key: cord-1016920-hwtm02y2
authors: Bruine de Bruin, Wändi; Carman, Katherine G.; Parker, Andrew M.
title: Mental associations with COVID-19 and how they relate with self-reported protective behaviors: A national survey in the United States
date: 2021-03-07
journal: Soc Sci Med
DOI: 10.1016/j.socscimed.2021.113825
sha: e573cfe3e10ebda395d1604195018c0d4183ce95
doc_id: 1016920
cord_uid: hwtm02y2

RATIONALE: To understand novel diseases, patients may draw comparisons to other diseases. OBJECTIVE: We examined whether mentally associating specific diseases with COVID-19 was related to self-reported protective behaviors early in the pandemic. METHODS: In March 2020, a national sample of 6,534 U.S. adults listed diseases that came to mind when thinking of COVID-19. They self-reported protective behaviors, demographics, and COVID-19 risk perceptions. RESULTS: Participants associated COVID-19 with common infectious diseases like seasonal influenza (59%), common cold (11%), and pneumonia (10%), or emergent infectious diseases like pandemic influenza (28%), SARS/MERS (27%), and Ebola (14%). Seasonal influenza was most commonly mentioned, in all demographic groups. Participants mentioning seasonal influenza or common cold reported fewer protective behaviors. Those mentioning pneumonia or emergent infectious diseases reported more protective behaviors. Mentioning pneumonia, SARS/MERS, and Ebola was associated with the most protective behaviors, after accounting for other generated diseases, demographics, and risk perceptions (e.g., for avoiding crowds, OR=1.52, 95% CI=1.26, 1.83; OR=1.28, 95% CI=1.13, 1.46; OR=1.30, 95% CI=1.11, 1.52, respectively). CONCLUSIONS: Early in the pandemic, most participants mentally associated COVID-19 with seasonal flu, which may have undermined willingness to protect themselves. To motivate behavior change, COVID-19 risk communications may need to mention diseases that resonate with people while retaining accuracy.

As COVID-19 spread across the world, people faced a novel health threat. To limit disease transmission, the U.S. Centers for Disease Control and Prevention started recommending protective behaviors such as hand hygiene and social distancing in March 2020 (New York Times, 2020) , adding non-medical mask use in April 2020 (National Public Radio, 2020a) . Mass adoption of these behaviors is especially important when pharmacological interventions are not yet available (Bruine de Bruin et al., 2006) .

Drawing comparisons with familiar threats has been advocated as a tool for communicating about novel risks (Edwards, 2003) . Such comparisons may include metaphors and analogies, which respectively identify specific features or relationships within a familiar domain that are like those in the novel domain (Gentner & Holyoak, 1997; Gentner & Markman, 1997) . This process should allow recipients to draw new inferences and improve their mental model of the novel domain.

However, the use of comparisons in risk communication may backfire. For example, comparing smallpox to chickenpox can not only facilitate people's understanding of smallpox transmission but also promote the misunderstanding that smallpox is a disease that affects mostly children (Bostrom, 2008) . Additionally, a 2003 UK newspaper analysis found that SARS was described as similar to seasonal influenza and pneumonia, but mentioning seasonal influenza was associated with describing SARS as less severe (Wallis and Nerlich, 2005) .

COVID-19 was also compared to seasonal influenza early on in the pandemic, perhaps to downplay the need for protective behaviors. For example, U.S. President Trump compared COVID-19 to seasonal influenza while arguing against business closures (National Public Radio, 2020b) . The World Health Organization (2020) warned that, despite similarities in symptoms J o u r n a l P r e -p r o o f Mental associations 2 and transmission routes, COVID-19 had greater rates of severe disease and mortality than seasonal influenza, and no licensed vaccines or therapeutics.

Comparisons between COVID-19 and other diseases also appeared in the public discourse. Policymakers and health care providers in Israel and the UK used comparisons to pandemic rather than seasonal influenza (Atkinson et al., 2020; Gesser-Edelsburg and Hijazi, 2020 ). An analysis of Flemish newspaper articles published early in the pandemic found that COVID-19 was compared with pandemic influenza (Spanish flu, H1N1), SARS, MERS, and Ebola (De Ridder, 2020) .

Early in novel outbreaks when uncertainty is greatest, which other diseases come to patients' minds could have implications for willingness to implement protective behaviors. In an exploratory study, we therefore examined the following research questions:

(1) Which diseases do participants identify as mentally associated with COVID-19?

(2) How are the diseases participants mention related to self-reported protective behaviors, even after taking into account demographics and risk perceptions?

Between March 10-31, 2020, 6,534 of 8,489 (77%) invited members of the University of Southern California's Understanding America Study (UAS) answered the questions analyzed here. The UAS is a panel of U.S. households who are regularly invited to participate in online surveys (https://uasdata.usc.edu/index.php). To obtain a nationally representative sample, UAS members were recruited from randomly selected U.S. addresses, invitations were written in English and Spanish, sampling probabilities were adjusted for underrepresented populations, and J o u r n a l P r e -p r o o f Mental associations 3 internet-connected tablets were provided to interested individuals if needed (Alattar et al., 2018) .

Following the survey literature (Valliant et al., 2013) , post-stratification weights were used to further align our sample to the U.S. adult population regarding age, gender, race/ethnicity, education, and location (https://uasdata.usc.edu/page/Weights).

Overall, 64% of participants were non-Hispanic white, 11% non-Hispanic African- There were no significant differences between invitees who completed the questions analyzed here and those who did not, in terms of percent male, non-Hispanic non-black minority, and living in worst-hit states (all p>.10). However, those who completed the survey included significantly more participants who were white (64% vs. 45%), 2 (2)=55.95, p<0.001, aged 65 or older (20% vs. 13%), 2 (2)=10.66, p<0.01, and with a college degree (34% vs. 25%), 2 (2)=15.26, p<0.01, as well as fewer African-Americans (11% vs. 22%), 2 (2)=39.19, 

The online survey was approved by U.S.C's Institutional Review Board. Survey and data are publicly available (https://uasdata.usc.edu/index.php; #230). The survey was available in English and Spanish, but 99% of participants selected English. The survey questions presented below were used in our analyses. In order, participants completed the questions that asked about risk perceptions, protective behaviors, and diseases mentally associated with COVID-19.

Participants were asked "When you think of the coronavirus , what other diseases come to mind?" followed by three text boxes. Using an inductive datadriven coding procedure, participants' open-ended responses were categorized as (1) seasonal, regular, common, or unspecified flu or influenza, influenza A or B; (2) cold or common cold, (3) pneumonia, (4) pandemic, Spanish, 1918/19, Hong Kong, Asian, H1N1, H5N1, avian, bird, swine flu or influenza, (5) Severe Acute Respiratory Syndrome (SARS) or Middle-East Respiratory Syndrome (MERS), or (6) Ebola. Codes were applied by a coder who was blind to our research questions, and reliability coding was conducted through an automated algorithm.

Cohen's Kappa was ≥.95 for each code.

Protective behaviors. Participants were asked: "Which of the following have you done in the last seven days to keep yourself safe from coronavirus in addition to what you normally do?" They indicated yes/no for the following recommended behaviors by the Centers for Disease Control and Prevention (New York Times, 2020): (1) "washed hands with soap or used hand sanitizer several times per day," (2) "avoided public spaces, gatherings, or crowds," (3) "avoided contact with people who could be high-risk", and (4) "canceled or postponed air travel for work" and "canceled or postponed air travel for pleasure", for which responses were J o u r n a l P r e -p r o o f Mental associations 5 combined. Non-medical mask use was not included because it was not recommended by the Centers for Disease Control and Prevention until April 2020 (National Public Radio, 2020a).

Participants answered "On a scale from 0 to 100%, what is the chance that you will get the coronavirus in the next three months?" and "If you do get infected with the coronavirus, what is the chance you will die from it?" on a validated visual linear scale ranging from 0% to 100% (Bruine de Bruin & Carman, 2018).

Demographics. Demographic information was already on record, including at-risk age over 65 (yes=1; no=0), male gender (yes=1; no=0), race/ethnicity (non-Hispanic white, non-Hispanic African-American, Hispanic/Latinx or other non-Hispanic minority; yes=1; no=0 for each), college degree (yes=1; no=0), household income below the US Department of Health and Human Services' (2020) federal poverty level (yes=1; no=0 for each), and residing in a state that was worst hit at the time, including Massachusetts, New Jersey, New York, and Washington (Table 1) , with chi-square tests (Table 1 ) and logistic regressions (Table 2) examining demographic differences in generating specific diseases. Phi correlations examined which diseases tended to be mentioned together (Table 3) . To examine how the diseases participants mentioned related to self-reported protective behaviors (research question 2), we computed the percent of participants engaging in each behavior and mean risk perceptions, by whether or not each specific disease was generated (Table 4 ). We conducted logistic regressions that predicted each protective behavior from whether or not each specific analogy was used while accounting for demographic variables and risk perceptions (Table 5 ). To further understand the role of risk perceptions, we conducted linear regressions that predicted risk perceptions from whether or not each specific analogy was used, while accounting for demographic variables (Table 6) . We also examined whether main conclusions held when considering only the first disease participants generated (Online Supplement).

In response to the open-ended question about which diseases come to mind when thinking of COVID-19, participants generated common infectious diseases such as seasonal influenza (59%), common cold (11%), and pneumonia (10%), as well as emergent infectious diseases such as pandemic influenza (28%), SARS or MERS (27%), and Ebola (14%) ( Table 1) .

Overall, 86% of participants mentioned at least one of these diseases.

In each demographic group, seasonal influenza was the most commonly mentioned disease, with pandemic influenza and SARS/MERS completing the top three (Table 1) .

Race/ethnicity was the only demographic variable that showed a significant difference in J o u r n a l P r e -p r o o f Mental associations 7 mentioning each disease (Table 1) , which remained significant after accounting for other demographic variables (Table 2) . Specifically, relatively more non-Hispanic white participants mentioned seasonal influenza and the common cold, relatively more non-Hispanic African-Americans mentioned pneumonia, relatively more Hispanic/Latinx participants mentioned Ebola, and relatively more participants from other minorities mentioned SARS/MERS (Table 1) .

Correlations among mentions of specific diseases were small to moderate (Table 3) .

Common infectious diseases were more likely to be mentioned with other common infectious diseases, and less likely to be mentioned with emergent infectious diseases. Similarly, emergent infectious diseases were more likely to be mentioned with other emergent infectious diseases, and less likely to be mentioned with common infectious diseases.

Across participants, 90% indicated washing hands, 57% avoiding public spaces or crowds, 58% avoiding high-risk individuals, and 37% canceling or postponing travel. With few exceptions, mentioning (vs not mentioning) seasonal influenza or the common cold tended to be associated with lower likelihood of reporting protective behaviors, while mentioning (vs. not mentioning) pneumonia or emergent infectious diseases (pandemic influenza, SARS/MERS, and Ebola) tended to be associated with greater likelihood of reporting protective behaviors (Table   4 ).

When considering the independent contributions of mentioning specific diseases while also accounting for demographics and risk perceptions, we found that mentioning the common cold was associated with lower likelihood of reporting two of the four protective behaviors (Table 5) (Table 5) . When only considering diseases that were mentioned first, mentioning pneumonia predicted greater likelihood of implementing all protective behaviors, SARS/MERS two, and Ebola one (Online Supplement, Table S4 ). Subsequent analyses suggested why controlling for COVID-19 risk perceptions had little to no effect on relationships between diseases participants mentioned and their self-reported protective behaviors. Although COVID-19 risk perceptions have been positively associated with protective behaviors (Bruine de Bruin & Bennett, 2020), they showed only limited variation with the diseases participants mentioned (Table 4 ). First, we examined associations of mentioning specific diseases with perceived risk of getting infected with COVID-19. Mentioning (vs. not mentioning) seasonal influenza was associated with perceiving slightly lower risk of getting infected, while mentioning (vs. not mentioning) pandemic influenza or SARS/MERS was associated with perceiving slightly greater risk of getting infected (Table 4 ). In linear regressions accounting for other diseases mentioned and for demographics, the positive relationship of pneumonia, pandemic influenza, and SARS/MERS with perceived risk of getting infected held (Table 6 , Model 1). When considering diseases that were mentioned first, only SARS/MERS was associated with risk perceptions of getting infected with COVID-19 (Table   S5 , Model 1). Second, we examined associations of mentioning specific diseases with perceived risk of dying if getting infected with COVID-19 Mentioning (vs. not mentioning) seasonal influenza, common cold, pandemic influenza, and SARS/MERS were associated with perceiving slightly lower risk of dying if infected (Table 4 ). In linear regressions accounting for other diseases mentioned and for demographics, all of these relationships held, while pneumonia became associated with perceiving slightly greater risk of dying if infected (Table 6 , Model 2).

When only considering diseases mentioned first, mentioning any of the diseases was associated with lower risk perception of dying if infected -with the exception of pneumonia (Table S5, Model 2).

When faced with a novel disease such as COVID-19, people may try to draw comparisons to various other diseases (Atkinson et al., 2020; Gesser-Edelsburg and Hijazi, 2020; De Ridder, 2020) . Such comparisons could have implications for their motivations to implement protective behaviors (Edwards, 2003) . In a nationally representative U.S. sample, we therefore examined which diseases came to mind when thinking of COVID-19, and relationships with reported protective behaviors. Participants mentioned common infectious diseases such as seasonal influenza, common cold, and pneumonia as well as emergent infectious diseases such as pandemic influenza, SARS/MERS, and Ebola. Across demographic groups, seasonal influenza was mentioned by far the most. Additionally, common infectious diseases tended to be mentioned together, as were emergent infectious diseases.

Generally, mentioning common infectious diseases like seasonal flu and the common cold was associated with lower likelihood of reporting protective behaviors, and mentioning pneumonia or emergent infectious diseases was associated with greater likelihood of reporting protective behaviors. Independent of which other diseases were mentioned, pneumonia, SARS/MERS, and Ebola were the diseases that were associated with the most (three of four) reported protective behaviors. When only considering the diseases that were listed first, mentioning pneumonia remained the best predictor of reported protective behaviors. These 

Like any study, the present study has limitations. One main limitation is that our exploratory study was cross-sectional and does not warrant causal conclusions. To examine causal effects on protective behaviors, confirmatory research is needed in which participants are randomly assigned to COVID-19 risk communications that mention different diseases.

Additionally, our study focused on a U.S. sample, and the diseases that people generate when thinking of COVID-19 may vary between countries, cultures, and languages. Mask use was not included as a protective behavior, because the CDC did not yet recommend mask use in March 2020 (National Public Radio, 2020a). We did not use a holistic assessment of COVID-19 risk perception, which may have been better at capturing experiential and cultural factors (Dryhurst et al., 2020) . We also did not assess participants' full mental models of COVID-19 and other infectious diseases, leaving it for future research to examine why diseases such as pneumonia, SARS/MERS, and Ebola were generated or how they motivated participants to engage in protective behaviors.

COVID-19 risk communications that aim to promote protective behaviors may be more effective if they avoid drawing comparisons to seasonal influenza or common cold, and instead mention pneumonia, SARS/MERS, or Ebola. Because symptoms of COVID-19 are more similar J o u r n a l P r e -p r o o f Mental associations 11 to those for pneumonia and SARS/MERS than to Ebola, those may be the more appropriate diseases to mention.

J o u r n a l P r e -p r o o f 

Seasonal influenza -Common cold 0.23 *** -Pneumonia 0.14 *** 0.09 *** -Emergent infectious diseases Pandemic influenza -0.30 *** -0.12 *** -0.14 *** -SARS/MERS -0.12 *** -0.12 *** -0.10 *** 0.16 *** -Ebola -0.12 *** -0.11 *** -0.09 *** 0.09 *** 0.11 *** -*** p<0.001; ** p<0.01; * p<0.05. Pearson correlations (r) between dichotomous variables represent phi correlations.

J o u r n a l P r e -p r o o f 

An introduction to the Understanding American Study internet panel

Understanding the policy dynamics of COVID-19 in the UK: Early findings from interviews with policy makers and health care professionals

Lead is like mercury: Risk comparisons, analogies, and mental models

Relationships between initial risk perceptions about the novel coronavirus and self-reported protective behaviors: Evidence from a national Survey

Measuring subjective probabilities: The effect of response mode on the use of focal responses, validity, and respondents' evaluations

Expert judgments of pandemic influenza risks

When the analogy breaks: Historical references in the Flemish news media at the onset of the COVID-19 pandemic

Communicating risks through analogies: BMJ, 749

Reasoning and learning by analogy: Introduction

Structure mapping in analogy and similarity

When Politics Meets Pandemic: How Prime Minister Netanyahu and a Small Team Communicated Health and Risk Information to the Israeli Public During the Early Stages of COVID-19

president-trump-says-cdc-now-recommends-americanswear-cloth-masks-in-public

Wondering about social distancing?

QuickFacts United States. Population estimates

Practical Tools for Designing and Weighting Survey Samples

Disease metaphors in new epidemics: The UK media framing of the 2003 SARS epidemic

Proclamation on Declaring a National Emergency Concerning the Novel Coronavirus Disease (COVID-19) Outbreak

Proclamation-Suspension of Entry as Immigrants and Nonimmigrants of Certain Additional Persons Who Pose a Risk of Transmitting

Influenza and COVID-19 -similarities and differences

01; * p<0.05. a FPL=Federal Poverty level (U.S. Department of Health and Human Services, 2020) b Worst-hit states in March 2020 included Massachusetts