Vaccine willingness: Findings from the COVID-19 effects on the mental and physical health of Asian Americans & Pacific Islanders survey study (COMPASS)

Willingness to get the COVID-19 vaccine is crucial to reduce the current strain on healthcare systems and increase herd immunity, but only 71% of the U.S. public said they would get the vaccine. It remains unclear whether Asian Americans and Pacific Islanders (AAPI), a population with existing inequalities in COVID-19 infection and mortality, are willing to get the vaccine, and the factors associated with vaccine willingness. Given this imperative, we used data from a national, cross-sectional, community-based survey called COVID-19 Effects on the Mental and Physical Health of AAPI Survey Study (COMPASS), an ongoing survey study that is available in English and Asian languages (i.e., Simplified or Traditional Chinese, Korean, Vietnamese) to examine vaccine willingness among AAPI. A total of 1,646 U.S. adult AAPI participants completed the survey. Self-reported vaccine willingness showed the proportion who were “unsure” or “probably/definitely no” to getting the COVID-19 vaccine was 25.4%. The odds for vaccine willingness were significantly lower for were Native Hawaiians and Pacific Islanders (vs. Asian Americans), Korean Americans (vs. Chinese and Vietnamese Americans), women (vs. men), heterosexuals (vs. non-heterosexuals), those aged 30–39 and 50–59 (vs. aged < 30), and those who reported having any vaccine concerns (vs. no concerns). AAPIs’ willingness to get COVID-19 vaccine varied by groups, which underscores the need for disaggregated AAPI data. A multi-pronged approach in culturally appropriate and tailored health communication and education with AAPI is critical to achieve the goal of health equity for AAPI as it pertains to COVID-19 mortality and morbidity.


Introduction
As the number of Americans infected and dying from coronavirus disease 2019  continues to increase (33,470,212 and601,808 respectively, as of June 29, 2021 (Centers for Disease Control andPrevention, 2021), it is vital that the U.S. population receive the COVID-19 vaccines to effectively mitigate this pandemic. In December 2020, the Food and Drug Administration (FDA) authorized Pfizer and 10,000) from COVID-19 as of July 2020. (Rubin-Miller et al., 2021) Based on the National Center for Health Statistics' data on attributable death from COVID-19 from February-October 2020, AAPI had higher attributable COVID-19 deaths than Whites among age groups 45-84; NHPI also had higher attributable mortality among those aged 15-24. (Chu et al., 2021)In San Francisco, Asian Americans had a four times higher case fatality rate vs. the overall population (5.2% vs. 1.3%) a few months into the pandemic. (Yan et al., 2020)The sparse disaggregated data for AAPI revealed that some groups had disproportionate COVID-19-related burdens. For example, based on data from 85,328 patients who tested for COVID-19 at New York City's public hospital system, South Asians had the highest rates of positivity and hospitalization among all Asians, which was only second to Hispanics for positivity and Blacks for hospitalization; and, Chinese patients had the highest mortality rate of all groups and were about 1.5 times more likely to die from COVID-19 (vs. Whites). (Marcello et al., 2020) Without important data about COVID-19 vaccine willingness, it remains unclear whether health disparities for AAPI, a population with existing inequalities in COVID-19 infection and mortality, (Marcello et al., 2020) may be exacerbated. As such, this study examined the proportion and factors associated with COVID-19 vaccine willingness among AAPI using data from a national ongoing, multilingual survey, COVID-19 Effects on the Mental and Physical Health of AAPI Survey Study (COMPASS).

Study Eligibility, recruitment and procedures
COMPASS is a cross-sectional, community-based national survey that assesses the COVID-19 effects on AAPI. To be eligible, participants must self-identify as AAPI; be able to read English/Chinese/Korean/ Vietnamese; be ≥ 18 years old; and reside in the U.S. The survey is available online (https://compass.ucsf.edu/), phone, and limited inperson. This paper reports on 1,646 participants who completed the survey from October 24-December 11, 2020, which was selected as the cutoff date for this analysis since it was the first day that the FDA authorized a COVID-19 vaccine. (Food and Administration, 2021)Mean survey completion time was 21.6 (standard deviation (SD) = 15.5) minutes. There were 100 incomplete surveys. Single persons (vs. married) were significantly less likely to be incomplete responders. There were no significant differences by other background/social characteristics or by vaccine willingness. Each participant was provided an option of receiving a $10 gift card upon survey completion.
Convenience sampling was employed in this study. Participants heard about COMPASS through organizations who serve AAPI, personal/professional networks, social media, email/listservs, flyers, and ethnic media. COMPASS also leveraged the Collaborative Approach for AAPI Research and Education (CARE) Registry (Collaborative Approach for Asian Americans and Pacific Islanders Research Education. CARE Registry for Asian Americans and Pacific Islanders., 2021) to recruit participants.
Participants provided e-informed consent (n = 1,535) or verbal phone consent (n = 111). The survey used the Research Electronic Data Capture hosted at the University of California, San Francisco (UCSF). (Harris et al., Apr 2009;Harris et al., 2019)The World Health Organization's process of translation and adaptation of instruments (World Health Organization. Process of translation and adaptation of instruments., 2018) was used to guide the translations of the study materials that were not already available in the targeted language(s).

Dependent variable
Willingness to Get the COVID-19 Vaccine Item was modified from another survey. (Neumann-Böhme et al., 2020)Participants were asked, "If a vaccine becomes available for COVID-19, would you get it?" Likert response options included: definitely yes; probably yes; unsure; probably no; and, definitely no.

Independent variables
Socio-demographic items were drawn from CARE. (Collaborative Approach for Asian Americans and Pacific Islanders Research Education. CARE Registry for Asian Americans and Pacific Islanders., 2021)Variables included race (Asian; NHPI; American Indian/Alaskan Native; Black/African American; White/Caucasian), cultural group (Asian Indian/Chinese/Huaren/Filipino/Korean/Japanese/Native Hawaiian/ Samoan/Taiwanese/Vietnamese/Other/Mixed), sex (Female/Male; Other/Prefer not to answer), sexual orientation (Heterosexual/Not heterosexual/Prefer not to answer), year of birth, nativity (country of birth), years lived in the U.S., marital status (Single; Married; Living with a partner; Separated/Divorced; Widowed), employment (Fulltime/Part-time/Homemaker/Unemployed/Retired/Other), education (option ranged from grade to Graduate school), and household income (options ranged from ≤$25,000 to ≥$200,0001). Participants were asked how well they could speak/read/write English (very well/well/ some/a little bit/not at all).
Participants completed several existing surveys related to COVID-19 including the Coronavirus Impact Scale (CIS) related to changes in family income/employment (School, 2020) and COVID-19 status (Cawthon et al., 2020) (yes, no, unsure diagnosis). COVID-19 Vaccine Concerns (Neumann-Böhme et al., 2020) included the following response options (check all that apply): 1) I do not have any concerns; 2) I'm concerned about potential side effects; 3) I think COVID-19 vaccine may not be safe; 4) I do not think that COVID-19 is dangerous to my health; 5) I am against vaccination in general; 6) The best way is for nature to take its course; 7) I believe natural or traditional remedies; 8) I'm afraid of injections; 9) Religious reasons; and, 10) Other.
Participants answered duration of Shelter-in-Place (SIP) questions based on region, per the Census Bureau's definition of region (Midwest/ Northeast/South/West), which was obtained by converting the zip code, or internet protocol address in the case of missing zip codes (n = 188). SIP and Perceived Severity of COVID-19 items were developed by COMPASS.
General Health was measured by asking participants to indicate their health "today" on a scale from 0 (worst) to 100 (the best health you can imagine) using the EQ-5D (EuroQol, Dec 1990; Rabin and de Charro, Jul 2001) item, which was categorized into quintiles.

Statistical analysis
Chi-squared tests were used to examine the association between vaccine willingness, modeled as a 5-level categorical variable (definitely yes/probably yes/unsure/probably no/definitely no) and hypothesized factors associated with vaccine willingness, specifically race, cultural group, sex, sexual orientation, age, nativity, marital status, employment, education, household income, vaccine concerns, length of shelter in place, perceived severity of COVID-19, effect of coronavirus on family income, region, and general health status. Participants who responded that they could speak/read/write English less than very well ("some", "a little bit", or "not at all") were categorized as having limited English proficiency (LEP). (Census Bureau, 2021)COVID-19 concerns were first classified as none, side-effects only, unsafe only, and multiple reasons for descriptive purposes. Then, it was further categorized as a binary variable (none or any concerns) in subsequent regression analyses to examine the association between vaccine willingness and any type of vaccine concerns.
This study used multinomial logistic regression to model the association between vaccine willingness, categorized as "definitely yes" (reference group), "probably yes", "unsure/probably no/definitely no", and these same factors. The "definitely yes" and "probably yes" responses differed conceptually. "Definitely" implied commitment whereas "probably" implied ambivalence and unwillingness to commit, which have differential associations with behavioral outcomes. (Baca-Motes et al., 2013;Rice et al., 2017)Consistent with previous research on vaccine willingness (Brandt et al., 2021;Ernesto, 2020)and guided by the data revealing differences in the two response groups in their proportion of participants with no vaccine concerns (44.1% vs, 9.7%, p < 0.01), the two responses were treated as separate outcome categories. The 'unsure,' 'probably no,' and 'definitely no' were combined into a single outcome category due to sample size considerations and the 3 groups appear to be similar in their associations with other factors. This study first conducted bivariate multinomial logistic regression to identify candidate variables for the final model. Since it was important not to miss variables that may become significant in the presence of others, this study retained variables that that attained a p-value of < 0.10 in bivariate analyses in the final modelLin et al., 2020-12-17 202014 (Zoran Bursac et al., 2008;Ranganathan et al., 2017). All statistical tests were two-sided. Hosmer-Lemeshow goodness of fit test (Hosmer and Lemesbow, 1980) indicated an acceptable fit of the final model (p = 0.27). The variables included in the final model were race, cultural group, sex, sexual orientation, age, nativity, marital status, employment status, income, vaccine concerns, and COVID severity. Statistical analyses were conducted using SAS Software. (SAS Institute Inc, 2014)

Human subjects protection
This study was approved by UCSF's Institutional Review Board (protocol 20-31925).

Sample Characteristics
The mean age of participants (N = 1,646) was 40.6 years (SD: 15.8) and ranged from 18 to 88. Participants included 62.6% female, 90.0% heterosexuals, and 97.6% Asian Americans. Major cultural groups included ethnic Chinese (including persons from Hong Kong and Taiwan; 37.1%), Vietnamese (29.0%), and Korean (20.5%). Overall, 61.5% of participants were foreign-born who had lived in the U.S. an average of 47.8% of their lives (SD: 24.0%), 20% spoke limited English, and 73.3%completed the survey in English (Table 1)

COVID-19 analyses
As shown in Table 1, 44.0% of participants said, "definitely yes", 30.6% "probably yes", 17.9% "unsure", 5.0% "probably no", and 2.5% "definitely no" to getting the COVID-19 vaccine. About 24% reported no vaccine concerns. One in three (34.1%) were only concerned about side effects, 6.0% were only concerned about vaccine safety, and 36.4% reported multiple concerns. >64% reported that the COVID-19 had impacted their family income and employment. Over half (52.1%) indicated their SIP was ≥ 3 months; 45.0% perceived the COVID-19 severity of where they lived was "somewhat" to "a lot" more than other areas. Few (2.0%) said they had tested positive for COVID-19, 90.2% had tested negative, and 7.9% who were unsure or did not provide a response.

Bivariate analyses
In bivariate analyses, race, cultural group, sex, sexual orientation, age, nativity, marital status, income, vaccine concerns, and COVID-19 severity were significantly associated with vaccine willingness, but LEP, employment status, education, length of SIP order, effect on family income/employment, region, and self-reported health status, and caregiver status were not (Table 2).
Sex and sexual orientation remained significantly associated with vaccine willingness in the adjusted model, with males (vs. females) and non-heterosexuals (vs. heterosexuals) less likely to be "unsure/probably no/definitely no" versus "definitely yes".

Discussion
This paper reports willingness to receive the vaccine and factors related to willingness among 1,646 AAPI from a national multi-lingual survey before the safety and efficacy data were widely known, thus future research may investigate the potential changes in attitudes and factors associated with vaccination willingness. Results highlight the heterogeneity in vaccine willingness for AAPI. The study sought to provide these data to inform public health officials, community partners, and health care systems of potential outreach targets to improve vaccine willingness and thereby provide a defense against COVID-19 that has disproportionately affected AAPI. (Rubin-Miller et al., 2021;Chu et al., 2021) From a health equity lens, the study provides disaggregated data for AAPI, which is significant because such data are scarce. These findings are among the first to reveal that AAPI groups are heterogeneous in COVID-19 vaccine willingness. This study found that NHPI were significantly less willing to receive the vaccine (vs. Asian Americans). This finding is critical as several U.S. states with large numbers of NH-PI report higher rates of COVID-19 infections (vs. other racial/ethnic groups). (Kaholokula et al., 2020)Low vaccine willingness and high infection rates highlight the need for concentrated efforts to address COVID-19 health inequities for NHPI. The study also found that Korean Americans were less willing to receive the COVID-19 vaccine (vs. both Chinese and Vietnamese Americans), consistent with a global online survey that found that 25% of South Koreans (vs. 20% of respondents from China) indicated little/low vaccine intention. (Ipsos. Global attitudes on a COVID-19 vaccine: Ipsos survey for The World Economic Forum., 2021)The study found that Vietnamese Americans were more willing to receive the COVID-19 vaccine (vs. Chinese Americans), and such differences could not be explained by vaccine concerns or other correlates examined in this study. Results suggest targeted efforts for outreach to NHPI and Korean Americans, who were less willing to get the COVID-19 vaccine. Outreach in these languages and in areas where NHPI and Korean Americans are located may help to increase vaccine uptake. Public health campaigns with spokespersons from these cultural backgrounds may also help. Providers serving NHPI, Korean Americans, women and age groups that were less willing to receive the vaccine, could educate their patients about the importance of COVID-19 vaccination.
Overall, 25.4% of participants reported that they were unsure/unlikely to receive the vaccine, which closely mirrors the proportion (27%) reported in a national poll. (Kaiser Family Foundation, 2020)This study also found that those aged 30-39 (vs. < 30) were significantly less likely to be willing to get the vaccine, which is consistent with prior studies that found that being younger in age (Fisher et al., 2020) (e.g., 30-49) (Kaiser Family Foundation, 2020)was associated with vaccine hesitancy. However, the anomaly study finding was that those 50-59 (vs. aged < 30) were also less likely to be willing to get the vaccine; future research may investigate this association. Similar to prior findings, (Fisher et al., 2020;Callaghan et al., 2020;Kaiser Family Foundation, 2020;Reiter et al., 2020)those who reported having any vaccine concerns (vs. none) were less willing to receive the vaccine.
Consistent with another survey, this study found that men are more willing to get the vaccine than women. (Callaghan et al., 2020)This study also found that non-heterosexual AAPI were less likely to state that they were unwilling to get the vaccine (vs. heterosexual). Lesbian, gay, bisexual, transgender, queer, and intersex (LGBTQI + ) individuals experience health inequities that have become heightened with the pandemic. Reasons for greater vaccine willingness may be related to historical experiences of discrimination and health inequity such as lack of health insurance, (Kline, 2020)which may motivate individuals in sexual minority groups to access vaccination to reduce their risk for COVID-related morbidity/mortality. Greater vaccine willingness may also be related to the increased risk of infection that LGBTQI + individuals perceive with lower self-reported health status, higher rates of substance use, higher rates of HIV/AIDS (and compromised immune systems), and worse mental health (vs. general population). (Kline, 2020;Cancer, 2020;Pedrosa et al., 2020;Salerno et al., Aug 2020) Furthermore, in a poll of 174 adults who identified as LGBT, 75% viewed vaccination as a responsibility to protect others (vs. 48% of non-LGBT individuals). (Kaiser Family Foundation. The Impact of the COVID-19 Pandemic on LGBT People., 2021) It is possible that participants who indicated that they were "unsure/ probably no/definitely no" about getting the vaccine were waiting to receive additional information that could alleviate their concerns; this group can be considered the "wait and see" group. (National Academies of Sciences, 2020)A potential model that may be used as a guide to increase COVID-19 vaccine willingness is the World Health Organization BeSD Increasing Vaccination Model, (National Academies of Sciences, 2020)which stipulates "what people think and feel" (e.g., perceived risk) and the "social processes" (e.g., provider recommendations; social norms) lead up to the core of the model, "motivation" (e.g., willingness;  hesitancy). By addressing "motivation" and barriers (e.g., cost; availability), individuals may be more likely to get the COVID-19 vaccine.
(January, 2021)Culturally appropriate/tailored strategies are likely to play pivotal roles in vaccine uptake. For example, faith-based organizations may be important community partners for COVID-19 vaccine uptake, as they have been for other health promotion activities such hypertension prevention. (Kwon et al., 2017)Removing access barriers will be essential for the vaccine rollout. By providing multilingual helpline and resources and being geographically relevant (i.e., taking services to neighborhoods with higher density of Asian Americans), AAPI languages to provide information and support with booking vaccination appointments. (Center, 2021)Additionally, recommendations from research for other types of vaccine uptake (e.g., HPV vaccine among Asian Americans) may be useful to apply in COVID-19 vaccine intention/uptake for AAPI (e.g., culturally appropriate educational information; family involvement). (Vu et al., 2020)

Limitations
There are a few limitations. First, this study is not an etiological study; however, future work may include repeated measures to assess change over time. Second, due to feasibility, not all AAPI languages were available, thus, limiting the ability to generalize findings to all AAPI. Thirdthe online survey limits the potential reach to AAPI with limited access to digital technology; however, COMPASS does offer toll-free numbers in the target languages. Fourth, the sample size for certain AAPI groups (e.g., NHPI; Asian Indians) were small; COMPASS' present targeted outreach with NHPI and Asian Indian community partners may help to improve this. A strength of COMPASS is that community partners were engaged to help in the outreach/recruitment, and administration, which research has shown to be critical to successful AAPI. (Rabinowitz and Gallagher-Thompson, 2010;Grill and Galvin, 2014)

Conclusions
Strong willingness to get the COVID-19 vaccine is vital for personal, public, economic, and social health. (National Academies of Sciences, 2020) More than 25% of the participants indicated an unwillingness to get the COVID-19 vaccine; however, unwillingness significantly differed across AAPI groups, which underscore the need for data disaggregation, as an understanding of vaccine unwillingness (among a variety of other important outcomes) may be obscured by lumping AAPI together. A multi-pronged approach in culturally appropriate and tailored health communication and education is critical to achieve the goal of health equity for AAPI especially given the high burden of COVID-19-related morbidity and mortality among AAPI. (Marcello et al., 2020)Additionally, COMPASS was a multi-lingual survey available in multiple mediums (online, phone, and limited in-person) that allowed for a diverse sample of AAPI. Future research is encouraged to include non-English languages as well to help diversify their samples.

Role of the Funder/Sponsor
The funding source has no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; or decision to submit the manuscript for publication.

Data Sharing Statement
We will follow the NIH data sharing policy, https://grants.nih.go v/grants/policy/data_sharing/.

Funding
This study was supported by a COVID-19 administrative supplement grant from the National Institutes of Health/National Institute on Aging (NIH/NIA) (3R24AG063718-02S1). Part of < named co-author > effort was supported by NIH/NIA R01AG067541.

Declaration of Competing Interest
The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.