We conducted a cross-sectional survey using online social media to recruit HIV-positive individuals. The purpose of the parent study was to examine the barriers to and facilitators of adherence to antiretroviral therapy and to describe the use of mobile telephones and other technologies to improve adherence among HIV-positive individuals living in the United States. Here, we describe the methods used for participant recruitment using online social media.

We implemented a campaign approach where participants were recruited through online social media such as Facebook, Twitter, and other social media platforms, such as LinkedIn, Craigslist, and Tumblr, from May 7 to August 31, 2013. The online survey was programmed using Qualtrics Research Suite (Qualtrics, Provo, UT), an online survey tool that allows researchers to build, distribute, and analyze online surveys in real time. The survey contained 112 items on demographics, HIV clinical outcomes, and use of technology. The University of California, San Francisco (UCSF) Committee on Human Research approved this study in April 2013.

A schedule was devised by the research assistant to regularly update and maintain these social media sites for a total of 8 hours per week (20% of full-time effort).

Individuals who clicked on the study Web link were directed to the Qualtrics survey where they were required to consent online before being screened for eligibility. The survey was used to screen individuals based on the following inclusion criteria: age 18 years or older, HIV-positive serostatus, and currently living in the United States. We excluded individuals who had already taken the survey by limiting the number of survey attempts to 1 per Internet Protocol (IP) address and by asking participants if they had already taken the survey. Qualtrics was programmed to end the survey if the participant did not qualify. We ensured anonymity by not collecting any identifiers or personal health information and storing responses on encrypted and password-protected servers at UCSF.

Given the anonymous nature of the Web-based survey and lack of identifiers to verify repeat and fraudulent responses, monetary incentives were not offered in order to de-incentivize multiple survey attempts. However, in order to keep participants engaged, we motivated participants by inserting 5 medically interesting facts strategically throughout the survey titled “Fun Facts”, which were accompanied by visually stimulating and relevant pictures. Examples of these fun facts include: “Did you know that we exercise at least 30 muscles when we smile?” and “Did you know that your brain uses as much power as a 10-watt light bulb?” Additionally, to spark interest, we asked participants before beginning the survey if they knew of a natural substance that can be potentially effective against HIV. Upon completion of the survey, we embedded a link to a video about how bee venom is being studied by the Washington University School of Medicine as a potential drug to treat HIV [33].

We consulted the UCSF Center for AIDS Prevention Studies (CAPS) Community Advisory Board (CAB), which is comprised of stakeholders from Bay Area agencies and communities with the mission of channeling community input into the CAPS research agenda, providing advice on study methods, and improving recruitment methods. Based on the recommendations of the CAPS CAB, we modified the Facebook ad pictures to exclusively use photographs of people, the Facebook ad text to be more direct and motivating, and incorporated Tumblr as an additional online social networking venue for recruitment.

Three variables were created to estimate the relative value of each of the 3 main methods of recruitment: (1) “Facebook Page Interactions” to estimate user activity on the Facebook fan page, (2) “Twitter Interactions” to evaluate the overall activity on the Twitter feed, and (3) “Facebook Ad Clicks” to assess the overall success of the Facebook ad. The number of likes, comments, and shares on the Facebook page per recruitment day were extracted from Facebook Page Insights and summed to serve as an objective measure of Facebook Page Interactions. Data collected from Twitonomy (a Twitter analytics tool) included the number of retweets and mentions per recruitment day and were used to compute Twitter Interactions. Information on the number of clicks per day for the Facebook ad was obtained from Facebook Ad Reports. Using descriptive data from Qualtrics, we collected information on the number of survey clicks per day. We used Pearson’s pairwise correlation to examine the relationship between the number of survey clicks per day and Facebook Page Interactions, Twitter Interactions, and Facebook Ad clicks per day. Correlation values of 0 to .3 were deemed weak correlation, .3 to .7 indicated moderate correlation, and >.7 were considered strong correlation. The significance level of each correlation coefficient was also estimated. A P value of <.05 was deemed statistically significant.

As part of the survey, participants were also asked about their method of recruitment. These methods were categorized under one of the following categories: Facebook, Twitter, word-of-mouth, email from a listserv, and other (including LinkedIn, Craigslist, Tumblr, or other). We used 2-way frequency tables to examine the distribution of the method of recruitment based on participants’ age (categorized as 18-29.9, 30-39.9, 40-49.9, 50-59.9, and ≥60 years), race/ethnicity (White/Caucasian, African-American/Black, Latino, and other), male sex at birth, and sexual orientation (homosexual/gay, heterosexual, bisexual, and other). A chi-square test was used to assess the null hypothesis that the categorical variables are independent.

Last, to estimate research retention in the absence of monetary incentives, we compared the total number of participants who responded to the first survey question with the total number of respondents who answered the last survey question. All statistical analyses were conducted using Stata, version 13.1 (StataCorp, College Station, TX).

A total of 2034 individuals clicked on the survey Web link, averaging 18 clicks per day. Of those who clicked, 1977 people consented to take the survey, 1404 met the inclusion criteria and initiated the first question of the survey, and 1221 responded to the last question of the survey (86.97% retention). A total of 43 respondents answered affirmatively when asked if they had taken the survey before and were therefore excluded from the survey.

The study received a total of 10,006 Facebook Ad Clicks, 278 Facebook Page Interactions, and 161 Twitter Interactions during the recruitment period. Although each of these factors contributed to the total number of survey clicks per day, only the number of Facebook Ad Clicks was moderately correlated with the number of survey clicks per day based on the Pearson pairwise correlation (.52, P<.001). The Facebook Page Interactions and Twitter Interactions were weakly correlated with Facebook Ad Clicks (r=.17, P=.06; r=.18, P=.06, respectively). Figure 1 illustrates the number of survey clicks relative to the number of Facebook Ad clicks, Twitter Interactions, and Facebook Page Interactions for the month of June 2013. This serves to visually demonstrate the correlation between each of the 3 major methods of recruitment and the total number of survey clicks per day.

On average, participants spent 31 minutes to complete the survey. Table 1 encapsulates key demographics of participants by self-reported method of recruitment. According to these data, participants were primarily 40-49.9 years of age, identified as White/Caucasian, male at birth, and homosexual. Facebook was the most common method of recruitment across all reported demographics. With increase in age, participants were more likely to report being recruited through non-Facebook methods. Similarly, heterosexual individuals were more likely to report non-Facebook methods of recruitment compared to bisexual or homosexual participants. Email from a listserv was the second most commonly reported method for those aged 60 years or older and African Americans. Racial/ethnic categories were independent of self-reported method of recruitment.

A total of US $5,021 was spent on the Facebook ad (mean CPC = $0.64), which targeted a total of 14 million Facebook users and accumulated 1404 initiated surveys. Therefore, the estimated mean cost per survey initiation was approximately US $3.56. The reach of the Facebook ads included participants from 48 states plus the District of Columbia but did not include participants from either Alaska or Wyoming. From the Facebook fan page, the research assistant “liked” 490 other fan pages while 53 other organizations and individual Facebook users “liked” the study’s fan page. A total of 141 Facebook messages were sent to community leaders, organizations, and individuals requesting to have the survey link reposted.

Over the course of the recruitment period, the research assistant tweeted 572 times, received 199 Twitter followers, and followed 1092 Twitter feeds. In addition, the research assistant contacted a total of 16 LinkedIn groups, added 3 Tumblr sites, and posted on 3 Craigslist forums.

Our study results indicate that the unique integration of online social media recruitment, Web-based survey, and non-financial incentives is an efficient and streamlined strategy for survey research with HIV-positive individuals. The restricted resources of US $5,000 in a 4-month recruitment timeframe with 20% research assistant’s time resulted in recruiting 1404 qualified participants in 48 states and the District of Columbia, averaging 18 survey responses per day, making this an efficient solution for participant recruitment. This is similar to other studies not focused on an HIV-positive population that have demonstrated the efficiency of Facebook ads [24,27]. Furthermore, we were able to access a difficult-to-reach population that may experience stigma and other barriers to research participation.

Even though Facebook has been used to effectively recruit HIV-positive individuals [34], our study is novel in that we utilized a campaign approach—using several different online social media platforms at once to benefit from various approaches and to maximize diversity. Compared to other methods of online recruitment, the Facebook ad was by far the most successful and least time consuming. The number of survey clicks per day and Facebook ad clicks was moderately correlated and highly statistically significant, illustrating the importance of a paid Facebook advertisement. We believe that setting a high CPC bid and increasing the number of interest groups as recommended by the Facebook representative, as well as using ad images of women and ethnic minorities [32] and more direct and motivating ad text as proposed by the CAPS CAB were associated with improved ad performance.

Although we observed a weak correlation between the number of survey clicks per day and both the Facebook Page and Twitter Interactions, these platforms were essential in establishing and maintaining a rapport with the study population and community. Such relationships were cultivated by liking and following relevant pages and groups of underrepresented populations, posting comments on walls and reposting relevant material, and sending personal messages. These techniques were particularly helpful given that the level of compliance with our requests was contingent upon the relationship and rapport between the researcher and the person or institution being contacted. For example, mass-produced messages and random posts on other Facebook pages or Twitter feeds were not as useful as personalized messages sent directly to a community leader or an HIV advocate with an established relationship. This highlights the social phenomenon known in economics as reciprocity, which correlates one’s likelihood of compliance with their familiarity with the one making the request [35]. Therefore, future researchers should not rely exclusively on online social media and underestimate the powerful nature of existing social networks to assist with recruitment efforts.

Retaining participants for Internet-based studies has been known to be difficult [36], particularly among African American and Latino MSM [31,37]. In response, our study catered ads for these demographics and implemented a novel incentive structure by adding interesting facts throughout the survey to motivate participants. We used this approach as opposed to one with pecuniary incentives to minimize duplicate and false responses. Given that the average response time for survey completion was 31 minutes and 86.97% of those who initiated the first question of the survey completed the survey, we believe this is a tactical approach that can be used in future online research with anonymous participants.

Since there are more Facebook users than any other social media site, it was not surprising that most participants identified Facebook as their primary method of recruitment. Moreover, the paid Facebook ad and the selection of targeted groups likely contributed to a higher percentage of Facebook recruits. However, it is unknown why older age and non-homosexual orientations were more likely to be associated with non-Facebook methods of recruitment. Future research should further examine possible reasons for these patterns. As such, the use of online social networks other than Facebook for recruitment remains a strong area of interest as venues for future research [38]. It is likely that sites such as LinkedIn, Craigslist, and Tumblr would be more useful for research within professional networks, other interest groups, or longitudinal research.

The self-reported demographics elucidated in Table 1 are not representative of the population of those living with HIV in the United States [39]. Responses revealed a higher proportion of individuals who identify as homosexual but fewer African Americans and Latinos. This may be attributed to HIV-specific reasons such as stigma, the necessity of having access to the Internet to participate in the survey, the need to be using social media to view recruitment advertisements, and the requirement for English proficiency to take the survey. Such drawbacks may be generalized to other study populations [24,40] and are important factors when considering the use of online social media for recruitment.

Another limitation of our study is the possibility of duplicate responses. A study using online social media also found this issue to be problematic, especially when participants were provided a financial incentive without the use of restrictive software [8]. However, attrition rates for online surveys in the HIV-positive MSM population are high [41], urging researchers to provide some form of incentive. With awareness of this issue, we implemented safeguards to protect the integrity of the survey data by introducing non-financial incentives, automatically disqualifying participants who used duplicate IP addresses, and asking participants whether they had already participated in the survey. This inquiry was a necessary precaution owing to the fact that 43 respondents admitted to taking the survey more than once. Future researchers might consider investigating the extent to which repeat online survey attempts are denied due to duplicate IP addresses using comparable online survey instruments and non-financial incentives. Even though we believe the likelihood of duplicate responses is low, it is not possible for us to determine if users circumvented these measures by accessing the survey through multiple devices or by providing inaccurate responses.

In terms of Facebook ad pictures, future research may be directed toward studying the effectiveness of different ad images by measuring the number of ad clicks and survey responses, which was a limitation of this study. Another limitation was that we were unable to verify the HIV serostatus of participants. However, given the lack of pecuniary incentives, it is unlikely that individuals would misrepresent themselves as being HIV-positive. Last, given that email from a listserv and directly contacting individuals using Facebook messages were valuable tools, it would have been beneficial for us to have asked participants about the specific listserv or the particular Facebook route (ie, Facebook ads vs personal message) in order to distinguish between various efforts.

Online social media is an indispensable tool for recruiting participants because it has the potential to be cost-effective and efficient. Social media has the unique ability to transcend barriers to study recruitment such as physical distance, transportation, and limited time and financial resources. It also affords users the added benefits of privacy and anonymity, which may ameliorate the effects of perceived stigma in difficult to reach populations. Moreover, non-financial incentives in the form of trivia are a viable alternative to monetary incentives that yields high rates of retention while minimizing the chances of duplicate and fraudulent responses. Facilitating this process using online data collection software can further reduce cost and time associated with data collection. In sum, as Internet-based research tools become more practical, researchers must embrace these methods not only to maximize efficiency but also to enhance the scientific validity and generalizability of their findings. As such, online social media, alternative types of incentives, and Web-based survey software are well poised to become staple methods of recruitment and engagement in research.