In this prospective survey-based study, the public was solicited to fill out an internet survey on the REDCap system using a public internet link disseminated through the Western Michigan University Homer Stryker MD School of Medicine Facebook account with a post introducing the research project. The survey was open from August 1, 2021, to August 30, 2021.

Consent was requested at the beginning of the survey after details on the nature and objectives of the study were made available to the participants. Data were collected from all survey results received through the REDCap system. Participants were under no obligation to complete or submit the survey.

The primary outcome was whether respondents chose to take the vaccine. Respondents were allocated into a group of participants that indicated they chose to take the vaccine if, on the survey, they selected any of the following: (1) they received the COVID-19 vaccine, (2) were scheduled to take the vaccine, (3) had taken the first dose and are waiting for the second dose, or (4) were fully vaccinated. Respondents were allocated into a group of participants that indicated they have not chosen to take the vaccine if, on the survey, they selected that they may take the vaccine or will not take the vaccine. Since the survey was at a single point in time, a portion of undecided participants may have later chosen to be vaccinated.

A media source use score was calculated using data from US adults from a survey conducted from October 29, 2019, to November 11, 2019, by the Pew Research Center, which solicited respondents on which media outlets they used and whether they identified as Democrat or Republican [24]. Our survey asked participants, “What news organizations do you use to get COVID information?” and listed the same media outlets from this study, which is freely available on the Pew Research Center website [24]. The question was phrased in this manner to appreciate that people may consume a media company’s information through a variety of sources, whether that is television, social media, radio, mobile apps, emails, newspapers, and more. Each media outlet was assigned a political leaning score based on the Pew Research Center’s data as a difference between the percentage of the audience who identified as Liberal or Democrat and the percentage identifying as Conservative or Republican. For example, if a respondent indicates using CNN (20), New York Times (37), and the Rush Limbaugh Show (80), the respondent will have a media source use score of (20+37+80)/3=−7.67. We note that the creation of this score has not been validated as a measure of the ideological profile, and some of the news sources indicated in the survey (Al Jazeera, Buzzfeed, CDC, etc) do not have an ideological profile measure and were scored as 0.

Additional covariates included political affiliation; presidential candidate choice; frequency of interaction with digital information; and demographic data (gender, education, employment status, and poverty screening). The survey also included a series of statements on a 7-point Likert-type agreement scale that assessed conception of COVID-19, vaccination, public health behavior, and the influence of their personal physician and those in their personal sphere of influence like friends and family (Table 1).

Categorical data are reported as frequencies (percentages) and continuous data are reported as means (SDs). The prevalence of COVID-19 vaccination use was estimated using the total number of respondents indicating they have taken the vaccine out of the total number of survey respondents. A 95% CI was used to measure the precision of the estimate of the prevalence of COVID-19 vaccination.

To evaluate the association between media use source and vaccination, a logistic regression model was used—the media use source score was the independent variable, and vaccination choice was the dependent variable. The odds ratio was reported as a measure of the association between the media use source score and vaccination choice. The odds ratio indicates the multiplicative increase in the odds of choosing the vaccine as the media use score increases by 1 point toward a more Liberal or Democrat stance. The association between other demographic characteristics and vaccination choice was measured using univariate logistic regression. Significant univariate predictors were included in an adjusted model, and odds ratios and P values were reported. Significance was declared at α=.05. SAS v9.4 was used for data analysis. A full logistic regression model was built using variables found to be significant in the unadjusted models.

We also considered a subtractive regression model and found that the results were identical to the original outcome using backward stepwise selection.

This study was reviewed and approved by the Institutional Review Board (IRB) of the Western Michigan University Homer Stryker MD School of Medicine, IRB number WMed-2021-0754, before data collection. The survey began with a statement on informed consent including that the study data are anonymous with no solicitation of identifying information. No compensation was provided for participation in or completion of the survey.

In this study, we examined factors associated with individuals’ decisions to receive the COVID-19 vaccine, with a specific focus on media sources and political leaning. We also explored the influence of the patient-physician relationship and close personal networks.

The relationship between demographic factors and vaccine uptake was not surprising. Regarding whether employment status was affected by COVID-19, no statistically significant difference was noted between those who chose to get the vaccine and those who did not. Yet, part-time employees and the unemployed were at greater odds of choosing the COVID-19 vaccine compared to full-time employees. Evidence on the relationship between employment and vaccine hesitancy is conflicting, with some showing unemployment being an impetus for vaccination as a prerequisite for job acceptance during the pandemic [17]. This contradicts research where the unemployed have reported lower influenza and COVID-19 vaccine acceptance compared to the employed, which could be confounded by education and income [25]. The US Bureau of Labor Statistics shows that as the years of education increase, income increases while unemployment rates decrease [26]. This is consistent with our survey results revealing statistically significant (P<.001) differences between those with a bachelor’s degree or higher compared to those without a bachelor’s degree. Social disadvantages, such as lower levels of income and education lead to poor access to accurate information, which has been cited as a driver for low confidence in the COVID-19 vaccine [27].

Our results on gender differences showed no significant difference in vaccination rates between male and female participants, differing from studies where women have reported a lower acceptance of the vaccine than men, especially for young women and pregnant or breastfeeding women, which was a distinguishing factor we did not explore [28]. Of note, most of our participants were at or approaching middle age and recipients of college education, which are 2 factors that predict greater rates of vaccine acceptance. This may have resulted in no significant gender difference.

Our results showing older respondents to be more likely to choose the vaccine compared to younger respondents is consistent with studies demonstrating that younger age groups tend to have a lower risk perception of COVID-19 [27].

Our results showed that those who consumed more Liberal or Democratic media and those who indicated a more Liberal or Democratic stance on the political spectrum were more likely to receive the vaccine. This is consistent with previous research and Centers for Disease Control and Prevention reports have shown that vaccination rates are higher in Democratic counties compared to Republican counties [29].

Unsurprisingly, our respondents heavily interacted with digital information, with a majority indicating its use multiple times a day (1267/1755, 72.19%), if not multiple times each hour (201/1755, 11.45%). As a measure of the influence that COVID-19 information had on our respondents, we asked them if they would consult a doctor upon the emergence of various symptoms including new loss of taste or smell, with 69.78% (1226/1757) indicating that they would. New change or loss in sense of smell is a hallmark symptom of the COVID-19 infection [30]. After a popular New York Times article linking loss of smell to the onset of a COVID-19 infection was published early in the pandemic, Tweets and Google searches regarding anosmia peaked for a week after the article was published at a frequency about 7 times higher than normal. The frequency of Tweets and Google searches related to anosmia was maintained at over 2.5 SDs from the typical frequency for weeks after the New York Times article publication [31]. These findings and our data suggest that media-informed knowledge of COVID-19 symptoms influences the public’s perceptions of the disease, which may affect their willingness to seek medical care.

The results of the Likert-type agreement scale showed that respondents who were more likely to take the COVID-19 vaccine had a stronger belief in vaccine safety and efficacy in general and, as such, in the COVID-19 vaccine (Table 1). They were also more hesitant to seek medical care due to the possibility of contracting COVID-19 at a medical facility, which suggests these respondents viewed the vaccine as a prophylactic measure. In addition, respondents were more likely to take the vaccine if it was in line with the beliefs, encouragement, and positive or negative experiences of their family and friends. This is consistent with evidence confirming a strong relationship between perceived vaccination social norms and the intention to get vaccinated, which declines as the referenced social group grows larger and more heterogeneous [32]. Respondents who chose not to get the vaccine were more strongly influenced by personal medical conditions and religious or moral reasons when compared to those who received the vaccine.

Most of the respondents (1531/1757, 87.14%) had consulted their doctor within the past year—a quarter of which were in the past month. All participants moderately agreed that they had a good relationship with their doctor (5.51 and 5.23, respectively, on the 7-point Likert-type agreement scale). However, no statistically significant difference was observed in vaccine choice between these 2 groups. Notably, only 48.36% (828/1712) of respondents indicated that their doctor encouraged them to get the vaccine and 32.49% (555/1708) talked about the possible harms and benefits. Regardless of whether the respondents spoke with their doctor on the issue, the influence physicians have on their patients’ decisions is challenged by the influence of social media, which the public may rely more heavily on for their sources of information, especially during public health crises when risk perception is elevated [33]. However, these 2 sources do not have to be mutually exclusive as physicians may also communicate with the public through media. However, a study published in 2021 found that less than 10% of tweets surrounding the COVID-19 vaccine came from medical professionals [34].

The respondent demographic may not have necessarily reflected the state or national vaccination status. Additionally, due to constraints set by our institution’s social media outreach policy, the survey was only posted on the institution’s Facebook page. During the survey period of August 2021, our survey group included 89.58% (1574/1757) of respondents who received at least one vaccine dose. This differed from the national vaccine status at the time, which was 61.8% (205,026,070/3,317,573,395) for the first dose and 52.4% (174,121,529/332,292,994) for series completion [35]. The vaccine status for Michigan was 55.2% (5,516,637/9,993,908) for the first dose and 50.5% (5,043,602/9,987,331) for series completion during this period [35].

The sample size was predominantly White race. This is important as racial and ethnic minorities in the United States are shown to have been more likely to report uncertainty or unwillingness to undergo vaccination [36]. Also, this study did not further distinguish between child-bearing-age and pregnant women as studies have shown pregnant or breastfeeding women to have greater vaccine hesitancy [37]. In the Likert-scale question on trusting in local health experts, both doctors and the hospital were grouped together to guard against survey fatigue; however, a split between the 2 categories may have better informed the patient-provider relationship.

Additionally, the political inclination of our respondents leaned more liberal, which was reinforced by more than two-thirds (1212/1737, 69.78%) of respondents indicating they voted for the Democratic candidate in the 2020 presidential election. Therefore, our sample size does not necessarily reflect the political preference of the country where 51.31% (81,268,924/158,383,403) voted for the Democratic candidate [38]. Meanwhile, the media score (c=0.77) was slightly less predictive than the political leaning scale (c=0.81). When both the media score and the political leaning scale are included in the model together, they remain significant, indicating that both measures account for variation in vaccine choice independent of one another.

In our measure of the frequency of the number of times respondents interact with digital information, it is important to note that this data may be difficult to generalize as some may do so for an extended period resulting in more interaction than those interacting multiple times an hour for brief moments. Additionally, some respondents may spend a greater proportion of time interacting with one media source versus another. Each participant’s media source use score is the average of their media sources’ political leaning score. We acknowledge that the media source use score is most accurate when participants interact with their media sources for an equal amount of time and is less accurate when participants favor some media sources more than others.

We recognize that by administering the survey through the medical school’s social media, we may be getting a unique group of respondents consisting of health-conscious individuals. However, we also know that this survey was open to the public during a time when social media engendered interaction with users who held both positive and negative perspectives about the vaccine [39].

Despite these limitations, conducting a survey such as this to gauge the social and media environment of the population served by the medical school fosters ongoing communication and outreach that aligns with the population’s needs. The insights gained from this study can be used to shape and improve how health care professionals and organizations approach patient care and prevention initiatives in the community. Conducting this study in local communities can yield insightful perspectives from the population, allowing physicians to tailor their communication approach accordingly.

The COVID-19 pandemic has grown into a politically charged public health issue fueled by a media infodemic that affected various health behaviors, such as the decision to get vaccinated. When we examine our model that includes the variables of media score, political leaning, and the interaction between the two, we observe a significant interaction suggesting that media use impacts the effect political position has on vaccine choice.

Additionally, the results from our survey show that participants are influenced by their family and friends. In contrast, despite most respondents having a good relationship with their personal physician, there was no significant difference in vaccination decision likely indicating that physicians had less influence on their choice.

Future research could investigate specific messages and persuasive techniques that media outlets use to influence their audiences regarding health decisions, including algorithms on social media platforms. It could also help illuminate our findings as to why physicians are not as influential as the media, political leaning, and a person’s family and friends during a global pandemic. Additionally, further work could improve upon our ideological profile model by creating a more validated measure of a person’s political preference based on their choice of media companies. Future research can investigate how physicians might be better trained in social media competency and how physicians with a strong and positive public media presence operate. Physicians have the opportunity to interact with their patients not only in their practice but also through the media, which can improve their reach in countering false information. We urge other health institutions to undertake similar efforts to assess the information sources used by their local communities and adjust their communication approach accordingly.

The decision to receive the COVID-19 vaccine is a complex and multifactorial process that is influenced by various situational and socioeconomic factors. This is further complicated by the widespread availability of information about the vaccine, both accurate and inaccurate, which can make it difficult for individuals to make informed decisions about whether to take the vaccine. The mass media plays a critical role in disseminating this information as it can both inform and misinform the public about vaccine safety and efficacy. Effective communication is critical in ensuring that accurate and reliable information about the COVID-19 vaccine is available to the public to help individuals make informed decisions about whether to get vaccinated.