key: cord-270669-hng8ivz2
authors: Li, R.; Yang, B.; Penn, J.; Houghtaling, B.; Chen, J.; Prinyawiwatkul, W.; Roe, B.; Qi, D.
title: Perceived vulnerability to COVID-19 infection from event attendance: Results from Louisiana, USA, two weeks preceding the national emergency declaration
date: 2020-04-06
journal: nan
DOI: 10.1101/2020.04.01.20049742
sha: 
doc_id: 270669
cord_uid: hng8ivz2

In response to the mounting threat of COVID-19, we added questions to an ongoing food preference study held at Louisiana State University from March 3-12 of 2020. We asked 356 participants: (1) In your opinion, how likely is it that the spread of COVID-19 (the coronavirus) will cause a public health crisis in the United States? (2) How concerned are you that you will contract COVID-19 by attending events on campus? Participants' estimates of an impending national health crisis increased significantly during the study's second week (March 9-12) while concern about personally contracting COVID-19 from attending campus events increased only marginally during the study's final days. We find those expressing a higher likelihood of an impending national crisis were more concerned about contracting COVID-19 by attending campus events, suggesting a possible transmission from perceptions of national-level events to perceived personal vulnerability via local exposure. However, about 30% of participants perceived that COVID-19 would likely cause a public health crisis yet did not express concern about contracting COVID-19 from event attendance. These participants were significantly more likely to be younger students who agreed to participate in response to recruitment using same-day flyer distribution. Women expressed a higher likelihood of an emerging national health crisis, although they were not more concerned than men that attending campus events would result in virus contraction. Other groups (e.g., white, students younger than 25, highest income group) displayed similar concern about a national-level crisis, yet were significantly less concerned about contracting COVID-19 from attending campus events than others. Also, participants randomly assigned to information emphasizing the national impacts of food waste expressed significantly greater concern of contracting COVID-19 by attending campus events. These results provide some initial insight about how people perceived national and personal risks in the early stages of the COVID-19 crisis in Louisiana.

official announcements were made regarding the cancellations of any on-campus activities 1 1 0 before the end of our last study session (2: 00 PM March 12 th ) [9] . At 4:00 PM on March 12 th , on-campus classes starting from the week of March 16 th [10] , and then announced the cancellation of non-class activities involving 30 people or more immediately at 11:30 AM on 1 1 4 March 13, 2020 [11] . For reference, a national emergency was declared in response to COVID- 1 1 5 19 the afternoon of March 13, 2020 [12] . The sample includes the 356 participants enrolled from March 3 through March 12, 2020. Individuals were recruited via pre-existing email recruitment lists, flyers circulated on campus, 1 1 9 advertising announcements on classes, and advertisements in university locations. Inclusion Results are analyzed in Stata (version 16). The focal variables relating to COVID-19 1 2 3 were captured using a 5-point Likert scale. When more convenient for exposition or analysis, 1 2 4 these responses are simplified into binary variables (very or moderately likely/concerned = 1; all 1 2 5 other responses = 0). We also define the variable National, Not Local to equal one when 1 2 6 participants think a national crisis is very or moderately likely but are neither very nor 1 2 7 moderately concerned about contracting the virus by attending campus events. Personal 1 2 8 characteristics included in the analyses include sex, age, student status (=1 if enrolled in 1 2 9 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint University classes, =0 otherwise), household income, race, health insurance status, recycling the study (e.g., March 3, March 4, etc.) is also controlled in all analyses. Descriptive statistics 1 4 2 for the variables appear in Table 1 . To model the Likert-scale response to the two COVID-19 perception questions, an 1 4 4 ordered logit regression model is estimated with the aforementioned explanatory variables. The improvement measure is used as the splitting criteria [13] . Three participants are omitted from 1 4 8 several analyses because of item non-response on at least one variable, leaving an effective 1 4 9 sample size of 353. Statistical significance was set at the 5% level with results at the 10% level CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint experimental element descriptions (S1 appendix and S2 appendix). Ethics Statement 1 5 9 This study was approved by the Louisiana State University AgCenter and Ohio State University Institutional Review Boards. All participants signed informed consent forms after nationally (line graph, right axis) and in Louisiana (bar graph, left axis) for the study period, while Figure 2 traces the daily averages among study participants for the two COVID-19 1 6 8 questions, while highlighting key events in the evolution of COVID-19 timeline for Louisiana. Specifically, the gray bar depicts the percent who responded that COVID-19 was likely 1 7 0 (moderately or very) to cause a national public health crisis while the black bars capture the percent that were concerned (moderately or very) that attendance at campus events would cause . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint Figure 1 shows the national case count went from less than 100 on the first day of the 1 7 4 study (March 3) to more than 1600 cases by the last day of the study (black line). Louisiana crisis timeline. No cases were identified and reported in Louisiana until the second 1 7 7 week of the study (bars, Fig 1) and LSU communications stated that no cases had been identified 1 7 8 on campus [9] . However, a lack of testing in the United States and in Louisiana likely 1 7 9 underrepresented the prevalence of COVID-19 at the time [19] . . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . respond moderately or very likely that COVID-19 will cause a national public health crisis and 1 8 8 black bars are the percent who respond moderately or very concerned that about contracting 1 8 9 COVID-19 from attending campus events. Public announcements occurred after daily study 1 9 0 hours, which ended by 2 PM central. Information sources: Centers for Disease Control and (binary version), and National, not Local over the experimental timeframe. National Likelihood 1 9 5 increased steadily through the study period, though even on the final day of the study, more than 1 9 6 10% of participants did not agree that a national crisis was likely. National Likelihood was 1 9 7 statistically greater than the first day of the study from March 9 to March 12, i.e., the entire Questions: (1) moderately or very likely that COVID-19 will cause a national public health crisis covariates. **, * denotes a statistical difference of the value on this date from the value for the 2 0 8 same variable on the first day of the study at the 5% and 10% level as determined by regression ( Table 2 ) that controls for personal and experimental factors. the United State and globally and expected more cases to be detected across the country, including more instances of person-to-person spread in more states [20] . Local Vulnerability cases increased more than 7-fold across the United States and participants' perceived National Likelihood increasd about 20 percent points. Local Vulnerability featured marginally significant CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint pattern (agreeing a national crisis was likely but not expressing concern about attending campus 2 2 2 events) stayed relatively constant over the period and featured no significant differences from the 2 2 3 first day of the study. Vulnerability variables in their Likert scale form (1 = very unlikely/unconcerned, …., 5 = very likely/concerned) and binary logit model was estimated for the National, not Local variable. author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint (food waste information, provision of compostable plates) were marginally significant.

Local Vulnerability was significantly and positively associated with National Likelihood

such that a participant's concern with contracting COVID-19 from attending campus events was Local Vulnerability question was asked immediately after the National Likelihood question. years or older (regardless of student status) and identifying with a race other than white or black.

Those in the highest income category ($100,000 or more) displayed significantly lower Local Vulnerability than those earning less than $15,000 per year. Participants that were randomly assigned the food waste information treatment (rather than the screen time information Only two variables were significant in the regression model for National, Not Local. Older (≥25 years), non-students were less likely to feature this response pattern than younger there any significant differences by the day of the study. . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint .

Rather than making predictions based on ceteris paribus regression coefficients, the 2 5 9 classification tree categorizes subjects based on splits from various predictor variables [21] . Vulnerability and the percent of participants falling into the branch. The bottom row features a 2 6 5 bar graph of the proportion in that branch that were very or moderately concerned about Local 2 6 6 Vulnerability and the number of participants in that branch. For example, 20% of participants contracting COVID-19 from attending campus events was 0.31. 4 shows the pruned classification tree with a misclassification rate of 41% (i.e., 59% prediction 2 7 2 accuracy). The first split is between those identifying as white versus other racial and ethnic those identifying as white and non-Hispanic, the study week in which they participated is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . 1 7 next branching variable, with those participating on March 10 th -12 th being sorted based on 2 7 6 whether they reported income less than $15,000 with the proportion in that lowest income group 2 7 7 expressing Local Vulnerability being twice as high. For those attending March 9 or earlier, those 2 7 8 randomly assigned to receive compostable plates were 72% more likely to express Local 2 7 9 Vulnerability than those randomly assigned to other treatments. Among those identifying with groups other than white, non-Hispanic, self-reported 2 8 1 recycling frequency was the next classification variable, with those reporting that they never food waste information on any date were also in a branch with high Local Vulnerability as were 2 8 4 those not receiving food waste information so long as it was on March 12 th , the final day of the As Poletti et al. [22] noted, the spread of epidemics can be dramatically delayed or population. Our analysis suggests that for LSU students, staff and faculty participating in a participants thought a national public health crisis from COVID-19 was moderately or very 2 9 5 likely even on March 3, the first day of our data collection, a date upon which the number of cases nationally was reported to be 80 [14] . Nine days later, when the number of reported 2 9 7 national cases exceeded 1600 [14] , nearly 90% of participants displayed National Likelihood, 2 9 8 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint .

with a significant jump in this perception compared to the first study day occurring at the 2 9 9 beginning of the second week of the study (March 9). National Likelihood that align with the previous literature, e.g., women perceive a national 3 0 2 public health crisis as more likely than men [23] [24] . Other significant associations have no those desiring to eat healthier have been exposed to wide-spread nutrition misinformation in the media, which has recently included unfounded claims that healthy eating or certain supplements products has resulted in federal warnings to several companies promoting unproven nutrition- based remedies and preventatives for COVID-19 [25] . This might indicate reduced perceived 3 1 3 risk of acquiring infectious disease among healthy eaters in our sample, based on these factors, 3 1 4 although further investigation is required. of risk perception to self-prophylaxis measures, such as social distancing, which slows 3 1 7 community spread. While we did not elicit explicit measures of such behaviors, we did assess 3 1 8 participants' perceived vulnerability to contracting COVID-19 from attending campus events, 3 1 9 (Local Vulnerability), which may signal a willingness to undertake social distancing and other 3 2 0 beneficial behaviors and be a behavioral precursor. The first insight from observing the raw data

CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint plot in Figure 2 is that Local Vulnerability persistently lags National Likelihood, and does not 3 2 2 significantly exceed the 50% mark until the last day of the study, which is the first day after the 3 2 3 state of Louisiana had declared a public health emergency, but before LSU had cancelled classes 3 2 4 or campus events. those identifying with races other than white and black are more likely to express Local 3 2 9 Vulnerability, while those in the highest income category expressed lower Local Vulnerability significantly higher risk from aggressive driving tactics than did younger drivers, while Lo [27] 3 3 3 found that higher income respondents expressed less concern about environmental risks, which 3 3 4 he hypothesized to stem from a heightened sense of material risk faced by those with lower 3 3 5 incomes. Flynn, Slovic and Mertz [28] found respondents identifying as white, particularly 3 3 6 white men, registered significantly lower environmental risk perceptions, hypothesizing that 3 3 7 socio-political factors including power and status may influence risk perceptions. Other characteristics typically identified in the literature (e.g., sex) are not significantly 3 3 9 associated with expressed Local Vulnerability. Interestingly, participants randomly assigned an we cannot provide a definitive explanation of this relationship given post-hoc design constraints, 3 4 3 we note the food waste information treatment was the only information treatment to emphasize 3 4 4 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint national level and household level implications of individual behavior (e.g., food waste causing 3 4 5 $161 billion of losses at the national level and $1500 of losses in an average household). Further, 3 4 6 the classification tree finds that, similar to the randomly assigned information about food waste, 3 4 7 participants who were randomly assigned compostable paper plates and the participant's 3 4 8 recycling habits also work as significant determinants of Local Vulnerability. One conjecture is 3 4 9 that participants who link the implications of individual behaviors to issues of sustainability may There is a persistent group consisting of about 30% of participants who, for the entire 3 5 3 study period, including the final day, do not translate their perceived likelihood of a national 3 5 4 public health crisis into personal vulnerability from attending campus events (National, not 3 5 5 Local). These are likely a critical group in terms of modeling diffusion of COVID-19, as Poletti 3 5 6 et al. [22] emphasize the role of translating perceived risk into preventative behaviors such as However, our analysis provides few insights into the characteristics associated with 3 5 9 National, not Local group. Regression analysis finds few significant associations other than the 3 6 0 fact that older non-students are less likely to feature this response pattern and that those who spontaneously attended the study in response to same-day receipt of flyers were more likely. The former suggests that younger people in academic settings may be diagnostic for predicting 3 6 3 this response pattern while the latter may be suggestive that certain personality traits have 3 6 4 predictive power. This lack of insight into the National, not Local group is likely due to the post-hoc nature 3 6 6 of the analysis, one of several study limitations. Specifically, the study was originally designed 3 6 7 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint we provide some insights into how participants drawn from one community in Louisiana were 3 8 0 perceiving the national and local implications of the public health crisis that was unfolding We find that perceptions during this critical time increased steadily and rapidly such that 3 8 4 nearly 90 percent of participants agreed that it was likely that COVID-19 would become a 3 8 5 national public health crisis by the final day of our study, which corresponded with the day that Louisiana declared a public health emergency. However, participants' views of their personal 3 8 7 vulnerability to contracting the virus from attending local events increased more slowly and, only 3 8 8 . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.01.20049742 doi: medRxiv preprint

Responding to Covid-19-A Once-in-a-Century Pandemic?

Covid-19: Are we getting the communications right?

Risk perception in epidemic modeling

The value of early response by surrounding areas of epidemic center Hubei during COVID-2019 outbreak in China: A quasi-experiment analysis. Available at SSRN 3548372

A meta-analysis of the association between gender and protective behaviors in response to respiratory epidemics and pandemics

Monitoring of risk perceptions and correlates of precautionary behaviour related to human avian influenza during 2006-2007 in the Netherlands: results of seven consecutive surveys

Perceptions and behaviors related to hand hygiene for the prevention of H1N1 influenza transmission among Korean university students during the peak pandemic period

Immediate psychological responses and associated factors during the initial stage of the 2019 coronavirus disease (COVID-19) epidemic among the general population in china

Message regarding coronavirus. LSU Coronavirus Updates and Information

LSU Coronavirus Updates and Information

LSU event canceled immediately. LSU Coronavirus Updates and Information

proclamation on declaring a national emergency concerning the novel coronavirus disease (COVID-19) outbreak. White House Proclamations

Classification and regression tree analysis in public health: methodological review and comparison with logistic regression

Cases in the US. coronavirus disease

State of Louisiana: Office of Governor. Reported Louisiana's first presumptive positive case of COVID-19

Confirmed 3 new presumptive positive COVID-19 cases. Office of Governor Newsroom

State of Louisiana: Office of Governor. Governor Edwards and U.S. Surgeon General giving latest update on coronavirus cases in LA

Centers for Disease Control and Prevention (CDC). CDC announces additional COVID-19 Infections. CDC Newsroom

Consumer research with big data: applications from the food demand survey (FooDS)

Risk perception and effectiveness of uncoordinated behavioral responses in an emerging epidemic

How much risk can you stomach? Individual differences in the tolerance of perceived risk across gender and risk domain

Gender differences in prescription opioid use and misuse: Implications for men's health and the opioid epidemic. Preventive Medicine

FDA send warning letters to seven companies about unsupported claims that products can treat or prevent coronavirus: Commission continues efforts to protect consumers from deceptive advertising. Press Releases

Age and gender differences in risky driving: The roles of positive affect and risk perception

Negative income effect on perception of long-term environmental risk

Gender, race, and perception of environmental health risks. Risk Analysis

Coronavirus outbreak: WHO sends message to young people "You're not invincible

to investigate a topic other than COVID-19 perceptions, hence logical experimental treatments 3 6 8 and additional questions about personal perceptions and behaviors relevant to understanding and 3 6 9 predicting the spread of COVID-19 were not included and the questions that were posed were 3 7 0 not motivated by theory. Another study limitation is that the sample is drawn from a single 3 7 1 academic institution, limiting the representativeness of the data geographically, demographically, 3 7 2 and socioeconomically. Finally, the data were acquired prior to the declaration of a national 3 7 3 emergency, and we would expect further evolution in how people in this location might respond 3 7 4 to these questions in the face of more dire national promulgations concerning the pandemic. By integrating questions focused on COVID-19 into an ongoing in-person experiment during the 3 7 8 two weeks prior to the major disruption in public activities in Louisiana and much of the country, 3 7 9