Individuals from Black, American Indian or Alaska Native, and Hispanic or Latino communities were disproportionately impacted by the SARS-CoV-2 virus that caused the COVID-19 pandemic. Multiple studies reported that Black, Hispanic, and American Indian or Alaska Native communities had higher hospitalization rates and death rates across the country [1,2]. In Kaiser Family data samples from >40 states, it was found that the rate of infection for Hispanic patients was >3 times the rate for White patients, and from March to July 18, 2020, hospitalization rates due to the COVID-19 pandemic for Black, American Indian or Alaska Native, and Hispanic patients were approximately 5 times higher than those for White patients [3,4]. In Wisconsin, compared with non-Hispanic White Wisconsin residents, Hispanic or Latino residents had 1.7 times higher case rates, Black residents had 2.1 times higher hospitalization rates, and American Indian residents had 1.4 times greater death rates [5].

Greater prevalence of comorbid conditions, such as diabetes, obesity, and cardiovascular disease, among Black, American Indian or Alaska Native, and Hispanic or Latino patients elevated their risk of hospitalization once infected [1]. Poor access to health care, low physician trust delaying help seeking and systemic racism and mistreatment in medical care likely worsened these hospitalization rates and death rates [1]. Furthermore, individuals who identify as Black, American Indian or Alaska Native, or Hispanic or Latino are more likely to live in socioeconomically disadvantaged neighborhoods [6,7], tend to rely more on public transportation, are more likely to live in multigenerational households, and are more likely to hold jobs that cannot be performed remotely [1,6], leaving them at a higher risk of exposure to SARS-CoV-2 virus. Notably, McCormack et al [8], in their study of >3 million persons, found that Black individuals were overrepresented in multiple industries with essential workers, including transportation, public administration, and health care. In many cities, these neighborhoods with communities of color have been segregated, which increased the concentration and spread of the virus in these neighborhoods and likely isolated its impact to these racially segregated communities [6]. The disparities in vaccination rates only exacerbated the issue. By May 25, 2021, a total of 44% of White residents of Wisconsin had received at least 1 dose of the COVID-19 vaccine. However, only 23% of Black residents, 29% of American Indian residents, and 31% of Hispanic or Latino residents had received a dose [9]. The reasons for lower vaccination rates were varied, including systemic racism [10], disparities in accessing care, lack of information about vaccination centers, misinformation and disinformation campaigns [11], and confusion about benefits and eligibility criteria [12].

During the COVID-19 pandemic, the lack of in-person interactions inhibited the use of traditional outreach approaches to address the misinformation, resulting in an increased use of social media [13]. As more individuals spend more time on online social networks, information and misinformation spread farther and faster than ever before, requiring the development of new approaches for using social media for public health promotion [14,15]. These approaches also needed to account for the mistrust in governmental and research organizations that communities of color might experience due to the structural racism, historical inequities, language of preference, and health literacy of the communities.

In light of the misinformation campaigns that were targeting communities of color [16], the community advisory board of the Wisconsin Alzheimer’s Disease Research Center made the recommendation that the team pivot and focus on providing accurate information to communities of color on the COVID-19 pandemic. After exploration with additional community partners, there was a high level of interest in a social media solution that sought to address the needs of Black, Latino, and American Indian or Alaska Native communities in the state of Wisconsin by creating culturally appropriate content that promotes medically accurate information about the COVID-19 pandemic.

Multiple psychological theories, such as cultivation theory [17] and attribution theory, can be used to explore how media messages, including social media messages, affect individuals’ perceptions and behaviors. Cultivation theory suggests that long-term exposure to media content shapes an individual’s beliefs, attitudes, and perceptions about the world. It argues that heavy viewers of television, for example, are more likely to adopt the values and perspectives presented in media programming, leading to a cultivation of their worldview [17]. In the context of social media, for example, if social media messages were to consistently portray smoking as a common and acceptable behavior, individuals who are heavy users of social media may develop a skewed perception of smoking norms, potentially impacting their attitudes and behavior regarding smoking cessation or prevention [18]. Therefore, by strategically incorporating positive health messages in social media content, public health campaigns can leverage the cultivation theory to influence people’s beliefs and behaviors in favor of healthier choices.

Increasingly, more Americans use social media platforms, according to a Pew Research Center Study in 2021, with vast demographic differences per platform [19]. A total of 81% of all respondents used YouTube (Google LLC) while 95% of adults aged 18 to 29 years used the platform. The study also reported racial and ethnic differences. Higher rates of WhatsApp (Meta Platforms, Inc) use were reported among Hispanic respondents, with 49% of them using WhatsApp compared with only 16% of White respondents. A total of 49% of Black respondents used Instagram (Meta Platforms, Inc) and 74% used Facebook compared with only 35% of White residents using Instagram and 67% of them using Facebook. Given that misinformation is reported to be more easily spread via social media [20] and Black, American Indian or Alaska Native, and Hispanic or Latino communities report heavier use of social media, public health professionals could harness cultivation theory effects to increase positive social media messaging that combat misinformation in social networks frequented by Black, American Indian or Alaska Native, and Hispanic or Latino individuals.

Attribution theory [21] focuses on how individuals attribute causes to the behavior of themselves and others. It suggests that people tend to make inferences about the underlying causes of behavior based on 3 main dimensions: internal factors (eg, personal characteristics), external factors (eg, situational factors), and stability (eg, whether the behavior is consistent or temporary). In public health messaging, attribution theory can play a role in understanding how people interpret health-related behaviors and social media messages shared on social media. Individuals may attribute the behavior of others, such as engaging in health activities or practicing preventive measures, to internal factors (eg, personal motivation or responsibility) or external factors (eg, societal norms of situational influences) [22]. Public health campaigns can leverage the attribution theory by highlighting positive health behaviors and framing them in a way that encourages attributions to internal factors (eg, personal agency) to motivate individuals to adopt healthier behaviors themselves.

Together, cultivation and attribution theories reflect how health professionals can develop more effective social media strategies to engage and influence their audience’s health behaviors and shape their perceptions related to health issues.

This implementation report aimed to describe the process used to develop social media content and deploy multimedia strategies to communicate with 3 different populations: Black, American Indian or Alaska Native, and Hispanic or Latino communities on digital platforms, as guided by these psychological frameworks. Furthermore, we present data from a comprehensive evaluation of our processes’ effectiveness. Specifically, we provide quantitative data obtained from social media metrics and qualitative data obtained from focus groups with community members from the 3 communities. This information provides evidence that addresses some of the gaps in knowledge about using social media in health promotion [13] and provides a blueprint for other organizations seeking to partner with communities of color to address the community’s informational needs.

This study was not considered human subject research by the University of Wisconsin–Madison Institutional Review Board and was classified as quality improvement or program evaluation not requiring an institutional review board approval in accordance with federal regulations, as defined under 45 Code of Federal Regulations Part 46.102(d). This report follows the recommendations of the iCHECK-DH (Guidelines and Checklist for the Reporting on Digital Health Implementations) [28] guidelines, and the checklist is available in Multimedia Appendix 2.

The research team contacted individuals to obtain informed consent and provide focus group details. During the introduction of the section, participants were reminded about the volunteer nature of participation and provided additional opportunities to opt out. Participants received US $60 to US $75 as a token of appreciation for participating in the 90-minute focus group; the specific amount changed across the communities based on the timing of the recruitment.

Between August 2021 and January 2023, we created 980 unique social media posts that reached 88,790 individuals and gathered >6700 engagements. Figure 2 describes the overall number of posts per community and platform and their associated social media metrics. Posts directed to members of the Black community showed an ARPP comparable with the other Facebook communities in this project (111.74 for Black or African American vs 119.46 for Hispanic or Latino vs 113.11 Oneida Nation) despite the lowest overall post count. Within the Latino community, Instagram posts had a higher ERR (9.2% vs 3.5% vs 3.2%; P<.001) despite the lower number of posts and reach. Multimedia Appendix 3 shows the demographics of the followers of each of the community platforms.

Some participants reported that they might not engage with a post because they want to avoid notifications or well-intended inquiries from others who can see their interaction.

Table 3 presents the social media metrics according to the type of posts of each platform for the Black community. On Facebook, posts with photos had the largest ARPP (132.85), while posts containing only text had the lowest; videos had the largest ERR (19.53%; P<.001). For Instagram, videos had the largest ARPP, and carousel posts had the largest ERR (P<.001).

All participants reacted favorably to the example video played [29]. Participants cited the following reasons why this video caught their attention:

Table 4 shows the social media metrics for the Facebook page “COVID-19 Information for Oneida Nation Community” according to the post type. Video posts had an ARPP that was statistically significantly larger than other posts (179.27; P<.001), and the largest ERR was seen on posts that had only text.

All Oneida Nation members who participated in the focus groups were positive about posts that used images or videos. Several participants said they preferred videos that were ≤2 minutes in length, with 1 participant checking the video length first, skipping portions of long videos, and watching just the ending:

Participants stated that it is especially important that a video be short if it is a topic that people are not following or not already interested in:

Everyone familiar with the COVID-19 Info for Madison’s Black Community Facebook page praised it for having reliable and relatable information:

Similarly, there was uniform favorability toward posts that offered resources and ideas to promote health. Participants cited the benefits to themselves, their families, and their communities:

Our study shows that social media communications for health promotion were favorably perceived by individuals from Black, American Indian or Alaska Native, and Hispanic and Latino communities when produced by reliable sources with whom they are already connected in real life. Similarly, there is evidence that social network sites are becoming essential to how people experience news, moving from a traditional news cycle that journalists primarily dominate to more complex information cycles that incorporate ordinary people within the process [30].

In our study results, individuals reflected that social media platforms are used for recreational purposes and staying up to date with family and friends. Prior research indicates that individuals from marginalized communities also use social media to find health information [31,32]. Given that racial and ethnic minority groups use these media platforms more than White individuals [13], our study developed processes to meet these populations on digital platforms in a way that promoted trusted and accurate dissemination of public health information. This model is distinct from prior studies using social media influencers as the community influencers were not merely tools for disseminating public health social media messaging created by officials and scientists [33-35].

As described in the Methods section, our community influencers determined what content was relevant to their communities and were empowered through skill building to be codrivers at every stage of the layperson translation process, from strategy to content creation to dissemination and analysis. This approach is similar to previous descriptions of tailored influencer marketing approaches that harness the social influencers’ experiences and competencies [36]. An additional key distinguishing feature of our model is detailed knowledge sharing between the scientists and the community influencers, which led to the self-renaming of the influencers to advocates as they began to take ownership and gain a proper understanding of the scientific content being shared on their pages. According to the attribution theory, these interactions within our team affected how the information was presented and ensured that the most relevant information for the audience was presented first, which might, in turn, influence the audience’s perception of the stability and controllability of health information, leading to a higher likelihood of message acceptance and behavior change.

Our results also highlight that there is not 1 unique strategy that works across various Black, American Indian, Alaska Native, and Hispanic or Latino communities. Health promotion social media messages must be tailored to the audience of interest and the platform used. Facebook audiences were mostly middle-aged individuals, while Instagram had more young adults and teenagers, which is consistent with the existing literature [37]. The Instagram account also reached a higher percentage of men than Facebook and even reached other communities of color in low- and middle-income countries (Colombia, India, Mexico, Nigeria, and Venezuela).

To increase the dissemination of social media messages, these must be easy to understand and attention grabbing to generate interest in audiences while being authentic without creating the sensation of spreading propaganda. The most successful strategies we identified to accomplish these goals were live streaming and videos. Videos had the advantage of overcoming literacy limitations, and when they displayed people, these created a sense of connection and authenticity with the audience, which partially addressed one of the limitations of social media: the lack of reliable sources [13]. These results are similar to those obtained by other studies that reflect that the content needs to be attractive and easy to understand to be effective [38].

The focus group data indicated that people use social media to keep in touch with people they already know or are connected to, which suggests that for future endeavors, it is paramount to identify an advocate who is highly connected to their community, as the audiences would be more likely to pay attention and trust the information presented.

Unsurprisingly, given the historical abuse endured by marginalized communities, trust and authenticity emerged as critical issues in our focus groups. The choice to empower the advocates to drive and create the content is a novel and distinguishing feature of our model and was instrumental in facilitating trust and ensuring authenticity [36]. This approach led to content production that aligned with the personal brand of each advocate and facilitated connectedness to each community, which was identified in the focus groups as important. These findings that social media communications are favorably perceived when produced by reliable sources align with cultivation theory, as consistent social media messaging from trusted sources could cultivate a perception of credibility and reliability in health promotion.

There are multiple studies describing strategies for health promotion that engage community representatives in the dissemination of information to improve trust and facilitate the dissemination of social media messages [39], and that crossposting across multiple platforms is necessary to combat misinformation [40]. However, our study is the first to show how different content types behave across platforms, emphasizing the need to design public health content according to the platform and audience.

Previous studies have used impressions [41,42], reach [41], and engagements [41-45] or specific elements (eg, shares and comments) to assess the dissemination of social media messages. These metrics in raw form are a reflection of the number of followers and might not be an accurate representation of the reach of the messages, and there is limited evidence of how these metrics relate to subsequent behaviors; however, previous studies suggest that these are related. A study by Nadarzynski et al [38] showed that Facebook advertisements could lead to increased traffic to websites promoting chlamydia testing and related sexual health information, resulting in a 41% increase in test requests, showcasing how increases in engagement in social media can be associated with new behaviors.

Selecting the right metrics to evaluate outreach efforts must be individualized to the campaign’s goal. On the basis of our results, we suggest using a combination of reach and engagement metrics, maximizing the use of tools available in social media professional accounts, and evaluating these metrics regularly to facilitate tailoring to the audience [14,37]. Contrary to common recommendations by social media influencers and marketing experts, in our experience, more posts did not lead to more engagement and reach. This might reflect the relevance of the social media messages to the audiences.

In our experience, producing large amounts of content at a fast pace does not allow enough time to create intentional and relevant social media messages and might affect individuals’ recognition of the messages. According to Vakratsas and Ambler [46], a minimum of 3 exposures is often necessary for a social media message to have an impact, with diminishing returns after 10 to 20 exposures. The frequency of these social media messages and the complexity of the message also affect the effectiveness in promoting a new behavior or increasing interest in a product; according to a white paper by Facebook from 2016, one to two social media messages per week might be enough to generate recognition [47], which support the frequency selected for our messages.

Although our study explores many perspectives, 1 limitation was that our focus groups engaged a small number of individuals across the communities, which might affect the richness of our results and potential generalizability to other communities and geographic locations. Considering that our study was performed in Wisconsin, it is important to acknowledge that context factors affect the perceptions and potential interactions that individuals have with our social media posts. For example, in Wisconsin, vaccination against COVID-19 was free to all individuals, and there was no requirement for proof of residence or US citizenship to access these services. This resulted in less fear of deportation and lower mistrust in vaccination social media messages compared with what could be seen in other localities in the United States, where these services were only available to US residents.

Another limitation of our study is that our metrics do not account for reach and engagements after our posts were shared. In July 2021, Facebook analytics changed how it collected and presented metrics for pages; since then, only the metrics of the page are shown, not accounting for engagement on third-party pages, resulting in lower overall numbers. Similarly, our results do not reflect what happened outside the social media platforms, such as sharing information in group chats, WhatsApp, or SMS text messages, which are platforms frequently used by our communities [13]. To ensure uniformity of the metrics, all data for the study period were downloaded in March 2023 and used for all the analyses.

This project created social media content that was medically accurate and culturally appropriate related to the COVID-19 pandemic and its physical and mental health consequences. One key aspect of our work is that we relied on building trustworthy relationships with partners and communities within and outside social media; this was achieved through the work of our advocates and team members who served as a connection with the communities. We identified that stories in videos, photos, or messages that contained only text created meaningful connections with individuals and provided trustworthy information with very encouraging results that we have been sharing with academicians and community organizations. However, our results showcase how analytics reports of the social media platforms (eg, share, likes, and views) remain inadequate metrics for measuring behavioral changes in the target audiences [48] and require the development of further population-health approaches to identify whether social media campaigns such as ours can translate on increased adoption and maintenance of health behavior and subsequently improved health outcomes. Future studies could explore whether social media engagements are associated with changes in beliefs and the acquisition of new health behaviors to promote health and well-being and how social media campaigns can build on community partnerships to address misinformation or disinformation at the population level in ways that are culturally appropriate and result in improved health outcomes.