This study, which was conducted between 2015 and 2020, used the results of 4 surveys with a total of 4,011 participants. A sample size of each survey participants was estimated to ensure the 95% confidence interval (CI), 0.025 desired margin of error, 0.2 behavioral response proportion of population. The 4,011 participants included 2,009 participants older than 19 years who were monitored during the MERS-CoV outbreak between June 2 and 25, 2015; and 2,002 participants, older than 18 years, who were investigated during COVID-19 outbreak between February 4 and April 2, 2020. All surveys were conducted using mobile (85%) or landline (15%) random digit dialing numbers in 8 regions (nationally representative). Participants were selected post-stratification by sex, age, and province and chosen independently by each survey. The total number of weighted cases in this survey equaled the total number of unweighted cases at the national level. The weights were normalized to calculate the proportions and ratios but not for estimating the subtotal populations. Trained interviewers conducted all interviews using computer assisted telephone interviewing. Surveys 3 and 4 began approximately 2 weeks after the index case occurred, while surveys 1 and 2 were conducted approximately a month after surveys 3 and 4. Survey 1 was conducted just 10 days before the last confirmed patient of MERS-CoV on July 4, 2015. However, survey 2 was conducted when there were more than 100 confirmed cases. The surveys were conducted by Gallup Korea, an affiliation of Gallup International. Details, including period, number of respondents successfully interviewed, and response rate for each of the four surveys are provided in Table 1.

Sex, age, occupation, self-reported household economic status, residential area, presidential job approval rating, party identification, and affective risk perception as participants' characteristics, were investigated to identify factors influencing non-pharmaceutical preventive behaviors. Age was classified into 5 levels (19–29, 30s, 40s, 50s, and 60 years and older). Occupation was classified into five levels (unemployed, self-employed including farming/forestry/fishery, blue-collar worker, white-collar worker, and full-time homemaker or student). Self-reported household economic status was classified into five levels (lower, lower middle, middle, upper middle, and upper). Participants were classified as either metropolitan or non-metropolitan residents. Presidential job approval rating was assessed using the following options: “approval,” “disapproval,” or “no opinion.” Support for party identification was assessed based on alignment either with the ruling party, opposition party, or no opinion. Affective risk perception was assessed using the options “worried” or “not worried.”19

The interviews were conducted on the two aspects of the non-pharmaceutical preventive behaviors, which are social distancing measures and transmission-reducing practices (Supplementary Data 1 and 2). Social distancing was assessed using the following four questions: 1) “Did you reduce or avoid outdoor activities or attend meetings this week because of MERS-CoV or COVID-19?”; 2) “Did you reduce or avoid using public transportation such as the bus or the subway this week because of MERS-CoV or COVID-19?”; 3) “Did you reduce or avoid using healthcare facilities such as the hospitals or public health centers this week because of MERS-CoV or COVID-19?”; and 4) “Did you reduce or avoid visiting crowded markets, departmental stores, or large discount stores this week because of MERS-CoV or COVID-19?” Transmission-reducing practice was assessed using the following two questions: 1) “Do you wash your hands more often than usual because of MERS-CoV or COVID-19?” and 2) “Have you ever worn a face mask because of MERS-CoV or COVID-19?” All the questions about the non-pharmaceutical preventive behaviors required “yes/no” responses. The development of the questionnaires on preventive behaviors had not gone through a validity procedure due to the urgency of the outbreak. We also imposed the survey items on existing questionnaire developed by Gallup Korea, an affiliation of Gallup International.

Response rates according to preventive behaviors were calculated according to participants' characteristics. Univariable analyses using χ² test were performed in the four surveys, entirely and respectively, to identify the relationships between practicing preventive behaviors and each demographic variable. Missing values of any outcome variable were ≤ 3.6%. Multivariable logistic regression analysis was performed to explore factors influencing preventive behaviors in the four surveys, entirely and respectively. We performed multivariable logistic regression model adjusted for sex, age, occupation, self-reported household economic status, residential area, presidential job approval, and party identification.15 Affective risk perception was excluded from survey 1 and survey 2 logistic regression models to attain comparability because no data existed for it in survey 2. On the avoidance of outdoors activities, extremely large number of events made the odds ratios (ORs) in survey 2 logistic regression model unstable; therefore, surveys 1 and 2 logistic models were not reported. Using logistic regression analysis for transmission-reducing measures and social distancing measures, “y = 1” was used when “yes” for preventive behaviors, otherwise “y = 0” was used. We analyzed with a 2-sided P value of less than 0.05 considered significant using SAS version 9.4 (SAS Institute Inc, Cary, NC, USA).

This study was reviewed and approved by the Institutional Review Board (IRB) of Seoul Metropolitan Government-Seoul National University Boramae Medical Center (IRB No. 20200403/07 - 2020 - 12/043). The need for informed consent was waived by the board.

Differences in participants' general characteristics between surveys 1 and 2 are shown in Table 2. Overall, the practice rate of avoiding outdoor activities in survey 2 increased 1.7-fold compared to that in survey 1. Depending on the general characteristic, avoiding outdoor activities' practice rate differed by as little as 28.2% (upper economic status) and as much as 54.6% (presidential job approval). Overall, avoiding public transportation practice rate in survey 2 increased 2.1-fold compared to that in survey 1. Depending on the general characteristic, avoiding public transportation' practice rate differed by as little as 29.3% (aged 30–39 years) and as much as 57.6% (no opinion of presidential job approval). Overall, avoiding healthcare facilities' practice rate in survey 2 increased 1.6-fold compared to that in survey 1. Depending on the general characteristic, avoiding healthcare facilities' practice rate differed by as little as 22.7% (opposition party identification) and as much as 49.9% (presidential job approval). Overall avoiding crowded places' practice rate in survey 2 increased 1.8-fold compared to that in survey 1. Depending on the general characteristic, avoiding crowded places' practice rate differed by as little as 4.4% (upper economic status) and as much as 49.3% (presidential job approval). There were no statistically significant differences between surveys with participants' characteristics except with occupation, self-reported household economic status, presidential job approval rating, and party identification. With occupation, higher proportions occurred in the unemployed and blue-collar workers in survey 2, while lower proportions occurred in white-collar workers and home makers or students. Of the self-reported household economic status, survey 4 had higher proportions in the ‘upper middle’ and ‘middle’ status, while lower proportions occurred in ‘low middle’ and ‘lower’ status. With respect to the presidential job approval rating, the percentage of participants who reported obtaining ‘approval’ increased in survey 2 compared to survey 1. Of the party identification, the proportion in the ‘ruling party’ in survey 2 were higher than that in survey 1.

Comparison of the general characteristics of the participants between surveys 3 and 4 are shown in Table 3. Overall, wearing of face mask rate in survey 4 had increased by more than 5-fold compared to that in survey 3. Depending on the general characteristic, the wearing of face mask rate differed by as little as 39.2% (upper economic status) and as much as 71.3% low middle economic status). Overall, the washing of hands rate in survey 4 increased by 1.3-fold compared to that in survey 3. Depending on the general characteristic, washing of hands rate differed by as little as 12.0% (aged 19–29 years) and as much as 37.5% (upper economic status). There were definitively, statistically significant differences in the wearing of face masks rate in all subgroups between surveys 3 and 4. No significant differences occurred in participants' proportions between surveys, except with occupation, self-reported household economic status, presidential job approval rating, and party identification. With occupation, survey 4 had higher proportions of the unemployed and blue-collar workers, while had lower proportions of the self-employed and home makers or students. Of the self-reported household economic status, survey 4 had a higher proportion of those in the ‘middle’ status, while had a lower proportion of those in the ‘low middle.’ With respect to presidential job approval rating, the percentage of participants who reported ‘approval’ increased when survey 4 was compared with survey 3. Of the party identification, the proportion in the ‘ruling party’ in survey 4 were higher than that in survey 3.

Table 4 reports the association between variables and non-pharmaceutical preventive behaviors, social distancing and transmission-reducing behaviors.

Of social distancing behaviors, generally, none of the factors (characteristics) consistently affected any kinds of social distancing behaviors (avoiding public transportation, healthcare facilities, and crowded places) in both surveys 1 and 2. The results showed that sex, presidential job approval rating, and party identification were significantly associated with social distancing behaviors in survey 1, but not in survey 2. Those aged 30 years and older were more likely to avoid public transportation in both surveys 1 and 2. Participants aged 30–39 years were more likely to avoid healthcare facilities in survey 2 only. Those aged 30–49 years were more likely to avoid crowded places in survey 1 only. Only in survey 2 were residents of metropolitan cities identified to have practiced avoidance of public transportation behaviors less.

In transmission-reducing behaviors, both surveys 3 and 4 reported that females were more likely to practice preventive behaviors (adjusted OR [aOR], 1.57–2.43; 95% CI, 1.04–3.78), which tended to be stronger in survey 4. The association of affective risk perception with transmission-reducing behaviors was also observed in both surveys 3 and 4. Participants who reported being ‘worried’ were more likely to practice both the wearing of face masks and handwashing (aOR, 3.24–4.81; 95% CI, 1.76–6.96). Those living in metropolitan cities more frequently wore face masks in both surveys 3 and 4. Participants aged 50 years and older practiced less wearing of face masks in survey 4 only. The results showed that presidential job approval rating and washing of hands were significant in survey 3, but not in survey 4.