations. In contrast, the metropolitan study recruited 30% (74/243) individuals of Black race and hence demonstrated greater racial diversity. Third, other settings may benefit from relational agents that are tailored for regional factors such as culture, traditions, digital services, and social and structural factors. We recognize the expansion of agent content as a priority for its implementation in other settings. Together, location, demographic composition, and relational agent design contribute to the limited generalizability of our trials. Finally, the remote design and conduct of assessments by telephone, albeit eliminating multiple obstacles, may be accompanied by a decreased opportunity for more extensive participant characterization and assessments. Remote trial investigators must balance the potential to eliminate participation barriers with the capacity to obtain more robust participant phenotyping and measurements.

We developed a decentralized, remote research program using a digital intervention. We successfully recruited and enrolled diverse participants that contrast with the relative geographic and social homogeneity of many clinical trials for AF. We intend for our program to provide a roadmap for attaining diverse study participation in digital interventions in decentralized clinical trials for chronic cardiovascular and noncardiovascular diseases.