The role of social circle COVID-19 illness and vaccination experiences in COVID-19 vaccination decisions: a representative online survey of the United States population

Design of the National Survey of COVID-19 Health Experiences

The survey instrument and recruitment protocol of the National Survey of COVID-19 Health Experiences were approved by the Institutional Review Board (IRB) of the author’s institution (file number: STUDY00006960, date of approval: November 17, 2021, name of IRB: Michigan State University Human Research Protection Program). All methods were carried out in accordance with relevant guidelines and regulations. The sample was obtained by Dynata, the world’s largest first-party data platform, and is representative for the US American population (7). The sampling using Dynata is based on opt-in sampling, respondents deliver high quality data, they are diverse and have community norms of honesty and accuracy (8). The survey was opened to the Dynata panel until the required number of responses was obtained from each category of the stratification variables age, sex, and income, as required for a balanced response set. With opt-in sampling there is no response rate as classically defined in survey research.

Development of questionnaire and pre-test

The questionnaire was developed in November 2021. A team that included a medical doctor and survey research specialist helped to validate the survey. The survey design was based on Shupp et al. (7). Of relevance are questions that ask respondents about the health status of people in their social circles. Shupp et al. (7) included a similar question in their survey but in the context of prescription drug abuse. A pre-test was conducted with 1,110 respondents December 6–9, 2021. The questionnaire was finalized using the responses from the pre-test.

The questionnaire is composed of five sets of questions: 1) questions about respondents’ experiences with COVID-19 illness, 2) questions about respondents’ experiences with COVID-19 inoculation, 3) questions about experiences with COVID-19 illness in respondents’ social circles, 4) questions about experiences with COVID-19 vaccination in respondents’ social circles, and 5) questions to obtain standard socioeconomic information, political affiliation, and views on COVID-19 policies, such as lockdowns and vaccine mandates. The questionnaire is provided in Supplementary Material 1.

Statistical analysis of the survey data

Means and standard deviations are provided for continuous variables, and absolute numbers (percentages in parenthesis) for categorical variables. Socioeconomic characteristics of survey participants were compared with those from the United States (US) Census and the US American Housing Survey (9, 10, 11) after adjustment for age and sex.

Logistic regression was used to identify factors associated with the chance of being vaccinated with at least one shot. The two primary independent variables of interest were: 1) knowing someone who suffered from the COVID-19 disease; and 2) knowing someone who has been injured by the COVID-19 vaccine. Adjustments were made for the following confounders: age, sex, political affiliation (Democrat, Republican, Independent), degree of urbanization using respondents’ self-assessment of whether they live in urban, suburban or rural areas, race (Caucasian, African American, Hispanic, Asian, Native American/Pacific Islander, Other), educational attainment as defined by the US Census (12), sources of information about COVID-19 (mainstream news, alternative news/other, peer-reviewed scientific literature, official government sources), COVID-19 illness problems in social circles, and COVID-19 inoculation problems in social circles. Social circles, as defined in the survey, include “family, friends, church, work colleagues, and social networks”. Among those in social circles who experienced health problems, respondents were asked to provide a description of the person they know best.

Comparing serious adverse events between publicly available data and the survey

Several steps are required to compare data on COVID-19 vaccine adverse events from the survey with publicly available government data. In the first step, public data on COVID-19 fatalities from the CDC (13) is combined with COVID-19 vaccine-related adverse events from VAERS (6) to create the ratio of COVID-19 vaccine-related fatalities to fatalities from the COVID-19 illness. The same ratio from the survey data is calculated so that a comparison can be made. To examine differences, the null hypothesis (H₀) is defined such that the True Ratio, X, is equal to the CDC ratio which is in turn equal to the survey ratio: X = CDC Ratio = Survey Ratio. The alternative hypothesis, H_a, is X = CDC Ratio < Survey Ratio. This hypothesis is tested using state-by-state VAERS data on reported COVID-19 vaccine fatalities and CDC data on COVID-19 illness fatalities. If there is a statistically significant difference, the two ratios can be used to estimate nationwide COVID-19 vaccine fatalities under the assumption that the survey is accurate:

Survey Ratio=\(\frac{SurveyCOVID-19VaccineFatalities}{SurveyCOVID-19IllnessFatalities}\) Pop. Ratio=\(\frac{y}{CDCCOVID-19IllnessFatalities}\)

Solving for y generates the estimated number of nationwide vaccine fatalities. Through the end of 2021, reported COVID-19 vaccine fatalities from VAERS (6) for the US states and the District of Columbia was 8,023, and the CDC (13) reported 839,993 fatalities attributed to COVID-19. These data were downloaded on January 16, 2022. The ratio of vaccine-associated fatalities to COVID-19 fatalities is \(\frac{\text{8,023}}{\text{839,993}}=0.0096\), or about 1%. A bootstrap method is used to obtain the 95% confidence interval, which is a non-parametric approach that does not assume an underlying distribution of the data. The procedure is as follows. First, resample the original dataset with replacement to obtain the same number of "pseudo-observations" where some of the original observations are counted multiple times. The new dataset serves as a pseudo-survey sample, which is used to recalculate the point estimate. This process is repeated 1,000 times to compute the 95% confidence interval.

In the second step, the fatality calculation from above is used to estimate the number of non-fatal adverse events. The ratio of estimated population-wide fatalities to reported fatalities in the survey is used to calculate nationwide adverse events, a, as per the two equations below. “Severe” and “less severe” adverse events are calculated separately.

Fatality Ratio=\(\frac{EstimatedPop.COVID-19VaccineFatalities}{SurveyCOVID-19VaccineFatalities}\) Adverse Event Ratio=\(\frac{a}{SurveyAdverseEvents}\)

Upon publication, the survey dataset and corresponding Stata code will be made publicly available in Supplementary Material 5.

Characteristics of survey participants representativeness of the survey

The National Survey of COVID-19 Health Experiences was administered online between December 18 and 23, 2021. A total of 2,840 participants completed the survey after removing the 216 respondents (6.5%) who opted out of the survey by not consenting to participate, 60 missing responses on age which is used to weight the data (1.9%), and 105 incomplete surveys (3.2%). Twenty-seven additional respondents did not answer the question about race; in portions of the evaluation where race is considered, there are 2,813 observations. Item non-response for the following variables is considered negligible: age 1.9% (age), 0.9% (race), and 0.28% (number of people in social circles). The other questions used in this evaluation did not have a single missing item.

The survey instrument is available in Supplementary Material 1. Table 1 provides descriptive statistics for the survey sample with comparison to data from the US Census (11, 14) and the American Housing Survey (15). Forty-nine percent of both the survey participants and the US population were male. Age of participants 46.9 (CI 95% ± 0.640) years. There were also some minor differences in political affiliation, race, degree of urbanization and education. The data on urbanicity are comparable to data from the American Housing Survey (15) with small differences in percent urban (30.8% vs. 27%), percent suburban (46.7% vs. 52%), and percent rural (22.5% vs. 21%). For educational attainment, the survey had a higher percentage with “some college” (35.4% vs. 27.6%) but a lower percentage of “college graduates” (18.9% vs. 22.1%), and a higher percentage with “more than a college degree” (14.2 vs. 12.7).

There are limitations with using a survey to collect COVID-19 health information, particularly for a politicized health issue. Respondents often interpret events with bias due to perceptions based on history, beliefs, culture and family background. For example, a respondent who self identifies as Republican may offer a report that is different than a person who identifies as Democrat. In addition, though a person may report that someone they know experienced a COVID-19 vaccine adverse event, it does not mean that vaccination was the cause of injury. As shown in the Table 4 and Supplementary Material 3, some respondents indicated that a person they know had a heart attack after being vaccinated, though the heart attack could have been unrelated to the inoculation. To address this issue, an estimate of the number of people within respondent social groups who are expected to die regardless of inoculation is calculated and subtracted from reported COVID-19 vaccine fatalities. Three commonly reported vaccine adverse events are heart attacks, strokes and other manifestations of blood clots. The average age of a person in the survey dataset who experienced these conditions after being vaccinated is about 40 years of age, and the average age of death is 48. The incidence of heart attacks (myocardial infarction) for people of age 48 is about 17 per 100,000, and the incidence of strokes and blood clots for this age group is very low, near zero (13). Heart attacks, strokes and blood clots are also commonly reported causes of COVID-19 vaccine death in VAERS. From the survey, about 51% of respondents reported being vaccinated. It is assumed that same proportion applies to those in respondents’ social circles. The estimated total number of people in respondents’ social circles is about 28,000. To calculate an estimated number of fatalities that might have occurred regardless of inoculation status, 17 is multiplied by the proportion of people who are vaccinated (0.51) and the proportion of people in social circles out of 100,000 (0.28). The estimated number of fatalities that might have occurred regardless of vaccination status is 17 x 0.51 x 0.28 = 2.43 people.

Direct respondent experiences regarding the COVID-19 illness or the COVID-19 vaccine are informative but incomplete because potential respondents who are very ill or died due to COVID-19 illness or the COVID-19 vaccine could not participate in the survey. For this study, the most important information comes from the questions about the experiences of those within respondents’ social circles because all these health experiences can be reported by survey respondents.

Descriptive statistics for primary endpoints

Table 2 presents summary statistics for the relevant questions answered of respondents with differences and p-values between those who had the COVID-19 illness and not, and those who were vaccinated and not. The survey questionnaire is provided in Supplementary Material 1. Twenty-three percent of respondents report have had the COVID-19 illness, of which 28% experienced lingering health issues; most indicated they had ongoing respiratory/breathing or taste/smell issues. About 8.6% of those who had health problems experienced more severe health problems resulting from COVID-19. Fifty-one percent of respondents indicated that they had been vaccinated of which 15% indicated that they experienced a health issue after vaccination, and 13% of those indicated that a severe adverse event had occurred. The respondents’ comments describing the nature of the COVID-19 illness and health issues and COVID-19 vaccine adverse events are available from the author upon request. There are statistically significant differences across groups, with notable differences across the vaccinated/unvaccinated groups in income ($70,919 vs. $48,903), knowing someone who experienced a vaccine adverse event (0.157 vs 0.277), as well as with the education, race, information sources, and political affiliation categorical variables.

Factors related to vaccination decision and vaccine injury

The Logit regressions for vaccination and knowing someone who experienced a vaccine adverse event are shown in Table 3, which reports the odds ratios with confidence intervals. All regressions are estimated using the unweighted data due to the inclusion of socio-economic controls used by Dynata to recruit a balanced sample. Starting with socioeconomic factors, age is positively associated with inoculation (OR: 1.025, 95% CI: 0.003–1.019), but negatively associated with knowing someone who has been injured from inoculation (OR: 0.979, 95% CI: 0.973–0.985). Higher income is also positively associated with inoculation (OR: 1.000005, 95% CI: 1.000004-1.000007). Relative to Democrats, those who self-identify as Republican have lower odds of being vaccinated (OR: 0.595, 95% CI: 0.067–0.477) and have greater odds of knowing someone who has experienced an adverse event (OR: 1.388, 95% CI: 1.089–1.769). Those who identify as Independent also have lower odds of being vaccinated (OR: 0.631, 95% CI: 0.066–0.514). There is evidence of an urban-rural divide, where rural residents have lower odds of being vaccinated (OR: 0.744, 95% CI: 0.090–0.587). Race is an important factor in vaccination status. African Americans (OR: 0.655, 95% CI: 0.513–0.835), Hispanics (OR: 0.647, 95% CI: 0.469–0.893), and Asians (OR: 0.599, 95% CI: 0.387–0.927) have lower odds of being vaccinated relative to the White population. African Americans are also more likely to know someone who has experienced a health problem post-vaccination (OR: 1.376, 95% CI: 1.066–1.776). Educational attainment is positively associated with inoculation. Those with doctoral (OR: 3.835, 95% CI: 1.759–8.358) or professional degrees (OR: 3.2821, 95% CI: 1.601–6.729) have higher odds of being inoculated. Those with doctoral (OR: 2.010, 95% CI: 1.143–3.536) or professional degrees (OR: 4.263, 95% CI: 2.009–9.043) also have higher odds of reporting that they know someone who has experienced a health problem after inoculation, respectively. Information sources are also associated with inoculation status. Those who report reliance on mainstream news and official government sources have higher odds of being vaccinated (OR: 1.394, 95% CI: 1.165–1.669). However, use of alternative news sources reduces the odds of inoculation (OR: 0.669, 95% CI: 0.557–0.802). Also, reliance on alternative news (OR: 1.481, 95% CI: 1.217–1.801) and peer-reviewed scientific publications (OR: 1.430, 95% CI: 1.143–1.789) increases the odds that a respondent knows someone who experienced a health problem post-vaccination.

Turning to the primary hypothesis, a respondent’s observations within his/her social circles have a significant influence on the decision to be vaccinated. Those who know someone who experienced a significant health problem from the COVID-19 illness have higher odds of being vaccinated (OR: 1.309, 95% CI: 1.094–1.566). Conversely, those who know someone who had a health problem following inoculation have lower odds of being vaccinated (OR: 0.567, 95% CI:

Comparison of serious adverse events between publicly available data and the survey

Table 4 presents a summary of COVID-19 illness and COVID-19 vaccine health experiences among respondents’ social circles. Thirty-four percent of respondents indicated that they knew at least one person who had experienced significant health problems from COVID-19, including 165 people who had died from COVID-19. Supplementary Material 2 provides a word-cloud of respondent descriptions of COVID-19 illness experiences in social circles along with respondent comments. Twenty-two percent of respondents indicated that they knew at least one person who experienced a health problem after COVID-19 vaccination. Fifty-seven people indicated that among the people they knew who had experienced a vaccine adverse event, the person they knew best had died. Supplementary Material 3 provides respondent descriptions of COVID-19 vaccine health problems in social circles in a word-cloud along with respondent comments. Respondents report a variety of problems including heart attacks and other heart related problems, blood clots and strokes, and neurological problems. Many of the descriptions such as “heart attack,” “stroke,” or “blood clot” are consistent with FDA (16) and Pfizer (17) documentation about the potential risks of the COVID-19 vaccine.

The ratio of COVID-19 vaccine deaths to COVID-19 illness deaths of the people respondents knew best who had health problems is \(\frac{57}{165}=0.345\), whereas the ratio of vaccine-associated fatalities to COVID-19 fatalities from government sources is \(\frac{\text{8,023}}{\text{839,993}}=0.0096\). The null hypothesis (H₀) that the true ratio, X, is equal to the CDC ratio which is also equal to the survey ratio: X = CDC Ratio = Survey Ratio. It is tested using state-by-state VAERS data on reported COVID-19 vaccine-associated deaths and COVID-19 illness fatalities. The alternative hypothesis (H_a) is: X = CDC Ratio < Survey Ratio. The mean (u) and standard deviation (σ) of the ratio of vaccine fatalities to COVID-19 fatalities from the state-by-state data are u = 0.0136 and σ = 0.0111. The probability that the Survey Ratio > CDC Ratio = X [P(CDC Ratio > 0.345)]. With P(CDC Ratio > 0.345) = 0 and a Z-score = 28.86, the null hypothesis is rejected.

Assuming the experiences captured in the survey represent the true ratio, the survey ratio is used to estimate nationwide COVID-19 vaccine fatalities: Estimated fatalities are 289,789 (95% CI: 229,319–344,319). Estimated nationwide deaths combined with other survey data on adverse events are also used to estimate total adverse events. “Severe” adverse events are estimated to be about one million nationwide, and “less severe” adverse events are about 2.1 million. Nationwide fatalities, “severe injuries” and “less severe” injuries tally to 3.4 million.

This evaluation is conducted under the assumption that the reported vaccine-related fatalities and injuries are caused by the COVID-19 vaccine but is now relaxed by reducing the number of reported fatalities by the fatalities due to other causes that would be expected to have occurred anyway. An estimated 2.43 fatalities might have occurred from heart attacks, strokes and blood clots within the survey sample regardless of vaccination status. Subtracting these fatalities from total estimated vaccine fatalities generates a nationwide estimate of 278,000 fatalities, which is 4.1% smaller. Estimated total adverse events are correspondingly reduced by 4.1%. Also, Supplementary Material 4 provides analysis of respondent bias as reflected by political affiliation and vaccination status. Estimated nationwide COVID-19 vaccine fatalities based on the Democrat, Republican and Independent subsets are 109,564, 463,444 and 247,867, respectively. With the vaccinated and unvaccinated subgroups, estimated COVID-19 vaccine fatalities are 110,942 and 659,995.