COVID-19 related messaging, beliefs, information sources, and mitigation behaviors in Virginia: a cross-sectional survey in the summer of 2020

Background Conflicting messages and misleading information related to the coronavirus (COVID-19) pandemic (SARS-CoV-2) have hindered mitigation efforts. It is important that trust in evidence-based public health information be maintained to effectively continue pandemic mitigation strategies. Officials, researchers, and the public can benefit from exploring how people receive information they believe and trust, and how their beliefs influence their behaviors. Methods To gain insight and inform effective evidence-based public health messaging, we distributed an anonymous online cross-sectional survey from May to July, 2020 to Virginia residents, 18 years of age or older. Participants were surveyed about their perceptions of COVID-19, risk mitigation behaviors, messages and events they felt influenced their beliefs and behaviors, and where they obtained information that they trust. The survey also collected socio-demographic information, including gender, age, race, ethnicity, level of education, income, employment status, occupation, changes in employment due to the pandemic, political affiliation, sexual orientation, and zip code. Analyses included specific focus on the most effective behavioral measures: wearing a face mask and distancing in public. Results Among 3,488 respondents, systematic differences were observed in information sources that people trust, events that impacted beliefs and behaviors, and how behaviors changed by socio-demographics, political identity, and geography within Virginia. Characteristics significantly associated (p < 0.025) with not wearing a mask in public included identifying as non-Hispanic white, male, Republican political identity, younger age, lower income, not trusting national science and health organizations, believing one or more non-evidence-based messages, and residing in Southwest Virginia in logistic regression. Similar, lesser in magnitude correlations, were observed for distancing in public. Conclusions This study describes how information sources considered trustworthy vary across different populations and identities, and how these differentially correspond to beliefs and behaviors. This study can assist decision makers and the public to improve and effectively target public health messaging related to the ongoing COVID-19 pandemic and future public health challenges in Virginia and similar jurisdictions.


INTRODUCTION
The early days of the coronavirus (COVID-19) pandemic (SARS-CoV-2) were characterized by conflicting messaging from nearly all levels of national and international mass media and government (Doran, 2020;Gaviria, Smith & PBS, 2020;Kolstoe, 2020).As public health and healthcare professionals attempted to quell the growing panic with science-driven narratives, conspiracy theories, misinformation, and disinformation continued to spread through social media platforms such as Facebook, Weibo, and Twitter, often undermining or contradicting the life-saving messages that scientists were trying to communicate (Garrett, 2020;Islam et al., 2020).This issue was further compounded by long-standing health, socioeconomic, and racist inequities as well as sharp decreases in funding to state and federal health agencies in the United States (Garrett, 2020).Throughout the pandemic, access to and acceptance of evidence-based messaging to prevent and respond to outbreaks of coronavirus disease (COVID-19) have been inconsistent across populations and subject to politicization (Jones et al., 2020;O'Shea & Ueda, 2021).Black, Hispanic, and Indigenous populations have been historically excluded from the United States' public health and medical institutions, often suffering disproportionately from many diseases and public health challenges (Hutchins et al., 2009;Krishnan, Ogunwole & Cooper, 2020).Given that that ethnic minority, low-income, low-education, and elderly populations are also overrepresented in COVID-19 related morbidity and mortality numbers, public health officials will need to effectively reach out to and target those particular groups (Alobuia et al., 2020;Holtgrave et al., 2020;Maroko, Nash & Pavilonis, 2020).
Several surveys have evaluated the awareness and concern that members of the public have experienced towards COVID-19 and local, state, and national government responses (Jones et al., 2020;Seale et al., 2020).Results showed that the majority of the general population wants to hear from public health and medical officials, and are likely to trust professional sources that have self-protective and pro-social messages that focus on positive ways to protect themselves and their loved ones (Banker & Park, 2020;Shelus et al., 2020).This includes demographic groups that are considered high-risk for COVID-19, like the elderly and low-income individuals from minority groups (Geldsetzer, 2020;Li et al., 2020;McFadden et al., 2020).
As vaccine coverage increases at different rates globally, the public health response to COVID-19 continues to necessitate coordination at all levels of government to ensure accessible and accurate testing, contact tracing, quarantine and isolation, treatment, and mitigation measures like social or physical distancing and mask wearing (Mobula et al., 2020).It is important that trust in public health information be maintained for these strategies to continue to be implemented effectively.Studies found that trust in public health officials and the information they provided allowed for successful messaging campaigns with past disease outbreaks, ranging from food safety incidents to worldwide polio vaccination campaigns (Li et al., 2020).As shown in past responses to foodborne disease outbreaks, demonstrating that public health measures and preventative strategies are in the best interests of the community overall is crucial to building and maintaining public trust that is essential to effective public health guidance (Henderson et al., 2020).High-risk populations may respond best when targeted with official messages that are consistent, credible, proactive, and also a mixture of self-focused and prosocial (Banker & Park, 2020).
Throughout the COVID-19 pandemic in the United States, social distancing, school closures, lockdowns, and targeted public health messaging have been sporadic and inconsistent.Many people obtain information from social media that can conflict with the messages from public health officials (Islam et al., 2020).In response, Facebook, Twitter, and online newspapers are now actively monitoring their own sites for inaccurate COVID-19 information that could mislead people into believing potentially dangerous rumors, stigmas, and conspiracy theories (Islam et al., 2020;Krause et al., 2020).The rapid development and rollout of COVID-19 vaccines have also been subject to false information (misinformation and disinformation) on social media platforms, with peernetworks exchanging large quantities of anti-vaccination posts that focus on adverse side-effects, misleading medical content, and unsubstantiated rumors (Puri et al., 2020).Similar methods have been used to undermine prior vaccine campaigns, and developing effective messaging to counter such false information will likely prove to be an important challenge for public health officials (Dror et al., 2020;MacDonald & Hesitancy, 2015).
Like many other large states, Virginia has had notable regional differences in case trends over time, with the more densely populated northern and central regions experiencing large case increases during the pandemic's initial wave in the spring of 2020, while the coastal eastern region and the more rural southwestern region experiencing their first large case increases mid-summer (Virginia Department of Health, 2021).Some Virginia college towns, such as Charlottesville, Blacksburg, and Harrisonburg, saw increases in local case counts when students returned in the late summer and mid-winter of 2020-21, showing that the movement of large groups of people can greatly affect community spread in less densely populated areas (Sidersky & Sauers, 2021;Smith, Hwang & Binkley, 2020).Given the continued need for effective evidence-based public health messaging, officials, researchers, and the public can benefit from exploring how people receive information they believe and trust, and how their beliefs influence their behaviors.In addition to not previously being of focus for this type of research, Virginia is a geographically and culturally diverse location, which allows for nuanced analysis of factors that influence messaging and behaviors within this state that can be generalized to other similar populations.Evaluating the differences in public health messaging and its effectiveness across difference socioeconomic and demographic groups can inform and improve future targeted messaging efforts.To gain better insight for understanding and developing effective messaging, we summarized and described the results from a cross-sectional survey administered during the summer of 2020 to examine COVID-19 related messaging, beliefs, information sources, and mitigation behaviors among adults in Virginia.We aimed to identify correlations between messaging, behaviors, and characteristics.

MATERIALS AND METHODS
We surveyed Virginia residents by distributing a link to complete the survey online through our professional and personal email listservs, via Facebook (including advertisements targeted to Virginia residents), and on flyers in select locations.Eligibility criteria included being 18 years of age or older and residing in Virginia.Participants provided electronic informed consent prior to beginning the survey questions.The survey collected sociodemographic information, including gender, age, race, ethnicity, level of education, income, employment status, occupation, changes in employment due to the pandemic, political affiliation, sexual orientation, and zip code.Participants were asked about their perceptions of COVID-19, risk mitigation behaviors, messages and events they felt influenced their beliefs and behaviors, and where they obtained information that they trust.The full survey, developed and administered using Qualtrics, is available in the supplement.To limit people from completing the survey more than once, participants were able to save and continue the survey and the Prevent Ballot Box Stuffing setting was selected.Responses were completely anonymous.

Analysis strategy
We conducted exploratory analyses by calculating descriptive statistics of survey responses and investigated correlations between information sources, perceptions, beliefs, and risk mitigating behaviors related to the COVID-19 pandemic.Figures presenting these comparisons are used to visualize these comparisons and differences greater than 5% are reported.We also investigated correlates of the fundamental risk mitigating behaviors mask wearing and social/physical distancing in unadjusted and adjusted analyses using logistic regression with robust variance estimates.Statistical significance was taken at the 0.025 level to account for these two primary outcomes using a conservative Bonferroni adjustment and a nominal type I error rate of 0.05.We adjusted for race, political identity, gender, age group, income, reporting national science and health organizations as an information source, believing in alternative messages, and living in southwest Virginia to identify the independent effects of these characteristics on risk mitigation behaviors.These variables were selected a priori as known correlates of COVID-19 beliefs and incidence (Alobuia et al., 2020;Azlan et al., 2020;Barari et al., 2020;Benham et al., 2021;Bonyan et al., 2020;Christensen et al., 2020;McCaffery et al., 2020;McFadden et al., 2020;Sherman et al., 2021;Wolf et al., 2020).Survey responses were excluded only if none of the questions beyond eligibility were answered.
All analyses were conducted using Stata/SE 16.1 and Microsoft Excel.This work was conducted by the Community and Collaborative subgroup of the integrated Translational Health Research Institute of Virginia (iTHRIV), a collaboration between Virginia Tech, University of Virginia, Inova, and Carilion Clinic.This study was approved by the Virginia Tech institutional Review Board (IRB number: 20-353) and the Inova Institutional Review Board (IRB number: U20 05-4056), prior to initiation of study activities at the respective sites.

Respondent characteristics
The survey was open from May 19th to July 19th, 2020.Of the 3,694 individuals who started the survey, 3,678 (99.6%) self-reported as eligible and of these 190 (5%) did not answer any survey questions and were excluded.Of the remaining 3,488 respondents 3,367 (97%) fully completed the survey.Of the 3,488 included in this analysis, 70% completed the survey in May, 21% in June, and 9% in July of 2020.Participants were represented throughout Virginia (Fig. 1), with the largest numbers of respondents residing in Montgomery County (home of Virginia Tech), Loudoun and Fairfax Counties (near Washington DC, home of Inova), and Wise County (home of UVA Wise), reflecting sites where survey recruitment began.
More women than men received information they trusted from local government leaders (48% vs. 41%) and more men than women received information they trusted from family/friends (31% vs. 25%), and federal government leaders (26% vs. 20%) (Fig. 2B).

Notes.
a Race/ethnicity and employment categories are not mutually exclusive.

Risk mitigation behavior changes
Ninety-eight percent of respondents completed the questions about changes in behaviors and 98% of those reported changing their behavior in some way in response to the pandemic (Fig. 9).More than half of respondents reported one or more of the following behavior changes: practicing social/physical distancing (95%), wearing a mask when in public (90%), washing hands more often (90%), shopping for groceries and other essentials less often (86%), washing hands for 20 s (86%), being more careful not to touch their face in public (82%) and/or with unwashed hands (80%), using hand sanitizer more often (79%), Figure 8 Percent that believed one or more alternative messages*, by participant characteristics.Percent of respondents who selected they believed in one or more alternative message when answering the question: ''The following messages are related to the coronavirus/COVID-19 (not all are true).Please check all that apply if you have heard, believe, and/or changed your behavior based on each message'' by (A) gender, (B) age-group, (C) race/ethnicity, (D) political identity, (E) education level, (F) income level, and (G) Virginia region.*Alternative messages response options include: COVID-19 ''was developed as a bioweapon,'' ''was developed to lower social security payments to seniors,'' ''is a sign of the apocalypse/end times,'' ''is a hoax,'' ''can be treated with natural remedies,'' ''was developed for population control,'' and ''was developed to increase sales of cleaning supplies.''.Full-size DOI: 10.7717/peerj.avoiding public spaces (73%), cleaning frequently touched surfaces (68%), stocking up on supplies (62%), and started working from home (52%).Wearing a mask in public was reported by more women than men (92% vs. 84%) (Fig. 10A), and the proportion of respondents who reported wearing a mask increased by age from 80% of those 18-24 years old to 95% of those 70 years and older (Fig. 10B).Non-Hispanic White respondents were less likely to report mask wearing (90%) compared to Hispanic (95%), Asian (95%), and Black (94%) respondents (Fig. 10C).Additionally, more Democrats than Republicans and others (97%, 77%, and 87%, respectively) (Fig. 10D) reported wearing a mask, and mask use increased with greater education from 76% of those with a high-school education or less to 94% of those with a doctoral degree (Fig. 10E), and more higher-than middle-and lower-income (92%, 89%, 87%) (Fig. 10F).Those in Southwest Virginia reported less mask wearers (86%) than other regions (91%-95%) (Fig. 10G).Distancing was more common than masking in all groups but showed similar demographic trends as wearing a mask.
In adjusted logistic analyses, we found that the odds of reporting not wearing a mask in public was greater than their comparative groups for those living in Southwest Virginia vs.

DISCUSSION
In our sample of adults residing in Virginia, we found many differences in where people received information that they trust, what they believed, and how their behaviors changed in response to the COVID-19 pandemic by socio-demographics, political identity, and  geography within Virginia.Respondents who identified as non-Hispanic white, men, Republican, other political identity, younger age, income <$100,000, did not report national science and health organizations as a trusted source, reported believing an alternative message, and/or living in Southwest Virginia had greater odds of not wearing a mask than their comparative groups in both unadjusted and adjusted logistic regression.Differences were also observed for physical distancing for these same variables, but at a lower magnitude as distancing was more likely than masking across all groups so differences were less pronounced.
The most consistently listed trusted information sources included national health and science organizations like the NIH and CDC, state and local health departments, and healthcare professionals.Additionally, CDC and gubernatorial recommendations and, more strongly, mandates, were often reported as strong influencers of COVID-19 related beliefs and correlated with masking and distancing in public.These results emphasize the importance of these entities to communicate accurate and timely information and responsibly issue recommendations and mandates to mitigate the pandemic.
Our study was subject to several limitations.First, complete demographic and socioeconomic information was missing from 9% of respondents included in this study.Second, the political identity response options were limited to Republican, Democrat, independent, and other, resulting in individuals identifying as ''independent'' and ''other'' being grouped together, although these individuals may hold extremely diverse political views.Finally, this internet-based convenience sample is not representative of the generalized Virginia population: females, non-Hispanic white, those with at least a bachelor's degree, those with a higher income (United States Census Bureau, 2021), and democrats were overrepresented (Pew Research Center, 2014;Virginia Department of Elections, 2023).However, while the summary descriptive statistics are not representative of the target population of Virginia residents overall, we were able to make valid comparisons between subgroups thanks to our large sample size, many of which have been observed similarly in less detailed surveys conducted in the United States and internationally, discussed in more detail below (Azlan et al., 2020;Barari et al., 2020;Baum et al., 2020;Benham et al., 2021;Bonyan et al., 2020;Carey, 2021;Christensen et al., 2020;Lee et al., 2020;MacDonald & Hesitancy, 2015;McCaffery et al., 2020;O'Shea & Ueda, 2021;Roozenbeek et al., 2020;Sherman et al., 2021;Wolf et al., 2020).Additionally, because this is a cross-sectional survey, our results may not reflect conditions at other time points given that the survey was conducted in the summer of 2020.Data collection began on May 19th, 2020 during the Governor's executive order that directed Virginians to stay home except for essential services, bans crowds of more than 10 people, closed recreation, entertainment, and personal care businesses; and limited restaurants to only takeout and delivery services only (Office of the Governor, 2020a;Office of the Governor, 2020b).This was just prior to the racial justice protests that began on May 26th in Minneapolis and continued throughout the United States (Taylor, 2021).People's behavior may have been altered based on their cost-benefit analyses of COVID-19 risk and the risks associated with racial injustice over the course of our data collection period (Godoy, 2020;Huang & Aubrey, 2020).Finally, our survey did not investigate an important feature of messaging observed in multiple other studies: consistent messaging focusing on positive ways to cope with lockdowns and other COVID-19 mitigation measures were more effective than messaging focused only on compliance in promoting long-term behavioral changes like staying at home, mask-wearing, social distancing, and hand-washing (Azlan et al., 2020;Barari et al., 2020;Benham et al., 2021;Wolf et al., 2020).Other cross-sectional survey studies from early in the COVID-19 pandemic (spring to summer 2020) found similar overall results describing the link between evidence-based messaging and behaviors (Alobuia et al., 2020;Azlan et al., 2020;Barari et al., 2020;Benham et al., 2021;Bonyan et al., 2020;Christensen et al., 2020;McCaffery et al., 2020;Sherman et al., 2021;Wolf et al., 2020).For example, studies in Australia, Malaysia, Italy, Canada, the United Kingdom, Arab countries, and other regions of the United States found that evidence-based COVID-19 messaging significantly influenced respondents' beliefs and risk mitigation behaviors.Studies also showed that while older adults were generally more concerned and had higher anxiety levels about potential COVID-19 infection, they were also less concerned than younger adults about the short-and long-term economic instabilities caused by the pandemic (Barari et al., 2020;Benham et al., 2021;Bonyan et al., 2020;Carey, 2021;Christensen et al., 2020;Wolf et al., 2020).Other studies also found that women, racial/ethnic minorities, and those with lower socioeconomic status experienced more COVID-19 anxieties compared to men, ethnic majorities, and those of higher socioeconomic status (Alobuia et al., 2020;Christensen et al., 2020;McCaffery et al., 2020;Wolf et al., 2020).In multiple countries, people identifying as politically conservative and those with lower health literacy and education level were more likely to report not following recommended COVID-19 precautions and believing that people were overreacting (Azlan et al., 2020;Bonyan et al., 2020;Christensen et al., 2020;McCaffery et al., 2020;Roozenbeek et al., 2020).People identifying as politically liberal and those with higher health literacy and education were more likely to follow public health guidelines and believe that their governments were not doing enough to stop the pandemic (Bonyan et al., 2020;Christensen et al., 2020;McCaffery et al., 2020;Wolf et al., 2020).Multiple studies showed that false information exposure and beliefs were consistently higher among younger people, ethnic minorities, and those who identified as politically conservative (Baum et al., 2020;Christensen et al., 2020;Lee et al., 2020;McCaffery et al., 2020;Roozenbeek et al., 2020).Several preliminary studies have also shown that false information and mistrust in government entities and/or the vaccine development process are major contributing factors to vaccine hesitancy, especially among minority populations and people with low education levels, socioeconomic status, and low perceived risk of contracting COVID-19 (Gatwood et al., 2021;Khubchandani et al., 2021;Troiano & Nardi, 2021).Our study supports and adds to this knowledge describing how information sources considered trustworthy vary across these different population believes and behaviors.By focusing on Virginia, we were able to collect a large amount of data quickly during a key moment early in the pandemic across a geographically and culturally diverse area within the United States.
This study can assist decision makers and the public in developing more effective, targeted, public health messaging for both the ongoing COVID-19 pandemic and for future public health challenges in Virginia and similar settings in the United States.Future

Figure 1
Figure 1 Map of survey respondents by county in Virginia (N = 3,307).Number of respondents residing in each county in Virginia.This figure was generated using ArcGIS.Full-size DOI: 10.7717/peerj.16714/fig-1

Figure 9
Figure9Behavior changes reported in response to the pandemic.Survey responses to the question: ''How (if at all) have you changed your behavior in response to the coronavirus/COVID-19? (Check all that apply)'' for all respondents.Full-size DOI: 10.7717/peerj.16714/fig-9

Messages that respondents believe and/or affected their behaviors
Silverman et al. (2024), PeerJ, DOI 10.7717/peerj.1671413/27 (N = 3,445).Survey responses to the question: ''The following messages are related to the coronavirus/COVID-19 (not all are true).Please check all that apply if you have heard, believe, and/or changed your behavior based on each message'' for all respondents.Full-size DOI: 10.7717/peerj.16714/fig-7

Figure 11 Percent of respondents reporting an information source as trustworthy by if they reported wearing a mask and distancing or not in public
. Survey responses to the question: ''How (if at all) have you changed your behavior in response to the coronavirus/COVID-19? (Check all that apply).''Percent of respondents reporting an information source as trustworthy by if they reported (A) wearing or not