Variables Associated with Coronavirus Disease 2019 Vaccine Hesitancy Amongst Patients with Neurological Disorders

Introduction: Given that the success of vaccines against coronavirus disease 2019 (COVID-19) relies on herd immunity, identifying patients at risk for vaccine hesitancy is imperative—particularly for those at high risk for severe COVID-19 (i.e., minorities and patients with neurological disorders). Methods: Among patients from a large neuroscience institute in Hawaii, vaccine hesitancy was investigated in relation to over 30 sociodemographic variables and medical comorbidities, via a telephone quality improvement survey conducted between 23 January 2021 and 13 February 2021. Results: Vaccine willingness (n = 363) was 81.3%. Univariate analysis identified that the odds of vaccine acceptance reduced for patients who do not regard COVID-19 as a severe illness, are of younger age, have a lower Charlson Comorbidity Index, use illicit drugs, or carry Medicaid insurance. Multivariable logistic regression identified the best predictors of vaccine hesitancy to be: social media use to obtain COVID-19 information, concerns regarding vaccine safety, self-perception of a preexisting medical condition contraindicated with vaccination, not having received the annual influenza vaccine, having some high school education only, being a current smoker, and not having a prior cerebrovascular accident. Unique amongst males, a conservative political view strongly predicted vaccine hesitancy. Specifically for Asians, a higher body mass index, while for Native Hawaiians and other Pacific Islanders (NHPI), a positive depression screen, both reduced the odds of vaccine acceptance. Conclusion: Upon identifying the variables associated with vaccine hesitancy amongst patients with neurological disorders, our clinic is now able to efficiently provide ancillary COVID-19 education to sub-populations at risk for vaccine hesitancy. While our results may be limited to the sub-population of patients with neurological disorders, the findings nonetheless provide valuable insight to understanding vaccine hesitancy.


Introduction
While the United States (US) Federal Drug Administration (FDA) has approved several vaccines to address coronavirus disease 2019 (COVID- 19), only an estimated 58-69% of US adults plan to get vaccinated [1]. Given that a vaccine's success relies on extensive uptake within the community, there is impetus to conduct public outreach and vaccine education for patients at risk for vaccine hesitancy [2][3][4]. To efficiently address hesitancy, a comprehensive understanding of populations at risk across major sociodemographic and disease strata should first be developed.
Given Hawaii's unique status as a minority-majority state, with the US's largest share of multiracial citizens, the population serves as an ideal backdrop for identifying the drivers of vaccine hesitancy amongst historically underserved patients (i.e., Asians, Native Hawaiians and Other Pacific Islanders (NHPI), etc.) [5]. Moreover, regarding disease subsets, with neurological disorders being the leading cause of years of life lost and years lived with disability, as well as being associated with high risk for severe COVID-19, there should be heightened efforts to protect such a vulnerable subgroup [6][7][8]. Hence, to judiciously expend clinic resources in providing vaccine education and outreach, a quality improvement (QI) survey was conducted at a large Hawaii multidisciplinary neuroscience institution, with the goal of identifying the patient subsets at risk of vaccination hesitancy.

Methods
For this QI study, a telephone survey of Hawaii Pacific Neuroscience (HPN) adult (18 years and older) patients was conducted between 23 January 2021 and 13 February 2021 to identify populations at risk for COVID-19 vaccine hesitancy or declination-patient subsets requiring greater HPN clinic resources for vaccine counseling. Deemed a QI survey, institutional review board exemption was attained from the University of Hawai'i at Mānoa, Office of Research Compliance. At survey onset, participants provided verbal informed consent after the disclosure of survey objectives, risks, and benefits, as well as assured anonymity; all data were deidentified. No incentive for participation or survey completion was provided. The survey followed reporting guidelines of the American Association for Public Opinion Research (https://www.aapor.org/Publications-Media/AAPOR-Journals/ Standard-Definitions.aspx, accessed on 22 January 2021).

Survey Instrument
The survey was developed after consultation with a cross-functional work group of patients, clinicians, and ancillary healthcare providers. Survey questions emphasized sociodemographic and medical data readily attainable by HPN staff, from electronic medical records or via routine in-clinic pre-appointment questionnaires (i.e., surrogate variables which may readily identify high-risk patients for vaccine hesitancy/declination, therefore requiring time-investment by HPN for auxiliary COVID-19 vaccine counselling). The ten-minute survey explored variables potentially predictive of vaccine hesitancy, based on prior research or emerging speculation amongst the consulted work group [9][10][11][12].
Participants responded to a structured and scripted survey of 13 questions, including: whether the patient had been counselled on COVID-19 vaccination by a physician; the primary source of COIVD-19 information; perceptions of vaccine safety and severity of COVID-19 illness; whether the patient believes herself/himself to have a medical condition making COIVD-19 vaccination unsafe; history of annual influenza vaccination; history of testing positive for COVID-19; self-identified race/ethnicity; work status; highest level of education; marital status; and political views (Appendix A). Cases characterized as complete interviews required a 100% response rate to the crucial question (Do you plan on getting the COVID-19 vaccine?) and 80% for all other questions; partial interviews differed only in that 50-79% of other questions required responses; break-off was defined as either nonresponse to the crucial question or less than 50% response to all other questions [13]. Only data from complete and partial interviews were included for statistical analysis. Participants were provided with the opportunity to terminate the survey at any time and decline to answer any question. Primary caregivers were permitted to assist in participant interviews when appropriate.

Study Population and Data Collection
Participants represented a random sample of the patients who had visited HPN at least once between 1 January 2019 and 1 January 2021. With four campuses (Honolulu, Kailua, Waikele, and Kona), the entire state of Hawaii serves as the patient catchment area for HPN (one of state's largest multidisciplinary neurosciences clinical care and research

Statistical Analysis
Primary analysis utilized nonparametric testing, as assumptions of normality were not met by quantile-quantile plots and histograms. Continuous variables were assessed by the independent Wilcoxon rank sum test, while categorical variables by either the Pearson's chi-squared test or the Fisher's exact test of independence, with Haldane-Anscombe correction [31][32][33][34][35]. Nonparametric continuous variables were presented as the median and interquartile range (IQR, 25th percentile and 75th percentile). Categorical data were expressed as the odds ratio with the 95th percentile confidence interval; for a particular variable's strata, each odds of the odds ratio represented the odds of accepting vaccination compared to declining it. Univariate and multivariable logistic regression, with Firth's correction, was performed to identify variables independently associated with vaccine acceptance [36]. After regression diagnostics, variables for the multivariable analysis were chosen by stepwise selection using the Akaike Information Criterion (AIC), with the final model selected by the McFadden's pseudo-R 2 and the lowest AIC [37][38][39][40]. All tests were two-tailed and used an alpha level of 0.05 for deeming statistical significance. Analyses were conducted through R Statistical Software (R Foundation for Statistical Computing, Vienna, Austria) [41].

General Sample Characteristics
From the 1494 randomly telephoned patients, 915 were non-contacts and 363 respondents (357 complete responses, two partial, and four break-offs; Figure 1). Including partial surveys, there was a response rate of 0.24, a cooperation rate of 0.62, a refusal rate of 0.147, and a contact rate of 0.388 [13]. Demographic breakdown of participants (complete and partial surveys) and non-participants can be found in Table S1.

Survey Responses
Participants whose primary source of COVID-19 information was from traditional media had a greater odds of vaccine acceptance (1.82, IQR: 1.02, 3.28; p = 0.042), contrary to those whose primary source was social media (0. 26 (Tables 8 and 9).

Patients with Neurological Disorders: Entire Cohort
To judiciously allocate clinic resources for COVID-19 vaccine counseling, our neuroscience center sought to first identify patient populations exhibiting vaccine hesitancy. From the 359 patients with neurological disorders surveyed, 81.3% accepted vaccination in our cohort. Fifteen variables were found to be associated with vaccine hesitancy: age, insurance type, income quartile, dyslipidemia, illicit drug use, the presence of a musculoskeletal disorder, CCI, employment status, education level, political views, annual influenza vaccination status, source of COVID-19 information, perception of COVID-19's illness severity, concerns about vaccine safety, and apprehension regarding a preexisting medical condition adversely interacting with the vaccine.

Race
Although the general cohort analyses did not reveal trends regarding race, subgroup analysis did. Male NHPI patients were at reduced odds of vaccination, while themselves, NHPI patients with a positive depression screen or history of a solid tumor were at reduced odds for vaccination. Given the inherent health disparities secondary to structural inequalities, enhanced outreach efforts should be extended to NHPI patients to ensure equitable opportunities for vaccination, particularly amongst those who are PHQ-2 positive, with a tumor history, or male [42,43].

Age
While patients with vaccine hesitancy were overall significantly younger, upon stratification, the trend was only observed amongst females and Whites. Hesitancy amongst younger females may reflect concerns regarding the COVID-19 vaccine adversely interacting with pregnancy, given the population's lack of inclusion in COVID-19 vaccine clinical trials-in spite of recommendations that the vaccination is not withheld from pregnant patients [44][45][46][47][48][49]. Therefore, amongst young patients with neurological disorders, those who are female or White should be targeted for vaccine counseling if appropriate.

Insurance Type
Likewise, Medicaid patients exhibited reduced odds of COVID-19 vaccination, paralleling observed trends for other vaccines, where patients on public insurance have reduced vaccination rates [50,51]. Medicaid patients represent a financially disadvantaged population, who experience reduced healthcare utilization secondary to not affording copayments, hence the lower COVID-19 vaccination odds may arise from financial concerns [52,53]. Upon demographic stratification, only amongst male Medicaid beneficiaries was vaccine acceptance reduced. Consequently, emphasizing the absence of cost for COVID-19 vaccination amongst the Medicaid population-with particular focus on males-could increase vaccine acceptance amongst the community.
Further subgroup analyses demonstrated trends amongst females and Asian patients. For females, Medicare patients were found to have significantly greater odds of vaccination, while among Asian patients, those with military insurance had reduced odds of vaccination, corresponding with reports of greater vaccine hesitancy amongst military personnel, arising secondary to distrust of the vaccine development process and concerns regarding vaccine safety [54]. Therefore, to increase vaccine acceptance amongst military members of Asian heritage, more resources should be expended to educate about the safety of the COVID-19 vaccine and authenticity of the FDA approval process.

Income Quartile, Work Status, and Education Level
In contrast, patients in the third income quartile-the historical middle class-exhibited the greatest odds of vaccination [55]. The third quartile likely represents patients not only with greater COVID-19 exposure risk (i.e., work in healthcare or in contact with the general public), but also greater health literacy and reduced barriers to vaccination [56,57]. Accordingly, neurological patients not able to work (males particularly), thus having greater likelihood of isolation from the general public, or with only a high school education (specifically Whites and both sexes), were at reduced odds of vaccination, while those with a graduate degree exhibited greater odds of vaccine acceptance (particularly females and Whites). Therefore, limited resources would likely be best expended on counseling patients not able to work or with only a high school education.

Information Source: Traditional Media and Social Media
As education level and health literacy correlate with ability to discern misinformation, patients acquiring knowledge from sources prone to false information may be less inclined to vaccinate [57][58][59]. Indeed, patients-males in particular-utilizing social media as a primary source of COVID-19 information had reduced odds of vaccination, contrary to those relying on traditional media. Given the pervasiveness of misinformation on social media and that social media use is highly predictive for believing vaccines are unsafe, clinicians should seek to address a patient's false misconceptions or direct patients towards reputable information sources [57][58][59][60].

Concerns of Vaccine Safety and Adverse Interaction with Preexisting Medical Conditions
Notwithstanding the information source, concerns regarding vaccine safety independently yielded a significantly reduced odds of vaccine acceptance; vaccine safety was the only variable to be statistically significant amongst all demographic strata (females, males, NHPIs, Asians, and Whites). Likewise, patients with the self-perception of a medical condition making vaccination unsafe (most evident amongst both sexes, Asians, and Whites) were at significantly reduced odds of vaccine acceptance, despite the CDC Advisory Committee on Immunization Practices (ACIP) authorizing COVID-19 vaccination for those with underlying medical conditions without contraindications (i.e., immediate allergic reaction to any vaccine components or severe allergic reaction to first dose) [61]. Hence, public health campaigns and physician counseling sessions should focus on alleviating vaccine safety concerns, as well as any individual patient concerns on vaccine interaction with suspected preexisting medical conditions.

Medical Comorbidities
While there were reduced odds of vaccine acceptance amongst those with the selfperception of a preexisting medical condition making vaccines unsafe, patients with more clinically diagnosed comorbidities (per CCI) were at greater odds for vaccine acceptancein particular, amongst females and Whites. Independently, patients (Whites specifically) with dyslipidemia or musculoskeletal disorders were at greater odds for vaccination; while among Whites alone, hypertension increased odds of vaccine acceptance. Patients with dyslipidemia and hypertension potentially represent a cohort of patients who are already engaged in preventative practices (i.e., diet modification, statins, anti-hypertensives, etc.), and thus have an appreciation for the benefits that preventative healthcare can provide; these patients are therefore more willing to vaccinate against a preventable illness. Similarly, patients with musculoskeletal disorders suffer from physically debilitating illnesses which often impactfully respond to medications or lifestyle modifications, and therefore greater vaccination acceptance may represent these patients' first-hand positive experiences with healthcare interventions.
Overall, healthier patients (lower CCI) are potentially demonstrating vaccine complacency, where the risk of vaccine-preventable disease is perceived as low and vaccines are therefore viewed as unnecessary [2]. Within the HPN cohort, the role of vaccine complacency was directly demonstrated, in that patients who did not believe COVID-19 to be a severe illness exhibited reduced odds of vaccination-in subgroup analyses, such was most evident amongst Whites, as well as both sexes. Therefore, vaccine education efforts should address vaccine complacency, particularly amongst healthier patients.
In contrast to the trend of patients with greater illness severity seeking vaccination, amongst the male subgroup, history of a solid tumor reduced odds of vaccine acceptance (likewise observed amongst NHPI patients). In studies of hematological cancers, vaccine hesitancy was most attributed to concerns that the vaccines were not appropriately tested among cancer patients, notwithstanding expert oncologist opinions advocating vaccination and the CDC omitting cancer as a contraindication [61,62]. Therefore, greater outreach should be undertaken by oncologists to advocate vaccination if appropriate, by allaying misconceptions amongst cancer patients [62].

Illicit Drug Use
Contrary to the other medical comorbidities, patients with illicit drug use (Whites specifically) demonstrated reduced odds of vaccine acceptance. Similar trends have been observed for the influenza vaccine and cancer screening, where patients with substance abuse disorder are less likely to attain preventative healthcare [63][64][65][66]. As a marginalized population, patients with illicit drug use are often detached from and mistrust the healthcare system-by extension, these patients may be more reliant on illegitimate information sources [67,68]. Therefore, when counseling patients with illicit drug use, emphasis should be placed on building trust and providing accurate COVID-19 information [67].

Influenza Vaccination Status
One variable which can be efficiently extracted from electronic medical records to identify patients requiring COVID-19 vaccine counseling is annual influenza vaccine status. Patients who did not receive the influenza vaccine in the past year were at reduced odds for COVID-19 vaccine acceptance (the trend evident amongst both sexes, Asians, and Whites). Hence, predictors of influenza vaccine hesitancy may be similar to those of COVID-19, including vaccine complacency or concerns regarding side effects [2,69].

Political Views
Regarding political views, neurological patients identifying as liberal were at greater odds for vaccine acceptance-amongst subgroup analyses, the trend was notable only among females. Such results demonstrate that, even in Hawaii, one of the more liberal states, COVID-19 vaccine acceptance remains highly politicized-as with the rest of the nation [70]. Therefore, patient education should seek to utilize neural apolitical sources for vaccine endorsement [70].

Strongest Predictors of Vaccine Acceptance
After conducting the univariate analysis, multivariable logistic regression was utilized to identify the strongest predictors of vaccine acceptance. For the overall cohort, seven variables were recognized: primary information source (social media use), concerns regarding vaccine safety, belief of COVID-19 to be a severe illness, self-perception of having a pre-existing medical condition making vaccination unsafe, education level (some high school), smoking status (current smoker), and history of a cerebrovascular accident. Current smokers and patients without a high school diploma were identified as having reduced odds of vaccination, and thus such populations require targeted intervention to mitigate potential health disparities from a lack of vaccination [71]. In contrast, patients who had experienced a cerebrovascular accident had greater odds of vaccine acceptance-these patients may be more inclined to engage with preventable health measures, secondary to having personally suffered a potentially avoidable life-altering illness [72].
Upon subgroup analysis, several unique trends were identified. After multivariable regression, only amongst females, Whites, and Asians, did concerns relating to vaccine safety, as well as self-perception of having a preexisting medical condition making vaccination unsafe, result in reduced odds of vaccine acceptance-therefore, if focused demographicoriented interventions are applied, education on vaccine safety and side effects may be most impactful for females, Whites, and Asians. For both sexes, not having received the annual influenza vaccine significantly reduced odds of COVID-19 vaccine acceptance, yet only amongst males did the perception of COVID-19 as a non-severe illness or those identifying as politically conservative significantly reduce odds. Meanwhile, there were several variables resulting in reduced odds of vaccine acceptance, which were exclusive to certain racial groups: for Whites, having only a high school degree; for Asians, a greater BMI; and for NHPIs, a positive depression screen. Therefore, to conduct public outreach or patient counseling efficiently, there may be utility in focusing on specific variables depending on the sociodemographic group of interest.

Limitations
While the findings of this investigation may be extrapolated to other subpopulations, the results should be considered in the context of several limitations, including subset sample sizes. Inherently, generalizability could be restricted, as our population represents patients with neurological disorders from a single institution, as well as from a minoritymajority state with unique sociocultural dynamics and differentially impacted by COVID-19. In Hawaii specifically, around the time of the survey (1 February 2021), the state had reported a cumulative 25,943 cases, 410 deaths, and 5.1% of the population had been vaccinated [73]. On 1 February 2021, there were 91 new cases, zero deaths within seven days, and a 2.1% positivity rate [73]. Given the lower rates of new cases at the time of surveying (relative to 21 August 2021: 671 daily cases, 8.3% test positivity, and nine new deaths within seven days) and the progression of COVID-19, our results may underestimate the current percentage of patients seeking vaccination [73,74]. Likewise, the variables representing disease hesitancy at the time of surveying may also have shifted and be dependent on daily case numbers and deaths [73,74].
Moreover, as participation was restricted to patients with a phone, the results may have been influenced by selection bias, in that more vulnerable subgroups (i.e., financially disadvantaged) could have been excluded. Finally, given the potentially polarizing nature of some survey questions, social desirability bias may have yielded participants providing responses which would be extrinsically viewed positively by others.

Conclusions
This QI survey provided our institution with actionable data, permitting for the efficient utilization of limited clinic resources, in providing vaccine counseling to at-risk patients with neurological disorders. In particular, patients with the following characteristics were recognized for being at risk of vaccine hesitancy (Tables 10 and 11): not having received an annual influenza vaccine, a younger age, a higher CCI, illicit drug use, Medicaid insurance, and social media use for COVID-19 information. Meanwhile, uniquely reducing odds were observed amongst Whites, Asians, and NHPIs, for the following respective variables: high school degree, military insurance, and a positive depression screen. Moreover, amongst all subgroups, vaccine hesitancy appears to be associated with concerns that vaccines are not safe and the self-perception of a preexisting medical condition making the vaccine unsafe (expect among NHPIs). Therefore, focused counseling on allaying patient fears of comorbidity contraindications or vaccine safety may be most impactful. In summary, the investigation not only identified variables that increase the odds of vaccine hesitancy, but also recognized that amongst different demographic strata, there are unique variables at play.