Enhancing Efficiency and Reach Using Facebook to Recruit Breast Cancer Survivors for a Telephone-Based Supportive Care Randomized Trial During the COVID-19 Pandemic

Abstract

PURPOSE

Evidence supporting social media–based recruitment of cancer survivors is limited. This paper describes how we used Facebook during the COVID-19 pandemic to augment our recruitment of breast cancer survivors for our two-site telephone-based randomized clinical trial (RCT) at Dartmouth-Hitchcock Medical Center and the University of Alabama at Birmingham.

METHODS

Originally a two-site RCT of a telephone-delivered breast cancer survivorship intervention, we extended our clinic-based recruitment to Facebook. Participant characteristics, geographic reach, and baseline outcomes were compared across recruitment sources (ie, two clinics and Facebook) using descriptive statistics and effect sizes.

RESULTS

Enrollment rates (20%-29%) were comparable across recruitment sources. The 21-month Facebook marketing campaign accounted for 59% (n = 179/303) of our total sample and had the greatest geographic reach, recruiting women from 24 states. The Facebook campaign reached a total of 51,787 unique individuals and cost $88.44 in US dollars (USD) per enrolled participant. Clinic samples had a greater proportion of women who were widowed (8% v 1%; P = .03) and Facebook had a higher proportion of women with a household income over $40,000 USD (83% v 71%; P = .02). There were no statistically significant differences between Facebook and the two clinics on baseline survey scores.

CONCLUSION

Augmenting traditional recruitment with Facebook increased our RCT's geographic and sociodemographic reach and supported meeting recruitment goals in a timely way. In the wake of the COVID-19 pandemic, cancer survivorship researchers should consider using social media as a recruitment strategy while weighing the advantages and potential biases introduced through such strategies.

INTRODUCTION

Restrictions imposed by the COVID-19 pandemic limited in-person clinical research trial recruitment.¹ This was especially true in oncology research as clinicians sought to protect immunocompromised patients, research staff, and themselves. Hence, in 2020, most clinical trials were curtailed and nearly 85% of research laboratories were affected.² Research teams scrambled to pivot to fully remote recruitment, capitalizing on recent trends to use telehealth¹ and social media as clinical research tools.³

CONTEXT

• Key Objectives

• To gain insight into how Facebook recruitment compared with clinic-based recruitment for a clinical trial and how it might be used or adapted to meet the needs of oncology researchers.

• Knowledge Generated

• Facebook extended the geographic reach of our study and accounted for the highest proportion of participants in our sample with comparable baseline characteristics and outcome scores as participants recruited through clinics. Cost per participant enrolled via Facebook was similar or less than previously reported costs of social media recruitment.

• Relevance

• Facebook recruitment can support meeting recruitment goals for oncology clinical trials while maintaining external validity. This strategy may be particularly appropriate for telephone-based oncology trials where the intervention purpose is easily communicated, and eligibility criteria do not require medical testing. Enhanced recruitment efficiency using Facebook may further benefit oncology researchers by reducing cumbersome, time-consuming recruitment methods, shown to lead to job dissatisfaction among clinical research coordinators.

We began recruitment for a two-site clinical trial 6 months before the pandemic began. We were testing a telehealth rehabilitation intervention using goal setting and problem solving to enhance participation in valued activities among women experiencing functional limitations in the first year after breast cancer treatment.⁴ Both the intervention and attention control conditions were delivered via telephone; however, our recruitment procedures depended on in-person clinic interactions.

When pandemic protocols limited in-person clinic recruitment, we decided to augment recruitment with a direct-to-consumer marketing approach via social media. We advertised on Facebook, the most widely used social media platform for research recruitment, where, in the United States, an estimated 264 million people are active monthly users.^5,6 Although recruitment through social media has been shown to be feasible and cost-effective,^3,7 only one clinical trial focused on cancer survivors had reported the outcomes of social media–based recruitment efforts.⁸

As researchers consider strategies for conducting clinical trials in the wake of the COVID-19 pandemic, and in the context of an increasingly digitally connected global population, a greater understanding of social media–based recruitment methods for oncology clinical trials is needed. Before launching our social media recruitment campaign, we wondered how the campaign might affect the external validity of our study. For example, reviews in nononcology populations have suggested samples recruited via social media were younger and had little racial diversity.^9,10 We describe a retrospective analysis of our recruitment process, which explored the following three questions: (1) To what degree did each recruitment source (ie, two clinics and Facebook) contribute to our target sample of 300 women? (2) What was the cost per enrolled participant recruited via Facebook? (3) To what degree did participants' baseline characteristics and geographic location vary by recruitment source? Our goal was to illuminate how similar recruitment strategies might be deployed in other telephone-based oncology clinical trials.

METHODS

Study Overview

This two-site randomized clinical trial (RCT) tests a remotely administered supportive care intervention informed by the behavioral therapies of behavioral activation^11,12 and problem-solving treatment.^13,14 The telephone-based intervention was designed to foster recovery by helping breast cancer survivors set and attain weekly goals regarding self-selected activities. This intervention and an attention control condition comprising cancer survivorship education were delivered over nine telephone sessions.⁴ The goal of this baseline data analysis was to compare the processes and productivity of the clinic versus social media recruitment sources. This analysis was conducted before completing primary data collection and analysis; primary outcomes of the RCT will be reported elsewhere. This study received institutional review board (IRB) approval; Massachusetts General Hospital IRB serves as the IRB of record for Dartmouth Hitchcock Health and University of Alabama at Birmingham.

Sites

The original recruitment plan was to be implemented at the breast cancer clinics within two National Cancer Institute–designated comprehensive cancer centers: Dartmouth Cancer Center at Dartmouth-Hitchcock Medical Center (DH) in New Hampshire and the University of Alabama at Birmingham (UAB). Because of lack of racial diversity in the DH catchment area, UAB was intentionally selected to recruit a more diverse breast cancer survivor sample. Each site had its own clinical research coordinators (CRCs) who managed recruitment activities and its own randomization scheme stratified by time since completing breast cancer treatment (<6 months v >6 months) and treatment received (chemotherapy v no chemotherapy).

Participants

Eligibility Criteria

Inclusion criteria included female participants age 18 years and older; within 1 year of completing locoregional treatment and/or chemotherapy with curative intent for stages I-III breast cancer; and self-reporting functional limitations as defined by a score of ≥10 on the Work and Social Adjustment Scale (WASA), a CRC-administered five-item screening measure.¹⁵ Exclusion criteria included non–English-speaking; noncorrectable hearing loss; moderate-severe cognitive impairment (scoring <3 on a six-item cognitive screener)¹⁶; history of severe mental illness (eg, schizophrenia and bipolar disorder); current major depressive disorder; and active suicidal ideation or active substance misuse.

Recruitment Procedures

Clinic-Based

Our original recruitment procedure, managed by CRCs, included (1) screening clinic schedules (at DH and UAB) or querying a data repository (at UAB) to identify potential participants, (2) confirming potential eligibility with clinicians, (3) delivering a study brochure to the potential participant during a clinic appointment, and (4) formally screening for eligibility and obtaining informed consent. We had enrolled 19 participants via this method when COVID-19 restrictions prevented CRCs from being physically present in oncology clinics. Thereafter, we modified our clinic recruitment procedure by having CRCs mail potential participants a study brochure and letter signed by the treating oncologist and then contacting them by telephone to screen for eligibility and initiate informed consent. Our IRB-approved remote consent procedure provided women with a consent form by mail or email, which the CRC reviewed by telephone. Women then signed the consent form electronically and were provided a dually signed copy for their records.

Social Media

We consulted High Level Marketing, LLC, a nationwide digital marketing company, to develop and manage a Facebook marketing campaign. The campaign goal was to advertise the study and allow interested and potentially eligible breast cancer survivors to securely submit their contact information (ie, full name, telephone number, and email address) to the study team who would follow-up with more information. Two advertisements were developed and deployed (Fig 1). Initially, advertisements were restricted to the two study sites catchment areas (ie, Northeast and Southeast United States). On the basis of early observations of enhanced recruitment via Facebook and anticipating that pandemic restrictions would continue for the foreseeable future and hoping to meet our original recruitment time frame, we expanded the campaign to the entire east coast. We limited advertisements and recruitment to the Eastern and Central Standard Time zones to enhance practicality of DH and UAB study staff communicating with participants. The campaign lasted for 21 months (August 2020 to April 2022) and cost $15,831 in US dollars (USD).

FIG 1.

Facebook ads. UAB, University of Alabama at Birmingham.

CRCs from both sites responded to women who answered our Facebook advertisement (lead) by email and telephone, making at least three attempts to contact each woman. Eligible women recruited via Facebook who consented to enroll were randomly assigned according to the randomization scheme at the study site of the CRC who completed the consent process.

Data Collection

DH CRCs, blind to randomization assignment, administered baseline outcome assessments by telephone, entering data into an online Research Data Capture database in real time.¹⁷ Baseline outcome measures are detailed in the protocol paper⁴ and in Table 1.

TABLE 1.

Baseline Outcome Measures

Analysis

To explore the recruitment outcomes by recruitment source (ie, DH clinic, UAB clinic, and Facebook), we created a recruitment flow diagram. We compiled descriptive statistics of social media performance including clicks (ie, clicks on the advertisement to view more study information), leads (ie, women who provided their contact information through the Facebook ad), click through rate (CTR; ie, the number of leads generated over the number of clicks), and cost per lead (ie, total cost/number leads) and cost per enrollment (ie, total cost/number enrolled).

To explore recruitment reach, we plotted states and zip codes represented in the samples from each recruitment source. We assessed participants' rurality (ie, whether they lived in a rural or urban setting) by classifying participants' zip codes by Rural-Urban Commuting Area codes.³³ SAS software (SAS Institute, Cary, NC) and Power BI (Microsoft Press, Redmond, WA) were used to classify zip codes and map individual participants' geographic locations.^34,35

To compare the sample's characteristics and baseline scores according to clinic-based or Facebook recruitment sources, we used measures of effect size (Cohen's V for categorical variables and r-squared for continuous variables). Cohen's V was considered small at 0.07, medium at 0.21, and large at 0.35. R-squared was considered small at 0.01, medium at 0.06, and large at 0.14. The level of statistical significance was established at P < .05. To further explore our sample, we also compiled descriptive statistics on the baseline characteristics for the three recruitment sources (ie, DH clinic, UAB clinic, and Facebook). Analyses were performed using SAS 9.4 statistical package (SAS Institute, Cary, NC).³⁵

RESULTS

Enrollment Outcomes by Recruitment Source

Figure 2 describes recruitment flow for each of the three sources: DH clinic, UAB clinic, and Facebook. Three-hundred and eighty patients were identified as potentially eligible from DH, of whom 124 (33%) were ineligible, primarily because of scoring low on the WASA. Of 256 potentially eligible women, 190 (74%) did not enroll and 66 (26%) did. Notably, some women declined before completing the WASA, which prevents us from ascertaining whether they would have been eligible for the study. The 66 women recruited via DH accounted for 22% (n = 66/303) of total enrolled.

FIG 2.

Recruitment flow. DH, Dartmouth Hitchcock; FB, Facebook; UAB, University of Alabama at Birmingham; WASA, Work and Social Adjustment Scale.

At UAB, 497 patients were identified and 297 (60%) were ineligible, most often because of being beyond 1 year since treatment or having a cancer stage of 0 or IV. The relatively high proportion of ineligible women was an artifact of querying a deidentified data repository (v reviewing clinic schedules as done at DH) to identify potential participants at this site. Of 200 potentially eligible from UAB, 142 (71%) women did not enroll and 58 (29%) did. Once again, eligibility of those who declined before completing the WASA screening could not be ascertained. UAB accounted for 19% (n = 58/303) of the total enrolled sample.

Facebook advertisements yielded 1,119 leads. Of those, 222 (20%) were ineligible, primarily because of being beyond 1 year since treatment. Of 897 potentially eligible, 718 (80%) women did not enroll and 179 (20%) did. Facebook accounted for 59% (n = 179/303) of the total enrolled sample.

Facebook Performance and Cost

During the Facebook campaign, advertisements were displayed 19,431 times/month, reaching a total of 51,787 unique individuals. It was possible for individuals to see the same ad more than once; on average, each unique individual saw the ad 4 times over the course of the campaign. The ads received a collective 20,663 clicks and the overall CTR was 1,119/20,663 = 5.42%. The total cost of the Facebook ad campaign was $15,831 (USD), which translated to $0.77 (USD)/click, $14.15 (USD)/lead, and $88.44 (USD)/enrolled participant.

Geographic Reach, Participant Characteristics, and Baseline Survey Scores by Recruitment Source

Geographic Reach

Figure 3 shows participant zip codes by recruitment source. Participants came from 26 states. Not surprisingly, Facebook had the greatest reach, recruiting women from 24 states, followed by DH (nine states), and UAB (five states). Thirty-four percent of the total sample was classified as living in a rural zip code. The Facebook and combined clinic samples differed significantly as only 18% of the Facebook sample lived in a rural setting, whereas 54% of the combined clinic sample lived in a rural setting (effect size V = 0.39; P < .0001). The Appendix Table A1 (online only) reveals that this difference was driven by DH sample having the highest proportion of rural participants (82.2%) compared with UAB (13.8%) and Facebook (17.5%).

FIG 3.

Site map (RUCA). Urban RUCA codes-1.0, 1.1, 2.0, 2.1, 3.0, 4.1, 5.1, 7.1, 8.1, 10.1. Rural RUCA codes-4.0, 4.2, 5.0, 5.2, 6.0, 6.1, 7.0, 7.2, 7.3, 7.4, 8.0, 8.2, 8.3, 8.4, 9.0, 9.1, 9.2, 10.0, 10.2, 10.3, 10.4, 10.5, 10.6. DH, Dartmouth Hitchcock; RUCA, Rural-Urban Commuting Area; ZIPA, zipcode tabulation areas.

Participant Characteristics

Of the 303 consented participants, 17 withdrew before baseline assessment and two were withdrawn when found to be ineligible, resulting in 284 participants completing baseline assessments. Table 2 shows characteristics and baseline outcome scores for these 284 participants. Overall, participants' mean age was 56.1 (10.2) years (Facebook, 55.7 [9.1] years; combined clinics, 56.5 [11.5] years). Most were White (82.4%), employed (53.9%), and had at least a college education (67.4%). Most participants were married (66.1%), had private health insurance (63.7%), and an income of at least $40,000 USD/y (77.7%). Most participants had either stage I or II disease (81.7%) and almost all had undergone surgical treatment (99.3%). The only significant differences in the samples recruited from Facebook versus the combined clinics sample were that clinic samples had a greater proportion of women who were widowed (8% v 1%; P = .03) and Facebook had a higher proportion of women with a household income over $40,000 USD (83% v 71%; P = .02). Appendix Table A1 reveals that this latter difference was driven by UAB having a higher proportion of women with lower income.

TABLE 2.

Baseline Participant Characteristics

Baseline Surveys

There were no statistically significant differences in baseline PROMIS Satisfaction with Social Roles and Activities scores—the RCT's primary outcome—between sources. The average scaled score for each source was below the population norm (ie, standardized T-score) of 50, which was expected, given that we intentionally screened to identify women experiencing functional limitations. Overall, participants' average anxiety score was 9.2 (4.3), which is considered borderline elevated anxiety.³² There were no significant differences between the Facebook and the combined clinic samples on any of the baseline survey scores.

DISCUSSION

Facebook accounted for the highest proportion of participants (59%) in our sample. The campaign reached over 50,000 individuals and yielded over 1,000 leads. Although 80% of these leads did not enroll, we found that Facebook recruitment processes were efficient. One reason for the efficiency was that CRCs did not need to perform time- and effort-intensive medical record screening to identify potential participants. Additionally, the Facebook leads were self-populated by individuals who were interested in learning more about the study, essentially eliminating a barrier often encountered by CRCs when making unsolicited approaches to potential participants to gauge interest in the study. This recruitment strategy may be particularly appropriate for telephone-based oncology clinical trials where the intervention purpose is easily communicated, and the eligibility criteria do not rely upon medical testing. These efficiencies may have numerous benefits for oncology clinical trials as cumbersome protocols and overwhelming schedules have been shown to lead to oncology CRCs' job dissatisfaction.^36,37 Hence, social media recruitment may also support staff retention, a common source of financial burden for research laboratories.³⁸

Our cost per enrolled Facebook participant ($88.44 USD) is aligned with previously reported costs of social media recruitment methods⁶ and was substantially lower than the cost per person reported by Juraschek et al⁸ in their oncology clinical trial ($794 USD), owing to our higher CTR (5.42% v 0.5%) and enrollment rate. Although our higher rates may have been influenced by increased social media use during the pandemic, the efficiency and reasonable cost of our experience provide additional support for the use of social media recruitment for oncology clinical trials.

Facebook recruitment extended the geographic reach of our study, and it did not appear to bias our sample relative to baseline survey scores or sociodemographic characteristics. Although Black and Hispanic women were relatively small proportions of the Facebook sample (17% and 3%, respectively), our advertisements did not directly target minority populations, and compared with a recent review of minority representation in oncology clinical trials, our Facebook recruitment of Black participants was higher than other breast cancer RCTs (17% v 7.2%) and equal to that of Hispanic breast cancer participants (3%).³⁹ Furthermore, the 17% of participants recruited via Facebook who identified as Black women mirrors the proportion of Black individuals living in the United States (13.6%).⁴⁰ These results suggest this recruitment method may provide opportunities to address the common problem of limited racial diversity in clinical trials,⁴¹ especially if implemented by sites that lack racial diversity. Although evidence suggests that targeted social media campaigns may enhance recruitment of minorities and marginalized populations,⁴¹ precautions must be taken to avoid harm or to propagate health disparities related to limited e-health literacy, access, or stigma.^42-45 Future studies may consider using microtrials—a method developed by Facebook using a proprietary algorithm to target under-represented groups through tailored advertisements and targeted zip codes⁸—to enhance minority population recruitment. Although researchers should weigh potential biases introduced through social media recruitment,⁴⁶ our results suggest that this strategy can produce a sample aligned with the study's recruitment goals.

Since our team had not previously used social media as a recruitment method, working with an experienced media firm to conceptualize and conduct our campaign was critical. The firm facilitated a streamlined navigation of regulatory and security issues, minimizing administrative burden for our team—a commonly perceived disadvantage of social media recruitment^47,48 Aside from collaborating with such a media firm, the recent review by Darko et al,⁶ published after we had met our recruitment goal, provides valuable recommendations for best practices on social media recruitment including use of study-specific recruitment webpages and secure data collection landing sites. We suggest that study teams review these recommendations and a useful checklist proposed by Gelinas et al⁴⁹ before initiating social media recruitment to avoid privacy and confidentiality ethical challenges and to avoid delays IRB approvals.

The parent study was not designed to test the efficacy of the different recruitment sources. This is an observational, retrospective analysis, limiting the generalizability of our results. Furthermore, we are unable to estimate and compare the cost of in-clinic recruitment to Facebook recruitment because we had the same team of CRCs managing both sources and we did not have them track their time with that distinction. Additionally, the costs of the social media campaign presented here do not include the labor cost of the CRCs' time spent in recruiting and consenting participants. Finally, it is important to note that we were not able to access the medical records to verify the Facebook sample's diagnosis—a major limitation of that recruitment strategy that may make this an inappropriate method in some studies. Other researchers have noted the importance of using techniques to identify individuals who many be misrepresenting themselves to participate in research that compensates them for their time.⁵⁰ Although we did remunerate participants for completing each of the four study survey batteries ($105 USD total if participants completed all four time points), our prolonged personal engagement (ie, nine intervention sessions focused on recovery from cancer treatment) made patient misrepresentation less likely than it might be in a cross-sectional survey study.

In conclusion, as oncology researchers consider strategies for conducting clinical trials in the wake of the COVID-19 pandemic, augmenting traditional recruitment efforts using social media appears to be an efficient and useful strategy. Our study adds to limited evidence that using paid advertisements through Facebook can increase telephone-based oncology clinical trials geographic reach and can achieve timely meeting of recruitment goals at a reasonable cost while maintaining external validity. Future research should further assess the effectiveness of using social media to recruit cancer survivors for clinical trials including benefits and barriers of other social media platforms, direct comparison of costs to traditional recruitment methods, and barriers and facilitators to implement these strategies more broadly.

APPENDIX

TABLE A1.

Sample Characteristics and Baseline Scores According to Recruitment Source

AUTHOR CONTRIBUTIONS

Conception and design: Deborah B. Ejem, Stephen Wechsler, Sarah Gallups, Courtney J. Stevens, Mark T. Hegel, Marie Bakitas, Kathleen D. Lyons

Financial support: Kathleen D. Lyons

Administrative support: Sarah Gallups, Jazmine Coffee-Dunning, Daphne Ellis, Stacey A. Ingram, Danielle Cloyd, Kathleen D. Lyons

Provision of study materials or patients: Jazmine Coffee-Dunning, Kimberly Keene, Mary Chamberlin, Daphne Ellis, Stacey A. Ingram, Marie Bakitas, Kathleen D. Lyons

Collection and assembly of data: Deborah B. Ejem, Sarah Gallups, Jazmine Coffee-Dunning, Kimberly Keene, Mary Chamberlin, Daphne Ellis, Stacey A. Ingram, Valerie M. Lawhon, Tiffany Gilbert, Kali Morrissette, Karen Thorp, Megan Codini, Jennifer Echols, Danielle Cloyd, Sarah dos Anjos, Colleen Muse, Susan Goedeken, Kristen Elizabeth Laws, Jennae Herbert, Marie Bakitas, Kathleen D. Lyons

Data analysis and interpretation: Deborah B. Ejem, Stephen Wechsler, Sarah Gallups, Sarah Khalidi, Jazmine Coffee-Dunning, Aoyjay P. Montgomery, Courtney J. Stevens, Gabrielle B. Rocque, Andres Azuero, Maria Pisu, Jamme Morency, Robin Newman, Marie Bakitas, Kathleen D. Lyons

Manuscript writing: All authors

Final approval of manuscript: All authors

Accountable for all aspects of the work: All authors

AUTHORS' DISCLOSURES OF POTENTIAL CONFLICTS OF INTEREST

Enhancing Efficiency and Reach Using Facebook to Recruit Breast Cancer Survivors for a Telephone-Based Supportive Care Randomized Trial During the COVID-19 Pandemic

The following represents disclosure information provided by authors of this manuscript. All relationships are considered compensated unless otherwise noted. Relationships are self-held unless noted. I = Immediate Family Member, Inst = My Institution. Relationships may not relate to the subject matter of this manuscript. For more information about ASCO's conflict of interest policy, please refer to www.asco.org/rwc or ascopubs.org/op/authors/author-center.

Open Payments is a public database containing information reported by companies about payments made to US-licensed physicians (Open Payments).