Social Media Recruitment for Mental Health Research: A Systematic Review

Abstract

Background:

Social media holds exciting promise for advancing mental health research recruitment, however, the extent and efficacy to which these platforms are currently in use are underexplored.

Objective:

A systematic review was conducted to characterize the current use and efficacy of social media in recruiting participants for mental health research.

Method:

A literature review was performed using MEDLINE, EMBASE, and PsychINFO. Only non-duplicative manuscripts written in the English language and published between 1/1/2004–3/31/2019 were selected for further screening. Data extracted included study type and design, participant inclusion criteria, social media platform, advertising strategy, final recruited sample size, recruitment location, year, monetary incentives, comparison to other recruitment methods if performed, and final cost per participant.

Results:

A total of 176 unique studies that used social media for mental health research recruitment were reviewed. The majority of studies were cross-sectional (62.5%) in design and recruited adults. Facebook was overwhelmingly the recruitment platform of choice (92.6%), with the use of paid advertisements being the predominant strategy (60.8%). Of the reviewed studies, substance abuse (43.8%) and mood disorders (15.3%) were the primary subjects of investigation. In 68.3% of studies, social media recruitment performed as well as or better than traditional recruitment methods in the number and cost of final enrolled participants. The majority of studies used Facebook for recruitment at a median cost per final recruited study participant of $19.47. In 55.6% of the studies, social media recruitment was the more cost-effective recruitment method when compared to traditional methods (e.g., referrals, mailing).

Conclusion:

Social media appears to be an effective and economical recruitment tool for mental health research. The platform raises methodological and privacy concerns not covered in current research regulations that warrant additional consideration.

1. INTRODUCTION

A perennial issue in mental health research recruitment is the trade-off between sample size, time to recruitment, cost, and representativeness compared to the general population. Traditional recruitment strategies (e.g., flyers, print, and television ads) often generate cohorts of questionable interval validity due to self-selection bias [1]. Recruiting for mental health research is even more complicated, as participation may be hindered by stigma or fear of negative consequences from self-disclosure. Over the last three decades, participation in population-based research has been steadily declining [1]. Developing new strategies to increase recruitment for mental health research is essential to addressing the field’s most pressing problems.

The debut of Myspace in 2003 and Facebook in 2004 fueled a proliferation of online platforms designed to increase social interconnectivity. Facebook, Twitter, YouTube, and Instagram, among others, have rapidly become an avenue for the daily consumption and dissemination of information. More individuals in the US consult social media for news than print newspapers, with 74% of Facebook users checking the website or application daily [2, 3]. While usage is highest among the 18 to 25-year-old demographic, the generational divide is rapidly closing. Approximately 46% of US seniors age 65 and older disclose lifetime Facebook use [2]. The socioeconomic technology gap is slower to close, although decreasing technology costs have made web-enabled smartphones more accessible. A recent study of homeless youth revealed that 90% had a Facebook profile [4]. Studies also suggest increases in racial/ethnic diversity among social media users [5].

In a recent systematic reviewof 30 studies using Facebook for health research recruitment, Topolovec-Vranic et al. reported Facebook performing as well or better than traditional recruitment in 50% of studies [6]. Other reported benefits included relatively low advertising costs, reduced time to meeting recruitment targets, and increased study retention [6–10]. With billions of active users, social media also affords increased access to hard-to-reach, low prevalence populations such as rare diseases and gender and sexual orientation minorities [11, 12].

To date, no systematic reviews have examined the use of social media for research recruitment in mental health research studies. Here, we address this gap by summarizing how social media is currently used in mental health research recruitment and assessing the effectiveness of social media recruitment compared to traditional methods in regard to final sample size and cost [13].

2. METHODS

2.1. Search

For the purposes of this study, social media was defined as websites and/or applications that enable users to create and share content with a network of people. MEDLINE, EMBASE, and PsychINFO were searched using the following terms in the title or abstract: (Facebook OR Twitter OR Instagram OR YouTube OR Tumblr OR Myspace OR Snapchat OR “social media” or “social networking”) AND (recruit* OR advert*) in combination with keywords related to mental health/disorders and MeSH topics relevant to social media, patient selection, and mental health/disorders. The full search strategy is outlined in Supplemental Figure 1. The references of extracted manuscripts were manually reviewed for additional citations. Results were filtered by English language and publication from 1/1/2004–3/31/2019. Studies using pay-for-participation platforms, such as Amazon Mechanical Turk (MTurk), or paid web search results or ads, such as Google, were not included as neither are social media platforms. Geospatial dating applications (e.g., Grindr, Tinder) were also excluded.

2.2. Screening

After removal of duplicates, titles and abstracts were screened. Manuscripts were included if a study: 1) used social media as a recruitment tool, 2) recruited for mental health research, 3) appeared in a peer-reviewed journal, and 4) reported primary data rather than secondary analysis. General well-being survey studies were included if the investigations incorporated mental health measures. Meta-analyses, systematic reviews, editorials, dissertations, and conference abstracts were excluded (although references were scanned for additional citations). The text of the remaining studies was then screened. Studies were excluded if they: 1) failed to name specific social media platforms, 2) described secondary analysis of data reported elsewhere, or 3) failed to describe the social media recruitment strategy.

2.3. Data Extraction

Data extracted from the relevant studies included study type and design, participant inclusion criteria, social media platform, advertising strategy, final recruited sample size, recruitment location, year, monetary incentives, comparison to other recruitment methods if performed, and final cost per participant. Currencies were converted into US dollar equivalents using the current exchange rate (Supplemental Table 2). Missing metrics were calculated if sufficient data were available.

3. RESULTS

As shown in Figure 1, 2,316 studies were identified through the initial search following removal of duplicates. An additional 24 citations were gathered from reference list review. From these results, manuscripts were excluded by title and abstract screening if unrelated to mental health research, failing to use social media for study recruitment, lacking original data, and/or not published in a peer-reviewed journal. Abstracts ambiguous as to social media use (e.g., “subjects were recruited online”) were retained for full-text screening. The full-text screening of the remaining 355 manuscripts resulted in 179 additional exclusions. Data from the 176 eligible studies were extracted by two reviewers, with a third reviewer utilized to resolve discrepancies.

FIGURE 1:

LITERATURE SEARCH WORKFLOW

As shown in Figure 2A, social media recruitment has been steadilyincreasing over the last decade. Table 1 summarizes the characteristics of the included studies (full details for each study are available in Supplemental Table 2). Studies were primarily conducted in the US (51.7%), Australia (22.7%), and United Kingdom (7.4%) (Figure 2B). The type of research that individuals were recruited to was: 62.5% cross-sectional studies, 20.5% randomized controlled trials, 11.4% non-controlled trials, 5.1% prospective studies, and 0.6% retrospective. Adults were the primary age demographic of interest in 71% of studies, while 5.7% focused exclusively on recruiting adolescents and 18.8% recruited a mixed range. Most studies (79.6%) recruited all genders, 13.1% recruited women exclusively, 5.7% recruited men only, and 1.7% recruited transgendered individuals specifically. Incentives, either monetary compensation or prize drawing entry, were offered in 41.5% of recruitment efforts, with no compensation offered or reported in the others.

FIGURE 2.

YEAR AND COUNTRY DISTRIBUTION OF STUDIES USING SOCIAL MEDIA RECRUITMENT FOR MENTAL HEALTH RESEARCH, 2009–2018.

TABLE 1:

SUMMARY OF SOCIAL MEDIA MENTAL HEALTH RECRUITMENT STUDIES

Age of recruited participants	N	Total	%
Adult only	125	176	71.0%
Both adults and children/adolescents	33	176	18.8%
Children/adolescents only	10	176	5.7%
Not reported	8	176	4.6%
Gender of recruited participants	N	Total	%
Any gender	140	176	79.6%
Female only	23	176	13.1%
Male only	10	176	5.7%
Transgender only	3	176	1.7%
Study design	N	Total	%
Cross-sectional	110	176	62.5%
Randomized controlled trial	36	176	20.5%
Non-controlled trial	20	176	11.4%
Prospective	9	176	5.1%
Retrospective	1	176	0.6%
Incentives (e.g., cash, gift card drawings)	73	176	41.5%
Psychiatric diagnostic research areas	N	Total	%
Substance use disorders	77	176	43.8%
Mood disorders	27	176	15.3%
General mental health	21	176	11.9%
Trauma & stress-related disorders	15	176	8.5%
Suicidal ideation & self-injurious behaviors	8	176	4.5%
Eating disorders	7	176	4.0%
Anxiety disorders	5	176	2.8%
Impulse control disorders	5	176	2.8%
Mental healthcare delivery	4	176	2.3%
Personality disorders	3	176	1.7%
Pharmacology	2	176	1.1%
Developmental disorders	1	176	0.57%
Psychotic disorders	1	176	0.57%
Social media platform(s) used	N	Total	%
Facebook only	121	176	68.8%
Facebook & Twitter	24	176	13.6%
Facebook & Instagram	9	176	5.1%
Facebook & multiple other	9	176	5.1%
Twitter only	6	176	3.4%
Weibo only	3	176	1.7%
Instagram only	1	176	0.6%
Myspace only	1	176	0.6%
Myspace & Mi Gente	1	176	0.6%
Tumblr only	1	176	0.6%
Advertising strategies	N	Total	%
Paid ads	107	176	60.8%
Free posts to relevant groups/pages	44	176	25.0%
Study page or profile	40	176	22.7%
Posts to personal profiles	11	176	6.3%
WebRDS	7	176	4.0%
Direct contact with users based on content	7	176	4.0%
Comparative metrics	N	Total	%
Social media only recruitment used	76	176	43.2%
Equivalent/high final participant rate	28	41	68.3%
More cost effective than some/all other methods	10	18	55.6%

Facebook was the main social media recruitment platform of choice: exclusively (68.8%), or in combination with Twitter (13.6%), Instagram (5.1%), or other platforms (5.1%). Recruitment methods included paid advertisements (60.8%), free posts on relevant pages/groups (25.0%), study social media pages or profiles (22.7%), posts to the personal social media accounts of the researchers (6.3%), web-based respondent driven sampling (WebRDS) (4%), or direct contact with eligible individuals through the social media platform based on their posted content (4%). Social media was the sole form of recruitment used in 43.2% of studies versus being coupled with traditional recruitment methods (e.g., print ads, clinic or hospital referrals).

Areas of investigation spanned the mental health spectrum: 1) substance use disorder (43.8%), including alcohol, tobacco, cannabis, methamphetamine, MDMA, and opiates, among others [12, 14–88]; 2) mood disorders (15.3%), including bipolar disorder, depression, postpartum anxiety/depression, bereavement [89–115]; 3) general mental health (11.9%) [116–136]; 4) stress, trauma, and/or PTSD (8.5%) [137–150]; 5) suicidal ideation and self-injurious behaviors (4.5%) [151–157]; 6) eating disorders (4%) [158–164]; 7) anxiety disorders (2.8%) [165–167]; 8) impulse control disorders, including gambling and gaming disorder (1.7%) [168–172]; 9) mental health delivery (2.3%) [11, 173, 174]; 10) personality disorders (1.7%) [175–178], 11) pharmacology (1.1%) [179, 180]; 12) developmental disorders (0.6%) [181]; and 13) psychotic disorders (0.6%) [182].

Of the 41 studies that directly compared the performance of social media to other methods, social media performed equally as well or better in 28 (68.3%) in terms of the number of final participants recruited. As Facebook was the primary social media platform utilized (68.8%), we calculated the estimated recruitment costs overall and across study design types for the 49 studies with data available (Table 3). The median cost per study participant recruited through Facebook was $19.47 for all studies, $6.25 for cross-sectional studies, $35.68 for non-controlled trials, and $42.82 for randomized controlled trials. Only a single prospective study had data available, costing $149.64 per final participant.

4. DISCUSSION

Social media offers a wide variety of recruitment solutions that differ in cost and level of interaction with eligible participants. In some studies, researchers directly contacted users based on review of their profile content. Tan et al. screened posts on Weibo (a popular Chinese micro-blogging website equivalent to Twitter) for text suggestive of suicidal ideation and messaged eligible users to participate in a cross-sectional survey [154]. Kelleher et al. similarly identified and contacted depressed users on Tumblr by searching posts for variants of “#depressed.” In other studies, investigators posted study information to social media pages or groups related to the demographic or subject area of interest (e.g., new mothers, college students, cannabis) [11, 16, 24, 31, 39–41, 47, 65, 75, 76, 89, 97, 99, 101, 106, 111, 116, 117, 121, 123, 126, 131, 132, 134, 137, 139–141, 143, 147–149, 156, 158, 162, 164, 169, 177, 181, 183, 184]. This method appeared particularly effective when the platform restricted ads with certain text or images (e.g. cannabis leaf). Some investigators recruited high profile “influencers” (i.e. users with many followers) to post the study to their accounts [118, 157]. Study-specific hashtags (e.g., #depressed) and sharable images were also used to increase secondary recruitment [111, 135, 136]. Less frequently, investigators post the study to their own social media accounts [65, 76, 116, 118, 127, 143, 144, 171, 176]. More common was creating an official social media account for the study [11, 49–51, 57, 63, 70, 73–75, 77–85, 88, 112–115, 128, 129, 131, 133, 136, 145, 149, 157, 161, 167, 172, 179, 185]. Study accounts facilitated promotion to relevant groups, created a centralized area to display more extensive study information, and provided a contact point for questions from interested individuals [117, 140, 159, 176]. Private, closed groups accessible only to enrolled participants were also used to improve retention for longitudinal studies or for the delivery of interventions [57, 58, 136].

Most studies used paid Facebook ads for recruitment. Ads were targeted for display to individuals within the desired range of demographics (e.g., age, location) and contextual interests (e.g., pages/groups to which participants indicated interest, searches). For example, Borodovsky et al. limited their Facebook advertisements to US users, 18 years or older, with cannabis-related interests, such as “liking” cannabis-related organizations (e.g., Marijuana Policy Project) or cannabis-related magazines (e.g. High Times) [16]. Daniulaityte et al. additionally targeted Twitter users who searched cannabis-related keywords or hashtags (e.g., #marijuana, #weed, #legalizeit) [26].

The median cost per participant for Facebook recruitment for any type of study in the current analysis was $19.47. This cost is comparable to two prior studies that reported an average cost per participant of $17.48 and $19.77 for Facebook recruitment across different types of health studies. Where cost comparisons were made between social media versus traditional recruitment methods, social media was frequently (55.6%) the more cost-effective option [11, 19, 21, 22, 30, 81]. In several cases, investigators indicated recruitment speed using social media was also noted to be faster, although direct calculations were made in a limited number of studies [35, 81, 85]. We generally found a highly variable degree of reporting of social media and comparative traditional recruitment practices. Additionally, we focused exclusivelyon English-language manuscripts. As such, incomplete retrieval may limit our assessment of the overall perceived benefits to this method.

Cost per participant is highly dependent on the desired final sample size, which is further contingent on study design and the rarity of potential participants within the general population. Post-recruitment statistical adjustment and weighting may impact final cost recruitment. For example, Bauermeister et. al recruited participants for a cross-sectional study of alcohol and illicit drug abuse in young adults. The cost per participant increased from $25.98 to $108 after weighting the cohort to reflect expected general population distributions [86]. A potentially cost-saving solution is WebRDS (Web Respondent Driven Sampling), in which seed individuals are selected who reflect the desired composition of the final sample. Seed individuals then refer from within their social networks with the supposition that referrals should closely resemble seeds in desired characteristics. A number of studies used this method [48, 59, 86, 88, 136, 186].

Comparable to prior reviews, social media recruitment facilitated increased access to potentially harder-to-reach, hesitant, and/or vulnerable populations, including combat veterans [49–52, 155], sexual or gender minorities [17, 120, 136, 146], African-Americans with OCD [184], natural disaster survivors [139], and rural adolescents [135]. This is quite promising since deriving large study samples of these groups via single or even multi-academic center traditional referral methods can prove challenging given lower prevalence rates. Traditional recruitment methods often require higher motivation (e.g., phone call, in-person appointment) and longer wait times for entry than the click of a mouse. The immediacy of contact may also reduce ambivalence toward participation. In addition, the perceived anonymity and privacy of online interactions may reassure hose individuals fearing disclosure. Additionally, social media’s impersonal yet connected nature can be theorized to foster a less anxiety-provoking entry for potential study participants.

Not all studies indicated overwhelming success with social media recruitment. Targeted advertising is effective only if the desired population is present on social media or self-discloses sufficient demographic (e.g., age, relationship status) or contextual (e.g., interests, associations) information for identification by the targeting algorithm. The wording and design of ads can also impact recruitment rate. In a depression awareness campaign, Huiet al. found that a happy face illustration garnered significantly more click-throughs than a sad illustration [96]. Choiet al. reported that in attempting to recruit men, advertisements worded toward strength outperformed those centered around happiness or resilience [122]. Schwinn et al. even found that costs and recruitment efficacy varied based on the day of week and month of the year [61]. To devise a strong advertising campaign, expert consultation may be necessary, a potentially costly additional recruitment cost.

Additional barriers to successful cohort recruitment with social media arise from the potential for misrepresentation by those responding to ads and the need to filter multiple submissions by the same individual respondent. Some studies checked internet protocol (IP) addresses to establish uniquenessof identity [35, 111, 187]. Others reviewed individuals’ social media profiles to verify identity [27]. In recruiting US military veterans, Teo et al. as well as Pedersen et al. employed the use of specific domain knowledge, such as pay grades, to validate individuals [51, 155].

Representativeness of the social media-recruited sample is an additional consideration. Gender, age, and education level are most likely to be imbalanced in social media-recruited cohorts [8, 9]. Batterham et al. compared the representativeness of two surveys’ samples recruited using Facebook ads compared to two postal-recruited samples as well as a national survey and census data (Australia). The Facebook samples tended to have an overrepresentation of younger and female respondents as well as higher rates of self-reported depression and anxiety[151]. Thorton et al. found Facebook-enabled recruitment of a cohort with a higher severity substance abuse compared to non-Facebook methods [188]. Bauermeister et al. reported that webRDS-enabled recruitment of a representative sample of young adults with substance use estimates comparable to those found in the National Survey of Drug Use and Mental Health (NSDUH) [86]. The demographic skew of a social media platform may be of consideration depending on the topic of study. Guillory et al. found Twitter ads more effective at recruiting younger, more frequent abusers of e-cigarettes than a Qualtrics online panel [35]. In a study of postpartum depression participants recruited via Facebook versus face-to-face referral, the authors found no differences in sociodemographic, general health, and mental health characteristics between the two samples [119].

Another major challenge associated with social media recruitment is translating and abiding by the federally-mandated requirements for the privacy and security of patients’ protected health information (PHI). Penalties for Health Insurance Portability and Accountability Act (HIPAA) violations, even inadvertent ones, can be high and devastating to a healthcare institution and its research enterprise. Civil penalties cannot be imposed in cases where there is no willful neglect if the violation is corrected within 30 days. However, it is unclear how this applies to online data, where retracted material may remain discoverable in perpetuity. The terms of service (TOS) of most social media platforms require their users to grant the company non-exclusive, transferrable license to their posted content[189].

User data may also be sold by the social media platform to interested third-party vendors. Although a user can delete their account, posted information may continue to exist in these purchased repositories. The TOS typically absolves the platform of responsibility for content shared with a third-party vendor prior to the user deleting their account. The lack of control over information exchanged through social media is a critical consideration for researchers planning a recruitment campaign. Social media accounts are common targets for hackers looking to acquire and sell highly lucrative personal information. An estimated 160,000 Facebook accounts are compromised each day [190]. In a data breach, PHI exchanged between researcher and participant may be exposed and rapidly disseminated with little recourse available to either party to facilitate removal.

Finally, the privacy of the research clinicians is also to be considered. In only 5% of studies reviewed here, studies were advertised on the personal profiles of the researchers. Clinicians may have varying attitudes about and comfort levels with their own accessibility to the public and prospective participants through social media. Disclosures of a clinician’s personal beliefs, political leanings, and even vacation or family photos may influence researcher-participant interactions in unanticipated, and potentially adverse, ways. For example, increased access to research staff may be inappropriately used as a means for help-seeking, presenting additional challenges and liability considerations for study staff.

Several limitations and biases are important to note to the current review. First, the characteristics of subjects recruited through social media compared to those recruited from traditional sources was not addressed. Second, as there is relatively little literature from non-English-speaking Western countries, particularly in Europe, it is unclear whether the findings could be generalized to this population. Third, publication bias — specifically researchers not publishing studies which had poor social media-based recruitment — may inflate the estimates of cost efficiency. Fourth, cost estimates may not be accurate given both currency exchange rates fluctuate over time and the potentially significant differences in research infrastructure between the countries studied. Finally, there are limitations of the search strategy, including: a) Although we specified seven of the most popular (past and current) platforms and used “social media” MeSH/Emtree and explicit search for “social media” and “social networking” — which would aggregate lesser-known platforms as well — popular discussion forums, such as Reddit, were excluded. These popular discussion forums are often labeled as “social media.” However, the mechanisms of how users interact and share information on discussion forums is quite different from networking-based platforms; b) Other forms of internet and digital advertising (e.g., Google AdWords and YouTube videos) were excluded from the search. The decision to limit the reviewto social media was made due its unique characteristics in comparison with other forms of digital/internet advertising, as social media affords to the opportunity for direct engagement with potential participants, free options for advertising/contacting potential participants, as well as passive paid advertising. Additionally, social media involves the aspect of “sharing”/snowball recruitment that other forms of digital advertising do not. Furthermore, Google AdWords, YouTube, and other internet/digital advertising do not have some of the unique privacy/HIPAA considerations as social media platforms. A systematic reviewof all forms of digital advertising and the underlying mechanics, reach, efficacy, and privacy considerations would be interesting, but beyond the scope of this review.

5. CONCLUSION

Social media presents a tremendous opportunity to address some urgent challenges in mental health research. In some instances, social media cohorts may allow for a higher degree of generalizability without compromise of internal validity. Other benefits reported by investigators using these methods include: 1) low cost, 2) increased access, especially for harder-to-reach populations or rare disorders, 3) less staff hours, and 4) reduced time to meeting recruitment targets. However, as social media platforms are commercially owned, there exist unique privacy and ethics considerations that are not addressed by current federal guidelines on human subjects research. Given the trend of increasing reliance on this method, there is an urgent need for further investigation and regulation in this area.

TABLE 2:

COSTS PER STUDY TYPE USING FACEBOOK RECRUITMENT

Study type	N	Median cost/participant	Range
All	49	$19.47	$0.13—339.50
Cross-sectional	27	$6.25	$0.13—267.67
Non-controlled trial	10	$35.68	$8.80—339.50
Prospective	1	$149.64	NA
Randomized controlled trial	7	$42.82	$4.53—112.48

HIGHLIGHTS.

• Social media is increasingly employed in mental health research recruitment.

• Compared to traditional methods, social media recruitment offers advantages in cost, speed, and efficiency.

• Minimal federal guidance exists in using social media for recruitment to avoid privacy violations.