Participants were recruited through advertisements published on Facebook. From September 2011 to November 2013, advertisements were set to randomly appear to Facebook users who were (1) female; (2) between the ages of 18 and 25 years; and (3) residing in Victoria, Australia.

The advertisements contained a brief headline (eg, “Women’s Health Matters”), a generic picture of young women and a brief caption (Figure 1). The advertisements were subsequently modified by changing the text to target women who had not received the 4vHPV vaccine, and made visible to Facebook users from November 22, 2013 to June 2015 (Figure 2). The decision to modify these advertisements arose mid-way through the study as we wished to understand the determinants of women eligible for, but not accepting, the HPV vaccine.

Respondents could click through to the secure VACCINE Study website to read more about the study and register an expression of interest (EOI). Potential participants were contacted and screened by telephone to assess their eligibility and willingness to comply with the study requirements.

Participants who verbally consented to the study were invited to provide electronic consent and to complete a self-administered, password-protected questionnaire hosted by the online survey tool SurveyMonkey. The survey questions related to demographic characteristics, sexual history and knowledge, attitudes and practices regarding HPV, the HPV vaccine, and cervical cancer screening. Participants were also requested to provide their HPV vaccination status and written or electronic consent for their HPV vaccination history to be verified with the NHVPR.

Descriptive statistical analyses were conducted using Stata 13.1 (StatCorp LP, College Station, TX, USA). The demographics of our cohort were compared with general population data sourced from the Australian Bureau of Statistics 2011 census data [22]. Socioeconomic status was assigned using the Postal Area of Relative Socio-Economic Disadvantage 2011 [22]. The Chi-square test was used for all demographic comparisons between our sample of study and the general population, as well as within study comparisons between females recruited by general advertisements and those recruited by targeted advertisements. A 2-sided P value <.05 was considered statistically significant. Data were treated as missing if a question was skipped or “prefer not to answer” was selected.

Participants who received all 3 doses of the HPV vaccine were considered “vaccinated”; those who had received 1 or 2 doses were considered “under-vaccinated” and women who never received the HPV vaccine were recorded as ”unvaccinated“.

Targeted Facebook advertising led to increased recruitment of young women who had not received the HPV vaccine, without employing other recruitment methods. The rationale behind over-recruiting unvaccinated women was to allow us to reliably measure any difference in the prevalence of high-risk HPV between unvaccinated and vaccinated women. We also show that recruiting through Facebook is cost-effective given that the cost per participant in this study was AUD $24.02.

Evidence from previous studies has shown that targeted Facebook advertising is effective in recruiting participants into health research [10,23,24]. In these studies, recruitment was targeted based on broad demographic characteristics such as gender, age, and location, to maximize generalizability [10,23,24]. However, there have been few studies in which Facebook advertising has been used to recruit participants with more specific characteristics [25,26]. For example, young adults who were cigarette users were sought in a study of tobacco and substance use [9,27].To attract their target audience, investigators developed Facebook advertisements which were shown to users whose profile pages contained tobacco- or marijuana-related keywords drawn from their listed interests, activities, job titles as well as the Facebook pages they ”liked“ or groups to which they belonged [9]. Our study differed slightly from this approach. Instead of using keywords to define the people to whom our advertisements were displayed, we relied primarily on Facebook users to read and respond to the text in our customised advertisements.

Our study sample compared well with the general population in age. The significantly greater proportion of women born outside of Australia recruited via targeted advertisements compared with those recruited through general advertisements is intuitive as overseas students are not eligible for the HPV vaccine. Although young women who were born in Australia and/or had completed tertiary education were over-represented in the total sample; these biases are common in population-based studies [10,24,28]. This is perhaps because highly-educated people are more likely to be aware of health issues. Therefore, highly-educated people may choose to participate in health research to address their personal health concerns and/or because of altruistic motives such as the desire to contribute to medical knowledge and improving the health of others [29]. The difference in the distribution of women living in inner regional areas in our study sample was statistically significant but small compared with the general population (2%, P<.001); a larger sample size is required to determine whether this is meaningful.

There is an inherent risk of introducing sampling bias when targeting Facebook advertising for specific characteristics. We found that our sample was reasonably representative of the general population, except for country of birth and education level. Another potential cause of bias associated with this recruitment method is snowball or chain-referral sampling, whereby users exposed to these advertisements share information about the study on their Facebook profile page with their friends, or relatives, who may then submit an EOI. We found that 78% heard about the study from the Facebook advertisements, whereas approximately 17% either read a post on their friend’s Facebook wall or were told by a friend to participate. These 17% referrals were constant pre and post changes to the advertisements. Given the small proportion who submitted an EOI without having seen the Facebook advertisements, we contend that this did not have a significant impact on our results.

We have demonstrated the utility of paid, targeted Facebook advertising as a contemporary and effective recruitment method. The ability to specifically target individuals with particular characteristics by tailoring Facebook advertisements enables researchers to recruit specific groups of individuals of interest into health studies.