COVID-19 vaccine acceptance among adults in four major US metropolitan areas and nationwide

This study assesses attitudes towards COVID-19 vaccination and the predictive value of COVID-VAC, a novel scale, among adults in the four largest US metropolitan areas and nationally. A 36-item survey of 6037 Americans was conducted in mid-April 2021. The study reports factors for COVID-19 vaccine acceptance among: (1) already vaccinated; (2) unvaccinated but willing to accept a vaccine; and (3) unvaccinated and unwilling to vaccinate. More than 20% were unwilling to vaccinate, expressing concerns about vaccine efficacy and safety and questioning the disease’s severity. Poverty, working outside of the home and conservative political views are predictors of unwillingness. Conversely, those who either personally tested positive for COVID-19, or had a family member who did so, were more likely to accept vaccination. Majorities of all respondents supported vaccination mandates for employees and university students. Respondents preferred to receive vaccines in their doctor´s office. Lower income and conservative ideology, but not race, were strongly associated with vaccine unwillingness. The predictive value of COVID-VAC was demonstrated. While vaccination mandates are likely to be accepted, additional effective, targeted interventions to increase vaccine uptake are needed urgently.


National and metropolitan area vaccination rates and intent to vaccinate. Vaccination rates
were significantly (P < 0.001) higher in three metropolitan areas (Chicago, NY, LA) than either for Dallas or the national sample ( Table 1). The top reasons to vaccinate were protection from COVID-19 for oneself (national 42.3%, metropolitan areas 38.5-50%; Table 2), or family and friends (national 30%, metropolitan areas 24.4-33.5%) and to "help end the pandemic" (national 23.4%, metropolitan areas 19.7-25.5%). More than a third of the respondents who had already received the vaccine said they encountered difficulties in making an appointment. The top reasons given were lack of proximity of available sites, or not knowing how to sign up for an appointment. Approximately half of the respondents not yet vaccinated still plan to do so. Of this group, a plurality indicated a preference for receiving the vaccine in their doctor's office compared to other sites. The majority of respondents who have been vaccinated or plan to do so felt confident the vaccine would protect against all variants of COVID-19 (Supplemental Table S6).
More than one-fifth (21.4%) of the national respondents indicated an unwillingness to vaccinate (Table 1). With the exception of the Dallas metropolitan area (19.7%), unwillingness to vaccinate was significantly lower in NY (10.1%), LA (11.5%), and Chicago (11.2%) than the national average (P < 0.001). The main reason for unwillingness was "waiting to see if there are no serious complications" ( Table 2). When asked what would increase their readiness to accept vaccination, almost half of those who expressed unwillingness said that "nothing would change their minds" ( Table 2).
Unwillingness to vaccinate was significantly lower among respondents age 60 and older compared to younger respondents (P < 0.0001; Table 3) in the national sample, but was comparable among women and men, different racial groups and educational levels. Except in Dallas and Chicago, where findings were mixed, unwillingness to vaccinate was significantly greater among lower-income individuals (P < 0.001). Unwillingness was also uniformly greater among individuals with moderate (25.5%) to conservative (26.7%) political views, compared to political liberals (10.9%), a trend that held across all metropolitan areas and the national sample (P < 0.0001). Unwillingness also correlated significantly with whether the respondent worked from home, worked outside the home or was unemployed (P = 0.007), and whether the individual or a household member had not received a positive test for COVID-19 in the past, compared to having tested positive (P = 0.001). Compared to similar respondents from the other three metropolitan areas, both White and Black residents of the Dallas metropolitan area were significantly less willing to vaccinate, as were those with less than a college degree (P < 0.001).
The statement that "COVID-19 is a dangerous health threat" (Table 4) generated the highest degree of agreement (nationally 84%, metropolitan areas 82.3-90.4%). However, more than a quarter (25.8%) of the national respondents did not believe that the dangers of COVID-19 exceed those of the vaccine, a level of skepticism found less frequently in the metropolitan areas. One in five (21.7%) national sample participants also disbelieved that COVID-19 can be prevented by vaccination (metropolitan areas 15.6-19.3%); more so among participants who reported holding conservative views (Supplemental Table S3). However, around three-fourths of responents did trust in the science behind the vaccines (72.9% nationally, metropolitan areas 72.7-85.2%, Table 4).
A slightly lower but still solid majority believed that COVID-19 vaccines would be distributed fairly "to everyone and everywhere" (69.6% nationally, metropolitan areas 70.2-76.2%, Table 4). Greater doubts about fairness were found among those with lower incomes (35.7%), worked outside the home (38.3%), or did not state their political views (45.2%). However, these factors showed no consistent associations across the metropolitan areas (Supplemental Table S3).  Table S4). Findings were even stronger with respect to universities requiring students to be vaccinated, but less so with respect to government-required vaccination. Respondents overall expressed the highest level of approval for requiring proof of vaccination for international travel. However, approval for this policy was substantially lower among respondents who were unwilling to vaccinate or who reported conservative political views (Supplemental Figure S1). www.nature.com/scientificreports/ www.nature.com/scientificreports/ Preferred vaccination locations among those not vaccinated. Amongst currently unvaccinated respondents, the doctor's office was the preferred choice for receiving the vaccine. The local pharmacy was a distant second ( Table 2).

Sources of vaccine information.
There was a high level of agreement that the CDC is the leading source of information about COVID-19 vaccination (Supplemental Table S5), followed at some distance by local health providers. The respondents' preferred media for obtaining COVID-19 information were cable TV and local news. This finding was consistent across racial and ethnic groups (Data not shown in table).
Top priorities and rationales for vaccination. Among respondents who were vaccinated or planned to do so, the top priority was "getting everyone vaccinated as soon as possible" (Supplemental Table S5), followed by "getting children back to school" and "getting people back to work". Among the minority of respondents who said they were unwilling to vaccinate, the top priority for them to even consider doing so was "getting people back to work".

Discussion
This study on COVID-19 vaccine acceptance in the United States found sharp contrasts in vaccination willingness among unvaccinated respondents in the national sample compared to the four metropolitan areas studied. Unwillingness to vaccinate was not uniformly distributed by age or racial group in any of the five populations surveyed. Those expressing COVID-19 vaccine hesitancy were unlikely to change their mind, most likely to have less education and lower income, and to work outside of the home. Respondents who reported that they or another household member had tested positive for COVID-19 were strongly motivated to accept vaccination. Recent studies report levels of COVID-19 vaccine hesitancy in the United States ranging from 22% to 42.4% 14,28,31,32 , similar to the range we report. One study reported even lower vaccination intent in the New York and Chicago regions for the Department of Health and Human Services 13 . Our finding that half of those Table 4. COVID-VAC score and score components. % are weighted to the geographic population; COVID -VAC as a predictor of unwillingness to vaccinate was evaluated using weighted logistic regression (outcome = unwillingness to vaccinate). Original response options ranged from strongly agree = 1 to strongly disagree = 5 for the presented item. OR odds ratios, CI confidence intervals. www.nature.com/scientificreports/ respondents who expressed unwillingness to vaccinate said that "nothing would change their minds, " should be a serious concern. Reassurance from family members or recommendations from doctors were reported "unlikely" to change this opinion, a level of recalcitrance that must be addressed in future policies and interventions.
It is important to consider key factors that differentiate those who are vaccine hesitant compared to those who are adamantly opposed to vaccination. The literature on COVID-19 vaccine hesitancy identifies lower education, income, and age; low perceived threat of infection; racial differences (Black or Latino/a vs White or Asian); political affiliation; living in a rural or generally remote area; and low trust in government, health authorities and scientific sources as key associations 14,28,31,32 . COVID-19 vaccine-specific concerns, assessed in COVID-VAC below, suggest a higher perceived risk of complications from the vaccines 13,32 . This risk perception is further amplified by the relatively lower conviction of dangers of COVID-19 disease among the sub-cohort of vaccine resisters.
Poverty was most frequently associated with unwillingness to vaccinate, with the exception of the Chicago metropolitan area. Hesitancy was also high among those who work outside the home, some of whom continued to perform essential but low-wage work even after lockdown 33,34 . In the Dallas metropolitan area, unemployment was also associated with unwillingness to vaccination. Limited economic resources are associated with conflicting life priorities that prevent individuals from prioritizing vaccination. Workers outside the home who have not yet been infected may have developed a perception of low risk from COVID-19 14,17,35,36 .
Many sources cite lower average vaccine acceptance among Blacks compared to other racial and ethnic groups 22,[37][38][39] . Recent polls, including this study, indicate a relative increase in COVID-19 vaccine acceptance among Black respondents, however 40 . Our results found no consistent racial patterns to vaccine hesitancy. In the NY and LA metropolitan areas, white respondents were more likely than Black respondents to express COVID-19 vaccine unwillingness. The reverse was true in the Chicago and Dallas metropolitan areas. These findings caution against demographic generalizations regarding vaccine acceptance, and emphasize the importance of assessing local context when tailoring interventions and messaging. Stereotyping racial attitudes could even lead to disenfranchisement of communities of color that are increasingly well-motivated to participate in vaccination programs 37,41,42 .
The primary motivators to get vaccinated were "protecting myself against COVID-19", followed by "protecting my family and friends"; these elements should remain at the core of vaccine communications messages. Testing positive for COVID-19 was a particularly strong motivator of vaccine acceptance. Continuing to encourage testing will counter misconceptions that the spread of the pandemic is over. As people continue to test positive across age, race, and ideology, this may encourage them to accept vaccination as a necessary protection.
The four metropolitan areas were chosen to cover all four US geographical regions with a significant combined proportion of the US population, their combined numbers of COVID-19 cases as a percentage of the US total, and their disproportionate influence on the national economy, cultural and informational trends 43,44 . Their high population density, greater reliance on public transport and status as major hubs of vacation and business travel also put them at particular risk for pandemic spread. Residents of these metropolitan areas have higher educational attainment and median income than national averages, but share equal or higher levels of poverty and unemployment; their racial and ethnic mix tends to favor higher representation of people of color (Supplemental Table S1) 37,45,46 .
Greater acceptance of vaccination may be higher in the New York, Chicago and Los Angeles metropolitan areas due to the more liberal political leanings of their residents. The politicization of the COVID-19 vaccine and other response efforts is well documented in the United States 47,48 . Conservative political affiliation is strongly associated with COVID-19 vaccine hesitancy 18,49 as well as for other prevention measures such as facemask use, physical distancing, and stay-at-home orders 50 . When evaluating a decision to vaccinate, US conservatives may undervalue the risk to physical health posed by COVID-19 in relation to other perceived risks 7,51 , such as violation of individual autonomy or misconceptions of risks posed by vaccination 52,53 . In our study, holding conservative political views was associated with higher hesitancy across all metropolitan areas, whereas in the national sample moderate and conservative respondents were almost equally unwilling to accept a vaccine. Conservative respondents were more likely to believe falsely that vaccination would not prevent COVID-19 (with the exception of the LA metropolitan area).
Effective motivation of those holding conservative views should link vaccine coverage with returning to work, the top priority expressed by respondents with conservative views. Support from vaccine advocates and media channels with conservative views may mitigate hesitancy among this audiences 54 .
COVID-VAC showed external and internal validity, and predicted (from six times as often in LA to ten times as often nationally) the likelihood of vaccine acceptance. Agreement that COVID-19 is a dangerous health threat was high nationally and in the four metropolitan areas, even among those unwilling to vaccinate; in contrast, trust in the government's ability to distribute vaccines fairly was uniformly lower for all respondents. This scale can be used to predict vaccination behaviors and as a comparative tool.
General and COVID-19-specific mistrust in the government and concerns about vaccine safety and efficacy are commonly cited as drivers of vaccine hesitance 6,38,55,56 . Trust in government regulation of vaccine development has significant influence on public confidence in vaccine safety and efficacy 57 . However, although trust in public health authorities is often cited as a source of confidence, it may not always play a significant role in the decisions of some Black people to accept a COVID-19 vaccine 58 .
Respondents from all study settings were more likely to support required proof of vaccination for access to international travel than domestic indoor events. Respondents were also more likely to accept employer vaccination mandates than government ones. As expected, those holding conservative poltical views were least likely to accept employer mandates and even less so a government one 59 . Though each state has the authority, none currently requires adult immunization against any disease 60  www.nature.com/scientificreports/ Although states require routine childhood immunization 62 , a pattern of increased vaccine hesitancy has been observed over the past decade 63 . Colleges and universities typically require state-mandated childhood vaccines and often require additional vaccines, such as hepatitis A and meningitis B 64 . A relatively high number of respondents in this study agreed that university students should be required to take the COVID-19 vaccine, a finding that has not yet been explored in COVID-19 literature.
In a CDC survey conducted in May 2020, public support for stay-at-home orders and closure of non-essential businesses was higher in NY (86.7%) and Los Angeles, CA (81.5%) than the national average (79.5%), with similar proportions reporting that state restrictions were not too restrictive 65 . Our results indicate a similar high level of support for vaccination requirements in these metropolitan areas, suggesting that risk communication strategies emphasizing individual empowerment and civic responsibility may be effective 66,67 .
All respondents tended to choose the CDC as their preferred source of vaccine information. Vaccine hesitant respondents were also likely to select other options, such as community leaders and local and state governments, suggesting that deploying community and local political leaders could be an appropriate strategy to reach them 68,69 .
In our study, the preferred media for respondents to obtain vaccine information were the internet and social media, followed by cable and local news. Both preferences can be double-edged swords. The internet and social media are sources of misinformation and disinformation, most commonly promulgated by political conservatives, including anti-vaccine activists 70,71 . Cable and online news organizations often supplant sensational crises or political content 72 . For example, viewers of one cable news provider were persuaded to not comply with social distancing recommendations 73 . Those who are hesitant may be more receptive, not only to online claims of misinformation, but to sensationalised reports on adverse effects and vaccine risks. Trusted sources should take proactive roles in delivering balanced, accurate information in internet, social, cable news and local TV media.
Most respondents preferred to receive COVID-19 vaccination in their doctors' offices, fewer at local pharmacies or hospitals. Very few chose sports arenas, places of worship or mobile units. Opportunities for vaccination may be missed with routine health visits when the doctor's office is ill-equipped to vaccinate 74 . As the first wave of enthusiasts is vaccinated, there will be a substantial decline in demand for vaccinations outside the routine healthcare setting 75 . Moving forward, ensuring proper storage and administration capability at primary healthcare facilities may be the way to accommodate public preferences 76,77 .
This study provides an in-depth understanding of perceptions underlying COVID-19 vaccine acceptance in the US, and describes motivational factors that could be used to sway those who are unsure or opposed to vaccination. Lower income and conservative ideology were strongly associated with vaccine hesitancy, as well as mistrust in vaccine efficacy and safety and underestimation of the risk of COVID-19. Non-immunization among these populations may delay or prevent reaching the level of herd immunity required to end this pandemic. Effective, targeted interventions are needed urgently. The COVID-VAC tool could be applied to specific populations, for example vaccine hesitant and vulnerable groups, to inform and further refine policies and strategies domestically and globally.

Methods
Sampling frame. A sampling frame of 242,127,140 US residents 18 years and older was employed. More than 80% of the US population live in urban areas 78 and in 2019 the four largest metropolitan statistical areas had an estimated total population of approximately 50.5 million, representing approximately 15% of the total US population 79 . Five stratified proportionate randomized population based samples were drawn, one for each of the four most populous metropolitan statistical areas in the US: Northeast (New York-Newark-Jersey City, New York-New Jersey-Pennsylvania), Midwest (Chicago-Naperville-Elgin, Illinois-Indiana-Wisconsin), West (Los Angeles-Long Beach-Anaheim, California) and South (Dallas-Fort Worth-Arlington, Texas), and a fifth nationwide sample covering urban, semi-urban and rural geographies.

Survey instrument design.
A 36-item instrument (Supplemental File S1) was created by an expert panel following a comprehensive literature review of COVID-19 vaccine acceptance studies 28 . The instrument included a 6-item COVID-VAC scale representing perceptions of risk, trust, safety, and efficacy. The domains were adapted from vaccine acceptance predictors identified in earlier general vaccine confidence studies 6,80-89 .
Multi-mode data collection design. Personalized short message service (SMS) text messages sent to each respondent. included a link to the survey and an opt-out option. Respondents could reply to the text with any queries and live operators were available to follow up with participants, also via SMS.
Interactive voice response (IVR) aka robo-poll messages were sent to landlines using a voice recorded IVR platform. An online opt-in panel of respondents was provided by Consensus Strategies and utilized to offset deficiency in the sample design with younger participants.The platforms were available in English and Spanish. Data were collected April 10-14, 2021.
Data analysis. Demographic weights, developed based on the 2019 American Community Survey singleyear estimates 90 , were applied: self-identified gender (male, female, prefer not to say and "other" categories), age (18-29, 30-39, 40-49, 50-59, 60-69, 70 +); race and ethnicity (African American/Black, Asian, Caucasian/ White, American Indian, Native Hawaiian, Hispanic, and multiple/other); education (High School or less, some college, college degree, post graduate); census regions were used for the four metropolitan statistical areas and the Northeast, South, Midwest and West census regions were used for the national study.
The study reports COVID-19 vaccine acceptance among the following self-reporting sub-groups: (1) already vaccinated; (2) unvaccinated but willing to accept a vaccine; and (3) unvaccinated and unwilling to vaccinate. www.nature.com/scientificreports/ Trends were analysed using weighted descriptive statistics and chi-squared tests in the four metropolitan areas compared to the national averages. Associations between unwillingness to vaccinate and sociodemographic and political factors were assessed using weighted logistic regressions. The paper also examined the predictive value of COVID-VAC, a novel scale that averages the reversed scores for survey items 1-6, based on responses ranging from strongly agree = 1 to strongly disagree = 5. Dimensionality of the scale was assessed using a principal component analysis (PCA) and reliability using Cronbach's alpha. Weighted logistic regressions were used to evaluate its predictive capacity of perceptions most associated with vaccine acceptance. Statistical analyses were conducted using SAS version 9.4. Statistical significance was set at ∝ = 0.05. Ethics statement. This study was approved by the Emerson College Institutional Review Board (protocol number 20-023-F-E6/12), original expiration date of 11 June 2021. A renewal request was accepted on 12 April 2021 to extend the expiration date to 11 June 2022. The online questionnaire was administered after obtaining respondents' informed consent on the landing page, which contained information about the survey and the purpose of this project. To comply with ethical compensation standards, equitable compensation per survey was applied (approximately US$ 2 per completion). Respondents' identities were verified using IP addresses or mobile phones to ensure that each participant was real and unique upon initial registration, though this data, nor any other personally-identifiable data, were stored, as recommended by the British Psychological Society Ethics Guidelines for Internet-mediated Research and in adherence to the principles of the Declaration of Helsinki. www.nature.com/scientificreports/