Targeted Vaccine Messaging to Promote COVID-19 Vaccines for Children and Youth

Data were collected through the Voices of Child Health in Chicago (VOCHIC) Parent Panel Survey, a triannual survey of Chicago parents from all 77 neighborhoods in Chicago on topics related to child, adolescent, and family health. Parents were originally invited to the panel via address-based sampling mailers. Some parents remain in the panel for multiple survey waves, others complete only one survey. Parent eligibility criteria included being ≥18 years old, living within the city of Chicago, and being the parent or guardian of at least 1 child aged 0 to 17 years who lived in the household. The survey was administered by NORC at the University of Chicago in English and Spanish, either online or by telephone between October 18 and November 23, 2021. The sampling approach was multipronged, using returning VOCHIC panel members, NORC’s AmeriSpeak panel members, respondent-driven sampling, and nonprobability samples from other vendor panels. Respondents were compensated $5 if they were returning panel members and $15 if they were joining the VOCHIC panel for the first time because first-time respondents completed a longer demographic survey. Non-probability panel respondents were compensated independently by their home panel. The Institutional Review Boards at Ann & Robert H. Lurie Children’s Hospital and NORC approved all study activities. Samples of parents recruited through the Voices of Child Health in Chicago using the same recruitment methods have informed other previously published research.^14–16 

Parents were asked if they had already vaccinated any of their children against COVID-19, and any parent who responded that they had at least 1 child who was not yet vaccinated was randomly assigned to read 1 of 4 distinct messages about COVID-19 vaccines. Previous research regarding messages to promote MMR vaccination revealed that communications that either corrected vaccine misconceptions or used fear appeals (eg, highlighted the dangers of diseases prevented by the MMR vaccine) were ineffective at promoting pediatric vaccination intentions among parents.¹¹  Therefore, the current study tested alternative message types familiar from pediatric and public health approaches: (1) the COVID-19 vaccine is well-tolerated by children, with few side effects, (2) the COVID-19 vaccine is safe and tested, (3) trusted parents are vaccinating their children against COVID-19, and (4) a control condition that simply provided information about the anticipated timeline for authorization of vaccine in children (Table 1). Respondents were unaware of other experimental conditions. After reading the vaccine message, parents were asked: “How likely are you to get your child(ren) vaccinated against COVID-19?” Response options were “very likely”, “somewhat likely”, and “not likely”.

Survey items were written at a 7.4-grade reading level on the basis of the Flesch-Kincaid readability test and were pretested by experts and parents; feedback from parent pretest respondents was used to enhance readability and ensure that questions were interpreted as intended. All survey items can be found on the Voices of Child Health in Chicago Open Science Framework page (osf.io.cjz82).

Parents also provided demographic information (Table 1). For ease of analysis, household income was combined into 3 groups on the basis of federal poverty level¹⁷  (FPL; <100% FPL, 100% to 399% FPL, 400%+ FPL). Parents’ self-reported race and ethnicity were combined into 4 groups (Black, white, Latinx, and Asian/other race). Appreciating that race/ethnicity is a social construct, the investigators included parents’ self-reported race/ethnicity because of the strong association of COVID-19 illness with race/ethnicity in the United States and the association of childhood vaccination patterns with race/ethnicity.^18,19  Parent education was combined into 3 levels (high school education or below, some college, college degree or higher). Parent age was collapsed into 2 groups (18–35 years and 36+ years).

Parent participants were first recruited from the probability-based VOCHIC panel and NORC’s AmeriSpeak panel at which a 29.4% response rate was observed (708 responses from 2409 eligible invitees). To ensure sufficient sample size to permit subgroup analyses across the randomized exposure groups, the probability sample was augmented by calibration-weighted, nonprobability-based responses obtained through other vendor panels. A response rate (ie, denominator) for the nonprobability sample was unable to be measured because these panels administer opt-in online surveys.

Survey response data were weighted to be representative of the population of Chicago parents. For the probability-based sample, base sampling weights are adjusted to account for nonresponse via a raking ratio method to American Community Survey 18+ Chicago parent population totals associated with the following topline sociodemographic characteristics: age, sex, education, race/ethnicity, and Chicago Community Side, and the following sociodemographic interactions: age × sex, age × race/ethnicity, and race/ethnicity × sex. The combined probability sample and nonprobability sample weights were derived by using small-area estimation methods.²⁰  These methods are frequently used by the US Census Bureau and national survey research organizations because of their efficiency and effectiveness.^21,22  Child-level survey weights were created by starting with the calibrated parents’ survey weights and by using a raking ratio method to align the weighted children’s distribution to the American Community Survey Chicago 0- to 17-year-old population totals associated with the following sociodemographic characteristics: age, sex, and race/ethnicity.

Rao-Scott χ² analyses and logistic regression were used to compare differences in the proportion of parents who were very likely to vaccinate their children against COVID-19 (vs less than very likely) between the randomly assigned messaging groups and by demographic factors. χ² analysis also was used in subgroup analyses to assess racial/ethnic disparities across the four message groups, and in subgroup analyses to explore message effects separately for parents who were already vaccinated and those who were not. We estimated predicted margins of intention to vaccinate by the message type (ie, trusted parents, safe and tested, well-tolerated, and control) and parent race/ethnicity derived from the adjusted logistic model and after holding each covariate at the observed values. Average marginal effects were the differences between the predicted margins.

Point estimates and confidence intervals (CIs) were adjusted for the clustering of children within households and, therefore, by parent respondent. Analyses were performed in SAS software version 9.4 (SAS Institute, Inc. Cary, North Carolina) and R version 2022.12.0 (RStudy Vienna, Austria) with α significance level set at P <.05.

In total, surveys were completed by 1142 parents who reported on 1977 children in surveyed households. Analyses in the present research are based on responses from 898 parents about 1453 children (73.4%) who had not yet received a COVID-19 vaccine at the time of the survey. Parent and child demographics did not differ significantly between message randomization groups (Table 2).

Vaccine intentions were highest for parents who received the trusted parents message (53.3% very likely, 95% CI: 43.1% to 63.4%), followed by the safe and tested message (48.9%, 95% CI: 39.5% to 58.2%). Lower proportions of very likely to vaccinate responses were found for the well-tolerated message (41.5%, 95% CI: 32.7% to 50.3%) and the control condition (37.5%, 95% CI: 29.5% to 45.4%) (Table 3). Pairwise comparisons indicated that the trusted parents and safe and tested messages were associated with significantly higher vaccination intentions compared with the control condition (P = .02 and P = .03, respectively). The well-tolerated condition did not differ significantly from the control condition (P = .62). Significant differences were not observed in pairwise comparisons between safe and tested and trusted parents (P = .65), well-tolerated and trusted parents (P = .15), or well-tolerated and safe and tested (P = .16).

Results of a logistic regression model that included parent and child demographic variables indicated that the trusted parents message was associated with higher odds of being highly likely to vaccinate a child compared with the control condition (adjusted odds ratio [aOR] = 1.98, 95% CI: 1.14 to 3.44), but the safe and tested and well-tolerated messages were not associated with higher odds of being highly likely to vaccinate a child than the control condition (aOR = 1.59, 95% CI: 0.92 to 2.73; aOR = 1.24, 95% CI: 0.76 to 2.01, respectively). Child and parent demographics also were associated with the odds of being highly likely to vaccinate a child (Table 4).

Overall, Asian/other parents were most likely to report they were very likely to vaccinate their children (62.3%, 95% CI: 46.5% to 78.1%), followed by white parents (61.6%, 95% CI: 54.8% to 68.4%), Latinx parents (42.8%, 95% CI: 36.1% to 49.5%), and Black parents (30.0%, 95% CI: 19.8% to 40.1%) (omnibus P <.001).

We examined whether there were associations between vaccine intentions and parents’ self-reported race and ethnicity because the COVID-19 pandemic has had a disproportionately negative impact on historically marginalized racial and ethnic groups, and we wanted to understand if some message types were differentially effective at reducing inequities in COVID-19 vaccination intentions. For these analyses, the Asian/other-race group was removed because for this subsample the cell sizes were too small within each randomization group. However, the results were consistent when the Asian/other-race group was included or removed.

In this analysis, health inequity is evidenced by differences in intentions to vaccinate children by race and ethnicity. There were significant differences in vaccination intentions by parent race and ethnicity in 2 conditions: the well-tolerated message and the control condition. In the control condition, 63.2% of white children were very likely to be vaccinated (95% CI: 50.8% to 75.6%), compared with 32.6% of Latinx/Hispanic children (95% CI: 21.5% to 43.6%), and 15.0% of Black children (95% CI: 3.7% to 26.3%; P <.001). Similarly, in the well-tolerated condition, 65.2% of white children were very likely to get vaccinated (95% CI: 49.9% to 80.4%), compared with 39.4% of Latinx/Hispanic children (95% CI: 25.9% to 52.9%), and 20.5% of Black children (95% CI: 6.5% to 34.5%; P <.01) (Table 5).

However, in the trusted parents condition and the safe and tested conditions, the difference in vaccination intentions by race and ethnicity was no longer significant. In the trusted parents condition, 61.1% of white children were very likely to be vaccinated (95% CI: 49.2% to 73.1%), followed by 51.6% of Latinx/Hispanic children (95% CI: 35.7% to 67.5%), and 49.9% of Black children (95% CI: 24.6% to 75.2%; P = .86). In the safe and tested condition, 56.9% of white children were very likely to be vaccinated (95% CI: 42.6% to 71.2%), followed by 49.6% of Latinx/Hispanic children (95% CI: 36.8% to 62.4%), and 40.7% of Black children (95% CI: 19.0% to 62.4%; P = .73).

To explore whether there were differences in message effects by parent race/ethnicity, we calculated the average marginal effect of each message type compared with the control message within each parent race/ethnicity group, holding all other covariates at their observed values. The average marginal effect (AME) among Black parents for the trusted parents message was AME = 0.37 (95% CI: 0.12 to 0.63; P <.01), indicating that, among Black parents, receiving the trusted parents message was associated with an increase in intention to vaccinate of 37% compared with the control condition. The average marginal effect among Black parents for the safe and tested condition was AME = 0.28 (95% CI: 0.02 to 0.55; P <.05). These results indicate that the trusted parents message and the safe and tested messages were effective among Black parents. No other average marginal effects were significant.

To understand whether vaccine message type was associated with vaccination intentions among a group of parents who may have been more vaccine-hesitant in general, we examined the effect of message type separately for parents who were not yet vaccinated themselves against COVID-19 and for those who were already vaccinated separately. Among parents who were not yet vaccinated against COVID-19 at the time of data collection (26.8% of parents), there was a significant effect of message type. Higher intentions to vaccinate children were seen in the safe and tested condition (37.7% very likely to vaccinate, 95% CI: 22.1% to 53.3%) and the trusted parents condition (27.8%, 95% CI: 14.4% to 41.1%) and lower intentions were seen in the control condition (16.9%, 95% CI: 7.4% to 26.4%) and the well-tolerated condition (10.4%, 95% CI: 4.3% to 16.6%; omnibus P <.01). Pairwise comparisons among only unvaccinated parents indicated that the trusted parents and safe and tested messages were associated with significantly higher vaccination intentions compared with the control condition (P = .02 and P = .02, respectively).

Among only parents who were already vaccinated, the effect of message type was not significant: 63.7% of parents in the trusted parents condition were very likely to vaccinate (95% CI: 51.8% to 75.7%), followed by 62.4% in the well-tolerated condition (95% CI: 52.1% to 72.6%), 59.0% in the safe and tested condition (95% CI: 48.3% to 69.6%), and 55.0% in the control condition (95% CI: 45.1% to 65.0%; omnibus P = .59).

This study compared the effect of 4 distinct vaccine messages on parents’ COVID-19 vaccine intentions for their children in a large, socioeconomically diverse US city, using a population-based survey experimental method that has been used for other research questions involving parents’ intentions related to their children’s health care, including vaccination against other diseases.^11–13  Overall, parents’ intentions to vaccinate children were higher if a parent received the trusted parents message compared with a control condition. Intentions to vaccinate children among parents who received the well-tolerated message did not differ significantly from the control condition. In an adjusted model, the trusted parents message was associated with higher odds of intending to vaccinate even after controlling for demographic characteristics. The safe and tested message resulted in a higher unadjusted proportion of parents very likely to vaccinate their child compared with the control condition, but in the logistic regression, the point estimate for this message type, although relatively large, was not significant. This suggests that the safe and tested message may be more effective than the control, but further research is needed to determine this more definitively. Together, these findings illustrate the importance of specific messaging strategies, and raise questions about whether combining the 2 message types, trusted parents and safe and tested, would have an additive effect on intentions to vaccinate children against COVID-19.

Racial and ethnic differences in intentions to vaccinate their children were evident in the control condition and in the well-tolerated condition; however, these differences were not present in the trusted parents and the safe and tested conditions. Findings from the marginal analysis suggest that the trusted parents and the safe and tested messages were particularly effective among Black parents. Among parents who were not yet vaccinated themselves, higher child vaccine intentions were seen in the trusted parents and the safe and tested conditions, compared with the control condition, but not in the well-tolerated condition. Taken together, the findings with respect to race/ethnicity and not-yet-vaccinated parents suggest that specific vaccine messages may be particularly effective at encouraging child vaccination among parents who are vaccine-hesitant.

Among parents who were already vaccinated, none of the message types in the current study resulted in higher child vaccine intentions compared with the control condition. The majority of vaccinated parents were very likely to vaccinate their children in each of the message conditions; therefore, it is possible that vaccinated parents already held positive views of vaccines and that the message types in this study did not elicit a strong enough reaction to further increase their vaccine intentions.

Our findings must be interpreted in the context of potential limitations related to our study context and approach. First, our methodology does not permit within-persons analyses to examine whether parents’ intentions to vaccinate their children increased after reading a message compared with before they read the message. Future research would benefit from evaluating these messages using a pre- post- design to determine the precise change in parents’ intentions to vaccinate their children.

Second, parents in our sample came from a major US city, and it is possible that their attitudes about pediatric COVID-19 vaccines differ from parents who reside in suburban or rural settings. However, the demographics of Chicago are similar to those of the United States more broadly, strengthening the generalizability of these results to broader populations at the national level.²³  Third, the current findings reflect parents’ stated intentions to vaccinate their children against COVID-19, rather than actual vaccination behaviors. We do not know whether parent respondents ultimately did or did not vaccinate their children. However, because we asked parents about their intentions at a high level of specificity (eg, intentions for each unvaccinated child in the household), the intentions that we captured may more closely align with vaccination behavior because there is evidence that more specific attitudes and intentions are better predictors of behavior than general attitudes.²⁴  Further research is needed to determine if the message types identified as most effective in this study result in increased vaccination rates in clinical settings. Finally, this survey was administered between October and November 2021, when the Food and Drug Administration’s emergency use authorization for COVID-19 vaccines in children 5 to 11 years old was still new and the emergency use authorization for COVID-19 vaccines in children <5 years old had not yet occurred. The landscape of COVID-19 vaccines for children is continuously evolving, and it is possible that parents’ concerns about the COVID-19 vaccine may change as vaccines move through the Food and Drug Administration approval process. Therefore, continued research on parents’ concerns and vaccine messaging is critical to continue to encourage vaccination as pediatric vaccines become available to wider audiences.

In summary, the results of this study indicate that the most effective messages to encourage parents to vaccinate their children against COVID-19 highlight that other trusted parents have decided to vaccinate their children against COVID-19, and messages that highlight that the vaccine is safe and thoroughly tested may also be effective. Importantly, these results also indicate that messages that focus on the COVID-19 vaccine being well-tolerated by children were less effective at promoting parents’ vaccine intentions for their children. In public health contexts, these findings suggest that it may be beneficial to highlight that other trusted parents have decided to vaccinate their children, and the safety and thorough testing of the vaccine. These findings also may help to guide clinicians’ discussions about child COVID-19 vaccinations with parents. In a clinical context, it may not be feasible for clinicians to highlight that trusted parents are vaccinating their children, but when it is feasible, clinicians may be more successful at encouraging vaccination if they highlight messages about trusted parents and, potentially, vaccine safety and the thorough testing process, rather than how well children tolerate vaccine side effects. This is important given the short amount of time clinicians often have with patients and families; it is critical to focus that time on messages that are most effective at promoting vaccination. Given the continuing pandemic and the centrality of effective vaccination among children in controlling future waves of COVID-19 illness at the population level, such messages may be some of the most important communications that pediatricians are currently providing.

AME

average marginal effect

aOR

adjusted odds ratio

CI

confidence interval

COVID-19

coronavirus disease 2019

FPL

federal poverty level

MMR

measles, mumps, and rubella

VOCHIC

Voices of Child Health in Chicago