Breastfeed4Ghana: Design and evaluation of an innovative social media campaign.

Abstract Although targeting health behaviour change through social media campaigns has gained traction in recent years, few studies have focused on breastfeeding social media campaigns. Within the context of rising social media utilization and recent declines in exclusive breastfeeding practices in Ghana, we implemented Breastfeed4Ghana, a Facebook‐ and Twitter‐based breastfeeding social media campaign. This study determined feasibility of implementing Breastfeed4Ghana and evaluated its impact on breastfeeding knowledge in Ghana. Key performance indicators of the campaign were monitored on social media platforms, Facebook and Twitter. An online cross‐sectional survey conducted across three time points (n = 451) assessed breastfeeding knowledge, campaign exposure, and understanding and acceptability of Breastfeed4Ghana among Ghanaian adults. Modified Poisson models were used to assess the relationship between campaign exposure and breastfeeding knowledge, adjusting for survey time point, sex, and parenthood status. The campaign acquired 4,832 followers. Based on follower demographics collected from Facebook and Twitter analytics, the target population was successfully reached. Campaign exposure among survey participants was 42.3% and 48.7% at midline and endline, respectively. Campaign acceptability was high (>90%), and >44% of those exposed to the campaign also shared the campaign with others. However, 61.0% of those exposed did not know or could not remember the purpose of the campaign. Campaign exposure was not associated with higher breastfeeding knowledge (APR [95% confidence interval] = 0.96 [0.73, 1.26]). Breastfeed4Ghana was highly feasible. However, campaign understanding yielded mixed findings and may explain the limited impact on breastfeeding knowledge.


| INTRODUCTION
The health benefits of breastfeeding to both the child and mother are well established (Victora et al., 2016); yet in Ghana, the prevalence of exclusive breastfeeding among infants under 6 months old declined from 63% in 2008 to 52% in 2014 (Ghana Statistical Service, Ghana Health Service and I. International, 2015). In response, a committee of Ghanaian breastfeeding experts and stakeholders led by the University of Ghana implemented the Becoming Breastfeeding Friendly (BBF) Initiative. The BBF Initiative provides countries with a toolbox for assessing their breastfeeding scale-up environment based on the breastfeeding gear model and developing recommendations to strengthen their breastfeeding environment (Perez-Escamilla, Curry, Minhas, Taylor, & Bradley, 2012;Aryeetey et al., 2018). Key recommendations from the BBF process included strengthening advocacy, harnessing support for maternity protection laws, and providing more effective dissemination of accurate actionable information about breastfeeding (Aryeetey et al., 2018). Additionally, the BBF committee emphasized utilizing cost-effective approaches to act on such recommendations, including through social media (Carroll et al., 2019).
Health behaviour change interventions facilitated through social media are on the rise, with a concentration of research within the context of high-income countries (Webb, Joseph, Yardley, & Michie, 2010, Laranjo et al., 2015, Maher, Ryan, Kernot, Podsiadly, & Keenihan, 2016, Yang, 2017. Social media has penetrated communities across the globe, and usage has recently increased in developing economies (Pew Research Center, 2018). Although the typical social media user is younger, more educated, and wealthier, trends across Africa indicate rapid adoption is shifting towards the middle class (Pew Research Center, 2016).
Between 2015 and 2017, internet use among adults in Ghana increased from 25% to 39%, and social media penetration increased by 60%, from 20% to 32% (Pew Research Center, 2016;Pew Research Center, 2018). Some sectors in Ghana have recognized the potential of social media as a tool to reach a wide audience for change (Pen Plus Bytes, 2016). However, the health sector, including specifically breastfeeding, has not effectively generated a social media presence.
In the Web 2.0 era, social media has become a prominent source of health information for the public (Moorhead et al., 2013;Giustini, 2006). Breastfeeding support is no exception. In a qualitative study among New Zealand mothers, most participants reported seeking breastfeeding advice online and "weak ties" on Facebook (i.e., low proportion of overlapping friend networks) were notable information sources (Alianmoghaddam, Phibbs, & Benn, 2019). Instagram also represents a platform for a supportive breastfeeding environment, gleaned from an analysis of 4,089 Instagram images and 8,331 comments on breastfeeding, which were primarily positive (Marcon, Bieber, & Azad, 2019).
Specifically in Ghana, 68% of breastfeeding mothers surveyed in Accra planned to use social media for breastfeeding information (Okasha, 2018). Furthermore, researchers deployed an online survey among breastfeeding mothers in the United States and found that many breastfeeding mothers are on social media while breastfeeding (Tomfohrde & Reinke, 2016). Thus, social media has been recognized at the global level as a potentially effective platform for breastfeeding promotion (Perez-Escamilla, 2012;Wolynn, 2012).
Thus, in the context of Ghana, where social media utilization is growing exponentially and exclusive breastfeeding rates have decreased, social media has potential to reach a wide audience through a campaign that targets the protection, promotion, and support of breastfeeding.
Social media-based campaigns are appealing because they offer a potentially more cost-effective option than traditional marketing campaigns, especially when combined with strategies to acquire high exposure and engagement. Such campaigns also have the ability to reach a wide target audience and include fathers and families, which have been identified by mothers as important sources for breastfeeding information (Brown, 2016). However, a breastfeeding social media marketing campaign has not been implemented in the context of West Africa before, and the feasibility of such an intervention in this context is unknown. Similarly, there is a lack of information regarding best practices for a campaign of this nature, including expected impacts and indicators of success, although an extensive list of key performance indicators has been published (Neiger, Thackeray et al. 2012).
Based on the BBF committee findings and recommendations (Aryeetey et al., 2018), a social media-targeted campaign on breastfeeding in Ghana was developed and tested through an iterative process between November 2017 and January 2018. For this campaign, three specific themes were designed to address key BBF recommendations, which highlighted protection, promotion, and support of breastfeeding. The objectives of the current study were to (a) determine feasibility of implementing a breastfeeding social media campaign in Ghana based on campaign reach, acceptability, and understanding and (b) evaluate the impact of exposure to the campaign on self-reported breastfeeding knowledge.

Key Messages
• Social media-based health behaviour campaigns are feasible and acceptable in sub-Saharan Africa and can serve as a reliable source of health information on the internet.
• Campaign acceptability was high; however, campaign exposure was not associated with improved breastfeeding knowledge.
• Campaign understanding, exposure, and engagement may be increased with a higher frequency of campaign materials being posted and more precise targeting.

| Description of campaign
Between March and September 2018, Breastfeed4Ghana was implemented as a 24-week-long breastfeeding social media campaign that targeted the protection, promotion, and support of breastfeeding in Ghana as derived from evidence-informed BBF recommendations (Aryeetey et al., 2018). There were 60 core campaign materials, each consisting of an image with a concise and pertinent educational message (see Appendix A). Figure 1

| Data: Campaign key performance indicators
Key performance indicator data for the campaign were collected To assess the page performance and users following the campaign on Facebook and Twitter, the number of followers were collected weekly. On Facebook, this also included the number of page likes, total reach, total engagements, total page views, and demographic information on followers and users who engaged with the content posted (i.e., age category, country, and city). Similar performance data were not available on Twitter. Exposure, reach, and engagement are three key performance indicators commonly used in understanding social media in the context of health promotion (Neiger et al., 2012).
Reach is most relevant in understanding page performance and can be measured by the number of page followers or page likes, as well as the demographics of the followers (i.e., age, sex, country, and city of residency). In this study, given the limitations of available data on Twitter, the primary metric for campaign page reach was weekly number of page followers.
For the data collection of core campaign materials (i.e., individual posts and tweets), key performance indicators were collected 1 day, 1 week, and 2 weeks after the material was posted and focused on exposure and engagement. Exposure was defined as Facebook reach (i.e., number of unique people who saw the content) and Twitter impressions (i.e., number of times a tweet appeared on a timeline and could be seen by a user more than once). Engagement included three subindicators calculated for each of Facebook and Twitter: applause rate (i.e., number of likes per number of followers), amplification rate (i.e., number of shares or retweets per number of followers), and conversation rate (i.e., number of comments or replies per number of followers). Therefore, Twitter impressions, Facebook reach, likes, shares, retweets, comments, and replies were collected for each post and merged with the corresponding weekly page data that included page followers. These data were collected at the same time points (i.e., 1 day, 1 week, and 2 weeks) for additional posts, such as blog, video, quote, and resource posts.

| Data: Online survey
An online cross-sectional survey was conducted at three time points: (a) baseline: 2 weeks prior to the initiation of the acquisition period, (b) midpoint: 13 weeks after the campaign launch, and (c) endline: at the end of the 24-week campaign. The aim of this survey was to assess Facebook and Twitter users' exposure to the campaign, awareness and perception of the campaign, and breastfeeding knowledge. A convenient sample of Ghanaian adults were recruited via Facebook and Twitter advertisements that were standalone from the campaign social media pages and did not specifically target campaign followers.
Given the novelty of the study in Ghana and lack of relevant literature to inform a sample size calculation for the survey, a target sample size of 300 women and 150 men across the three time points was established. This sample size was based on feasibility for online recruitment of this population and an interest to oversample women as we anticipated that the campaign would be more appealing to women. Individuals were eligible to participate if they were ≥18 years of age, residing in Ghana, and had not taken the survey in the past month. The survey was self-administered via Qualtrics®.
Self-reported breastfeeding knowledge was assessed from three multiple-choice questions and four true/false questions that focused on key elements of information shared throughout the campaign. Demographic characteristics and social media usage of participants was also collected.
All data were imported into Stata 13.1 (StataCorp, College Station, TX, USA) for cleaning and analysis.

| Analytical approach
Feasibility of campaign implementation was explored through descriptive analyses of (a) the platform performance based on reach and cumulative reach over time, measured from monitoring data, and (b) understanding of the campaign purpose and acceptability of the campaign based on the cross-sectional survey data.
To evaluate the impact of campaign exposure on breastfeeding knowledge, cross-sectional survey data were pooled across the three time points. Breastfeeding knowledge was calculated as a score of correct responses to the seven knowledge-based questions and dichotomized into "high knowledge" (≥6 questions correct) or "low knowledge" (<6 questions correct). Modified Poisson models were used to assess the relationship between campaign exposure and breastfeeding knowledge, adjusting for survey time point (i.e., baseline, midpoint, or endline), sex, and whether the participant was a parent of at least one child.

| Campaign feasibility and performance
During the acquisition period, a total of 3,061 followers were acquired on Facebook and 27 on Twitter, for a total of 3,088 followers at the campaign commencement. The small number of followers on Twitter was due to a delay in initiating paid advertising for the campaign on Twitter. The Facebook followers grew steadily over the initial 12 weeks and grew at a higher rate thereafter to 4,096 by the last week of the campaign. There were two spikes in Twitter follower growth due to targeted advertisement efforts, and the number of Twitter followers in the last week of the campaign was 736 (Figure 2). Therefore, the campaign reach grew from 3,061 to 4,832 followers by the end of the campaign period.
Accounting for weekly follower growth over time, the cumulative campaign reach totalled 92,502 follower-weeks across the 22-week campaign period.
Demographics of Facebook followers shifted significantly between the start (Week 2) and end of the campaign (Week 24), with a higher prevalence of followers residing in Ghana (Week 2 vs. Week 24: 73.6% vs. 93.2%) and female (39% vs. 44%) by the end of the campaign. At the end of the campaign, 91.9% of followers were between 18 and 54 years of age, which was similar to that at the start of the campaign (92.2%). Demographic data of Twitter followers were not available.   8% [26.6, 35.3] exposed to breastfeeding resources to 76.7% [72.5, 80.5] exposed to child health topics (   In a post hoc power calculation using the given exposure to the campaign of 0.32, we had sufficient power to detect~19% deviation in breastfeeding knowledge scores between those exposed to the campaign and those not exposed; thus, the study may have been underpowered.

| DISCUSSION
This study successfully demonstrated the feasibility of implementing a breastfeeding social media marketing campaign in Ghana. The number of campaign followers grew throughout the active campaign period, with a cumulative reach of 92,502 follower-weeks across 22 weeks.
Furthermore, the campaign was implemented as planned in terms of posted material and continuation across 6 months. Breastfeed4Ghana was the first social media campaign designed specifically to protect, promote, and support breastfeeding in West Africa. Therefore, no information was available on the feasibility, acceptability, or impacts  Many government, public health, and United Nation entities also have a social media presence that can be leveraged for health promotion purposes. Partnership with a reputable organization willing to house such a campaign could be an effective model to reduce the cost and time needed to acquire a campaign audience and may lead to a more sustainable campaign compared with the model conducted in the current study (Gough et al., 2017).
Paid advertising had a substantial impact on follower acquisition for Breasfeed4Ghana, as observed on both Facebook and Twitter.
Each platform has advanced targeting mechanisms that can be utilized to effectively advertise to specific populations. The demographics of the Facebook followers are evidence that a combination of the targeted campaign advertisements and the campaign material and branding designed for 18-to 49-year-old Ghanaian adults was effective in acquiring the desired follower demographic for this campaign.
It should be noted that the efficiency of the advertisements differs by platform and target population. For this campaign, the target population on Twitter was smaller compared with Facebook (i.e., less Ghanaian adults on Twitter than Facebook), and we found that paid advertisements were not as efficient at reaching the target population as seen on Facebook. It is also noteworthy that Twitter and Facebook have different algorithms for the content that appears on a user's feed and is dependent on how users engage with the different social media platforms (e.g., users who share specific content or like certain pages may receive more targeted advertisements as more user data are available on their interests; Facebook, 2019). Given the generally understood cost-effective nature of online campaign delivery and the interest in using social media for health promotion (Yang, 2017, Allom et al., 2018, it is crucial that more data be generated on the actual cost and cost-effectiveness of such campaigns (Korda & Itani, 2013, Johns, Lewis, & Langley, 2017 Mom" campaign solicited user-generated material by asking social media followers to share their stories, with the top stories "winning" the opportunity to have their story broadcasted (Ashley & Tuten, 2015). At the core of effective marketing is an understanding of consumer needs, motivations, and goals, for which Schmitt has proposed a conceptual model (Schmitt, 2012). Campaigns that target the consumer are likely to have greater success in achieving their desired impacts.
Based on the demographics of our Facebook page followers and those who took the online survey, there is a clear opportunity to reach a younger, perhaps preparenthood, population including both women and men. As noted by breastfeeding mothers in Brown's study, educating fathers, family society, and the next generation is key in breastfeeding support efforts (Brown, 2016 Given the campaign exposure (0.32) and proportion of high breastfeeding knowledge among those not exposed to the campaign (44%), the sample size would have needed to double (n = 892) to detect a difference of 10% at an α = .05 and β = .2; of course, utilizing a highly sensitive, and possibly continuous, indictor for breastfeeding knowledge and narrowing the target population would drive the sample size down. Therefore, additional studies designed to rigorously examine the impacts of such campaigns are needed. The demographic of the sample should also be noted. The participants were a population of young, highly educated, and heavy social media users, which is not representative of Ghanaian adults of child-bearing age, though possibly representative of social media users. Thus, the potentially biased sampling frame due to the internet-based survey approach must be considered when interpreting the survey results (Wilson & Laskey, 2003); and future campaign should consider the "digital divide" and corresponding health disparities in planning the campaign and evaluation (Chou, Prestin, Lyons, & Wen, 2013). Furthermore, it is possible that as social media utilization continues to rise across Ghana, social media content will influence caregiver decisions and behaviours.
Despite these limitations, this study provides important data for future health-focused social media campaigns in similar contexts. This study reveals that the first social media campaign in Ghana that aims to protect, promote, and support breastfeeding was highly feasible and acceptable. Such campaigns can provide a reliable and accurate source of information for the public in a sea of unreliable web-based information (Chou et al., 2013). Like Breastfeed4Ghana, such campaigns must generate evidence-informed content, and future campaigns should consider strategic investments in campaign promotion. It is also important to consider sustainability of a campaign, given evidence that online conversations around health promotion are not self-sustaining (Syred, Naidoo, Woodhall, & Baraitser, 2014). Social media is a platform for health promotion and interventions that can be utilized at a relatively low cost to achieve high coverage, representing a unique opportunity for health behaviour intervention (Yang, 2017). Along with the opportunity for impacting health behaviour through social media, there is a great need to fill the gaps in the literature regarding best practices and impacts of such interventions in order for the health sector to successfully uptake and utilize social media most effectively for health promotion and education.