The study was approved by the National Health Sciences Research Committee (NHSRC), an Institutional Review Board of the Ministry of Health of Malawi (NHSRC Protocol No.20/08/2598). This study conformed to the ethical guidelines in the Declaration of Helsinki. Written informed consent was obtained from all participants.

This was a cross-sectional survey that collected data through personal one-on-one interviews conducted in communities from nine districts in all three regions of Malawi: Blantyre, Lilongwe, Mangochi, Mzimba, Chitipa, Zomba, Chikwawa, Dowa, and Nkhatabay. The survey was conducted between 5 October 2020 and 25 October 2020.

Districts were selected for inclusion using the risk classification system adopted in the National COVID-19 Preparedness and Response Plan for the period March to June 2020.¹⁴ This risk classification system was based on the presence of an international airport and the volume of people traveling from high-risk transmission areas into the district of interest. Nine districts were included, each belonging to one of the three categories of risk. “Category 1” districts (Blantyre, Lilongwe, and Mangochi) were those with international airports and a potential of daily passengers from COVID-19-affected countries. “Category 2” districts (Chitipa and Zomba) were areas with a high volume of travelers using ground crossing routes from other affected countries. “Category 3” districts (Chikwawa, Dowa, and Nkhatabay) were areas classed to have the risk of local transmission only. Data collection officers were identified through the respective District Health Offices of each of the nine aforementioned districts.

Participants were resident adults of the communities in the selected districts who were 18 years and older, and who consented to taking part in the survey.

The Likert scale-based questionnaire used here was adapted from another one in a similar study conducted in Hong Kong.^12,15 This questionnaire had four components: i) socio-demographic characteristics; ii) risk perceptions towards COVID-19; iii) preventive behaviors; iv) willingness and ability to self-isolate.

Socio-demographic characteristics consisted of sex, age, ethnicity, marital status, caring responsibilities, area of residence, and socio-economic status (SES). SES was assessed using five indicators: education level, employment status, household income, savings, and household tenure.

Risk perceptions towards COVID-19 were measured by perceived susceptibility and perceived severity. Susceptibility was measured by asking respondents about the perceived likelihood of being infected with COVID-19 under the Government’s current preventive measures. Severity was measured by asking respondents about the perceived seriousness of symptoms if they were infected with COVID-19.

Preventive behaviors included information on perceived effectiveness and actual adoption of preventive behaviors (to protect oneself and others), to prevent both contracting COVID-19 and onward transmission, and were collected under three categories: i) hygiene practices (wearing a face mask, washing hands more frequently with soap and water, using hand sanitizer more regularly, covering nose and mouth when sneezing or coughing); ii) travel avoidance (travel to affected countries and travel to areas inside and outside Malawi, regardless of whether they were affected); iii) social distancing (avoiding public transport, social events, going out in general, going to the hospital or other healthcare settings, crowded places, and contact with people who have a fever or respiratory symptoms).

Data were collected through personal one-on-one interviews, where data collectors interviewed participants found in communities. Participants were identified using systematic sampling where the 3^rd participant was selected for enrolment.

Sample size calculations were based on the findings of a similar study from Kenya¹¹ which found that the probability of having knowledge of the novel coronavirus disease was 97%, with a 26% proportion of survey non-response. Thus, at a 95% confidence interval, a limit of precision of 5%, and a design effect of 1.0, the calculated sample size was 57 participants per district, giving a total of 510 participants.

The primary study objective was to investigate the knowledge, perceptions, and behaviors towards the COVID-19 pandemic in the Malawian community. The secondary objective was to investigate the barriers to the uptake of COVID-19 preventative measures in this community.

Statistical analyses were performed using R (version 4.0.0), 2020 (Vienna, Austria: R Foundation for Statistical Computing), and Stata. Participant distribution was described using mean and standard deviation for numerical data, and proportions for categorical data. Baseline characteristics were compared between males and females using Chi-square tests and Fisher’s exact tests. Perception was scored numerically from 1 to 5 denoting one of the following responses respectively: strongly disagree, disagree, neutral, agree, strongly agree. The scores were summed up and a multivariate regression analysis was used to examine factors associated with a high perception score (defined as a score of 45 to 60).

Table 1 describes the baseline characteristics of the participants who took part in the survey. A total of 521 participants were interviewed, 98.8% (507/521) of whom were Malawian nationals. The overall proportion of women was 53.9% (281/521). The commonest age group of the respondents was 25-34 years (43.1%, 224/521). There was no significant difference in the age distribution between male and female respondents.

Most of the participants (88.2%, 480/521) had at least some level of education, though only one in five participants (21.6%, 112/521) had had some tertiary level training. The proportion of respondents who were unemployed was 27% (137/521), and female respondents had a significantly higher proportion of unemployed individuals (36.1%, 97/281) compared to male respondents (16.9%, 40/240). Commonly, participants were unable to work from home (46.3%, 224/521), amongst whom there were significantly more males (55.2%, 123/240) than females (38.2%, 99/281).

Most participants (70.6%, 357/521) had very good self-reported health, with only one in five (19.3%, 99/521) participants reporting having a chronic non-communicable disease. The commonest non-communicable disease reported was hypertension (6.15%, 32/521) followed by asthma (4.2%, 22/521). There was no significant difference in the prevalence of self-reported chronic diseases between males and females, though males were more likely to report very good health (75.4%, 175/240) than females (66.5%, 181/281).

All respondents were aware of the ongoing COVID-19 pandemic. Table 2 summarises the sources of information on COVID- 19 among the study participants. The radio was the main platform [overall proportion of 32.9%, 95%, confidence interval (CI) 30.4-35.6%] through which participants gained information on COVID- 19 across all age ranges and sexes. Other common sources of information were family and friends (22.3%, 95% CI 20.1-24.7%), television (17.4%, 95% CI 15.4-19.7%), and social media (15.3%, 95% CI 13.4-17.4%). Less than 1% of all respondents (0.8%, 95% CI 0.4-1.5 %) would have heard about COVID-19 from either a health facility, their local church, an information car, or their local Chief. Figure 1 additionally shows the sex- and age-stratified information sources of COVID-19.

Figure 2 describes the knowledge of participants on COVID- 19 symptoms and interventions when symptomatic. Three-quarters of all participants (75.2%, 95% CI 71.4-78.7%) displayed knowledge of all three key symptoms of COVID-19 (cough, fever, and shortness of breath), though female participants aged 35 and below had higher proportions of those who know all symptoms than older age groups. The highest proportion of knowledge of all symptoms was displayed by male respondents aged 45-54 years (85.4%).

The majority of respondents (97%, 507/521) had an idea of existing interventions that could be sought upon developing any symptom suggestive of a COVID-19 infection. When asked about which intervention was appropriate, 51.6% (270/521) responded “stay-at-home and call a hotline” while 45.3% (237/521) suggested “visiting the nearest hospital”.

Responses on the perceived threat of COVID-19 to the study participants are summarised in Figure 3. Overall, 80.6% (420/521) of participants were worried about COVID-19, though only twothirds (63.3%, 330/521) of the respondents believed that under the current preventative measures, they would most likely get infected. With regards to the severity of the disease they would have if infected with the coronavirus, 50.3% (262/521) believed they may die, while only 34.9% (182/521) believed they may have asymptomatic disease.

Perception of COVID-19 was significantly associated with the district where the respondent resides and the participant’s gender (Table 3). The odds of male respondents having a high perception score was 6.3 (95% CI 1.9-21.0) times higher than female respondents. Participants from all other districts apart from Chitipa had higher odds of a high perception score than Blantyre. No other factor demonstrated an association with the perception of the disease.

Most of the participants believed that the prescribed preventative measures were effective in preventing COVID-19 (Figure 3). Practicing hand hygiene with soap was perceived by the highest number of participants (90.2%, 470/521) as an effective measure. The use of hand sanitizers had the highest number of participants (18.6%, 97/521) who were unsure about its effectiveness in preventing coronavirus transmission. Table 4 displays the number of preventative measures practiced by participants based on their cues to action. Only one participant reported not observing any COVID- 19 preventative measures. The majority of participants practiced all five preventative measures (160/521, 30.7%), and mostly reported adopting them for the reason of “preventing the spread of COVID-19 to others” (74/160, 46.2%). Participants who adopted preventative measures because “they comprehended the importance” or were “following friendly advice” had the least number of respondents who adopted all five measures [4/160 (2.5%) and 3/160 (1.9%) respectively]. When asked about adherence to COVID-19 home isolation if infected, 89.1% (464/521) of participants were willing to isolate (Figure 4). Conversely, various barriers to isolation were highlighted. Most [79.3% (413/521)] of the participants believed that isolation would lead to loss of income for the isolated individual, 66.8% (348/521) noted that isolation would make it difficult to get supplies such as food and medicines, 65.5% (341/521) noted that isolation may affect their mental health and 62.2% (324/521) noted that it would be difficult to separate self from others in the household.

This study has two main limitations. Firstly, the study tool predominantly featured closed-ended questions, therefore, it was difficult to adequately explore the depth of the responses given. Secondly, some participants may have given socially desirable responses in line with government recommendations rather than actual opinions. Additionally, the fast-changing landscape of the COVID-19 pandemic means it is difficult to accurately describe present-day perceptions of the disease. Further studies are thus required to explore attitudes and perceptions of the pandemic with the inclusion of novel measures of prevention and management of the illness.