Background

The novel coronavirus disease (COVID-19) has rapidly emerged as a pandemic, precipitating the largest global health and economic crisis in recent history (World Health Organization 2020a). Highly transmissible, with relatively high morbidity and mortality rates (Wilson et al. 2020), particularly among older adults (Bai et al. 2020; Chan et al. 2020; Rothe et al. 2020), COVID-19 has already inflicted myriad adverse health, social, and economic consequences, and exacerbated existing health inequities. While many states have enacted social distancing measures to slow viral transmission, there is limited empirical research with regard to people’s knowledge, beliefs, and perceptions related to COVID-19 that may affect compliance with recommended mitigation measures. Given the disproportionate burden of COVID-19 morbidity and mortality among older adults, age-specific data are also of great importance (Verity et al. 2020). Equally lacking are data on COVID-19-related anxiety, depression, and substance use, which may adversely affect adoption and sustainment of mitigation behaviors (Brooks et al. 2020). Social separation, particularly for prolonged periods, can promote poor mental health outcomes (Lu and Bouey 2020; Person et al. 2004; Shultz et al. 2016). Given the absence of a vaccine, preventing COVID-19’s more dire outcomes largely depends on compliance with preventive behaviors, modulated by the public’s perceptions of the pandemic (Arons et al. 2020; Fauci et al. 2020; Gandhi et al. 2020; Hartley and Perencevich 2020; Wong et al. 2007b). The development and widespread dissemination of effective risk communication strategies are critical for containment. To optimize the development of effective public health messaging, it is critical to identify gaps in COVID-19-related knowledge, beliefs, and mitigation behaviors.

To address this gap, we implemented an internet-based, nationwide survey of knowledge, beliefs, mental health, and behaviors related to the COVID-19 pandemic among US adults.

Methods

Participant recruitment

The sample was a self-selected non-probability sample of social media users aged ≥ 18 years and residing in the USA who responded to an anonymous web-based survey administered via Qualtrics (Provo, UT, USA) (Ali et al. 2020). To minimize redundant reporting, participants could only complete the survey once (based on IP address). Eligibility was confirmed through two screening questions (age and residence in the USA) at the start of the survey. Survey reporting followed the American Association for Public Opinion Research guidelines (AAPOR 2016). The [redacted] Institutional Review Board reviewed and exempted the study procedures.

Recruitment was facilitated via a social media advertisement campaign through Facebook and its affiliated platforms. Used by 69% of US adults, Facebook was chosen because it is the most popular social media platform among adults ≥ 65 years (Perrin and Anderson 2019; Smith and Anderson 2018). A growing body of evidence suggests that Facebook is a valid and effective recruitment tool in health research, resulting in lower costs and shorter recruitment periods than other methods (Whitaker et al. 2017). The advertisements ran for 11 consecutive days, from March 20–30, 2020.

Survey

The survey questionnaire was developed for this study (see Supplementary File S1 Questionnaire). The survey was informed by the Health Belief Model (HBM), which is a widely used model to explain preventive health behaviors based on knowledge, attitudes, and cues to action (Durham and Casman 2012; Najimi and Golshiri 2013). HBM has previously been utilized to guide surveys for other viral outbreaks, such as H1N1 and Middle East Respiratory Syndrome (MERS) (Al-Rabiaah et al. 2020; Hasan et al. 2018; Lau et al. 2005; Rubin et al. 2009; Vartti et al. 2009) and for Ebola in the USA (Painter et al. 2017; Painter et al. 2018). The survey was also informed by the World Health Organization (WHO) tool for behavioral insights on COVID-19 (World Health Organization 2020b). The survey included questions on COVID-19 knowledge, risk perceptions, and preventive behaviors, and psychometrically validated mental health scales adapted for this pandemic.

Knowledge of COVID-19 was measured by 13 binary response format (True/False) items, such as “Coronavirus is a contagious disease.” Responses consistent with information provided by the CDC as of March 11, 2020, were summed to create a composite knowledge score.

Awareness of COVID-19 protective practices was assessed by ten binary response format (True/False) items, that respondents may have considered preventive of COVID-19 infection, such as “Getting a flu shot.” Items were selected from previous surveys (Jalloh et al. 2018; Lau et al. 2007; Painter et al. 2017) and updated to reflect behaviors relevant to the current pandemic. Responses consistent with CDC recommendations as of March 11, 2020, were summed to create a composite COVID-19 prevention score.

Adoption of COVID-19 protective practices was assessed by 12 binary response format (Yes/No) items about specific evidence- and non-evidence-based behaviors that respondents may have considered preventive of COVID-19 infection, such as “Started wearing rubber gloves in public.”

COVID-19-related changes in tobacco and alcohol use were assessed by the question “Since hearing about the Coronavirus outbreak, has your smoking (tobacco products) and alcohol use behaviors changed?” Response options ranged from “Much more” to “Much less” and included “Not applicable.” The variables were re-coded as “more,” “less,” or “no change.”

Anxiety and depression related to COVID-19 were assessed with an adapted version of the four-item Patient Health Questionnaire (PHQ-4) (Jalloh et al. 2018; Lowe et al. 2010). The stem question was, “Over the last 7 days, how often have you been bothered by any of the following problems because of the Coronavirus outbreak?” Response options were rated on a 4-point Likert scale ranging from “Not at all” to “Nearly every day.” Total scores ranged from 0 to 12, with higher scores indicative of greater anxiety and depression symptoms. The score was dichotomized based on the clinical cutoff for possible depression and anxiety (< 6 and ≥ 6).(Lowe et al. 2010) The scale demonstrated internal reliability (alpha = 0.89).

The Impact of Event Scale (IES) and its abbreviated 6-item version (IES-6) are widely validated scales to measure subjective stress (Lau et al. 2005; Thoresen et al. 2010; Weiss and Marmar 1997). We adapted the IES-6 scale to measure stress associated with the COVID-19 pandemic. An item example was, “I thought about Coronavirus when I didn’t mean to,” with four response options ranging from “Not at all” to “Nearly every day.” The mean item response was calculated (possible range from 0 to 3), with higher mean scores indicative of more subjective stress. The sample was dichotomized at the median into high and low stress scores. The scale’s internal consistency was 0.86.

Pessimism was assessed by the question: “I am optimistic that the Coronavirus outbreak will be controlled in the next 3 months.” Response options on the four-point Likert scale ranged from “Strongly agree” to “Strongly disagree.” The variable was reverse coded and dichotomized into a Pessimism variable (Disagree/Strongly Disagree = 0; Strongly Agree/Agree = 1).

Demographics assessed included sex, race, age category (by decade), employment status, educational attainment, living with children < 18 years of age, state of residence re-coded by US Census region (U.S. Census Bureau 1984), urban/rural residence, and political party affiliation.

Statistical analysis

Given the potential importance of age as a demographic risk factor for mortality among people with COVID-19 infection, the sample was categorized into three age groups: (1) 18–39 years, (2) 40–59 years, and (3) ≥ 60 years. Descriptive statistics characterized respondents’ demographics and COVID-19-related knowledge, behaviors, substance use, and mental health for the total sample and the three age categories. Pearson’s chi-squared tests for categorical variables, and ANOVA tests for continuous variables, were used to examine differences by age categories. Post-hoc pairwise comparisons were also tested. Logistic regression analyses assessed changes in alcohol consumption and smoking associated with anxiety, depression, and stress, controlling for age and race. All analyses were complete case analyses. Statistical analyses were performed using Stata version 15.1 (StataCorp, College Station, TX, USA).

Results

Sample characteristics

The sample was comprised of 6391 respondents; 4998 respondents (78%) completed 100% of the survey, 1393 (22%) were partial completers (906 completed < 33% and 487 completed 50%–79% of the items). Respondents took a median of 10.9 min [interquartile range (IQR) = 9.0, 13.4] to complete the full survey.

Sociodemographic characteristics of the sample

Sociodemographic characteristics of respondents are presented for the entire sample and by age group in Table 1. Approximately half of the respondents were 50–69 years old and 58% were female. The majority self-identified as non-Hispanic White. Participants were almost uniformly distributed between three of the four major geographic regions in the USA (26–30% in each), with fewer respondents residing in the West (15%). Almost half lived in suburban areas. Approximately, three-fourths of adults under 60 years were employed, compared to a third of those aged 60 years and over.

Table 1 Characteristics of adult Facebook users in the USA during the COVID-19 outbreak, March 20–30, 2020

Knowledge of COVID-19

Respondents had high COVID-19 knowledge scores [mean = 12.0; standard deviation (SD) = 1.2; range = 0–13] (Table 2). Most were aware that people infected with Coronavirus could be asymptomatic (89%), and that a vaccine was unavailable (97%). One item assessing knowledge of the protective value of alcohol-based hand sanitizers, while high (84%), was lower relative to other knowledge items. Only 81% of respondents aged ≥ 60 years were aware of the protective value of alcohol-based hand sanitizers, lower than respondents aged 40–59 years (81%≥ 60 vs 86%40–59 years, p < 0.001).

Table 2 Accurate knowledge of COVID-19 facts and protective practices among 5960 Facebook users in the USA, March 20–30, 2020

Awareness of COVID-19 protective practices

Respondents, on average, correctly identified 90% (mean = 9.1; SD = 0.8; range 0–10) of protective practices (Table 2). However, some gaps were observed: only 84% and 51% of respondents, respectively, were aware that receiving an influenza vaccine did not confer protection against the novel Coronavirus, and that wearing a facemask was protective. Few differences were observed across age groups; with one exception: fewer respondents 60 years and older were aware that the influenza vaccine was not protective against the novel Coronavirus compared to 18–39 year-olds (82%≥ 60 years vs 86%18–39 years, p = 0.005).

Adoption of COVID-19 protective practices

Most respondents reported avoiding crowded places and using alcohol-based hand-sanitizers (Table 3). Only 19% of respondents reported purchasing facemasks. Overall, significant differences were observed across age groups; although the magnitude of differences was relatively modest. Differences across age groups were more pronounced with regard to non-evidence-based protective practices (e.g., taking hot baths and dietary supplements) than to evidence-based practices.

Table 3 Adoption of COVID-19 protective practices among a sample of US adult Facebook users (n = 5486), March 20–30, 2020

COVID-19-related changes in tobacco and alcohol use

Among self-identified smokers (n = 1359), about half indicated no change in their smoking behavior as a result of COVID-19, while 22% reported smoking more (Table 3). With respect to alcohol consumption, 53% reported no change in drinking behavior, while 25% reported drinking more. There was an observed age gradient in those reporting increased drinking: 34% of respondents aged 18–39 years reported increased drinking compared to 25% of those aged 40–59 years and 15% of those aged 60 years or more (all differences p < 0.001).

Mental health indicators: anxiety, depression, stress, and pessimism

COVID-19-related anxiety and depression symptoms experienced within the week prior to survey participation were high; 44% of respondents had a score consistent with depression and anxiety (PHQ-4 score of ≥6) (Table 4). Notably, 85% of respondents reported “feeling nervous, anxious, or on edge.” Stress levels were also high, with 52% reporting high scores (at or above the median of 1.33). Similarly, 47% of respondents reported “pessimism” about the COVID-19 pandemic being resolved in the next 3 months. Differences were observed between age groups for both anxiety/depression and stress: older age was associated with lower proportion of identified symptoms of anxiety, depression, and stress. Among 18–39-year-olds, 57% had PHQ-4 scores ≥ 6 compared to 33% of ≥ 60 year-olds (p < 0.001). Respondents aged 18–39 years were more likely to report high stress scores than adults aged ≥ 60 (67%18–39-year-olds versus 39%≥ 60 years, p < 0.001).

Table 4 Mental health related to the COVID-19 outbreak among a sample of U.S. adults (n = 5044), March 20–30, 2020

Associations between mental health, stress, and substance use

Adjusted analyses, controlling for age and race, showed that respondents reporting symptoms of anxiety and depression (PHQ-4 score ≥ 6) were 79% more likely than those below the clinical cutoff to report drinking more [adjusted odds ratio (AOR) = 1.79; 95% confidence interval (CI) = 1.49, 2.15, p < 0.001] and more than twice as likely to report smoking more (AOR = 2.17; 95% CI = 1.64, 2.88, p < 0.001) in response to the COVID-19 pandemic (Table 5). A similar pattern was observed with stress symptoms, whereby respondents with high stress levels were 80% more likely to report increased drinking (AOR = 1.80; 95% CI = 1.49, 2.17, p  < 0.001), and 75% more likely to report increased smoking (AOR = 1.75; 95% CI = 1.31, 2.33, p  < 0.001) relative to those with low stress levels.

Table 5 Substance use associated with depression, anxiety and stress during the COVID-19 outbreak

Discussion

Overall, respondents had high levels of COVID-19 knowledge and awareness of protective practices. Most were aware that a vaccine was unavailable and to avoid congregating in crowded places, a key COVID-19 mitigation strategy (Ferguson et al. 2020; IHME 2020). However, purchase of facemasks was low. This may be attributable to mixed public health messaging at the time of the survey (March 2020) about the effectiveness of facemasks as a personal protective practice, and the limited availability of masks to the public (Keshtkar-Jahromi et al. 2020). Another gap was the relatively low awareness about alcohol-based hand sanitizers as a protective practice. These findings suggest that while adults are largely cognizant of accurate information about COVID-19 and preventive practices, some gaps remain to be addressed.

As our scientific understanding of COVID-19 improves continuously with respect to transmission, treatment, and prevention, public attitudes towards COVID-19 will also likely evolve. Public health messaging will need to rapidly respond and adapt to both. Furthermore, COVID-19 is likely to require sustained public health action in the foreseeable future. Research on recent global emerging disease outbreaks, such as SARS and H1N1, has shown that public perceptions and behaviors can change significantly during the course of outbreaks, and several studies have highlighted the importance of understanding the public’s response (Lau et al. 2011; Lau et al. 2003; Leung et al. 2005; Wong et al. 2007b). Therefore, monitoring perceptions and behaviors over time in response to changes in the COVID-19 pandemic, and its containment strategies, is warranted, and can provide policymakers with useful information to ensure continued compliance with key recommended mitigation behaviors (Brug et al. 2009).

Of particular concern is the high prevalence of COVID-19-related depression, anxiety, stress, and pessimism, and increases in alcohol and tobacco use. While the impact of the COVID-19 pandemic has been predominantly measured in terms of morbidity and mortality (Dong et al. 2020), the myriad adverse societal consequences may be precipitating a “hidden epidemic” of mental illness that is not receiving adequate attention (Brooks et al. 2020). Indeed, the American Psychological Association has predicted the need for a scaled-up response to traumatic stress resulting from the pandemic (Chang et al. 2013; Horesh and Brown 2020). Outbreaks such as H1N1 negatively impacted the mental health not only of survivors (Mak et al. 2009) and healthcare workers (Chen et al. 2006), but also of communities (Elizarrarás-Rivas et al. 2010; Wheaton et al. 2012; Wong et al. 2007a). Furthermore, suicide rates increased by up to 6.4% following the 2008 financial crisis (Chang et al. 2013). Given that the COVID-19 pandemic is driving a global economic recession, similar trends may emerge.

Depression and anxiety already constitute the leading cause of disability worldwide (Friedrich 2017). Exacerbation of their burden in the USA may produce unforeseen public health and economic consequences long after the COVID-19 pandemic is controlled. Therefore, policymakers should allocate resources for psychosocial support to alleviate both the personal and societal burden of a mental health crisis. The observed associations between stress, substance use, and depression and anxiety suggest that these factors may be interrelated adverse consequences of the COVID-19 pandemic. Although there is limited research on substance use following viral outbreaks, increases in problem alcohol use were observed among hospital employees in China during the 2003 SARS outbreak (Wu et al. 2008). Thus, confronting these challenges may necessitate enhancing access to substance use treatment and mental health services. Given the scope and disruptive effect of COVID-19 relative to other outbreaks (Shultz et al. 2015), mental health and substance-use interventions utilized after natural disasters (Jaycox et al. 2010; Waelde et al. 2008) can provide valuable guidance. Resiliency from the short term mental, social, and economic impacts of COVID-19 will require intense study (Abramson 2020).

Even though many states have instated social distancing measures, which may preclude face-to-face counseling with healthcare providers, opportunities remain to utilize online and telehealth counseling services. These services are compliant with social distancing policies, and can play a critical role in coping with the adverse consequences of COVID-19, such as loneliness, income loss, and pessimism. One technological strategy to enhance self-care may include the use of smartphone applications to monitor and link users to services (Bakker et al. 2016, 2018). Distance telehealth counseling strategies are not only valuable, but may be essential to confront the mental health and substance-use threats posed by the COVID-19 pandemic.

During a pandemic, accurate and verifiable information is essential to dispel misconceptions and mitigate fears and ambiguity that could result in the adoption of non-evidence-based practices, maladaptive coping strategies, or use of unproven therapeutics. Non-evidence-based treatments are not solely ineffectual; they are dangerous. Complex medical information, such as COVID-19 virology, immunology, medical treatments, and epidemiology, needs to be clear, consistent, and understandable. Thus, there is a need to improve risk communications and increase the message dissemination channels. Risk and health communication theories and techniques can be utilized to enhance message clarity and impact (DiClemente and Jackson 2016). For instance, the use of dynamic visual depictions may enhance effective communication and facilitate accurate information transfer. The visual depiction of “flattening the curve” has been used to convey social distancing recommendations based on complex epidemiological modeling (Anderson et al. 2020). A dynamic depiction of the virus’ spread could show the increasing prevalence of COVID-19 across the USA over time. Diverse communication channels, strategies, and techniques are needed to facilitate message reach. News media outlets, predominantly those on television, are a major source of COVID-19 information. Other media, such as social media, blogs, and text messaging, can complement accurate news and may be harnessed to disseminate important information.

Strengths and limitations

Strengths of this study included a large sample size, the wide geographic distribution of participants, the use of theory-driven, validated scales and items, and the fact that data were collected during the peak of the pandemic, thus providing a snapshot of the perceptions and behaviors of the US population during a critical period of the crisis. However, the findings should be considered in the context of several limitations. First, given that the survey was opt-in, there is probably a bias towards respondents with greater interest in the subject. These respondents may differ in unmeasured, systematic ways from those uninterested in COVID-19. Second, the population from which the sample was drawn included only Facebook users. Although almost 70% of Americans use Facebook (of whom 74% use it daily), certain demographic groups, such as men and racial/ethnic minorities, are under-represented (Pew Research Center 2019). The demographic distribution of our respondents mirrors this disparity, with an over-representation of Non-Hispanic Whites and females. Consequently, this survey many not be generalizable to the US population and may not adequately reflect the perceptions and behaviors of other groups.

Conclusions

As the toll from COVID-19 continues to rise, there is a need to address both the community transmission and the concomitant mental health and substance-use consequences of the crisis. This requires a well-articulated, coordinated, and systematic pandemic control strategy; one based on science and effective behavior change and risk communication strategies. This study contributes to the vital evidence base needed to inform targeted public health interventions on disparities in knowledge, beliefs, and behaviors (including substance use) occurring during the COVID-19 pandemic. To increase the uptake and sustained practice of COVID-19 preventive behaviors requires utilizing theory- and empirically-based strategies with demonstrable evidence of effectiveness in meeting other health threats. Ultimately, reducing the transmission of COVID-19 will require an all-out effort (the public health equivalent of a 'full-court press' in basketball), with prevention messages aimed at addressing knowledge gaps, misconceptions, and practices, using the most effective social and behavioral science communications strategies.