key: cord-254027-rndu5ake authors: Geldsetzer, Pascal title: Using rapid online surveys to assess perceptions during infectious disease outbreaks: a cross-sectional survey on Covid-19 among the general public in the United States and United Kingdom date: 2020-03-17 journal: medRxiv DOI: 10.1101/2020.03.13.20035568 sha: doc_id: 254027 cord_uid: rndu5ake Background: Given the extensive time needed to conduct a nationally representative household survey and the commonly low response rate in phone surveys, rapid online surveys may be a promising method to assess and track knowledge and perceptions among the general public during fast-moving infectious disease outbreaks. Objective: To apply rapid online surveying to determine knowledge and perceptions of coronavirus disease 2019 (Covid-19) among the general public in the United States (US) and the United Kingdom (UK). Methods: An online questionnaire was administered to 3,000 adults residing in the US and 3,000 adults residing in the UK who had registered with Prolific Academic to participate in online research. Strata by age (18 - 27, 28 - 37, 38 - 47, 48 - 57, or >=58 years), sex (male or female), and ethnicity (White, Black or African American, Asian or Asian Indian, Mixed, or "Other"), and all permutations of these strata, were established. The number of participants who could enrol in each of these strata was calculated to reflect the distribution in the US and UK general population. Enrolment into the survey within the strata was on a first-come, first-served basis. Participants completed the questionnaire between February 23 and March 2 2020. Results: 2,986 and 2,988 adults residing in the US and the UK, respectively, completed the questionnaire. 64.4% (1,924/2,986) of US and 51.5% (1,540/2,988) of UK participants had a tertiary education degree. 67.5% (2,015/2,986) of US participants had a total household income between $20,000 and $99,999, and 74.4% (2,223/2,988) of UK participants had a total household income between GBP15,000 and GBP74,999. US and UK participants' median estimate for the probability of a fatal disease course among those infected with SARS-CoV-2 was 5.0% (IQR: 2.0% - 15.0%) and 3.0% (IQR: 2.0% - 10.0%), respectively. Participants generally had good knowledge of the main mode of disease transmission and common symptoms of Covid-19. However, a substantial proportion of participants had misconceptions about how to prevent an infection and the recommended care-seeking behavior. For instance, 37.8% (95% CI: 36.1% - 39.6%) of US and 29.7% (95% CI: 28.1% - 31.4%) of UK participants thought that wearing a common surgical mask was 'highly effective' in protecting them from acquiring Covid-19. 25.6% (95% CI: 24.1% - 27.2%) of US and 29.6% (95% CI: 28.0% - 31.3%) of UK participants thought it prudent to refrain from eating at Chinese restaurants. Around half (53.8% [95% CI: 52.1% - 55.6%] of US and 39.1% [95% CI: 37.4% - 40.9%] of UK participants) thought that children were at an especially high risk of death when infected with SARS-CoV-2. Conclusions: The distribution of participants by total household income and education followed approximately that of the general population. The findings from this online survey could guide information campaigns by public health authorities, clinicians, and the media. More broadly, rapid online surveys could be an important tool in tracking the public's knowledge and misperceptions during rapidly moving infectious disease outbreaks. Organization (WHO) has declared Covid-19 a "public health emergency of international concern" on January 30th 2020. (3) Given their connectedness to China and the world in general, both the United States (US) and the United Kingdom (UK) could be faced with sustained community transmission of Covid-19 within weeks. In fact, on February 25th 2020, the Centers for Disease Control and Prevention identified for the first time a person in the US who had no travel history or exposure to a known Covid-19 case. (4) The course of the Covid-19 epidemic in the US and the UK will likely be strongly impacted by how the population behaves, which in turn is influenced by what people know and believe about this disease. (5) A particular concern in this regard is the spread of disand misinformation about Covid-19 on social media sites, which has led the WHO to host a page with "myth busters" on their website and engage in discussions with social media companies. (6) Understanding what the general public knows about Covid-19 and which misperceptions they hold about the condition is important for US and UK public health authorities as well as the media to design effective information campaigns. This study, therefore, aims to determine knowledge and misperceptions of Covid-19 among the general adult population in both the US and the UK. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint This study is a cross-sectional online survey conducted on the research platform created and managed by Prolific Academic Ltd.. Prolific is an online platform that connects researchers with individuals around the world who are interested in participating in online research studies. (7) The platform's pool of participants numbers approximately 80,000 individuals of whom circa 43% reside in the UK and 33% in the US.(8) Researchers are required to pay participants a minimum of $6.50 per hour. For this study, Prolific selected a sample of 2,987 participants residing in the US and 2,978 participants residing in the UK who are representative by age, sex, and ethnicity (and each combination thereof) of the US and UK general population (using numbers from the latest census in each country), respectively. Prolific achieves representativeness by age, sex, and ethnicity by establishing population strata for the survey into which eligible participants in the online participant pool can enroll on a first-come first-serve basis. Participants had to have indicated that they are fluent in English when registering with Prolific to be eligible for this study. The data were collected between February 23rd and March 3rd 2020 using an online questionnaire (Text S1 and Text S2). Participants received US$1.50 (UK participants received the equivalent in pound sterling) for participating in the study. Following an informed consent form, the questionnaire asked participants about i) the cause, current state, and future development of the Covid-19 epidemic; ii) the risk of a fatal disease course; iii) knowledge of symptoms and recommended healthcare-seeking behavior; iv) measures to prevent a Covid-19 infection; and v) their perception of the risk posed by individuals of East-Asian ethnicity in their . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 17, 2020. . community. In order to investigate to what degree dis-and misinformation about Covid-19 has affected the general public's beliefs about the condition, participants were directly asked whether or not they believed several falsehoods listed on the WHO's "myth busters" website, (9) which the WHO selected because they had been circulating on social media.(10) Specifically, the questionnaire asked whether receiving a letter or package from China poses a risk of infection, and whether using hand dryers, rinsing your nose with saline, eating garlic, applying sesame oil to the skin, taking antibiotics, and vaccinating against pneumonia are effective in preventing a Covid-19 infection. The questionnaire was built using Qualtrics software. Participants had to answer a question to reach the next question. Numerical entry questions did not allow for non-sensical inputs (e.g., percentage questions were restricted to inputs between 0 and 100). For binary and categorical response options, I computed the percentage of participants who selected each response. For Bernoulli trials, I calculated two-sided 95% confidence intervals using the Wilson score interval. (11) The use of sampling weights was not required for age, sex, and ethnicity because, given the sampling strategy, the sample is self-weighting by these variables. I computed and used sampling weights to weight participants to the estimated distribution of total household income and educational attainment in the general population in the US and the UK in 2019. (12, 13) Six types of data quality checks were performed. First, participants who took less than two minutes to complete the questionnaire were excluded from the analysis because I judged this to indicate random clicking. This resulted in the exclusion of one participant. Second, to further investigate the robustness of the findings to possible random clicking by respondents, I ran all analyses for each question when excluding the 10% and 20% of participants who took the least . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint time to answer the question. Third, if some respondents used random clicking to obtain the $1.50 reward as fast as possible, then a bimodal distribution in the time taken to complete the survey might be expected (with one study population clicking as fast as possible and one reading the questions). I, therefore, plotted a histogram of the time taken to complete the survey. Fourth, I checked whether any participants always selected the same response option (e.g., always the second option in multiple choice questions), alternating response options (e.g., the first option followed by the second and then again the first), or steadily increasing or decreasing response options (e.g., the first, followed by the second, then the third option, and so on), which was not the case for any participants. Fifth, participants were asked at the end of the questionnaire whether they looked up any answers online ("It is natural to be tempted to look up the answer to a question, especially when it's only a click away. For approximately how many of the questions did you first look up the answer on Google or somewhere else before responding? The answer to this question will not affect your payment in any way.") and if so, for which question. Those who self-reported looking up an answer online for a question were excluded from the analysis for that particular question. This was the case for 81 US participants and 63 UK participants who reported looking up the answer online on a median of 1 (IQR: 1 -5) and 1 (IQR: 1 -2.5) questions, respectively. Sixth, based on the rationale that this may indicate having looked up an answer online, I ran the analysis for each question when excluding the 10% and 20% of participants who took the longest time to answer the question as a further robustness check. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . Out of 3,000 adults residing in the US and 3,000 residing in the UK who could participate, 2,987 and 2,978 adults, respectively, completed the questionnaire (response rate of 99.6% for US and 99.3% for UK adults). Approximately two thirds of US participants and one half of UK participants had a tertiary education degree ( Table 1) . 67.5% (2,015/2,987) of US participants had a total household income between $20,000 and $99,999. 74.4% (2,215/2,978) of UK participants had a total household income between £15,000 and £74,999. 17.3% (516/2,987) of US participants and 13.7% (408/2,978) of UK participants were current students. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) .5%) of participants thought that the number of fatalities from Covid-19 in their country will be 500 people or less by the end of 2020 (Figure 1 ). . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. When asked what percent of individuals infected with Covid-19 experience a fatal disease course, the median estimate given by participants was 5.0% (IQR: 2.0% -15.0%) among US participants and 3.0% (IQR: 2.0% -10.0%) among UK participants. The full distribution of responses, as well as a magnification of the distribution of responses among those who estimated a risk of death 10%, is shown in Figure 2 . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. When asked "when they have been infected, what age groups are most likely to die from the illness caused by the new coronavirus" and presented with the option to select 'children', 'young . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. adults with other health problems were more likely to experience a fatal disease course than those without any other health problems. Most participants in both the US and the UK recognized fever, cough, and shortness of breath as three common symptoms and signs of a Covid-19 infection (Figure 3) . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint Figure 3 . Percent of participants who replied with "yes" to whether each of seven symptoms or signs were common in a Covid-19 infection1 1 The horizontal black bars depict 95% confidence intervals using Wilson's method. (11) When asked "if you have a fever or cough and recently visited China, or spent time with someone who did, what would be the best course of action?", 64.2% (95% CI: 62.4% -65.9%) of US participants and 64.3% (95% CI: 62.5% -66.0%) of UK participants responded with the recommended care-seeking option of staying home and contacting their health system. About a third of respondents stated that they would either delay care-seeking, attend the hospital . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint emergency department unannounced, or take a taxi or public transport to their primary care provider (Figure 4) . More than 90% of participants in both the US and the UK thought that washing your hands, avoiding close contact with people who are sick, and avoiding touching your eyes, nose, and mouth with unwashed hands were effective in preventing a Covid-19 infection (Figure 5) . However, a substantial proportion of participants also thought that using a hand dryer, rinsing . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint your nose with saline, taking antibiotics, and gargling mouthwash were effective prevention measures. 37.8% (95% CI: 36.1% -39.6%) of US participants and 29.8% (95% CI: 28.2% -31.5%) of UK participants agreed with the following statement: "Consistently wearing a face mask is highly effective in protecting you from getting infected with the new coronavirus. For the purpose of this question, "highly effective" is defined as reducing your risk of getting infected by >95% and a "face mask" is a common medical mask." 1 The horizontal black bars depict 95% confidence intervals using Wilson's method. (11) . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. 6 ). . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . More than half of both US and UK participants estimated that the prevalence of Covid-19 among East-Asian individuals in their country was greater than 0.5% (Figure 7) . This percentage rose to more than 60% for both US and UK participants when asking about the prevalence of Covid-19 among "adults of East-Asian ethnicity in your neighborhood who wear a face mask". . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. 39.0% -42.5%) of UK participants responded with 'sometimes', 'often', or 'always' (Figure 8) . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint For none of the questions in the questionnaire did the results change to a degree that it substantively altered the conclusions when excluding the 10% and 20% of respondents who took the least and the most amount of time to answer a question. There was also no evidence of a bimodal distribution in the time taken to complete the survey (Figure S3) . . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint The general public in both the US and the UK holds several important misconceptions about Covid-19. Participants in both countries generally overestimated the probability of a fatal disease course among those infected with Covid-19 (the case fatality rate is currently believed to be between one and two percent among reported cases, (1, 14, 15) but could be substantially lower if there are many unreported and/or asymptomatic cases (16) ), thought that children were at an especially high risk of death from Covid-19 (which is currently not believed to be the case (1, 15-17) ), and believed that common surgical masks are highly effective in protecting them from catching a Covid-19 infection. Participants also vastly overestimated the prevalence of Covid-19 among East-Asian individuals in their communities. Relatedly, a substantial proportion thought that they should refrain from frequenting Chinese restaurants, stated that they would refuse Uber rides to individuals of East-Asian ethnicity, and perceived that receiving a package from China poses a risk of a Covid-19 infection. In general, differences in knowledge and misperceptions between US and UK participants were small. The findings of this study could be used to set priorities in information campaigns on COVID-19 by public health authorities and the media. Specifically, such information provision could emphasize the comparatively low fatality rate, the low risk posed by individuals of East-Asian ethnicity living in the US and the UK, and that children do not appear to be at a heightened risk of contracting or dying from Covid-19. It is also important to note that while the general public appears to be well informed regarding the common symptoms of Covid-19, over a third of participants selected a healthcare-seeking option that could lead to further transmission of Covid-19. Thus, clear messaging on the recommended care-seeking action when experiencing some of the core symptoms of Covid-19 will be crucial. In addition, to ensure that individuals focus their attention on those prevention measures that are most effective, this study suggests . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint that it will be important to inform the public about the comparative effectiveness of common surgical masks versus frequent and thorough handwashing and avoiding close contact with people who are sick. Public health information campaigns may also want to directly target some of the mis-and disinformation that appears to have circulated on social media. (6, 9, 10, 18) Such measures could include information that rinsing your nose with saline, using a hand dryer, taking antibiotics, and gargling mouthwash are not effective prevention measures, and that receiving a letter or package from China does not pose a risk of Covid-19 infection. These are all falsehoods listed on the WHO's "myth busters" website that a substantial proportion of participants in this study believed.(9) More broadly, this study underscores the need for the WHO and other public health bodies to continue working with social media campaigns to minimize the circulation of inaccurate information about Covid-19. In line with recent media reporting that this conspiracy theory has been actively spread on Twitter,(18) about one in five participants believed it to be 'slightly likely', 'moderately likely', or 'extremely likely' that SARS-CoV-2 is a bioweapon developed by a government or terrorist organization. Participants did not expect that a large number of individuals would die from Covid-19 in their country by the end of 2020. This finding may be surprising considering that fear-inducing headlines in the media may (at least up to a certain extent (19) ) result in more attention by readers than more emotionally neutral ones, which could result in a catastrophizing of the epidemic. Moving forward, information campaigns on Covid-19 may need to balance the messaging of two important facts about the Covid-19 epidemic that could be interpreted by the general public to stand in direct conflict with each other. That is, on the one hand, the case fatality rate of Covid-19 appears to be lower compared to other recent infectious disease outbreaks, such as Ebola, (20) the severe acute respiratory syndrome (SARS), (21) and the . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. The copyright holder for this preprint this version posted March 17, 2020. This study has several limitations. First, while the sample of participants is representative of the US and UK general population by age, sex, and ethnicity, and the distribution of participants by household income and education was similar to that in the US and UK general population (and weighted in the analysis to be exactly representative by household income and education), participants may still differ from the general population on a variety of other characteristics. These characteristics may be both correlated with their knowledge and perceptions of Covid-19 as well as with their decision to participate in the study and/or to create a profile with Prolific. However, conducting a nationally representative household survey in the time frame needed to yield useful insights for a rapidly developing infectious disease epidemic is infeasible. Similarly, random-digit-dial phone surveys tend to suffer from very low response rates (typically well below 10%(23)), which can be a major source of bias even when sophisticated weighting adjustments are undertaken. (24) Second, the estimates of discrimination against individuals of East-Asian ethnicity may be an underestimate because some participants may not have wanted to volunteer their discriminating tendencies to themselves or to the researcher. However, I as the researcher had no access to any identifying information about the research participants and participants were reminded of this fact prior to answering the question. In addition, such social desirability bias has been found to be lower in online surveys than in telephone or in-person surveys. (25, 26) Third, it is possible that some participants may have randomly selected responses to spend the least amount of time in order to earn the $1.50 reward. I believe this is unlikely to be an important source of bias because i) only one individual completed the survey in under two minutes (while it was physically possible to complete it in well under 90 seconds), ii) the results for each question were robust to excluding the 10% and 20% of survey respondents . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 17, 2020. . who took the shortest amount of time to complete the question, iii) there was no bimodal distribution in the time taken to complete the survey, and iv) $1.50 is a relatively small monetary incentive suggesting that earning the reward was probably not the main motivation for participating in the study. Lastly, it is possible that individuals looked up the answers to some of the questions online prior to answering, which may have biased the results from factual (rather than opinion-focused) questions. Participants were reminded of the importance not to look up answers online prior to taking the survey and were asked at the end of the survey (while being reassured that their payment is not influenced by whether they volunteer information on having looked up an answer online) for which, if any, questions they searched for an answer online prior to responding. Covid-19, which should be targeted in information campaigns organized by government agencies, information provision by clinicians to their patients, and news coverage furnished by the media and social media platforms. Ensuring that the general public is well informed about Covid-19 could reduce unnecessary anxiety as well as reduce transmission of the virus and, thus, ultimately save lives. PG was supported by the National Center for Advancing Translational Sciences of the National Institutes of Health under Award Number KL2TR003143. This study received an exemption for minimal risk research by the Stanford University Institutional Review Board on February 22 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted March 17, 2020. . https://doi.org/10.1101/2020.03.13.20035568 doi: medRxiv preprint Report of the WHO-China Joint Mission on Coronavirus Disease 2019 (COVID-19) An interactive web-based dashboard to track COVID-19 in real time World Health Organization. Statement on the second meeting of the International Health Regulations (2005) Emergency Committee regarding the outbreak of novel coronavirus (2019-nCoV) The Health Belief Model: A Decade Later How to fight an infodemic. The Lancet Prolific Academic Ltd. 2020;Pages. Accessed at Prolific Academic Ltd Fights a Pandemic Besides Coronavirus: an 'Infodemic'. The New York Times Probable Inference, the Law of Succession, and Statistical Inference Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72314 Cases From the Chinese Center for Disease Control and Prevention Clinical Characteristics of Coronavirus Disease 2019 in China Covid-19 -Navigating the Uncharted Millions of tweets peddled conspiracy theories about coronavirus in other countries, an unpublished U.S. report says. The Washington Post Selective Exposure to Health Information: The Role of Headline Features in the Choice of Health Newsletter Articles Ebola virus disease SARS and MERS: recent insights into emerging coronaviruses Coronavirus: covid-19 has killed more people than SARS and MERS combined, despite lower case fatality rate How Social Processes Distort Measurement: The Impact of Survey Nonresponse on Estimates of Volunteer Work in the United States Social Desirability Bias in CATI, IVR, and Web Surveys: The Effects of Mode and Question Sensitivity