Introduction

Clinical trials support the iterative advancement of modern medicine. Across trial designs and phases, rigorous evaluation of outcomes is critical, ensuring safety and efficacy prior to the translation of findings into clinical practice. During the past two decades, the number of registered trials on ClinicalTrials.gov has increased by over 1800%, and as of December 2024, 20,319 studies based in the United States (USA) are actively recruiting participants1. Furthermore, the World Health Organization observes that 78% of registered clinical trials are enrolling both males and females, with most trials seeking to enroll between 1 and 100 individuals2. However, as more trials are implemented, challenges in achieving population-representativeness or participant sampling commensurate with the burden of disease limit generalizability, reproducibility, and equitable distribution of the benefits and burdens of clinical research and healthcare advances3,4,5,6.

Historic overrepresentation of research participants of European descent and underreporting of social determinants of health (SDoH) data have been serially documented in clinical research7,8,9,10,11,12. For example, in 6680 genome-wide association studies (GWAS) published between 2005 and 2023, 94.5% of participants were reported to be of European descent13. Moreover, in a systematic review of COVID-19 clinical trials, SDoH data, including sex, race, and ethnicity, were routinely underreported14. Sex bias remains pervasive in clinical trials, though over/underrepresentation by sex varies across areas of biomedical research15. To address known discordance between trial participation and real-world population composition, the United States Department of Health and Human Services (HHS) published guidance on trial inclusivity (e.g., Diversity Action Plan and Enhancing the Diversity of Clinical Trial Populations) in addition to five overarching SDoH goals reported in the Healthy People 2030 initiative16,17,18. This national guidance contributes both awareness and structure towards advancing inclusive trial engagement. However, diverse enrollment challenges persist despite current recommendations, due in part to mistrust of medical institutions, uneven health literacy and healthcare access, as well as time and financial burdens19,20,21,22,23,24,25,26.

Digital recruitment and conduct of decentralized trials pose potential solutions to longstanding challenges in achieving population-representative recruitment and trial engagement27,28,29. Integration of digital health technologies (DHTs) in clinical trials may ease participant burden and increase reach among populations historically underrepresented in clinical research30,31,32,33. While fully decentralized research presents ethical concerns related to the digital divide, combining traditional and digital elements of recruitment and study implementation may offer an opportunity to achieve enrollment targets satisfying statistical power and inclusiveness amongst trial participants to promote generalizable findings28,34,35,36,37,38.

In this study, we present the recruitment strategies and cohort profile of the Engaging Adolescents in Decisions about Return of Genomic Research Results (EAS) clinical trial (NCT0448106), where traditional, targeted hybrid, and digital recruitment strategies with quota sampling were implemented to achieve population-representative enrollment of adolescents (ages 13–17) and young adults (ages 18–21). We aim to demonstrate how a multipronged recruitment approach can support more diverse trial engagement.

Results

Baseline characteristics

The EAS trial recruited participants for three years (July 2021–June 2024), during which 1524 individuals completed eligibility screening (Fig. 1). In this manuscript, we present findings for the 483 adolescent and young adult (AYA) participants who enrolled and completed study visits. Assenting adolescents ages 13–17 (n = 266, 55.1%) enrolled as dyads with a parent or legal guardian. Consenting young adults ages 18–21 (n = 217, 44.9%) could choose to enroll on their own (n = 181, 83.4%) or participate with a parent (n = 36, 16.6%). Quota sampling recruitment was implemented to balance enrollment by age, gender, and race, or ethnicity, averaging 54 participants in each of the nine eligible age groups (Supplement A). The racial and ethnic composition of the study cohort was assessed in comparison with Census data for the USA population and 100 mile eligibility radius of the academic medical center (Fig. 2). Comprehensive participant demographics are described in Table 1 following reporting recommendations from Coss et al. SDoH guidelines for DHT-enabled clinical research39. Seventeen individuals were excluded from further analysis due to an indeterminate recruitment pathway. Total counts are presented by recruitment class, as well as participant counts for known strategies and sub-strategies.

Fig. 1
figure 1

CONSORT flow diagram.

Full size image
Fig. 2: Race and Ethnicity Composition by Defined Populations.
figure 2

The research team retrieved USA race and ethnicity data from the 2022 American Community Survey (ACS) demographic and housing estimates report of Hispanic and Latino and race (https://www.census.gov/programs-surveys/acs/data.html). The same Census data source was linked with Social Explorer to retrieve demographic data for ZIP codes enclosed within a 100 mile radius, reflecting the eligibility criteria of the academic medical center situated in a tri-state area (https://www.socialexplorer.com). EAS trial data were self-reported by participants using a select all that apply questionnaire. Genome-Wide Association Studies (GWAS) Diversity Monitor (https://gwasdiversitymonitor.com/) hosts summary statistics from GWAS studies published between 2005 and 2023.

Full size image
Table 1 Baseline characteristics of the study population
Full size table

Recruitment Strategies

Traditional Recruitment (n = 165)

Recruitment enabled by offline channels including postal mailings, brick-and-mortar information sharing, and face-to-face interactions.

1. Community Research Advocate (CRA) Program (2.3%, n = 11): Research team collaboration and academic-community engagement partnership through which CRAs conducted outreach about the EAS trial at community centers and local events.

2. Institutional Biobank (2.3%, n = 11): Letters mailed to parents of institutional pediatric biobank participants (ages 13–17) or directly to biobank participants (ages 18–21).

3. Study Flyer (15.9%, n = 77): Printed flyers and flyers with tear-off contact information (Supplement B) designed by the academic medical center’s institutional marketing team.

3.1 Medical Location (11.6%, n = 56): Flyers hung around a single academic medical center and surrounding affiliated outpatient clinics.

3.2 Public Location (2.1%, n = 10): Flyers hung at local high schools, colleges, universities, and public community gathering locations.

3.3 Location Unspecified (2.3%, n = 11).

4. Study Team Community Engagement (1.0%, n = 5): Members of the research team attended community events, such as health fairs, to share recruitment information.

5. Word of Mouth (12.6%, n = 61): Information exchanged about the EAS trial directly from one individual to another.

5.1 Personal Network (8.3%, n = 40): Information about the study shared by family, friends, teachers, classmates, or colleagues.

5.2 Study Participant (1.4%, n = 7): An EAS participant shared information about the study.

5.3 Word of Mouth Unspecified (2.9%, n = 14).

Targeted Hybrid Recruitment (n = 112)

Recruitment enabled by both traditional and digital elements designed and implemented to engage a specified population.

1. Letters and Texts (23.2%, n = 112): The research team collaborated with clinics operated by the Division of General and Community Pediatrics at the academic medical center, serving a predominantly non-Hispanic Black or African American (70% of patients), low-income (80% insured by Medicaid, 10% self-pay) population, to identify prospective participants meeting EAS eligibility criteria. Epic Systems Electronic Health Record (EHR) SlicerDicer was used to query patients, specifically targeting patients between the ages of 13 and 21. In total, 5852 (sex distribution: 57.9% male and 42.1% female; ethnicity distribution: 97.0% not Hispanic or Latino and 3.0% Hispanic or Latino; race distribution: 0.05% American Indian or Alaskan Native, 1.06% Asian, 72.7% Black or African American, 0.07% Native Hawaiian or Other Pacific Islander, 20.2% White, and 5.93% unlisted race) patients who were potentially eligible participants were mailed personalized letters signed by the primary care physician affiliated with EAS (Supplement C). The letters were sent directly to 18–21 years old patients and to the parents of 13–17 years old patients, which included a QR code to the study FAQ and an interest survey. The study team used REDCap Twillio Programmable SMS capability to send up to three follow-up text messages with a link to a digital pre-screening survey to letter recipients’ phone numbers listed in the EHR.

Digital Recruitment (n = 189)

Recruitment enabled by cellular network and internet-based communication systems and online information40.

1. Email (7.5%, n = 36): Research spotlight emails sent to employee email addresses for an academic medical center with 19,514 staff and individuals who receive the institutional office of clinical research email newsletters.

2. Referral Code (5.8%, n = 28): After completing an EAS study visit, participants received a unique three-digit code to share with their social network. Participants earned $5 for each participant recruited using their unique code.

3. Social Media (8.9%, n = 43): Digital flyers developed by the academic medical center’s institutional research marketing team were posted on media platforms. Participants or unaffiliated individuals also posted information about the study on personal media accounts (Supplement D).

3.1 Facebook (5.2%, n = 25): Academic medical center’s research Facebook page.

3.2 Instagram ( < 1.0%, n = 1): Academic medical center’s research Instagram page.

3.3 Pinterest (0.0%, n = 0): Academic medical center’s research Pinterest page.

3.4 Snapchat (1.2%, n = 6): Study information shared on Snapchat by participants or unaffiliated individuals.

3.5 Friend’s account (1.4%, n = 7): Participants reported seeing EAS-related posts on a friend’s social media account, without specifying the social media platform.

3.6 Social Media Unspecified (0.8%, n = 4).

4.Virtual Word of Mouth (3.3%, n = 16): Family, friend, teacher, classmate, or colleague shared information about the study through digital platforms other than social media, such as a forwarded email or online messaging application.

5. Webpage (13.7%, n = 66): Study recruitment and contact information hosted on online platforms.

5.1 Hospital Webpage (6.2%, n = 30): Study recruitment information, contact information, and interest survey posted on the academic medical center’s clinical research website.

5.2 ClinicalTrial.gov ( < 1.0%, n = 1): Study recruitment and contact information posted on the ClinicalTrial.gov website.

5.3 Website Unspecified (7.2%, n = 35).

Recruitment analysis

The largest proportion of study participants enrolled through digital recruitment strategies (39.1%, n = 189), followed by traditional recruitment strategies (34.2%, n = 165), and the targeted hybrid strategy (23.2%, n = 112). At the trial outset, the research team intended to recruit participants through the institutional biobank and CRA program, which combined enrolled 4.6% (n = 22) of the total study population; thus, adaptive and multimodal recruitment approaches were deployed out of necessity during the remainder of the three-year recruitment window. With the addition of diverse time-delimited recruitment strategies, enrollment increased consistently across study years: 17.0% (n = 82) in Year 1 (2021 Q3–2022 Q2), 25.9% (n = 125) in Year 2 (2022 Q3–2023 Q2), and 41.6% (n = 201) in Year 3 (2023 Q4–2024 Q2) (Fig. 3). Although the targeted hybrid recruitment strategy using letters and texts had the shortest active timeline (1.5 years), this method accrued the largest number of participants from any sole strategy with a mean of 19 participants per quarter. Study flyers in a medical location, classified here as a traditional recruitment strategy, was the only strategy used by the study team to accrue participants in every quarter of the trial enrollment timeline, enrolling a mean of 5 participants per quarter.

Fig. 3: Participant Count by Recruitment Strategy Across Study Duration.
figure 3

During the three-year trial recruitment phase, 483 participants were recruited through: Traditional recruitment: CRA Program 2% (n = 11), Biobank 2% (n = 11), study flyer in medical, public, or unspecified locations 16% (n = 77), study team at community events 1% (n = 5), word of mouth—including unspecified 13% (n = 61). Hybrid recruitment: Letters and text messages 23% (n = 112); Digital recruitment: Email 8% (n = 36), referral code 6% (n = 28), social media via Facebook, Instagram, Snapchat, or unspecified or friend-referred sources 9% (n = 43), virtual word of mouth 3% (n = 16), webpage—hospital, ClinicalTrials.gov, or unspecified 13.7% (n = 66). Start and stop timelines were known for 85% (n = 408) of participants. The remaining 15% (n = 75) of participants were recruited through a method with an undeterminable timeline: Selection of multiple recruitment methods (n = 3), unspecified social media platform (n = 12), unspecified website (n = 35), unspecified study flyer location (n = 11), unspecified recruitment method (n = 14).

Full size image

Socioeconomic diversity was calculated using the Area Deprivation Index (ADI) from Federal Information Processing Series (FIPS) codes associated with participant self-reported primary address, wherein most study participants (409, 87.8%) reported addresses linked to an ADI greater than 25 on a national scale 1 (lowest deprivation) to 100 (highest deprivation)41. When comparing recruitment strategies using state-level ADI (scale: 1 = least deprived to 10 = most deprived), participants recruited through the targeted hybrid strategy lived in significantly more disadvantaged areas, with a median ADI of 7 (IQR 4–8). This was higher than those recruited via traditional strategies, who had a median ADI of 3 (IQR 1–5), and those recruited via digital strategies, with a median ADI of 2 (IQR 1–5). These differences were statistically significant in pairwise comparisons between targeted hybrid vs. traditional and targeted hybrid vs. digital strategies (p < 0.001, Fig. 4a).

Fig. 4: Participant Demographic and Recruitment Pathways.
figure 4

a Participants recruited via targeted hybrid recruitment had a higher level of deprivation compared with participants recruited via traditional and digital recruitment strategies (both p < 0.001), and participants recruited through the targeted hybrid were younger compared to the other groups (Kruskal-Wallis p < 0.001). b Gender was not a statistically significant measure for differences in recruitment strategies. c The proportion of participants from groups historically underrepresented in research differed significantly by recruitment method (p < 0.001). Participant reported confidence filling out medical forms was evaluated using a 5-point Likert scale, where significance is observed in traditional and targeted hybrid strategies (p < 0.001) and targeted hybrid and digital strategies (p = 0.0028). d The Sankey diagram depicts recruitment flow from one node to another beginning with consent type (consenting 18–21, or assenting 13–17), followed by recruitment class (Digital, Traditional, or Hybrid), recruitment strategy (the primary pathway of recruitment), and recruitment sub-strategy (the secondary pathway of recruitment).

Full size image

Age distributions also varied significantly across recruitment methods. The targeted hybrid strategy recruited a younger group, with a median age of 15 years (IQR 14–16.8). In contrast, traditional strategies and digital strategies recruited older participants, both with a median age of 18 years, and IQRs of 15–20 and 15–19, respectively (p < 0.001 for both comparisons with targeted hybrid recruitment).

Gender distribution did not differ significantly across recruitment strategies (Fig. 4b). The distribution of participation by gender is as follows: traditional strategies (man/boy 46.2%, n = 73; woman/girl 53.8%, n = 85), the targeted hybrid strategy (man/boy 52.8%, n = 57; woman/girl 47.2%, n = 51), and digital strategies (man/boy 44.3%, n = 81; woman/girl 55.7%, n = 102).

Racial and ethnic representativeness of the cohort was assessed for participants with known recruitment pathways. Participants identifying as White, non-Hispanic, who are typically overrepresented in genomic research, accounted for 48.7% (n = 227). Participants identifying with all other racial or ethnic groups, considered historically underrepresented in genomic research, accounted for 51.3% (n = 239). Representativeness was found to differ significantly by recruitment method (p < 0.001; Fig. 4c). The highest proportion of underrepresented participants was recruited through the targeted hybrid method (87.5%), and the lowest through digital recruitment (32.3%). Participant proportions by recruitment class, listed as overrepresented followed by underrepresented, were as follows: traditional (51.5%, n = 85; 48.5%, n = 80), targeted hybrid (12.5%, n = 14; 87.5%, n = 98), and digital (67.7%, n = 128; 32.3%, n = 61).

Health literacy was calculated using a 5-point Likert scale of self-reported confidence in filling out medical forms resulting in a median confidence of 4 “quite a bit” for each recruitment class, traditional (IQR 3–5), targeted hybrid (IQR 3–4), and digital (IQR 3.5–5). There was a significant difference (p = 0.0012) in medical form confidence by recruitment strategy group wherein participants recruited through the targeted hybrid strategy reported less confidence overall (Fig. 4c). A summary of participation by consent type (consenting 18–21 and assenting (13–17), recruitment class (traditional, targeted hybrid, and digital), recruitment strategy, and recruitment sub-strategy is depicted in Fig. 4d.

Discussion

The EAS trial successfully enrolled and conducted study visits with 483 AYA participants. Balanced enrollment by age, gender, and race and ethnicity was achieved using a multimodal recruitment approach consisting of five specified traditional recruitment strategies (biobank, CRA program, study team community engagement, study flyers, and word of mouth), one targeted hybrid strategy (letters and texts), and five digital recruitment strategies (email, webpage, referral code, social media, and virtual word of mouth). Of these recruitment classes, digital recruitment enrolled the largest proportion of participants, accounting for 39.1% of the total study population. However, digital recruitment also enrolled the largest proportion of trial participants self-identifying as White, non-Hispanic (56.4%, n = 128), an overrepresented group in clinical research. Traditional and targeted hybrid strategies recruited 34.2 and 23.2% of the study population, respectively, and garnered more participant diversity as compared to digital strategies. Most trial participants self-identifying as belonging to a racial or ethnic group underrepresented in genomic research were recruited through traditional and targeted hybrid strategies (74.5%, n = 178). These findings demonstrate a multipronged recruitment model for achieving recruitment targets necessary for statistical power of the study design and inclusive trial participation to counteract ongoing overrepresentation of participants of European descent in clinical research8,13,42,43,44,45,46.

Traditional, targeted hybrid, and digital recruitment strategies with quota sampling were implemented out of necessity to achieve clinical trial enrollment targets in a timely manner. Many clinical trials and clinical research initiatives focus on diversity in participant recruitment by prioritizing representation based on race and ethnicity, a critical step towards achieving population representativeness in clinical research for generalizable translation into clinical practice47,48,49,50,51,52. Moreover, stakeholder-engaged recruitment approaches in pediatric genomics research are helping to improve participant diversity, thereby supporting more generalizable findings53. Here, we expand upon these efforts with transparency about strategies for engaging historically underrepresented populations in a clinical trial through traditional, targeted hybrid, and digital recruitment strategies, and by incorporating additional SDoH features, including gender and age, in the quota sampling approach. Additionally, the targeted hybrid strategy (letters and texts) demonstrates the importance of leveraging pre-existing relationships that prospective research participants may value or trust (e.g., with a pediatrician’s clinic or a local hospital) to engage specific populations that are otherwise more challenging to recruit, in this case, younger, racially and ethnically diverse, and socioeconomically less advantaged individuals. While each additional strategy brings with it new learning hurdles and implementation challenges for the study team, these results suggest a promising future towards diverse and inclusive clinical research through multimodal recruitment reflecting parallel implementation of traditional, hybrid, and digital strategies54.

This analysis of recruitment strategies implemented in the EAS trial provides real-world evidence to inform future clinical trial design, encouraging the utilization of digital, traditional, hybrid, and targeted recruitment methods. However, limitations exist. The EAS trial was not designed to analyze recruitment approaches; thus, studies formulated to specifically assess reach, engagement, and enrollment of diverse study populations comparing traditional, hybrid, and digital recruitment strategies ought to inform future clinical trial recruitment plans. Importantly, individuals without decision-making capacity and with known or suspected genetic conditions were excluded from participation, which may limit representativeness of the cohort in relation to the population that presents for clinical genetic testing and better approximate representativeness for population genomic screening55,56. Study materials were only available in English, diminishing the prospect of participation by individuals with limited English language proficiency. While access to personal mobile and computing devices and internet access were not prerequisites for participation in the study, as participants could complete all study procedures in person at the academic medical center, it is possible that groups without digital access, experiencing housing insecurity, with complex health needs, or with social challenges may have faced barriers to participation57. Also, because both assenting adolescent and consenting young adult participants enrolled in the EAS trial, findings from recruitment for this clinical trial may not be generalizable to recruitment of other study populations. As such, future work is necessary to further understand the impact of recruitment modality on trial diversity.

In conclusion, these findings demonstrate the value of a multipronged recruitment approach. Digital strategies can help accrual to enable sufficiently powered statistical analyses by reliably enrolling populations historically overrepresented in clinical studies. Traditional and targeted hybrid strategies may better promote trust and connection with populations that have historically been less likely to participate in clinical research. In combination, traditional, targeted hybrid, and digital recruitment strategies with quota sampling can demonstrably facilitate enrolling a more population-representative and inclusive study cohort. As trials and technology advance, ongoing evaluation of recruitment strategies will be necessary to facilitate diverse representation in clinical research.

Methods

This study is a non-randomized, single-assignment clinical trial evaluating the use of an electronic decision-making tool to learn adolescents’ and parents’ preferences for return of genomic research results for preventable, treatable, and adult-onset conditions as well as carrier status for some autosomal recessive conditions (CONSORT Checklist Supplement E). More details about the primary aims of the clinical trial can be found in the trial protocol58. This manuscript presents findings from the secondary analysis of recruitment data.

Study Design and Setting

The EAS trial was conducted at an urban pediatric academic medical center located in a tri-state area. AYAs between 13–21 years of age and a parent /legal guardian (optional for 18–21 year-olds) were invited to participate in a genetic decision-making study. During an in-person or virtual study visit, participants were asked to complete an online decision tool to make choices about what, if any, genetic information they would like to receive about the adolescent. AYAs and a parent/guardian (if applicable) completed the digital decision tool independently, then jointly with a study facilitator, a member of the study team with clinical genetics or shared decision-making experience. AYAs who chose to receive personal genetic test results provided a blood or saliva sample for DNA extraction and variant analysis. The research team returned results based on participants’ choices. Participants could receive up to $180 in compensation for their participation in the trial. This research study was approved by the Cincinnati Children’s Hospital Medical Center’s Institutional Review Board.

Participant eligibility

AYAs were eligible to participate in the study if they were 13–21 years old, lived within 100 miles of the academic medical center, and could complete a study visit conducted in English. AYAs were not eligible to participate if they had a developmental disability that interfered with their ability to make decisions for themselves, had ever been followed regularly in a genetics clinic or received a molecular diagnosis for a genetic condition, or were a full biological sibling of another AYA EAS participant. Prior to starting a study visit, study staff provided an opportunity for prospective participants to learn about study procedures and provide assent or consent to participate via phone or video call using REDCap E-Consent tools59,60.

Sample size and composition

The trial aimed to achieve near equal numbers of participants in each eligible age group (13-21) and gender identity (boy/man, girl/woman), with no more than 65% of participants identifying as White, non-Hispanic. The clinical trialists proposed these thresholds for enrollment after findings from the Electronic Medical Records and Genomics Network Phase III (eMERGE III) cohort intervention study suggested differences in decisions by sex, age, and self-identified race. Specifically, Black and African American adolescent participants and their parents had higher rates of discordance in the type and amount of personal genetic information they would like to learn about the adolescent than other participants61. However, fewer than 17% of the eMERGE III cohort self-identified as Black or African American, therefore, the EAS trial was designed to purposively recruit AYAs from populations that have been underrepresented in genomic research25,61. The investigators hoped that garnering a greater proportion of participants who identified as belonging to a racial or ethnic minority group would generate additional insight and more generalizable results than prior research58.

Quota sampling methods were used with a goal of recruiting 240 assenting adolescent/parent dyads and 200 consenting young adults. The sample size for this study was determined by the primary objectives of the clinical trial. While many of the primary comparisons would be sufficiently powered with smaller sample sizes, enrolling 440 participants in this manner would allow the research team to closely examine age effects and account for covariates in analyses. All power calculations used alpha = 0.05. Details about the sample size calculation can be found on ClinicalTrials.gov for trial NCT04481061.

Recruitment strategies

The clinical trialists initially proposed two recruitment strategies, the institutional biobank and CRA program, to recruit the trial participants. They aimed to recruit similar numbers of participants from each source to compare decisions made by current biobank research participants and participants who may be unfamiliar with research participation who reside in communities that have historically been underrepresented in research. However, due to COVID-19 pandemic-related limitations on face-to-face community engagement and insufficient enrollment, additional strategies were required to accrue the proposed trial cohort of 440 participants. Therefore, three recruitment classes, Traditional, Targeted Hybrid, and Digital, were implemented throughout the remainder of the recruitment phase.

Demographic Survey and Participant Data

Participant demographic data was obtained via an online self-reported survey (Supplement F). The Neighborhood Atlas was used to derive ADI for each participant from 2022 American Community Survey (ACS) data associated with the 12-digital FIPS code for participant-reported primary place of residence via the Federal Communications Commission Census Block Conversions Application Programming Interface (v1.0.0)41. The recruitment strategy for each participant was determined in the pre-screening eligibility survey that prospective EAS participants completed independently online, over the phone, or video call with a study team member. Participants were asked “Where/how did you learn about this study?”, with a finite list of options and the ability to select as many as apply.

Statistical analysis

JMP statistical software was used to perform statistical analyses. Descriptive analyses included frequencies for categorical outcomes, means, and standard deviations for the normally distributed continuous measures, and median and interquartile range for non-normally distributed continuous data. To compare dichotomous outcomes between recruitment strategies, contingency tables and goodness of fit tests were performed. To compare continuous outcomes between recruitment methods, Kruskal-Wallis Tests were performed. In addition, pairwise comparisons were performed to gain insight about which recruitment strategies were driving overall differences.