Vaccine hesitancy was designated as one of the top ten threats to global health by the World Health Organization (WHO) in 2019 (WHO, 2019a). The urgency of understanding and addressing vaccine hesitancy has been amplified by the COVID-19 pandemic which has led to the largest overall decline in United States (US) life expectancy in more than seven decades (Arias et al., 2021). Minority populations have reported greater hesitancy to get the COVID-19 vaccine, with some communities of color being half as likely to be vaccinated as Whites (Khubchandani et al., 2021; Malik et al., 2020; Nguyen et al., 2021; Niño et al., 2021; Willis et al., 2021), but these gaps have narrowed over time (Daly & Robinson, 2021; Hamel et al., 2021). Vaccine attitudes are often perceived as being on a continuum ranging from complete refusal to active demand (Dubé et al., 2013). However, research defining hesitancy as a behavior, or both an attitude and behavior, faces the problem of conflating issues such as access with the attitude of hesitancy (Quinn et al., 2019). If someone reports being unvaccinated, it may or may not actually reflect their level of vaccine hesitancy (Dubé et al., 2013).
Emerging evidence suggests those who express hesitancy, or an intention to not receive the COVID-19 vaccine, may still get vaccinated. For example, one study shows nearly a quarter of individuals interviewed in January 2021 who reported they would definitely not get the COVID-19 vaccine or would only get it if required have now received at least one dose (Kirzinger et al., 2021). Yet, little is known regarding hesitancy among those who choose to be vaccinated. Two qualitative studies conducted prior to the emergence of COVID-19 focused on hesitancy among parents who accepted a vaccine for their children (Enkel et al., 2018; Walker et al., 2020). One study focused on Australian parents making decisions about flu and measles-mumps-rubella vaccinations for their children (Enkel et al., 2018), and the other focused on human papillomavirus vaccination for young adult children (Walker et al., 2020). Both studies demonstrate that hesitancy may co-occur with adherence (Enkel et al., 2018; Walker et al., 2020). However, many questions remain about “hesitant adopters.”
Most research on vaccine hesitancy has overlooked those who have been vaccinated, in part because of inconsistency in the operationalization of vaccine hesitancy (Dubé et al., 2013). A review as recent as 2015 found there was no established definition of vaccine hesitancy (MacDonald, 2015). Consistent with the Increasing Vaccination Model (Brewer et al., 2017; WHO, 2019b; WHO EURO, 2011), we view vaccine hesitancy as an attitude or motivational state (Brewer et al., 2017), whereas vaccination is a behavior which may or may not correspond with reported attitudes towards vaccination. Therefore, the behavior of being vaccinated is distinct but not mutually exclusive from the attitude of vaccine hesitancy (Dubé et al., 2013). Further examination of “hesitant adopters”— individuals who are hesitant but vaccinated—can provide insights into the process of overcoming hesitancy to receive a vaccination, which is critical to vaccine uptake.
Rather than assuming those who receive a COVID-19 vaccine are not hesitant, we asked the following questions: (1) How prevalent is hesitancy towards the COVID-19 vaccine among those who have recently been vaccinated? (2) How does this COVID-19 vaccine hesitancy among the recently vaccinated vary across sociodemographic groups? (3) How does this COVID-19 vaccine hesitancy among the recently vaccinated relate to other factors such as prior diagnosis of COVID-19, health care coverage, health literacy, and flu vaccination behavior over the past five years?
We examined these questions in the context of a state which has COVID-19 vaccination rates substantially lower than the national average. As of August 24th 2021, only 39% of the total population of Arkansas was fully vaccinated, compared to 52% of the US population (HHS & CDC, 2021). States with low vaccination rates have experienced a disproportionate burden of COVID-19 hospitalizations and deaths, especially since the emergence of the Delta variant (Leatherby & Walker, 2021a, b). Thus, understanding COVID-19 vaccine hesitancy in states such as Arkansas is particularly pressing.
Respondents were recruited at more than 30 vaccination sites across the state of Arkansas where the COVID-19 vaccine was given. Vaccination sites include drive-through clinic locations and vaccination events in the community and at faith-based organizations. Respondents had to be 18 years of age or older, be recently vaccinated, and speak or read English, Spanish, or Marshallese to be included in the study.
The consent and survey were administered during the fifteen-minute wait time after the COVID-19 vaccine was given using REDCap (Harris et al., 2009). Participants were seated after receiving their vaccination, provided information about the survey, and invited to participate. The study included an incentive of entry into a raffle for a $100 Walmart gift card. The consent information and survey were available in English, Spanish, and Marshallese. Respondents could document consent and complete the survey on their smartphone by scanning a QR code or utilizing a sterilized iPad provided by study staff. Respondents could also complete the study with assistance of a bilingual staff member asking them the questions verbally. Respondents had the ability to refuse (as indicated by skipping or selecting “prefer not to answer”) questions. Data were collected between April 22nd and July 6th, 2021. The study was approved by the University of Arkansas for Medical Sciences Institutional Review Board (IRB #262645).
The survey included sociodemographic questions, as well as questions to assess COVID-19 vaccine hesitancy, prior COVID-19 diagnosis, health care coverage, health literacy, and past history of flu vaccinations.
COVID-19 vaccine hesitancy
To measure the dependent variable of COVID-19 vaccine hesitancy, we modified an existing single-item measure of vaccine hesitancy to specifically capture attitudes about the COVID-19 vaccine. The original question asked, “Overall, how hesitant are you about getting vaccinations?” (Quinn et al., 2017, 2019). We modified this to ask respondents to think specifically about how hesitant they were to receive the COVID-19 vaccine. Respondents were asked, “Thinking specifically about the COVID-19 vaccine, how hesitant were you about getting vaccinated?” Response options included the following: “not at all hesitant,” “a little hesitant,” “somewhat hesitant,” and “very hesitant.”
Sociodemographic information, including age, sex, race, education, marital status, and whether non-English languages were spoken in the home, was collected. Sociodemographic information other than age was assessed using standard questions from the Behavioral Risk Factor Survey (CDC, 2020). Age was measured using a question from the General Social Survey as a continuous variable calculated from reported date of birth or respondents’ best estimates of their age (Smith et al., 2019). Sex was measured as a categorical variable of either female or male with the option to self-describe—responses from non-binary respondents were too few to include in the analysis. Respondents were asked to identify their race and could select all that apply from a list of valid responses: American Indian or Alaska Native (AIAN), Asian, Black/African American, Native Hawaiian or Pacific Islander (NHPI), White, or Other. These responses were combined with a response to a question about ethnicity to indicate whether each individual was non-Hispanic Black/African American, non-Hispanic NHPI, non-Hispanic White, non-Hispanic Multiracial or other racial group, or Hispanic or Latino. Responses from AIAN, Asian, and multiracial respondents were too few to keep these categories separate from others in analyses, so they were combined. Although combining these groups is problematic, we decided this option was preferable to excluding them from the analysis. To assess educational attainment, respondents selected the closest fit from an ordinal set of responses. Possible responses included “never attended or only kindergarten,” “grades 1 through 8,” “grades 9 through 11,” “grade 12 or GED,” “some college or technical school,” or “college 4 years or more.” There were few respondents who selected options below “less than a high school,” and these responses were combined into a single “less than high school” category. Respondents were presented a question about relationship status. For analysis, responses were coded into married and non-married respondents. Non-married respondents include any respondent who provided a valid response other than “married.” Respondents were asked if they spoke a language other than English at home, with possible responses of “yes” and “no” (American Community Survey, 2019).
Prior COVID-19 diagnosis
To assess whether individuals had been diagnosed with COVID-19 prior to their vaccination, respondents were asked, “Has a doctor or another health care professional diagnosed you with the coronavirus (COVID-19)?” Possible responses included “yes” or “no.”
Health care coverage
To assess health care coverage, we used a standard measure from the Behavioral Risk Factor Surveillance Survey (CDC, 2020). Respondents were asked, “Do you have any kind of health care coverage, including health insurance, prepaid plans such as HMOs, or government plans such as Medicare, or Indian Health Service?” Possible responses included “yes” or “no.”
A single-item screener of health literacy was used. Health literacy screeners have been validated among low-income and Spanish-speaking patients (Chew et al., 2008; Hadden et al., 2019; Ylitalo et al., 2018), and evidence suggests that a single item is effective at detecting limited to marginal health literacy with reasonable specificity (Wallace et al., 2006). This question asked respondents, “How confident are you filling out medical forms by yourself?” Possible responses included “extremely,” “quite a bit,” “somewhat,” “a little bit,” or “not at all.”
Flu vaccine behavior
An existing measure of flu vaccination behavior was used to provide insight into whether individuals typically received a flu shot over the past five years (Quinn et al., 2017). The question asked respondents, “How often in the past 5 years have you gotten a seasonal flu vaccine?” Possible responses included “every year,” “most years but not all,” “once or twice,” and “never.”
We present descriptive statistics for all variables in the study. Spearman correlations and Kruskal Wallis H tests were performed to assess bivariate relationships with COVID-19 vaccine hesitancy. Due to the ordinal nature of the dependent variable, a generalized ordered logistic regression with partial proportional odds (Williams, 2006) was used to analyze multivariable relationships with COVID-19 vaccine hesitancy. We made the decision to use a partial proportional odds model following a Brant test which revealed that age, health care coverage, and health literacy violated the proportional odds assumption in an ordered logistic regression. The generalized ordered logistic regression allowed us to relax the proportional odds assumption for age, health care coverage, and health literacy, while maintaining it for other variables (Williams, 2006). All tests were two-tailed, with a standard significance level of p < 0.05. Statistical analysis was performed on Stata SE/15.1.