Reaching Populations at Risk for HIV Through Targeted Facebook Advertisements: Cost-Consequence Analysis

Background An undiagnosed HIV infection remains a public health challenge. In the digital era, social media and digital health communication have been widely used to accelerate research, improve consumer health, and facilitate public health interventions including HIV prevention. Objective We aimed to evaluate and compare the projected cost and efficacy of different simulated Facebook (FB) advertisement (ad) approaches targeting at-risk populations for HIV based on new HIV diagnosis rates by age group and geographic region in the United States. Methods We used the FB ad platform to simulate (without actually launching) an automatically placed video ad for a 10-day duration targeting at-risk populations for HIV. We compared the estimated total ad audience, daily reach, daily clicks, and cost. We tested ads for the age group of 13 to 24 years (in which undiagnosed HIV is most prevalent), other age groups, US geographic regions and states, and different campaign budgets. We then estimated the ad cost per new HIV diagnosis based on HIV positivity rates and the average health care industry conversion rate. Results On April 20, 2021, the potential reach of targeted ads to at-risk populations for HIV in the United States was approximately 16 million for all age groups and 3.3 million for age group 13 to 24 years, with the highest potential reach in California, Texas, Florida, and New York. When using different FB ad budgets, the daily reach and daily clicks per US dollar followed a cumulative distribution curve of an exponential function. Using multiple US $10 ten-day ads, the cost per every new HIV diagnosis ranged from US $13.09 to US $37.82, with an average cost of US $19.45. In contrast, a 1-time national ad had a cost of US $72.76 to US $452.25 per new HIV diagnosis (mean US $166.79). The estimated cost per new HIV diagnosis ranged from US $13.96 to US $55.10 for all age groups (highest potential reach and lowest cost in the age groups 20-29 and 30-39 years) and from US $12.55 to US $24.67 for all US regions (with the highest potential reach of 6.2 million and the lowest cost per new HIV diagnosis at US $12.55 in the US South). Conclusions Targeted personalized FB ads are a potential means to encourage at-risk populations for HIV to be tested, especially those aged 20 to 39 years in the US South, where the disease burden and potential reach on FB are high and the ad cost per new HIV diagnosis is low. Considering the cost efficiency of ads, the combined cost of multiple low-cost ads may be more economical than a single high-cost ad, suggesting that local FB ads could be more cost-effective than a single large-budget national FB ad.


Undiagnosed HIV
Complications and risks from undiagnosed HIV infections continue to be a public health challenge [1]. In 2020, a total of 16% of people with HIV worldwide were unaware of their HIV status, and 27% were not receiving antiretroviral therapy [2]. In the United States, undiagnosed HIV has an estimated prevalence of 13.3%, with certain populations affected disproportionately [3]. The latest estimates from the Centers for Disease Control and Prevention (CDC) indicate that effective HIV prevention and treatment are not adequately reaching men who have sex with men, transgender persons, American Indian, Alaska Native, Native Hawaiian, other Pacific Islander, African Americans, Asians, Hispanics, and youth [4,5]. In 2019, an estimated 44.3% of people aged between 13 to 24 years were unaware of their HIV infection [5]. Moreover, during the COVID-19 pandemic and the resulting shutdowns, fewer HIV tests were performed [6], triggering concerns that unprecedented financial stressors to patients and health care systems and required modifications to health care delivery greatly disrupted HIV diagnosis and care [7].

Social Media as a Health Communication Solution
In the last decade, social media has provided health communication solutions for multiple health challenges, including patient outreach and education. Despite concerns about privacy and data use [8], social media platforms and medical crowdfunding websites have proven to be powerful tools for health communication in the social media era [9][10][11][12][13]. Compared with the beginning of the HIV epidemic, the COVID-19 pandemic highlighted social media solutions and challenges such as the rapid spread of misinformation [9][10][11][12][13][14][15]. Dis-and misinformation require social media companies to step up their stewardship by removing false information and redirecting users to reputable websites. In this study, we evaluated the potential benefits of social media by estimating the projected cost and efficacy of using Facebook (FB) advertisements (ads)-the biggest social network outreach solution-to reach populations at risk for HIV.
Meta Inc (Menlo Park) continues to lead the social media market with 2.91 billion active users monthly as of the fourth quarter of 2021 [16]. The FB ad platform allows promoters to target audiences that meet certain criteria, including specific gender, age, demographics, interests, and location [17]. Previous studies have tailored campaigns to target populations of interest, including underrepresented populations. Although some studies have evaluated the efficacy and cost-effectiveness of the FB platform [18][19][20][21][22][23], others have compared the cost-effectiveness of FB ads with other ads [24,25].

Previous Use of FB Ads to Reach Populations at Risk for HIV
The Chicago pre-exposure prophylaxis (PrEP) campaign (PrEP4L) was launched in 2016 and used the FB ad platform to disseminate nonstigmatizing health education to high-risk populations. PrEP4L garnered 6,970,127 views on FB and 1,719,446 views on Instagram. The average number of visitors to the PrEP4L website from this campaign was 182 per day, with a click through rate (CTR) of 0.06%, which is below the industry standard of 0.5% to 0.9% for social campaigns [22]. Another social media study recruited individuals identified as lesbian, gay, bisexual, and transgender (LGBT) through ads on FB and Instagram in 2016 and found that study enrollment was notably faster on social media than in-person recruitment in LGBT bars and night clubs [20]. However, gay women, bisexual men and women, and other gender minorities were easier to recruit than gay men via social media [20]. These studies highlight the feasibility of reaching LGBT populations on FB with an opportunity for improvement in terms of cost-effectiveness and reaching gay men.

Use of FB Ad Estimates During the Ad Creation Process
Researchers used FB-provided estimates during ad creation to demonstrate that FB ad audience estimates can be used to model regional variations in the prevalence of health conditions such as obesity [26][27][28]. During the process of FB ad creation and based on the selected target criteria, FB ads provide promoters with a projected potential reach that estimates the size of the audience matching the selected target criteria. The potential reach depends on the target criteria and the ad placement options selected while creating an ad [29]. On the basis of the user-adjusted budget, the platform estimates the daily reach and daily clicks.

Study Aim
In this US-centric study, we evaluated and compared the projected cost and efficacy of different FB ad approaches, simulating a 10-day video ad campaign that targets at-risk populations for HIV to estimate the costs for every resulting new HIV diagnosis using the health care industry conversion rate [30] and HIV positivity rates [31] in various regions and age groups.

Study Definitions
In this study, we refer to FB ad metrics as defined by FB [29] (definitions summarized in Table 1). We refer to the US regions as defined by the US Census Bureau and as referenced in the CDC report "Diagnosis of HIV Infection in the US and Dependent Areas, 2019" [3]. We created ads (without actually launching them) targeting only Puerto Rico and the US Virgin Islands and referred to these territories as US dependent areas. For our estimations, we used simulated FB ad estimates, HIV positivity rates provided by the CDC [31], and the average conversion rate for the health care industry on FB (11%) as reported by previous marketing research [30].

Definition Term
An option available during FB a ad b creation that allows promoters to define the group of people that will see an ad on FB. It may include other pages or ads users click on and activities they engage with on FB. Additional selection criteria include demographics such as age, gender, location, and network connection speed.
Detailed targeting A choice that promoters make when creating an ad to share with FB to influence the result they receive from the promotion. Promoters can also select an "automatic" goal by allowing FB to set the most relevant goal based on other ad settings.

Ad goal
The estimated maximum audience size that could see an ad based on the selected target criteria, ad placements, and how many people were shown ads on FB apps and services in the past 30 d. Potential reach is not an estimate of how many people will actually see the ad and may change with time. Estimates are not designed to match census population.

Estimated potential reach
The estimated number of people an FB ad will reach in a certain audience each day based on budget and ad bid. Ad bid is how much an advertiser is willing to pay for a specific action.

Estimated daily reach
The estimated number of link clicks that an FB ad will receive each day based on a campaign performance and estimated daily reach.

Estimated daily link clicks
The average cost to reach 1000 people with an FB ad; reach can be a more insightful metric than impressions, because it measures how many people were exposed to an ad and how efficiently an ad reached them.

Cost per 1000 people reached
The percentage of users who perform a desired action after clicking on an ad. Conversion rate a FB: Facebook. b ad: advertisement.

Ad Creation Process for Age Group 13 to 24 Years
We created an FB video ad for a 10-day period without launching it. First, we selected the FB ad goal (Figure 1). The ad's target criteria included participants being men, aged 13 to 24 years (the age group with the highest undiagnosed HIV rates), and with 1 or more of the following interests as defined by FB: "homosexuality, same-sex marriage, same-sex relationship, transgenderism, or LGBT community" (Figure 2). On changing the target geographic location, FB ads provide an estimated potential reach in that location. As estimates vary over time, we collected the potential reach for each state and for the United States as a whole for the same single ad on the same day (April 20, 2021).
FB ads allow promoters to choose a budget, and based on the budget, FB provides an estimated daily reach and estimated daily number of link clicks. For each state, we adjusted the budget to reach the total estimated target population in 10 days, which meant that we targeted an estimated 10% of the total potential reach per day. We adjusted our budget to the nearest US $10 to achieve this goal. We documented the budget, estimated daily reach, and estimated daily clicks for 2 manually adjusted budgets for each ad campaign. The first budget was adjusted to target the highest estimated reach over 10 days, and the second budget was adjusted to target the lowest estimated reach over 10 days (Figure 3).

Ad Estimates at Different Budgets
We collected the estimated daily reach and daily clicks for a single ad with the same criteria targeting the entire United States at multiple budget set points ranging from US $10 to US $1,000,000 to explore the differences in reach and link clicks at different budget levels.

Estimating Ad Cost per New HIV Diagnosis
Finally, for every state, we estimated the cost of ads for every new HIV diagnosis secondary to the campaign based on estimated daily clicks (provided by FB), the health care industry conversion rate [27], and previously reported new HIV diagnosis positivity rates [28] (Figure 4). We also compared the ad's cost for every new diagnosis using a single ad in the United States as a whole the ad's cost using multiple US $10 ads in different US regions or states and age groups (Multimedia Appendix 1 [32]).

Ethical Considerations
As all the data used in our study were publicly available, no institutional review board approval was required. All FB ad estimates used for modeling the cost per new HIV diagnosis were collected from the provided anonymous estimates on the FB ad platform without launching any ads. Hence, as no simulated study ads reached consumers, informed consent was not required. FB was not a collaborator in this study.

Results
On April 20, 2021, the potential reach for our criteria (men with 1 or more of the following interests as defined by FB: "homosexuality, same-sex marriage, same-sex relationship, transgenderism, or LGBT community") in the United States was approximately 16 million for all age groups and approximately 3.3 million for the age group 13 to 24 years.

Age Group 13 to 24 Years
California had the highest potential reach at 430,000, followed by Texas (360,000), Florida (210,000), and New York (200,000; Table 2 and Figure 5). The average estimated ad budget to reach the maximum potential audience on FB through a 10-day video ad campaign from April 20, 2021 Figure 6).
On the basis of the average estimated budget to reach the maximum audience for each state in the age group 13 to 24 years, the estimated average cost per 1000 individuals reached was the lowest in Alaska, Maine, and West Virginia (US $12-US $12.99), followed by Arkansas Figure 8).
Using a single ad with the same target criteria for the entire United States demonstrated an estimated average cost per 1000 of US $18.50 and an estimated average cost per new HIV diagnosis of US $257.51. Table 2. Estimated total reach, daily reach, and daily link clicks on the Facebook Ads platform based on the selected target criteria for age group 13 to 24 years and the manually adjusted budget to include the total reach on both extreme ends of the provided estimated range of reach.

Exploring Different Ad Budgets
When we created a single ad using the criteria for the entire United States at different budgets ranging from US $10 to US $1,000,000, the daily reach per dollar and daily clicks per dollar followed the cumulative distribution curve of an exponential function (Figures 9 and 10).     (Table  3).  (Table 3).

Evaluating Efficacy and Cost of FB Ads Reaching Population at Risk for HIV
Creating FB ads that target at-risk populations for HIV infection in US states, regions, and age groups while comparing different ad budgets revealed interesting and useful findings, including potential low ad cost per new HIV diagnosis and higher cost efficiency for low-budget ads. On April 20, 2021, the potential reach for our target criteria in the United States was approximately 16 million for all age groups and 3.3 million for the age group 13 to 24 years-the group that had the highest rate of undiagnosed HIV infection. The estimated ad cost per new HIV diagnosis averaged US $13.96 to US $55.10 for all age groups and was the highest in the age group 13 to 19 years at US $55.10. The estimated ad cost per new HIV diagnosis averaged US $12.55 to US $24.67 for all US regions and was the lowest in the South at US $12.55. Comparing different ad budgets, low-budget ads were more cost-effective than high-budget ads, as the daily reach per US dollar and the daily clicks per US dollar followed the cumulative distribution curve of an exponential function on the FB ad platform. We concluded that multiple small campaigns would generate similar results to larger campaigns at a lower cost.
Understanding the variables contributing to the cost and efficacy of social media ads for health promotion is important in the digital era [33]. Our study methodology proposes a framework ( Figure 4) for estimating the efficacy and cost of FB ads per new diagnosis before launching the ads. At our study time, examining social media metrics among the US states revealed the highest estimated potential reach and the highest estimated ad budgets in California and Texas (Figures 5 and 6). However, the average cost per 1000 reach on the FB ad platform ( Figure  7) was not the highest in California or Texas, which is likely owing to a proportionally lower ads cost in these 2 states at the time of our study. Moreover, the average FB ad cost per new HIV diagnosis (Figure 8) was the lowest in Texas and the highest in California. This difference was likely related to a difference in the ads cost between the 2 states and higher HIV test positivity rates in Texas than in California.
Although our framework focuses on FB ads, evaluating the platform's potential as a health communication solution should be the first step when considering social media for health promotion. Selecting the best social media platform for health promotion campaigns can be accomplished by understanding the available criteria for targeting, comparing reach estimates of the target audience on the social media platform to real-world estimates of the population at risk, and examining the penetration rates of different social media platforms among the population of interest. For example, in our study, among all age groups, we found the highest FB ad cost per new HIV diagnosis in the age group 13 to 19 years. This may be explained by the lower penetration of FB among this age group than other platforms such as TikTok and Snapchat. In addition, these estimates are subject to user interaction with the ad's design and content, which is beyond the scope of this study.  [5].
As public health efforts to reach HIV-positive individuals are hampered by financial constraints, an understanding of cost becomes crucial to investing in appropriate resources. In 2013, the CDC designed the "Start Talking. Stop HIV." page on FB (as part of a larger "ACT against AIDS" campaign) to reach and influence gay and bisexual men to spark conversations about HIV prevention and sexual health [36]. To date, this FB page has generated over 125,000 followers while only spending US $11,943 for ads between May 7, 2018, and February 7, 2022 [37]. In our study, the reach per dollar on FB ads followed a cumulative exponential distribution curve. We want to alert public health agencies that a single high-cost ad is more expensive and may have less impact than same-value multiple low-cost ads. For future implementation of our simulated models, ethical considerations targeting at-risk populations for HIV who are frequently marginalized and disadvantaged should include community advisory groups empowered by representatives of the proposed recipients of the ads. This will promote engagement in the development of the study design and ad's content to accomplish inclusion of the targeted participants for intervention while balancing the concern of representation that may have a further stigmatizing effect. In addition, future implementation designs should ensure the protection of participants' privacy while targeting vulnerable populations, which may generate an identifiable digital trail.
Since our study was conducted, multiple social media platforms have changed their policies to prevent targeting users younger than 18 years with disinformation. In addition, FB no longer allows targeting sexual orientation interests, which may hinder public health outreach efforts on the platform for sexual health education. Considering the challenge of misinformation spreading via social media platforms and the resulting societal harm [14], it stands to argue that social media platforms owe it to the society to counteract disinformation. The authors of this paper argue that social media platforms should partner with public health organizations to provide a free outreach window for spreading health education information similar to their response during the COVID-19 pandemic.

Limitations
Our study has several limitations that must be discussed. First, during the ad creation process for each state, the manually adjusted budget was rounded up to the closest US $10, which may falsely increase the estimated cost per click, cost per 1000 reach, and cost per new HIV diagnosis, especially in the states with the lowest potential reach. Second, we did not use the average health care industry CTR in our estimations, given that CTR is usually based on impressions, and the platform only provides estimated reach and estimated clicks during the process of ad creation. If estimated impressions were available during the creation of ads, we would have used the industry CTR to calculate estimated clicks from impressions and performed subsequent simulations. Third, potential reach, estimated daily reach, and estimated daily clicks were collected on the same day to avoid fluctuations that occurred based on different ads bidding and active FB users in the prior 30 days to data collection. All these estimates may change substantially when using a different time range given the fluctuations associated with the ad platform seasonality. Fourth, in our simulations, we used the positivity rates published in the CDC-funded HIV testing report from 2017, which has the most recently reported US data for test positivity rates for different age groups and US regions, and it may be different from the 2021 estimates. Fifth, we used the health care industry average conversion rate on FB as reported by WordStream based on a sample of 256 US clients in all industries between November 2016 and January 2017, which may also differ from the 2021 average health care conversion rate. Finally, our proposed feasibility model is subject to the limitations and algorithms of the social media platform because it is heavily based on estimates provided by the FB ad platform during the process of ad creation. FB ad estimates are based on multiple factors, including ad bids and active FB and Instagram users, may be intentionally misleading, and are not designed to match the census population.

Conclusions
Targeted FB ads have the potential to reach populations at risk for HIV and to facilitate education on exposure risks, HIV testing, and PrEP. This is especially critical in the Southern United States given the increased rates of new HIV diagnoses, high potential reach of FB, and low cost per new HIV diagnosis. Although FB allows targeted and granular location-based and interest-based ads at varying budgets (critical for public health interventions), our study found that multiple small-budget ads were more cost-efficient than 1 large-budget ad campaign. From a public health perspective, this translates into local FB ads being more economical than a single large-budget national FB ad. Therefore, our future efforts will leverage these findings by focusing on populations at risk for HIV in the Dallas-Fort Worth metroplex through different FB ads targeting strategies to delineate the best combination of interest-and location-based targeting criteria to improve HIV care.

Conflicts of Interest
RJM has received research grants from the Centers for Disease Control and Prevention and research funding from the Sergey Brin Family Foundation, the Verily Life Sciences, and the Texas Health Resources Clinical Scholars Program. None of the