Digital Strategies for Recruitment to Improve Diversity and Inclusion in Prospective Cardiovascular Clinical Research

Cardiovascular disease remains the leading cause of death globally, and significant disparities exist in outcomes based on sex, age, race, and ethnicity [1]. Randomized and observational prospective clinical studies are crucial for the development of evidence-based guidelines. However, women, older adults, and individuals from racial and ethnic minority groups are underrepresented in these studies relative to both their disease burden and the proportion of the general population [1]. Ensuring diverse representation in cardiovascular research is critical not only for the generalizability of findings but also for building trust within communities [2]. Addressing these gaps will require a multifaceted strategy, including novel approaches to participant recruitment and engagement [1].

Given the need for more representative cardiovascular research and the potential of digital health technologies, this study aims to assess the impact of two distinct digital recruitment strategies on participant diversity. We evaluated participant demographics in two large, prospective, multicenter cardiovascular clinical studies conducted within the Stanford Center for Clinical Research: the Apple Heart Study (AHS), which employed a largely siteless, app-centric model, and the Project Baseline Health Study (PBHS), which utilized a hybrid approach combining digital platforms with traditional site visits and intentional, targeted enrollment. The demographic composition of each study cohort was compared with United States (US) Census data from 2017, the year of initial enrollment for both studies, to evaluate their representativeness relative to the US population. By analyzing participant demographics in relation to 2017 US Census data, this study seeks to assess the potential of these digital tools and strategies to support more inclusive and representative clinical research.

This study employs a comparative observational design to evaluate participant demographics resulting from the digital recruitment strategies used in two large-scale cardiovascular studies conducted in the Stanford Center for Clinical Research: AHS and PBHS [3, 4]. Participant demographic data, including sex, age, race, and ethnicity, were extracted from each study's enrollment records. The primary comparison was the demographic distribution reported in the US Census data for the year 2017, which was also the initial year period for both AHS and PBHS.

The AHS (NCT03335800) was a prospective, single-arm, open-label study of 419,297 participants enrolled over 8 months and evaluated an irregular pulse notification algorithm on the Apple Watch [3]. It was a siteless trial with telehealth study visits and ambulatory electrocardiogram evaluation. Initial recruitment relied entirely on a digital, participant-initiated process. The AHS app was featured on the Apple App Store, and interested individuals downloaded the app. Potential participants self-declared that they met the primary inclusion criteria, including age 22 years or older, US residency, ownership of a compatible Apple Watch and iPhone, and proficiency in English. The informed consent process was conducted digitally within the app, where participants reviewed study information and provided an electronic signature. This onboarding was fully virtual, requiring no interaction with study staff or physical sites. A stated goal of the AHS team was to enroll participants from all 50 states, but no other participant demographic targets were prespecified.

An additional direct-to-participant, large-scale digital recruitment strategy was used on May 7, 2018, when a recruitment e-mail was sent to all Apple Watch users. This intervention represented a push to augment the study cohort beyond the initial app-based self-enrollment.

PBHS (NCT03154346) was a prospective, longitudinal cohort study of 2502 participants designed to establish a reference health state and analyze longitudinal multi-dimensional data [4]. PBHS investigators prespecified a goal to mimic US Census data for sex, race, and ethnicity, with overrepresentation for certain high-risk populations. PBHS utilized multiple channels to identify potential volunteers, including community advertisements, outreach after identification through the electronic health record (EHR) systems, and direct clinician referrals. All interested individuals were directed to the Project Baseline website for eligibility screening and enrollment. A call center was also available to answer questions.

Volunteers who passed the initial web-based screening were placed into an active digital waiting room. Within this platform, potential participants could explore various clinical research opportunities offered through Project Baseline, which maintained the engagement of potential participants while awaiting possible selection for the main cohort study. The digital waiting room was a virtual community for participants and an active database for researchers. Participants were provided information about ongoing studies and offered enrollment in studies for which they may be eligible. Researchers leveraged the data provided by waiting room participants to identify eligible patients for specific studies and also to pinpoint populations that matched targeted demographics and clinical characteristics. A critical distinction from AHS was the PBHS investigators' explicit, prespecified goal to recruit a cohort that reflected US Census demographics for sex, race, and ethnicity, and intentionally overrepresenting individuals with elevated risk for certain conditions, including cardiovascular disease, breast/ovarian cancer, and lung cancer. PBHS study staff actively selected participants from the digital waiting room pool based on these pre-defined demographic targets and risk profiles.

In two large, prospective, multicenter cardiovascular studies, the use of digital recruitment strategies led to a cohort of participants that was comparably aligned with the US Census. A recent 10-year analysis of randomized controlled trials used to support Food and Drug Administration approval of 35 cardiometabolic drugs found that of the total participants, 36% were women, 11% Hispanic, and 4% Black [5]. While the initial app strategy of AHS resulted in higher proportions of Hispanic and Black participants compared to these prior traditional cardiovascular studies, the direct-to-participant digital recruitment strategy resulted in increased enrollment and further improvements in racial, ethnic, and sex distributions while maintaining rates of comorbidities. Finally, the PBHS active strategy of a digital waiting room to recruit select volunteers to meet prespecified diversity targets led to a population with higher proportions of underrepresented groups as compared to AHS. This intentional, curated approach to enrollment in PBHS, facilitated by digital tools but driven by specific diversity objectives, represents a fundamentally different strategy than the open, passive recruitment model of AHS.

The comparison between AHS and PBHS suggests that the use of digital technology for recruitment is not enough. These digital tools need to be used in the service of the underlying, intentional recruitment strategy it enables. AHS's broad but largely passive digital strategy resulted in a cohort shaped largely by access to technology, whereas PBHS's active, prespecified, goal-directed digital selection process led to deliberate efforts to meet diversity objectives. Achieving diversity requires intentional design and strategic implementation of digital tools, not just their adoption.

Because this was an observational analysis, we are unable to draw definitive causal conclusions regarding the impact of specific strategies on enrollment. Both studies relied at least partly on self-reported data for demographics and health history, introducing potential for reporting bias. A limitation for AHS is the inherent selection bias introduced by the requirement for participants to own specific consumer electronic devices (Apple Watch and iPhone), which likely contributed to the underrepresentation of older and potentially lower socioeconomic status individuals. While PBHS aimed for broad representation, its reliance on four physical sites in two states (California, North Carolina) limits the geographic generalizability of its specific cohort compared to the nationwide reach of AHS. Although both approaches resulted in demographic distributions similar to the US Census, these methods may be better suited for forming general population cohorts rather than cohorts for clinical trials targeting more specific conditions or questions. Selection mechanisms that work well for inclusive, community-based recruitment may not translate directly to trials with narrower eligibility criteria. The potential for selection bias in both cohorts underscores the importance of aligning recruitment strategies with trial objectives, and highlights the need for further research into how novel digital and AI-enabled methods can be tailored for different types of trials and populations.

Digital recruitment tools may enhance diversity in participant enrollment in large cardiovascular clinical studies, potentially overcoming some limitations of traditional methods. However, this analysis suggests that prespecifying commitments to diversity and enacting active strategy in combination with digital recruitment tools can improve participant diversity. Ultimately, digital tools are enablers, but not the entire solution. The strategies evaluated here demonstrate potential scalability and applicability in both siteless and traditional site-based research paradigms and may guide the future design of more representative and equitable clinical trials of AHS.