Digital health interventions (DHIs) are rapidly reshaping health care, including the delivery of cardiovascular care. Health and fitness smartphone app utilization is increasing, with 84 million users overall in the United States in 2021 [1]. Racial and ethnic minorities have higher rates of smartphone use for health-related purposes than White individuals in the United States [2,3]. This represents a unique opportunity to reach diverse groups of patients outside hospital walls and empower them with disease management knowledge and tools. By engaging patients in self-management, DHIs have the potential to improve access, management, and ultimately health outcomes of patients with cardiovascular disease. There are more than 50,000 health-management apps available in the Apple and Android Google Play app stores [4,5], but the majority do not adequately engage patients [6]. This problem is likely multifactorial, with causes that include a lack of evaluation of patient needs and development of products without grounding within established behavior change techniques. To design successful DHIs, innovators would benefit from identifying challenges related to management of the disease from patient, caregiver, and clinician perspectives in the early phases of app development. Furthermore, to ensure that the end products are useful to patients across different age, race, ethnicity, sex, and socioeconomic groups, engaging diverse individuals in the design process is crucial. This approach has been supported by a randomized study in which app selection based on target users’ needs led to better usability and user satisfaction [7].

Human-centered design (HCD) is a core creative design methodology developed in the 1970s to facilitate innovative solution development processes. HCD methodology has been widely adopted in engineering and social sciences. More recently, HCD has been adopted for health care intervention development (eg, web-based health app co-design for refugee and migrant women, diabetes management app development, heart failure digital tool development, and personalized integrated care platform development for people with neurodegenerative disorders) [8-12]. To date, more than 100 design methodologies have been described that use different approaches to HCD, including participatory design and affinity diagramming, among other methodologies [9]. While there are differences in how various design methodologies are structured, many share the integral components of understanding end user needs through user engagement in the design process. DHIs developed with an HCD methodology have the potential to promote person-centered care, as they are designed in a way that respects patient needs, values, and preferences; allows care continuity outside the walls of the hospital; empowers patients with education; and provides emotional support. Overall, while general field guides for HCD are widely available, structured frameworks to guide innovators and investigators in designing disease management DHIs by applying HCD principles, including relevant stakeholders, and addressing the challenges in diverse patient engagement are limited. This provides an opportunity to describe our approach to applying HCD principles in the health care setting and contribute to the growing literature [13-15].

Atrial fibrillation (AF) is the most common cardiac arrhythmia. The lifetime risk of AF is 1 in 5 among Black individuals and 1 in 3 among White individuals [16,17]. AF is associated with poor quality of life [18], poor subjective health [18-20], and increased morbidity and mortality [21,22]. Scientific breakthroughs in therapy (including anticoagulation and rhythm control) and interventions (including advanced ablation techniques and left atrial appendage closure) have not been uniformly adopted in practice, especially among individuals from disenfranchised racial and ethnic groups [23-28]. In order for scientific advances to reach diverse groups of patients, there is a need to identify patient-, caregiver-, and clinician-level challenges and to engage all stakeholders in the process of creating innovative solutions. We previously developed a minimally viable AF DHI by tailoring the Corrie Health digital platform. This platform consists of the patient-facing Corrie app, connected devices, and a clinician dashboard, where actionable insights are stored to inform clinical decision-making. The Corrie Health DHI was shown to successfully improve outcomes among patients recovering from myocardial infarction [29-31]. Several features of the original Corrie app were based on behavioral change theories, such as social cognitive theory and the health belief model, which includes domains of perceived susceptibility, perceived benefits, barriers, cues to action, self-efficacy, enactive attainment, knowledge, and outcome expectations [32]. Behavior change techniques are an important addition to the patient feedback processes described above, as they promote patient engagement and uptake of DHIs.

We aimed to further develop the AF DHI on the Corrie Health platform to create a mature product that reflects the needs of diverse patients, caregivers, and clinicians. Furthermore, we aimed to summarize our findings from the process and to propose an approach to HCD tailored for DHI development in an inclusive manner: the Virtual Inclusive Digital Health Intervention Design to Promote Health Equity (iDesign) framework. This framework can be applied to the development and optimization of other DHIs with relevant stakeholder input. Our goal is to help guide clinical investigators and health app developers to create engaging, equitable, and scalable digital health solutions.

To conduct this qualitative research study, we recruited an expert designer from the Johns Hopkins Center for Health Equity (author NM) who had completed in-person training in HCD methodology. The expert designer led the design process along with a clinician (author NI) who had experience in caring for patients with AF and had also completed online training in HCD methodology. We also recruited 4 individuals to help capture the meeting discussions in real time, including 2 clinicians, 1 undergraduate student, and 1 postgraduate student.

The study was approved by the Johns Hopkins Institutional Review Board (IRB00230621).

We conducted purposeful sampling with the goal of having representation from multiple ethnic and racial backgrounds. While we were able to enroll White, Black, and Asian individuals, we were not successful in enrolling Hispanic or Native American individuals due to the very small number of eligible individuals in the source population. We recruited 8 patients with a diagnosis of AF and their caregivers or health partners (if the patient chose to include them) who were scheduled to see a cardiologist or electrophysiologist within 6 months of enrollment. All participants were recruited virtually via phone call, and consent was obtained by email. Design meetings were conducted on the Zoom platform. Virtual study procedures were chosen to allow for a safe study process in the setting of the COVID-19 pandemic. This also allowed us to increase access and engage eligible patients with a limited ability to attend in-person visits due to barriers such as work, household circumstances, and transportation [33].

Inclusion criteria were a diagnosis of AF, ownership of a smartphone (iPhone 5 or newer), residency in the United States, and proficiency in English.

Clinicians were recruited using flyers and emails with information regarding the study. Eligible clinicians included cardiologists, electrophysiologists, pharmacists, and advanced care providers. All clinicians had more than 3 years of experience caring for patients with AF. All participants provided informed consent virtually via email, as discussed above. Methods for each step of the iDesign process are outlined in the following sections and are also summarized in Figure 1.

We strived to be as inclusive as possible throughout the process, within the limitations of the COVID-19 pandemic, by (1) including a health equity expert (who was also an expert in HCD) to co-lead the HCD sessions, (2) inviting patients, caregivers, and clinicians with different backgrounds and expertise to drive the development of the components of the AF DHI, (3) offering virtual study visits that allowed for the inclusion of individuals who would not be able to take time off from work or travel to in-person visits, (4) offering one-to-one technology onboarding to allow the participation of individuals with different levels of technology literacy, and (5) ensuring that everyone contributed throughout the meetings by providing enough time for participants to speak and to reflect at the end of the session on whether their ideas were correctly captured. While the study team carefully considered providing a loaner phone with a data plan to patients who did not own that technology or have internet access, we decided at the time that device recirculation may have posed an undue risk of COVID-19 transmission.

Following step 1, a qualitative analysis of the meeting script was used to identify broad themes, which were converted into insight statements, as outlined above. We then constructed “how might we” questions from the insight statements. “How might we” questions are core components of the HCD approach and are meant to promote creativity [34]. We conducted a 1.5-hour meeting via Zoom with the goal of brainstorming and ideating solutions for the top challenges patients and caregivers expressed during the first meeting. We asked participants if the list of “how might we” questions generated from the transformation of emerging themes reflected what they shared during the first meeting. Participants were given the freedom to add or modify any questions. After confirming the accuracy of the “how might we” question list, we asked everyone to vote on the top-4 priority questions they wanted to brainstorm about.

Following the guidance of our expert human-centered designer, we decided to deploy the existing technology after the ideation phase. This was done to avoid bias in the thought process and facilitate generation of creative ideas without exposing participants to the existing technology. Additionally, we felt it was necessary for participants to have an understanding of existing capabilities and overall app design prior to engaging in prototyping.

We set up a Zoom meeting at a convenient time for each study participant to conduct individualized virtual onboarding to the minimally viable version of the Corrie app, which we named Corrie Afib. We followed a systematic approach for onboarding participants, which is described in Multimedia Appendix 1. The time required for onboarding each participant varied considerably, ranging from 15 minutes to 2 hours, based on the participant’s digital health literacy. Patients attended this session without caregivers.

During the third 1.5-hour Zoom meeting, we first asked the study participants to identify the top solutions they wanted to design from the list of solutions that were generated during the second meeting (discussed in the results section). Study participants then voted using the Zoom chat box on the top solutions they wanted to prototype. The participants were split into 3 groups, and each group was given the task of prototyping the intervention. Virtual prototyping was conducted using Google Jamboard, which lets study designers capture patients’ design ideas in real time and allows them to confirm their accuracy by continuously asking questions.

Additional ranking with study investigators and engineers was conducted, because the patients expressed their desire for other solutions to be incorporated in the DHI, although due to the time constraints of the virtual meetings, prototyping of only 3 solutions was feasible. Three study coinvestigator-clinicians and engineers met for discussion and ranked more than 30 patient ideas and more than 20 clinician ideas (Multimedia Appendices 2 and 3) that were not prototyped by patients; the ranking was based on feasibility, viability, and desirability and used a 3-point Likert scale [35]. Top-scoring solutions were prioritized for incorporation into the app, in addition to the ideas chosen by patients for prototyping. Our approach, like the HCD process, was flexible. When feasible, the ranking process can be combined with the ideation session. We then leveraged the Theory of Health Behavior Change as a conceptual model to guide further incorporation of the features in the design process [36,37]. The Theory of Health Behavior Change considers 3 major components driving individual behavior change: knowledge and beliefs, self-regulation skill and ability, and social facilitation (Multimedia Appendix 4) [36,37].

Once specific functionalities were prioritized, the engineering team began to integrate the selected features into the app.

The app functionalities informed by our iDesign process will be modified iteratively until no further major feedback is obtained from pilot clinical testing in a larger group of patients (N=100), who will be either Android or iOS users.

After one-on-one onboarding sessions with patients, we identified several learning points to further streamline the process for future studies. First, prior to connecting with the patient via Zoom, the study team member should send an introductory email explaining to the patient what they need prior to onboarding, including (1) their Apple ID and password or Google Play ID and password, (2) a stable Wi-Fi connection, and (3) a smartphone or, ideally, another device (such as a laptop, iPad, or computer) where they can watch an explanation of the onboarding process by the study team member while downloading the app to their phones and completing onboarding. Second, there is a need for 2 versions of onboarding videos, one for individuals who are more comfortable using technology and the other for those who need more guidance, to streamline the process while still addressing the needs of diverse individuals. The video directed toward individuals who are comfortable with technology should be brief and should only show each feature and give one example of using that feature, while the version directed toward individuals with low digital literacy should review the details of how to open an email from the study team and download the app, retrieve the passwords needed for app installation, set up an automated login (if desired), and give more than one example of using each feature of the app. Third, the study team should be available for questions and technical support.

At the end of the ideation meeting, we asked patients to vote on the top 3 ideas that they felt were most desirable to prototype. Table 1 shows the voting results.

Participants were then split into 3 groups, and the groups were given the task of prototyping each feature. Models of working prototypes were created with patient input for (1) providing information about AF triggers (Multimedia Appendix 5), (2) symptom and rhythm correlation for alerting the clinician team and family while experiencing AF (Multimedia Appendix 6), and (3) improved medication tracking (Multimedia Appendix 7).

A second round of ranking identified patient education, peer support, and clinician engagement in remote AF management as viable, feasible, and desirable functionalities to incorporate within the app. New or modified functionalities chosen to be incorporated were aligned with the components of the Theory of Health Behavior Change (Multimedia Appendix 4).

Engineering and development are in progress to expand the feature scope of the Corrie Afib app and the clinician dashboard based on HCD. Features include an expanded library for tailored AF education, risk factor tracking, rhythm monitoring for potential triggers, symptom and rhythm correlation for better therapeutic guidance, and community support (Multimedia Appendix 4).

Timeline for each step of the design process is included in Multimedia Appendix 8. Even after completion, this iterative process will continue with feedback obtained from patients enrolled in a pilot study of 100 participants.

In this exploratory co-design project, patients, their caregivers, and clinicians identified salient challenges in AF care, including inadequate education and guidance on AF management topics, such as lifestyle changes and guideline-directed medical therapy, inadequate resources to guide decision-making during an AF episode, and poorly structured patient-clinician communication during AF episodes. Similar qualitative and co-design experiences have not yet been described for AF management, but the challenges identified by our study participants overlap with known gaps in AF care outlined in the literature. In line with our participants’ identification of inadequate education and guidance on AF management, there is evidence that a significant proportion of individuals with AF have inadequate disease-specific health literacy [26]. Furthermore, anticoagulation therapy, a core component of AF management to prevent stroke, is only implemented in less than half of individuals who are at high risk for stroke [25]. Significant disparities in stroke prevention therapy also exist, as women and Black individuals receive anticoagulation therapy less often than men and White individuals [24,27]. In addition, Black and low-income individuals are less likely to receive rhythm control therapy, including catheter ablation [23]. Educating and empowering patients regarding these core principles of AF management, as suggested by the patients and clinicians, has the potential to improve implementation of guideline-directed therapies. We co-designed and coprototyped solutions together with patients, caregivers, and clinicians and identified top features that will inform further development of AF DHIs to address the identified challenges. Further feasibility of our AF DHI will be assessed in a pilot study and, subsequently, in a definitive clinical study to assess its efficacy in improving quality of life and cardiovascular outcomes.

To date, HCD concepts have been applied in the design of DHIs for management of diseases including diabetes, heart failure, and metabolic syndrome [10,38,39]. A company specializing in HCD was asked to design an app for Ascensia Diabetes Care Holdings AG with the goal of educating and monitoring patients with diabetes; however, a detailed description of the methods is not available [34]. Joshi et al [38] engaged 10 dietetic interns to define problems, ideate, and prototype a mobile app for patients with metabolic syndrome. While clinicians were engaged in the design process, patient-level input, which is critical to meeting the needs of end users, was missing. Zachari et al [39] designed a diabetes self-management app using participatory design, in which understanding patient challenges through empathy was at the center of the design process; however, clinician involvement was missing, as was the case for the design of a heart failure digital tool [10]. Ahmed et al [11] described an investigation targeting the need for development of an integrated care platform for individuals with neurodegenerative disease, incorporating feedback from patients, clinicians, and other stakeholders with a mixed methods approach; however, the ideation component of HCD was not reported in that study. Given the limited literature on using comprehensive methodologies to apply HCD principles specifically to DHI design, we used this paper to summarize our experience in developing an AF DHI and to outline our design framework, iDesign, as an approach to HCD (Figure 1). The iDesign framework is based on core HCD principles and can be used as a template for investigators planning to develop or optimize DHIs. Throughout the co-design process, we have made a number of findings that will help promote an inclusive design process and streamline the iDesign framework for future innovative projects (Multimedia Appendix 9). Future study teams may build on or modify individual components of the iDesign process as deemed appropriate for particular scenarios. The iDesign framework may serve as an overall guide for the process of DHI development for disease management.

First, while our study engaged a diverse group of patients, caregivers, and clinicians in the co-design process, our sample size was modest, and we may not have captured insights from participants who do not speak English or have uncommon backgrounds and experiences. However, our study included participants from various backgrounds, and both the patients and clinicians shared similar challenges and experiences, suggesting that we attained inductive thematic saturation. Furthermore, clinicians with years of experience caring for patients with a rich array of backgrounds were engaged as stakeholders, which may have counterbalanced this limitation. In a subsequent stage of testing the AF DHI, we will recruit a larger sample size with broader representation to solicit further feedback. Second, while the virtual co-design approach facilitated inclusiveness by overcoming barriers such as transportation or work schedules, a virtual environment may not be ideal for participants who have limited technology skills or do not have access to technology or the internet. By engaging patients with various levels of comfort with technology in our design process, we learned the significance of the onboarding process, in which skills can be taught to help overcome the technology-literacy challenge. To overcome a lack of access to technology and barriers to internet access, we plan to work with local communities, as well as policy makers, insurance agencies, and other stakeholders to facilitate both design and further implementation of the DHI.

We identified key recommendations from patients, caregivers, and clinicians with diverse backgrounds that informed the development of a DHI for AF management on the Corrie Health digital platform. We have extended our design process experience and formulated a structured inclusive framework, iDesign, which stems from components of HCD, for informing the development and optimization of DHIs. This framework adds to the body of knowledge on approaches to HCD, is flexible, and is transferable to the development and adaptation of DHIs.