What message appeal and messenger are most persuasive for COVID-19 vaccine uptake: Results from a 5-country survey in India, Indonesia, Kenya, Nigeria, and Ukraine

Effective strategies to encourage COVID-19 vaccination should consider how health communication can be tailored to specific contexts. Our study aimed to evaluate the influence of three specific messaging appeals from two kinds of messengers on COVID-19 vaccine acceptance in diverse countries. We surveyed 953 online participants in five countries (India, Indonesia, Kenya, Nigeria, and Ukraine). We assessed participants’ perceptions of three messaging appeals of vaccination—COVID-19 disease health outcomes, social norms related to COVID-19 vaccination, and economic impact of COVID-19—from two messengers, healthcare providers (HCP), and peers. We examined participants’ ad preference and vaccine hesitancy using multivariable multinomial logistic regression. Participants expressed a high level of approval for all the ads. The healthcare outcome–healthcare provider ad was most preferred among participants from India, Indonesia, Nigeria, and Ukraine. Participants in Kenya reported a preference for the health outcome–peer ad. The majority of participants in each country expressed high levels of vaccine hesitancy. However, in a final logistic regression model participant characteristics were not significantly related to vaccine hesitancy. These findings suggest that appeals related to health outcomes, economic benefit, and social norms are all acceptable to diverse general populations, while specific audience segments (i.e., mothers, younger adults, etc.) may have preferences for specific appeals over others. Tailored approaches, or approaches that are developed with the target audience’s concerns and preferences in mind, will be more effective than broad-based or mass appeals.


Introduction
vaccine concerns and increase acceptance [15]. Beyond the appeal, the messenger can also strongly impact the effectiveness of the message. Prior evidence has highlighted the importance of a healthcare professional vaccine recommendation as one of the strongest drivers of vaccine uptake [2].
Effective strategies to encourage COVID-19 vaccination uptake must consider how different aspects of health communication can be tailored to specific contexts. This gap in research informed our study aimed to evaluate the influence of three specific messaging appeal framings-COVID-19 disease health outcomes, social norms related to vaccination, and economic impact of COVID-19-of vaccination from two kinds of messengers, healthcare providers and peers, on COVID-19 vaccine acceptance in a diverse set of countries.

Ethics statement
The study protocol was approved by the Johns Hopkins Bloomberg School of Public Health Institutional Review Board (No:00018565) prior to study initiation. A brief consent statement appeared on the screen used for the survey. Informed consent was obtained online through the presentation of the consent message and asking a single question about whether the participant would like to participate in the study.

Participants
We conducted online surveys in five countries (India, Indonesia, Kenya, Nigeria, and Ukraine), using SurveyMonkey, a survey and panel research company. We chose these countries given their history with vaccine hesitancy, as well as related COVID-19 morbidity and mortality. For data quality checking for online panel participants, SurveyMonkey employs bot and fraud detection when recruiting panel participants [16]. For inclusion in our online panels, participants had to be located in the target country, over the age 18, and proficient in written English.

Procedures
As this was a descriptive and exploratory study, we did not have apriori hypotheses that we sought to test. We were interested in ad preferences among users. A total of 1,894 responses were collected. Gender balance was a parameter we applied to achieve approximately 50% representation of both males and females. To ensure high-quality results for analysis, we excluded responses from the analysis if there were indications of poor attention or "speed racing" responses. Exclusion criteria included unrealistically short completion times (answering the 55 survey questions in less than 5 minutes, n = 704), failure of attention checks (n = 217), failure to answer all questions, including open text responses (n = 20). Of the total 1,894 survey responses collected, 953 met our quality criteria for inclusion in the analysis (Fig 1). We undertook these quality checks to ensure that we were able to capture the highest quality responses. We determined our sample size based upon calculation of confidence interval width around an expected sample proportion of vaccine hesitancy across what is known about vaccine hesitancy in each country we included. We estimated vaccine hesitancy prevalence to range between 10-50%; a sample size of 500 participants will yield two-sided 95% confidence intervals with a width of approximately 0.05.
Each participant viewed six ads, which were broadly composed of two elements: a distinct messenger and a distinct appeal. The messengers included a healthcare provider image, which depicted a medical provider talking to a patient, and a peer image, which depicted two people speaking to each other. We included the following appeals: health outcome, which focused on the risk of COVID-19 disease and the protective effect of vaccination against disease; economic benefit, which focused on loss of work time and income due to COVID-19 infection and the protective effect of vaccination against economic loss; and social norms, which focused on how most people have received the COVID-19 vaccine and the protective effect of vaccination for the community. We chose these appeals after conducting a scoping review to determine which appeals may be most effective in nudging an individual to accept a vaccine.
After each ad, six questions covering participant level of agreement specific to each ad were asked, including relevance (Please indicate your level of agreement with the following statement: this ad was relevant to me), motivation to get the COVID-19 vaccine (Please indicate your level of agreement with the following statement: this ad motivates me to get the COVID-19 vaccine), motivation to get COVID-19 vaccination for their child (if applicable) (Please indicate your level of agreement with the following statement: this ad motivates me to get the COVID-19 vaccine for my child under 18 years of age), if the ad was designed for someone like the participant (Please indicate your level of agreement with the following statement: this ad was designed for people like me), and if the ad would prompt the participant to tell someone about the COVID-19 vaccine (Please indicate your level of agreement with the following statement: this ad would prompt me to tell someone about the COVID-19 vaccine). Participants were also asked to indicate which ad of the six would motivate them the most to get the COVID-19 vaccine (Which ad motivates you most to get the COVID-19 vaccine?).
Three questions were used to assess participant vaccine hesitancy. Participants were asked if they had ever delayed getting a recommended vaccine through a yes/no response (Have you ever delayed getting a recommended vaccine or decided not to get a recommended vaccine for reasons other than illness or allergy?). Two questions asked participants about their level of agreement related to COVID-19 vaccine safety concern (How concerned are you that a COVID-19 vaccine might not be safe?), and participant perception of vaccine effectiveness (How concerned are you that a COVID-19 vaccine might not prevent the disease?).
To further examine the reason why an individual may not want to receive the COVID-19 vaccine, participants were asked to rank six statements based on level of concern, with one being the most concerning, to six being the least concerning (There are multiple reasons why someone may not want to get the COVID-19 vaccine. Please rank these reasons in order, with 1 being the most concerning, to 6 being the least concerning.). Concerns included safety (I do not feel the vaccine is safe), vaccine effectiveness (I do not feel the vaccine is effective), trust in government (I do not trust the government), vaccine experience (People I know have not had a good experience getting the vaccine), and cost (It is cost prohibitive for me to get the vaccine), and belief in the existence of COVID (I believe COVID-19 is not real).
We constructed an ordinal scale from 0 to 3 to describe participants' level of vaccine hesitancy by assigning 1 point for each of three yes/no questions. We then examined the distribution of scores and established a cut-off to stratify participants into two groups (0-1 and 2-3), which we defined as lower hesitancy and higher hesitancy. Statistical analysis was performed in Stata 16.1 (Stata Corp, College Station, TX). The study received ethical approval from the Institutional Review Board at the (blinded for review).

Statistical analysis
We summarized participant characteristics overall, participant responses and assessed differences between countries using chi-squared tests. We collapsed ad preference responses into binary variables (strongly agree/agree vs. strongly disagree/disagree) and examined the proportion responding positively across constructs for each ad. Variables for some participant characteristics were collapsed due to small numbers in some categories, including age (collapsed to <40, �40 levels for regression models), gender (female and male levels used for regression models as the "Prefer not to say" option had only n = 3 observations).
We examined ad preference by asking participants which ad they most preferred and comparing responses by country and participant characteristics using multivariable multinomial logistic regression. Individual country models were used to assess within-country differences. Participants were pooled due to the identical components across the five countries, such as inclusion and exclusion criteria, recruitment methods, measures, intervention approach, and procedures of study implementation [17]. To explore further differences (India, Indonesia, Kenya, Nigeria and Ukraine), a pooled participant model was used to assess cross-country preferences. We estimated relative risks and 95% confidence intervals using ad 1 as the reference group (Health Outcome-Healthcare Provider ad) and a binary variable for vaccine hesitancy as our primary characteristic of interest. To examine associations for ads depicting healthcare providers and peers independently, we stratified on this condition and modeled the relationships in separate multivariable multinomial models. Included in models were participant characteristic data known from the literature to be associated with vaccine attitudes, including age and gender.
More than a quarter of participants (n = 288, 31.8%) reported having ever delayed or refused a recommended vaccination. Most participants were at least slightly concerned that the vaccine might not prevent COVID-19 disease (n = 512, 53.7%) or might not be safe (n = 595, 62.43%). The majority of participants reported more concerns (participants were either moderately or extremely concerned), when asked about the vaccine for pregnant women (n = 531, 55.7%) and children under the age of 18 (n = 540, 56.7%). Few participants had a score of 0 on our vaccine hesitancy scale (n = 77, 8.1%). The vaccine hesitancy scale distribution across the three questions for the remaining participants was 1 (n = 178, 18.7%), In general, participants agreed with the message (level of agreement > 90%) across all six ads ( Table 2). For all countries except Kenya, the majority of participants indicated a preference for the health outcome-healthcare provider ad. Participants in India (26.6%), Indonesia (31.9%), Nigeria (34.9%), and Ukraine (31.6%) reported the health outcome-healthcare provider ad as the ad most likely to motivate them to get the COVID-19 vaccine (Fig 2). Participants in Kenya reported a preference for the health outcome-peer ad (35.1%) followed by the health outcome-healthcare provider ad (28.9%). Across country pooled descriptive analysis indicated participants reported the health outcome-peer ad (92.3%) as the ad that would prompt them the most to tell someone about the COVID-19 vaccination. Participants found the health outcome-healthcare provider ad to be relevant to them (92.3%), motivated parents to vaccinate their children (67.5%), and motivated them to get the COVID-19 vaccine (91.9%). Multivariable multinomial regression models were used to examine individual countries with ad preference. Indonesian participants with a graduate degree were more likely to prefer the economic outcome-healthcare provider ad (RR = 5.16, P = .05, 95% CI [1.03, 25.79]). Indonesian participants (RR = 6.46, P < .01, 95% CI [1.78, 23.52]) with higher vaccine hesitancy reported a preference for the social norm-healthcare provider ad (S2 Table). Male Ukrainian participants reported a preference social norm-peer ad (RR = 0.21, P = .03, 95% CI [.05, .87]) (S5 Table). Kenyan male participants (RR = -1.25, P = .04, 95% CI [-2.30, -0.20]) and participants with moderate/high vaccine hesitancy (RR = -1.21, P = .02, 95% CI [-2.37, -0.06]) were less likely to prefer the social norm-healthcare provider ad (S3 Table). Nigerian and Indian multivariable multinomial regression models did not indicate significant findings (S1 and S4 Tables).

Discussion
As vaccination, not vaccines, saves lives, understanding how to improve persuasive communication to promote vaccine uptake is crucial. While both the message itself and the message appeal is critical to nudge individuals toward vaccination [18], the messenger also plays a key role in building trust in vaccines [19]. This is one of the first studies to examine and compare multi-country vaccine ad preferences in an online, English-speaking, adult population across a diverse set of low-and middle-income countries encompassing Asia, Africa, and Europe. This study has several key implications for informing persuasive messaging to improve vaccine uptake and provides important guidance for effectively engaging with distinct populations. First, across all 5 countries, almost one-third of participants reported having ever delayed or refused a recommended vaccine. There are clearly concerns about vaccines, and more so among vaccines for pregnant women and children, regardless of country, which is not new [20]. This finding indicates a need to continue to identify specific concerns across populations to develop and implement approaches to overcome such concerns [21,22]. Our results also point to the fact that across countries many people have concerns over the effectiveness of COVID-19 vaccines in preventing illness. When COVID-19 vaccine rollout began, there was a missed opportunity to frame these vaccines as a means in which to prevent severe disease or hospitalization rather than prevention of illness altogether. With refined messaging, these perceptions and concerns can be addressed. Correcting this misperception will also help manage expectations related to vaccine effectiveness [23].
Second, all six ads-all three appeals and both messengers-were received favorably across all five countries. These findings suggest that appeals related to health outcome, economic benefit, and social norms are all acceptable to diverse general populations, while specific audience segments (i.e., mothers, younger adults, etc.) may have preferences for certain appeals over others [24]. Our findings also showed that both health care providers and peers served as acceptable messengers. Given the decline in trust in health care systems globally [25], and hesitancy among health care providers themselves [26], it is prudent to identify a variety of trusted messengers, including non-health care messengers, in order to effectively promote vaccination. Understanding the role of non-health care messengers has been studied in the US [27,28], and studies have identified that faith-based leaders [29], for example, are credible and trusted messengers that could play crucial roles in promoting vaccination. Along with identifying additional trusted messengers, additional appeals for vaccination should also be tested in these countries to further aid effective audience segmentation [30,31]. While this work identifies promising communication strategies that appeal to diverse, multi-national audiences, it also identifies country-specific differences that may aid the effectiveness of immunization appeals on a national level in LMIC settings. Regarding ad preference by country, Ukrainian participants were more likely to prefer the ad with a peer messenger and a health outcome appeal. Studies have shown that Ukraine is plagued with misinformation [32] and that there is low trust in the Ukrainian government [13], which may explain why a peer is preferred over a health provider as a persuasive messenger for vaccination. Nigerian participants were more likely to prefer appeals related to social norms with both healthcare provider and peer messengers. This demonstrates that social norms may be powerful influences for improving vaccine uptake in Nigeria, suggesting that identifying individuals that are perceived as peers, rather than hierarchical, may be useful for vaccine promotion [33].
Encouraging pregnant individuals to accept COVID-19 vaccines where they are available and recommended has been challenging globally. As most COVID-19 trials did not include pregnant participants [34], conveying the safety and effectiveness of COVID-19 vaccines during pregnancy continues to be a challenge. Pregnant individuals across all 5 countries tended to prefer the peer messenger. This finding is interesting, given the literature that suggests that pregnant women rely on their health care provider for vaccine recommendations to inform their own vaccine behavior [35]. It would be prudent to test additional peer messengers as potential channels for promoting vaccination among this group [36].
This study has several limitations. First, we relied on an online panel to test our message appeals and messengers. Selection bias associated with online surveys has been well documented [37], as such approaches tend to underrepresent individuals who are older, lacking internet access, have lower income, and have less education. We did not design this study to be a population representative sample. We were interested in understanding ad preferences using in this descriptive study, using samples from diverse countries. Although online surveys have several limitations as a research tool, they also provide a promising tool for researchers to reach diverse audiences outside of the more traditionally accessible WEIRD populations (Western, Educated, Industrialized, Rich, and Democratic) [38]. While these tools are not perfect, they do allow us to reach populations that are traditionally under-reached. This study's cross-sectional nature allows us to capture a moment in time, even though vaccine attitudes shift over time, given a multitude of factors.
Despite these limitations, this study is one of the first to test appeals and messengers in across a variety of low-and middle-income countries. To increase vaccine acceptance, identifying preferences for appeals and messengers is paramount. Tailored approaches, or approaches that are developed with the target audience's concerns and preferences in mind, will be more effective than broad-based or mass appeals [39]. Vaccine behavior is complex with many determinants. To ensure global vaccine uptake remains adequate, it is important to meet people where they are and to respond to their concerns through trusted messengers relevant to specific audiences and appeals that are salient and relevant. We are hopeful that this work will help inform future messaging and will inspire researchers and practitioners across the globe to examine how they can more effectively promote vaccines given their target audiences.
Supporting information S1