We administered a cross-sectional survey. To evaluate the public’s health behaviour responses to COVID-19, we conducted an anonymous online survey. To increase the representativeness of the sample, a proportional stratified sampling that reflected age, gender, and population region in the sample quota ratios was used. The number of participants was set based on the composition of the registered resident population announced by Statistics South Korea in February 2020. A total of 1406 patients visited the online survey between April 14 and April 20, 2020.
In South Korea, the first confirmed COVID-19 cases occurred on 4 January 2020, which prompted the Korea Disease Control and Prevention Agency to implement the ‘New Normal Level’ and strengthen their surveillance . When the fourth confirmed case occurred on 28 January 2020, the KCDC scaled up the alert level and conducted publicity campaigns about taking preventive behaviours against infectious disease. During the month of February, the number of confirmed cases increased radically as the new infectious disease rapidly spread nationwide, even at the local community level. During this period, the health authorities conducted various campaigns on personal preventive behaviours through posters, digital images, and text messages. In particular, they suggested specific health behaviours according to place and time in order to induce people to adhere to health behaviours.
The data for the analysis were collected by an online research company named “TRUIS”, which maintains 420,000 online panels . According to the composition of the registration population announced by the National Statistics Office in February 2020, we set gender, age group, and regional quota ratios. Before beginning the survey, TRUIS set gender, age group, and regional quota ratios based on registration population data for impartial analysis. At first, 1406 people had accessed the online survey. A total of 102 participants exceeded the quota ratio, so we excluded them from the survey. For example, assuming that 1406 subjects were subject to the survey, 165 people were allocated to the quota when applying the ratio of the number of men in their 20s. Thus, if 200 men in their 20s answered the questions, then the results of the survey could be biased. This means that, in this example, 35 respondents would need to be excluded from the sample. Ultimately, a total of 1304 respondents completed the questionnaire; however, among them, 66 respondents did not complete the survey, and 31 respondents did not provide consistent responses. Therefore, we concluded that 1207 respondents were credible based on the quota ratios, which resulted in a response rate of 92.5%.
The final sample size was 1207, with a considerable margin error of 2.82% and a 95% confidence interval. Since this study analysed people’s autonomous actions or responses to COVID-19, which means that they needed to be able to decide their actions on their own, it was important to choose adults as the respondents. For this reason, the survey decided to provide comprehensive information about the adult population in the age bracket of 20–59 years. Prior to the survey, participants agreed to the provision that the contents and purpose of this study were understood and that they were willing to participate in the study. Anonymous participation was strongly mandated, and no identifiable information was collected from the respondents.
To evaluate the degree of the respondents’ adherence to COVID-19 preventive behaviours, we analysed their responses to the personal preventive measures recommended by the World Health Organization (WHO). The WHO developed a comprehensive strategy to control COVID-19 that is made up of a list of actions recommended for individuals, communities, governments, and international bodies to suppress the spread of the SARS-CoV-2 virus . Of these actions, we focused on the individual aspects of the preventive measures to assess the respondents’ beliefs and perceptions concerning preventive behaviours regarding COVID-19. Consequently, we used five items, namely, frequent hand hygiene, respiratory etiquette, wearing a mask, environmental cleaning at home, and self-quarantine. The answers were rated on a 7-point scale ranging from 1 = strongly disagree to 7 = strongly agree. The total value of the precautionary behaviours was calculated by averaging the scores of each of the questions. To measure internal consistency, a reliability analysis was carried out on the preventive behaviours scale comprising 5 items. The Cronbach’s alpha value for the survey was .75, which indicated an acceptable level of reliability.
By building upon the HBM from previous literature, we developed a total of eight categories of determinants that influenced the preventive behaviours taken towards preventing COVID-19. The structured variables covered sociodemographic information, perceived susceptibility, perceived severity, perceived benefit, perceived barrier, self-efficacy of preventive behaviours, and cues to take action. In particular, the sociodemographic characteristics of the survey participants included gender, age, education level, monthly household income, and marital status.
The second part of this study was based on the HBM. The study participants were asked to provide their opinions on specific statements. Perceived susceptibility, severity, benefits, and barriers were each evaluated. To measure the HBM factors, except for self-efficacy, the respondents are asked to answer the two separate questions. The final scores from each factor were obtained by averaging each score. If the final score was above the average score, this was considered indicative of each factor being at a high level. Perceived susceptibility refers to one’s belief regarding the possibility of being infected (e.g., If I do not take precautions, I think I will be more likely to be infected with COVID-19). Perceived severity refers to one’s belief in the seriousness of the infection (e.g., If I am infected with the SARS-CoV-2 virus, it will impact me severely) . Perceived benefits refer to the efficacy of preventive behaviours in reducing the risk of being infected by the SARS-CoV-2 virus (e.g., If I follow the preventive behaviours, doing so will reduce the risk of getting infected with COVID-19). In contrast, perceived barriers represent the obstacles that inhibit the implementation of preventive behaviours (e.g., It is annoying and uncomfortable to follow preventive behaviours) . The answers were scored on a scale ranging from 1 to 7 (1 = strongly disagree, 7 = strongly agree).
Self-efficacy refers to an individual’s confidence in successfully carrying out preventive health behaviours for the prevention of COVID-19 (e.g., I am able to follow the preventive behaviours) . The survey participants were asked to assess their self-efficacy through a question, and they were asked to indicate their level of agreement using a 7-point Likert scale.
Finally, the HBM assumes that people are set in motion through cues to take action. These cues to take action trigger individuals to take action by using various sources . We chose seven items to evaluate the survey participants’ trust cues that could affect their preventive behaviours. The respondents were asked to indicate how much they trusted the following sources of information with regard to the information provided about COVID-19: printed media, radio, television, health care providers, official government website, social networks, and family and friends. The answers were scored from 1 (do not trust at all) to 7 (trust completely). The scores were obtained by averaging the scores of the seven questions. To measure internal consistency reliability, we calculated the Cronbach’s alpha coefficient on items of the Health Belief Model and cues to take action. The Health Belief Model subscale consisted of 9 items and α = .71, while the cues to take action subscale consisted of 7 items and α = .79. Each of the Cronbach’s alpha values showed that the questions reached acceptable levels of reliability.
A descriptive analysis was conducted to illustrate the general characteristics of the study sample using the frequencies and percentages of the categorical variables. We conducted single and multiple linear regression analyses to identify the factors that affected the respondents’ health behaviours towards COVID-19 prevention. The data were analysed using IBM SPSS software version 22.