Consider “MyBehavior,” a mobile app designed at Cornell University and Michigan State University as part of a study to promote healthy behavior change through personalized, in-the-moment suggestions. The app leveraged a machine learning model to generate suggestions contextualized to the participant’s data on physical activity and dietary intake.26
Grounded in established behavioral theories, the app identified low-effort and frequent physical activities (e.g., a specific walk) that users engaged in as part of their daily lives and encouraged increasing these behaviors in attainable ways. For instance, to address stationary behavior, the app algorithms generated prompts such as walking three steps for every hour spent in a sedentary position. Such low-effort prompts were intended to push users into action even when their motivation was low. A preliminary evaluation of the research indicated that a significant percentage of the app users engaged in more physical exercise.27
Rooting campaigns in behavioral theories becomes even more important when addressing concerns that involve potential stigmas such as mental health, sexual health, and opioids use disorders. While internet- or mobile-based interventions are becoming more common methods to address such health concerns, audience uptake remains low as those with sensitive health issues typically prefer in-person interactions.28
Digital interventions that mimic human behavior can fill in the gaps in such cases that would otherwise require in-person intervention. For instance, AI-based conversational tools that mirror face-to-face human dialogue can reach out to audiences that might be reluctant to seek support. These automated nonhuman agents can act as a substitute to in-person interactions while providing a space for unrestrained dialogue.
A 2017 randomized controlled trial conducted at Stanford University tested the feasibility of an AI-based automated conversational agent to deliver cognitive behavioral therapy for reducing anxiety and depression in a sample of American college students aged 18 years and above. Apart from providing cognitive behavioral therapy, the automated agent was programmed to deliver empathic responses and tailored advice based on the mood of the user, encourage regular check-ins for greater accountability, send personalized prompts to facilitate engagement, and promote reflection by providing weekly charts exhibiting each user’s mood over time. As measured by a nine-item health questionnaire (PHQ-9), the study reported a significant reduction in depression after two weeks for students who engaged with the chatbot.29
Despite the efficacy that digital interventions promise, the human element remains an essential component of public health campaigns.30 For instance, involving community members in the development phase or engaging social influencers well-known in target communities for culturally sensitive interventions. In fact, one research study found that individuals with similar beliefs were more likely to engage intensely with their community members than with outside entities.31
Use human-centered design for a deeper audience connection
Human-centered design is an approach to problem-solving that incorporates human perspectives at all stages. It involves developing a deep empathy with the target audience, understanding their perspectives and barriers, collaboratively ideating, as well as rapidly prototyping and iterating those ideas with the users.
Human-centered design approaches to personalizing public health communication have been growing in popularity globally. Some campaigns enhance their influence by combining a human-centered design approach with a community-based participatory research approach in which researchers and community members collaboratively work on all research aspects ranging from study design to analysis. This method is designed to ensure audience voices are incorporated throughout the process.32
Community participation in developing public health campaigns increases responsivity to issues that are unique to each community, such as sociodemographic profile, cultural composition, and influence of local norms on behavior.33 Also, community involvement can empower audiences, help strategists gain first-hand knowledge of audiences’ perceptions and preferences, and instill an element of trust in the process.34 The “Layla’s Got You” campaign described earlier showcases an example of involving members of the community at every stage of the campaign design for targeted messaging.35
Partner with culturally relevant influencers
Social influencers have emerged at the forefront of many public health campaigns, helping to shift perceptions and health behaviors. From addressing eating disorders in adolescents to instilling positive attitudes toward the influenza vaccine, influencers have helped drive greater impressions and engagement using social media.36 Because influencers already have an established level of trust with their audiences, they can be uniquely suited to shape perceptions and behaviors. Public health messaging can gain credibility and further engagement by working with influencers to amplify a campaign’s message within their communities.37
While an influencer can be well-known, an effective messenger does not have to be a celebrity or prominent individual. Some public health campaigns rely on what are commonly referred to as microinfluencers, especially while targeting hard-to-reach communities. Microinfluencers, as the name suggests, have a niche group of followers in specific geographies and can build trusting relationships with their followers who are often concentrated in a specific community. Given that microinfluencers are likely to be viewed as a friend or a peer as opposed to a celebrity who seems out of reach, they can potentially have a sizable impact on audiences’ beliefs and perceptions.38 Online technologies are enabling strategists to identify microinfluencers with significant local followings across social media platforms.39
The Stop Flu campaign (2018) employed microinfluencers to build positive attitudes toward the influenza vaccine among African Americans and Hispanics residing in Kaiser Permanente service areas (Northern and Southern California, Colorado, Georgia, Hawaii, Mid-Atlantic States, Oregon, and Washington), i.e., areas in which the organization has an active presence. For this campaign, a microinfluencer was defined as someone with 500–10,000 followers on at least one social media account. With the help of user-generated content from these influencers, the campaign reached out to target audiences in the identified areas. A postcampaign evaluation suggested a significant positive shift in the audiences’ perceptions toward the influenza vaccine.40
Step No. 3: Assess impact throughout implementation
Advertising is often imprecise in its attempt to grab audience attention, requiring multiple iterations before finding messaging that demonstrably influence audiences. Even if a public health message is shared in the right context, it may or may not resonate with a particular audience at a particular moment in time. This is where digital methods shine: while a static message on television or radio has only one chance for success, digital messaging can be evaluated throughout implementation and rapidly and continuously improved upon based on audience reactions.
Monitoring outcomes during implementation is increasingly possible with the analytic capabilities that accompany digital methods. Advanced analytics, sometimes on a real-time basis, can assess how well messages perform with particular audiences, identify facets of messaging that need revision, or even spotlight messaging that is successful in surprising ways and worthy of replication elsewhere. The ability to continuously evaluate messaging performance toward campaign goals allows for course correction and may result in quality improvements to campaigns.
However, strategists should bear in mind two important things when using evidence to adjust interventions during implementation. First, strategists should select metrics that align with the campaign objectives. Second, they should conduct A/B testing, a marketing research method to test two hypotheses at the same time, which can help public health leaders to determine the most optimal campaign content and design.
Determine the right mix of metrics (modern and traditional)
Selecting appropriate metrics for evaluating campaign impact depends on the goals and objectives of the campaign. Does the campaign aim to raise awareness, address misinformation, change interactions, or influence behavior?41 (To know more about managing online misinformation, see sidebar “Tackling online misinformation.”) Awareness, for example, can be captured by the reach of the campaign (i.e., the number of people who view a campaign’s social media posts).42 Interactions between audience members or between the audience and the source (e.g., a social media channel) can be measured by the level of engagement.43
Evaluating behavior change using digital metrics is, however, particularly challenging. For instance, metrics such as likes, shares, page visits, and comments on social media platforms at best provide a directional indication of the audience’s intentions and may or may not equate to behavior change on the ground. If a public health campaign aims to change behavior, campaign strategists will need to specify the call to action at the outset to accurately assess outcomes later. For instance, a lung cancer screening campaign might ask its target audience questions such as: “Have you talked to your doctor about getting the new CAT scan test that can detect lung cancer early?”
Given that it’s difficult to measure behavior change on online platforms, using a mix of modern digital and traditional metrics can provide sharper insights into the overall impact. Focus group discussions, social media survey questionnaires, or omnibus surveys (surveys that collect and examine information on a number of unrelated subjects and can be used by multiple research clients) using website voting buttons can complement more modern metrics such as likes, shares, and comments, and can add to the accuracy of evaluating impact.44