Community engagement through social media: A promising low-cost strategy for rural recruitment?

Abstract

Purpose:

For the same reasons that rural telehealth has shown promise for enhancing the provision of care in underserved environments, social media recruitment may facilitate more inclusive research engagement in rural areas. However, little research has examined social media recruitment in the rural context, and few studies have evaluated the feasibility of using a free social media page to build a network of rural community members who may be interested in a research study. Here we describe the rationale, process, and protocols of developing and implementing a social media approach to recruit rural residents to participate in an mHealth intervention.

Methods:

Informed by extensive formative research, we created a study Facebook page emphasizing community engagement in an mHealth behavioral intervention. We distributed the page to local networks and regularly posted recruitment and community messages. We collected data on the reach of the Facebook page, interaction with our messages, and initiations of our study intake survey.

Findings:

Over 21 weeks, our Facebook page gained 429 followers, and Facebook users interacted with our social media messages 3,080 times. Compared to messages that described desirable study features, messages that described community involvement resulted in higher levels of online interaction. Social media and other recruitment approaches resulted in 225 people initiating our in-take survey, nine enrolling in our pilot study, and 26 placing their names on a waiting list.

Conclusions:

A stand-alone social media page highlighting community involvement shows promise for recruiting in rural areas.

Strong participant recruitment and retention underlie successful clinical trials. However, these tasks can be challenging and resource-intensive, particularly in rural areas. ¹ As social media use has gained popularity researchers have turned to these web-based, interactive platforms as another avenue for participant recruitment. ^2–4 However, few studies have examined the feasibility of social media recruitment in the rural context, a missed opportunity given the increased uptake of technology. ^5–8 Here we describe the rationale, process, and protocols of developing and implementing a social media approach to recruitment for a randomized control trial evaluating an mHealth intervention in Appalachia. We also evaluate the degree to which these strategies resulted in engagement with messages on the study’s social media page and directed potential participants to our study’s intake survey. This paper adds to existing research by (1) assessing the feasibility of using a free social media page for study recruitment in a rural setting and (2) evaluating the success of social media recruitment messages regarding their ability to generate interaction with potential participants.

Study Recruitment in Rural Areas

Recruitment in rural areas poses particular challenges, as these communities are likely to be geographically isolated, have low population density, and have particular cultural and social perspectives that may undermine study participation, including strong values of self-reliance and privacy and skepticism about science and sources of health information. ^1,7,9 Additionally, since most researchers are located in metropolitan academic medical centers, they often are perceived as outsiders, and participation may require costly and time-consuming travel. ¹⁰

These recruitment challenges are particularly salient in Appalachian Kentucky, where the present study was located. Of Appalachian Kentucky’s 54 counties, 36 are designated rural (nonmetro and not adjacent to small metropolitan areas), 14 are designated nonmetro (adjacent to small metropolitan areas), and only 4 are small metropolitan areas (population <1 million). ¹¹ Residents of rural Appalachian counties tend to be older than the national average, have less education, and are more likely to be unemployed and living in poverty. ¹² Despite these characteristics, which previously have been associated with hesitancy about study participation and adoption of new technologies, ^13–15 our formative research supported implementing social media approaches. ¹⁶

Recruitment Using Social Media

Social media refers to web-based platforms that allow users to connect with others and to create and share content. ¹⁷ In the last two decades, these platforms have gained popularity across the United States; today an estimated 72% of U.S. residents use at least one social media platform. The existing research suggests that recruiting study participants via social media has the potential to be less costly than traditional methods (e.g., newspaper, radio) and to identify study candidates more quickly. ^2,4 In rural areas, social media recruitment may provide additional benefits similar to telehealth. Like social media, telehealth has grown in popularity with increasing access to technology in underserved areas. ¹⁸ By limiting the need for extensive travel, telehealth has the potential to increase access to care in underserved environments. ¹⁹ Similarly, by reducing the need for in-person interactions and facilitating the distribution of messages across geographically dispersed audiences, social media recruitment has the potential to create more inclusive engagement in research. ⁴

Methods for Social Media Recruitment

Most research recruitment efforts that have deployed social media have used paid advertisements to distribute recruitment messages. ² This approach guarantees a large distribution of messages while maintaining the potential to target specific populations based on geography, demographics, and other characteristics. Although paid social media advertising costs less than newspaper or radio ads, ⁴ paid social media recruitment does not take advantage of the full potential of social media for research recruitment and community engagement. In contrast, stand-alone social media pages, which are free, provide a channel to engage community members in the research project by acknowledging community contributions and providing regular study updates. These pages also allow a project to build a network of followers and provide opportunities for interactions and feedback: community members can comment on, react to, and share study messages.

This second area of potential of social media recruitment – to provide a vehicle for community engagement – may be especially important for recruitment in rural areas. Incorporating community-based elements into recruitment strategies appears to increase their effectiveness, particularly in the rural context. ²⁰ Additionally, social networking sites, such as Facebook, can help build social capital and maintain social connections within rural communities and between communities. ²¹ Social media can also facilitate sharing information about community events and local organizations. ²² For this type of network-based recruitment, the network size (i.e., the number of friends or followers an account has) has important consequences. In general, the more followers an account has, the more likely its messages will be shared with other users. ^23–25

Network-based recruitment also relies on the degree to which people in the network interact with posts (e.g., reacting, commenting, sharing). These interactions increase message reach (the number of people seeing a given message). They both increase the message’s visibility to the person’s identified network and positively influences the proprietary algorithms that Facebook and other social media platforms use to distribute messages. They also, which are recorded by the platforms, provide a way to evaluate social media messages as they indicate messages have been noticed by audience members. ²⁶ Despite these potential benefits, few studies have examined social media recruitment in a rural context. The few existing studies yielded contradictory findings. ^5–8

Technology Access

Increasing access to the technologies, including home broadband and smartphones, which make social media use possible, also indicates a need to evaluate the feasibility of social media recruitment. This move toward online technologies was apparent in Appalachian Kentucky before the COVID-19 pandemic began in March 2020. Between 2016-2020, 83% of Appalachian households owned one or more computer device (desktop or laptop computer, smartphone, or tablet) compared to the national average of 91.9%.¹¹ Although a digital divide persists, most rural residents report having Internet access, and gains in access in rural regions have outpaced gains in urban and suburban areas, especially during the COVID-19 pandemic. ²⁷ In 2019, 63% of rural residents reported broadband access compared with 75% of urban and 79% of suburban residents, a gap of 14 to 16%.²⁸ By 2021, this gap had decreased to 5-7%, respectively, with 72% of rural residents reporting home broadband access, compared to 77% of urban residents and 79% of suburban residents. ²⁹ Smartphone ownership also increased among rural residents, from 71% in 2019 to 80% in 2021. ²⁷

For the purposes of this study, we define “rural” consistent with the definition suggested by the Health Services Research Administration (HRSA). ³⁰ This definition draws on the U.S.D.A.’s Economic Research Service’s Rural-Urban Commuting Area (RUCA) codes. ³¹ The RUCA codes for the area included in the pilot study are all nonmetropolitan and range from 7-9.

These gains in technology resulted in increasing levels of social media use among rural residents. For example, in 2021, rural residents reported using Facebook at levels (67%) similar to urban residents (70%).³² Furthermore, rural residents reported using social media for health information (73%) at rates similar to urban residents (79%).³³ In the coming years, access to these technologies is predicted to increase as federal initiatives expand access to and reduce costs for broadband internet access in rural areas. ³⁴

Objectives and Research Questions

To our knowledge, no previous research has examined whether researchers can use social media as a community engagement tool and, in so doing, create a network of followers to distribute information about a study. We assessed the degree to which a stand-alone social media page can serve as the main recruitment tool for an mHealth intervention in a rural region. We aimed to determine whether a stand-alone social media page can (1) attract followers and/or likes and distribute messages, (2) encourage interaction with study messages, and (3) direct potential participants to the study intake survey.

Methods

The study team engaged in a process to design the social media page and message strategy; disseminate study messages through social media using community connections; and evaluate the success of this strategy in terms of audiences reached through social media, intake surveys completed, and interactions with social media messages.

Design of Social Media Page and Message Strategy

To develop content for the social media page, the study team drew on formative research we had conducted to adapt Make Better Choices 2 (MBC2), an mHealth behavioral intervention, to rural Appalachia. ³⁵ This community-based participatory research included focus groups and key informant interviews with 54 community members. ¹⁶ The findings identified social media as a way that community members wanted to learn about the study and identified content for promoting the intervention, including aspects that appealed to residents. The messaging strategy, also leveraged principles of community-based participatory research³⁶ and drew from Goal System theory, which situates behavioral goals in associative cognitive networks and connected to multiple sub goals, such as increasing fruit and vegetable consumption plus eating more delicious food. Additionally, Goal Systems theory recognizes the importance of the local socio-cultural and physical environment. Based on these theories and our foundational findings, the study team developed a series of messages (Selling Points, see Table 1) highlighting how the program implemented technology, health coaches, and financial incentives to help people make behavioral changes.

Table 1:

The definition, frequency, and intercoder reliability of message features with example messages

Code	Definition	Descriptive Information	Intercoder Reliability	Example Message from Facebook
Selling Points	Message includes information about aspects of the study identified by formative research as appealing to the target audience, including technology (fit bits/app), health coaches, and financial incentives.	n = 86, 45.2%	α = 0.99	We love how the MBC2 app helps keep us on track. You can try the app, too, if you want to volunteer for our research study at the (omitted for peer review). To participate, you need to be age 18 or older and be willing to use a smart phone to track eating, exercise, sleep, and other health behaviors. To learn more, visit our website: www.makebetterchoices2.com.
Local Involvement	Message describes local community involvement in the study, including participating in study planning and design, perspectives of community partners, local study staff, and individual and local participants endorsing the program.	n = 40, 21.1%	α = 0.91	Jane Smith [pseudonym] of H. County helped us make MBC2 fit here in Eastern Kentucky. She makes better choices because it feels good. “I love feeling healthy and at my best at all times.” www.makebetterchoices2.com
Study Updates	Message provides information about study progress, study personnel, or changes in the study protocol. These messages seek to keep the community informed and involved in the project and demonstrate community responsiveness. Messages include upcoming events, new study procedures, expanded inclusion criteria, etc.	n = 100, 52.6%	α = 0.88	Thank you to everyone who filled out our survey. Based on your feedback, we’ve made some changes so that more people can participate. If you are still interested in this study, please fill out the survey again. www.makebetterchoices2.com
Request to Share	Message asks people to share, like, or follow information about the project with others on social media	n = 36, 19.0%	α = 0.97	We want to thank everyone who has signed up to be part of Make Better Choices 2! We will be in touch soon. And, if you know someone who is ready to change their health behaviors, please tell them about this opportunity. www.MakeBetterChoices2.com
Link	Message contains a link to the research study page	n = 120, 63.1%	α = 0.98	Balancing stress, exercise, diet, and sleep can be tough. We are looking for volunteers from Eastern Kentucky who are willing to use a health app and work with a local health coach. Learn more about this free program at www.makebetterchoices2.com

Note: Descriptive Information includes the number messages (n) containing the message feature followed by the percentage of the total messages in the data. Reliability is measured by Krippendorff’s alpha (α).

Participants in the formative research also identified the value of having local people and organizations involved in the project, both as a way to provide social and community connections and to avoid the negative association of “outsiders.” ¹⁶ Based on this finding, the study team developed a series of Local Involvement messages. These messages described how community members had contributed to study adaptation and discussed what these community partners valued about the study (see Table 1).

The study team developed a third series of messages that provided information about the study and study progress. These Study Updates introduced the project to the larger community, provided background about its development and success in an urban setting, ³⁷ and updated the community about changes to the project as they occurred.

Development of social media messages was iterative and continued throughout the study (May 15 to Oct. 5, 2021; 21 weeks). Some messages included multiple types of content (e.g., Selling Points and Study Updates). More than half of messages included a link to a study web page (63.1%, n = 120); in general, we included the study link in messages that focused on recruitment either by discussing benefits or announcing a change to recruitment. We did not include a link in all messages as some research has suggested that links reduce the likelihood that messages will be shared by other users. ^38,39

Almost all messages included an image (96.8%, n = 185). These images included the study logo, maps showing study location, and photographs. The photographs featured local attractions from across the study region, area residents who had contributed to study design, study locations, and other local scenes.

The Facebook page included an overview of the study, contact information, and a link to a study webpage, where potential participants could complete an intake survey. Individuals were not incentivized for completing the survey; however, they could receive incentives later if they completed additional screenings and enrolled in the study. All messages and study recruitment procedures were approved by the Institutional Review Board (omitted for Peer Review).

Dissemination of the Facebook Study Page and Messages

In May 2021, we published our Facebook page. To disseminate this page, we reached out to 42 community contacts via email and through Facebook’s messenger application and asked them to follow our page and like messages. Drawing on guidance from social media marketers, ⁴⁰ we posted messages 6 to 13 times a week, with the number of messages varying depending on study activities. We varied the content of posts, interspersing messages that highlighted recruitment and messages that highlighted community involvement. Study updates were posted as they occurred.

Evaluation of the Social Media Recruitment Strategy

To evaluate our social media recruitment strategy, we collected data from Facebook and our intake survey. From Facebook, we collected the daily count of page followers, the total reach of each message (i.e., the number of times a message appeared in users’ news feeds), and total engagement (i.e., the number of times users interacted with a message). From our intake survey, we collected the date individuals started the survey, the way they learned about the study, and their self-identified gender.

Message Distribution and Intake Survey

To evaluate our distribution strategy, we conducted a series of descriptive analyses that examined the growth of the page’s followers and the total reach of individual posts over time. We used this same strategy to evaluate the number of people who started the intake survey.

Message Strategy Evaluation

To evaluate our message strategy, we first conducted a content analysis⁴¹ to identify the type of information included in each posted message. Content categories included whether the post contained information about appealing aspects of the study (Selling Points, defined as mentioning technology, coaches, and/or financial incentives); a community engagement strategy (Local Involvement, Requests to Share, and Study Updates); or a hyperlink to the study research page. Code definitions and examples are presented in Table 1. After a brief training period, two researchers established intercoder reliability by coding 100 messages. Codes were distinctive but not mutually exclusive, meaning messages could be coded for multiple types of information. After establishing a high level of intercoder reliability (Krippendorf’s α ≥ 0.88, see Table 1), one researcher coded the remaining messages (n = 91). All reliability analyses were conducted in SPSS 28 using the KALPHA macro. ^42,43

We then modeled the effect of message content (Selling Points, Local Involvement, Requests to Share, Study Updates, and hyperlink) on the number of engaged users using a generalized linear model⁴⁴ with a specification of the negative binomial distribution to account for overdispersion. ⁴⁵ We used the total number of page followers on the day a message was posted to control for message exposure. We did not include Total Post Reach in the model, as this variable was strongly correlated with our outcome variable.

We opted for a negative binomial distribution after fitting a Poisson distribution with log-link function, as the negative binomial distribution had better fit as measured by deviance and scaled deviance measures. During the modeling process, we removed two variables (Requests to Share, Study Updates) that did not significantly contribute to the model (p > 5%). Model fit improved as a result of removing these variables. The standard 5% significance level was used for all hypothesis testing. SAS version 9.4 (TS1M1, SAS institute, Cary, NC) was used for all modeling analyses.

Findings

During the 21-week recruitment period, we posted 192 Facebook messages, an average of nine per week. All messages used in the analysis (N = 191) were distributed using the page’s network of followers.

Reach of Study Page and Messages

Followers of the study social media page grew from zero to 429. After a rapid spike to 206 followers in May 2021, page followers slowly increased throughout the pilot study (see Figure 1). During the first few weeks of the study, 20 Facebook accounts associated with the community members and organizations we had reached out to “liked” a message we posted and/or followed our page.

Figure 1:

Study Page and Message Reach

The number of accounts following (Followers) the social media page for the MBC2 study increased rapidly after recruitment began and then gradually increased over the course of the 21-week recruitment period (May 15 – Oct 5, 2021). The number of times individual messages in the newsfeeds of Facebook users (Message Reach) varied widely with some posts appearing more than 3,000 times and others appearing fewer than 100.

Study messages appeared in the news feeds of individual Facebook accounts 41,026 times. The number of times any individual post appeared ranged from a low of 16 to a high of 3,578. Approximately half of the posts had fewer than 100 appearances. This variable was highly correlated with message engagement (0.88) in that messages that accumulated more engagement actions also reached a larger audience.

Survey Initiation and Pilot Enrollment

Two hundred and twenty-five unique people started the intake survey. Of those answering a question about how they heard about the study (n = 166), a majority indicated that they had learned about the study through either Facebook alone or a combination of Facebook and word-of-mouth (see Table 2). In addition, large proportions of people indicated that they had learned about the study either through word-of-mouth alone or a university research website. This same pattern of more people indicating that they had learned about the study from Facebook and word-of-mouth was also apparent among those who eventually enrolled in the pilot study (n = 9) or who opted to be put on a waitlist for the main study (n = 26). We created the waitlist due to limitations in our technology and study staffing during the pilot study. Of those who filled out the intake survey and answered a question about gender (n = 141), most indicated that they identified as a woman (66.0%, n = 93).

Table 2:

Sources of Information about the Study (N=191)

	Intake Survey Completion (n= 166)	Pilot Study Participants (n= 9)	Wait List (n = 16)
Facebook Only	89 (53.6%)	3 (33.3%)	4 (25.0%)
Facebook and Word-of-Mouth	16 (9.6%)	1 (11.1%)	6 (37.5%)
Word-of-Mouth Only	28 (16.9%)	5 (55.6%)	3 (18.8%)
University Research Website	26 (15.7%)	0 (0%)	1 (6.3%)
Other Methods	9 (5.5%)	0 (0%)	2 (12.5%)

Note: Table shows the number of people indicating how they learned about the study for each recruitment method used and the percentage (%) of the total number of respondents (n) for each category. Data include people who answered a question about how they heard about the study.

Message Engagement

Individual Facebook users interacted with our messages 3,080 times. The number of users commenting, sharing, or clicking on a link for an individual post ranged from zero to 366. Only 10 posts had 50 or more engaged users.

The results from the modeling analyses show the degree to which message characteristics influenced message sharing on Facebook, holding all other predictors constant (see Table 3). Of the message content predictors (local involvement, selling points, and hyperlinks), including local involvement resulted in the most engagement with messages. Including a hyperlink and selling points also resulted in more engagement, but the effect was not as strong as highlighting local engagement. Requests to share a message and study updates did not contribute to a significant increase in engagement and were not included in the model.

Table 3:

Parameter estimates and their expected counts from a loglinear model with negative binomial distribution

Parameter	Estimate	SE	P-value	Expected Count	95% CI
(Intercept)	3.355	0.342	<.001	-		-
Link	−0.179	0.166	0.280	23.955	11.476	50.003
Local Involvement	1.439	0.189	<.001	120.736	61.498	237.034
Selling Points	−0.434	0.161	0.007	18.556	9.052	38.038
Follow numbers	−0.003	0.001	0.007	28.568	14.637	55.761
Dispersion	1.023	0.101	-	-	-	-

Note: The results predict the degree to which a parameter influenced message sharing on Facebook, holding all other predictors constant. Higher estimates indicate that a message feature was associated with increased message engagement. The Expected Counts show the degree to which including a message feature or adding an additional follower increased the expected number of engaged users.

Abbreviations: SE = standard error, CI = confidence interval

Discussion

The results from this study suggest that a standalone social media page that relies on community and network connections for distribution can direct potential participants to an mHealth intervention study in a rural area. While this finding runs counter to stereotypical images of people in rural areas and Appalachia as reluctant to adopt new technologies, recent data indicate a significant uptake and embrace of technology use by rural residents. Some of this increased use of technology may have been precipitated by the COVID-19 public health emergency. ^46,47 While the proclamation of telehealth as rural healthcare’s lifeline may be overstated, ^48,49 such technologies do appear to play an important role in connecting rural residents to programs and providers who otherwise may be out of reach or simply to each other for information sharing.

In addition to providing greater access to our study, the specific type of information we included in our social media messages influenced whether people noticed and interacted with the messages. Our messages about local involvement in the study featuring local communities and community members – rather than messages that emphasized the desirable aspects of the intervention or study – created the most engagement and attention. This interaction was key to distributing messages on Facebook, as our study messages that received the most interaction also had the largest potential audiences (i.e., the largest reach). This connection between reach and interaction stemmed partly from the proprietary algorithm that Facebook uses to distribute messages across user accounts. This algorithm favors messages with more positive interactions. ⁵⁰

The value of local involvement for distributing messages across Facebook audiences underscores the importance of weaving community engagement and personal connectivity into online recruitment methods in rural areas. ²⁰ Featuring locally known people and places may resonate with our potential participants because of cultural features like close-knit communities and community pride. This may be especially important in times of social isolation brought on by a public health emergency like COVID-19.

Like many previous studies of social media in general, ²³ we found that having more followers predicts broader reach of our messages. This finding re-enforces the importance of building local networks when recruiting in a limited geographic area on social media.

Conversely, including more details about the study (technology, coaches, and/or financial incentives) contributed to interaction with our messages, but to a lesser extent than local involvement. Although recommended by our Community Advisory Board and our formative research, such information did not increase interaction as robustly as community engagement messages. Updates on the study and requests for social media viewers to share our messages did not contribute to message interaction.

Creating the social media page and the associated messages was low-cost because we did not pay Facebook to distribute the messages. However, this recruitment method did account for time-related costs, as study personnel developed social media messages, took pictures for social media posts, and spent time planning social media strategies and multiple messaging. Future research should compare the time-related costs of a standalone social media page to other recruitment methods, including in-person recruiting.

Finally, like many studies that have used social media recruitment, our social media messages generated many views but few actual participants. ^7,8 Although our messages were interacted with more than 3,000 times and appeared in news feeds 41,000 times, only 225 people started our intake survey. Still, given this approach’s relatively modest cost and effort, we answer the question in our title, “Community engagement through social media: A promising, low-cost strategy for rural recruitment?” with “yes.”

Limitations

Our findings must be understood in light of several limitations. First, when we started the pilot study, we limited participation to a small geographic area and to people who used Android phones. Because of a coding error in our intake survey, we initially did not collect information from people who may have qualified when study participation was expanded. Second, eligibility to participate in this study involved strict diet and physical activity criteria and a significant time commitment which may have constrained our ability to recruit participants. Another limitation involves the gender skew of our potential participants. This skew may be partly a result of our recruitment method and our study type, as women tend to use social media more than men (78% compared to 66%³²) and have been overrepresented in previous research using social media recruitment^51,52 and previous intervention research. ⁵³ Additional attention should be paid to ensuring gender equity in recruitment. Additionally, we did not compare social media methods to traditional recruitment methods (e.g., mailed flyers, newspaper advertisements). As a result, we cannot evaluate how social media recruitment compared to other recruitment strategies. Finally, since we limit our study population to rural Appalachian residents, we cannot extrapolate our findings to other rural populations.

Conclusion

Despite these limitations, we conclude that using a standalone social media page that emphasizes community engagement is a feasible, low-cost method for recruiting participants in rural areas. Future research should continue to evaluate this method for engaging geographically dispersed communities in research and comparing it to traditional methods.