key: cord-0982302-7zo6pz90
authors: Banda, J.; Dube, A. N.; Brumfield, S.; Crampin, A. C.; Reniers, G.; Amoah, A. S.; Helleringer, S.
title: Controlling the first wave of the COVID-19 pandemic in Malawi: results from a panel study
date: 2021-02-23
journal: nan
DOI: 10.1101/2021.02.21.21251597
sha: d3adc42a8e9e3a3a0c8723fc734869f1dec3eb73
doc_id: 982302
cord_uid: 7zo6pz90

Many African countries have experienced a first wave of the COVID-19 pandemic between June and August of 2020. According to case counts reported daily by epidemiological surveillance systems, infection rates remained low in most countries. This defied early models of the potential impact of COVID-19 on the continent, that projected large outbreaks and massive strain on health systems. Theories proposed to explain the apparently limited spread of the novel coronavirus in most African countries have emphasized 1) early actions by health authorities (e.g., border closures) and 2) biological or environmental determinants of the transmissibility of SARS-CoV-2 (e.g., warm weather, cross-immunity). In this paper, we explored additional factors that might contribute to the low recorded burden of COVID-19 in Malawi, a low-income country in Southeastern Africa. To do so, we used 4 rounds of panel data collected among a sample of adults during the first 6 months of the pandemic in the country. Our analyses of survey data on SARS-CoV-2 testing and COVID-related symptoms indicate that the size of the outbreak that occurred in June-August 2020 might be larger than recorded by surveillance systems that rely on RT-PCR testing. Our data also document the widespread adoption of physical distancing and mask use in response to the outbreak, whereas most measured patterns of social contacts remained stable during the course of the panel study. These findings will help better project, and respond to, future waves of the pandemic in Malawi and similar settings.

In December 2019, a novel coronavirus, the severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), emerged in Hubei province, in Central China (1) . Since then, SARS-CoV-2 has been detected in 191 countries. By the end of 2020, over 1.7 million deaths had been attributed to COVID-19, i.e., the disease caused by SARS-CoV-2 (2) . The World Health Organization (WHO) declared this novel coronavirus a public health emergency of international concern on 30 January 2020 and classified its spread as a global pandemic on 11 March 2020. By comparison, the transmission of other coronaviruses (e.g., SARS-CoV-1, MERS-CoV) has been much more limited (3, 4) .

SARS-CoV-2 has however not spread evenly throughout the world. Europe and the Americas (e.g., USA, Brazil, Mexico) constitute the most affected regions (5) . In these settings, the recorded incidence of the novel coronavirus has often remained high since the first few months of 2020, causing significant excess mortality relative to previous years (6) (7) (8) (9) . Other large outbreaks have been documented, for example in south Asia (10) as well as in Iran (11) . In contrast, the incidence of SARS-CoV-2 appears low in African countries, which have recorded fewer than 4% of the global cases and deaths in 2020, despite accounting for more than 15% of the world population (12) .

This limited burden of COVID-19 on the African continent contrasts with expectations and projections formulated at the beginning of the pandemic. At that time, there were strong concerns that overcrowding in slum neighborhoods of large cities, limited access to water and sanitation, inadequate human resources for health, and the high-prevalence of multigenerational households might create a particularly fertile ground for the spread of SARS-CoV-2 (13, 14) . A model from the WHO regional office for Africa thus projected 37 million symptomatic cases in 2020, requiring high levels of hospitalization and life support throughout the African continent (15) . Whereas some African countries have indeed experienced multiple waves of the pandemic and high levels of excess mortality (e.g., South Africa), most African countries have not reported the predicted levels of strain on healthcare systems, and associated increases in morbidity and mortality (16, 17) .

Several theories have been proposed to explain the lower-than-expected burden of in most African countries in 2020. Some theories have emphasized the limited links between the continent and other world regions where SARS-CoV-2 first spread. African countries might thus have experienced a delayed onset of SARS-CoV-2 outbreaks due to the low volume of air travel to the region (18) . Urbanization is also more limited in African countries than in other parts of the world (19) . Lower population densities in rural areas limit opportunities for spreading SARS-CoV-2 in large regions of African countries (20) . Other theories have suggested that the transmissibility of the novel coronavirus might be lower on most of the African continent. This might be because of climate factors, since SARS-CoV-2 spreads less efficiently in warm weather (21) . It might also be attributed to (partial) protection from infection with SARS-CoV-2 due to prior exposure to other coronaviruses (17, 21) , genetic differences in susceptibility to viruses (22) , or the non-specific effects of some vaccines (e.g., BCG) frequently administered to local populations (23) . Furthermore, youthful age pyramids in African countries leave a smaller proportion of the population at-risk of severe COVID-19 and death compared to populations with an older age structure (20, 24, 25) .

In addition, rapid policy responses might have played a key role in limiting the spread of SARS-CoV-2 in African countries. For example, several African countries closed their borders when COVID-19 cases were growing rapidly in China and in Europe. This might have delayed and/or reduced the importation of SARS-CoV-2. Restrictions on schools and gatherings and/or lockdowns might have reduced the scale on which the novel coronavirus could spread during the early phase of the pandemic. In South Africa, a lockdown and its associated travel restrictions might have limited the growth of the epidemic between March and June 2020 (26) .

Recent experience with other epidemic diseases (e.g., Ebola, HIV) might have better prepared local health actors for engaging local communities, and implementing required information campaigns, quarantines and contact tracing (27, 28) .

Beyond these oft-discussed factors, the limited burden of COVID-19 on the African continent might also be an artifact of incomplete data collection systems (12) . As in other regions of the world, the recording of cases by epidemiological surveillance systems in African countries depends on the extent of testing for SARS-CoV-2 through RT-PCR. In several countries, the laboratory capacity to conduct such tests was not immediately available at the start of the outbreak and had to be established. In other countries, few reference laboratories had the capacity to carry out RT-PCR tests. These laboratories might have been initially overwhelmed by the volume of tests to be conducted to support pandemic response. Subsequently, the testing capacity has been greatly expanded in several countries, for example by mobilizing GeneXpert machines commonly used to test for tuberculosis (29) . Nonetheless, African countries have so far conducted fewer tests per 1,000 people than elsewhere (30) . The proportion of COVID-19 cases (and deaths) that are not recorded in case-based surveillance systems might thus be larger in African countries than in other parts of the world (12) . Most African countries also lack death registration systems that are sufficiently complete to allow detecting peaks of excess mortality due to the pandemic, even in the absence of extensive testing for SARS-CoV-2 (31) .

. CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint Even though governments were often quick to adopt recommendations from the WHO and the African CDC for the prevention of SARS-CoV-2 spread in local communities (e.g., physical distancing, enhanced hand hygiene, use of facial masks), few studies have investigated the contributions of such behavioral changes to epidemic control in African countries. Surveys conducted early in the pandemic in African countries, e.g., in the first few weeks after the introduction of SARS-CoV-2 in a country, have assessed the level of knowledge, and the perceptions of risk, about COVID-19 among various population groups (32) (33) (34) (35) (36) . They have highlighted widespread concerns about getting infected with SARS-CoV-2, gaps in understanding of the modes of transmission of SARS-CoV-2, as well as varying levels of early adoption of protective behaviors (e.g., mask use, physical distancing). Subsequent surveys have focused predominantly on the impact of the COVID-19 pandemic on several socioeconomic dimensions including, for example, poverty, education, healthcare utilization, mental health and family planning (37) (38) (39) (40) .

In this paper, we analyzed data from a panel study conducted in Malawi during the 6 months that followed the introduction of SARS-CoV-2 in the country. We used these data to 1) evaluate the potential extent of under-reporting of COVID-19 cases in case-based surveillance systems, and 2) explore the pace at which protective behaviors (e.g., physical distancing, mask use) were adopted in response to the spread of SARS-CoV-2.

The potential for a pathogen like SARS-CoV-2 to spread within a population depends on the interactions between features of the virus and the characteristics and behaviors of its (human) hosts. This is summarized in an indicator called the basic reproduction number, ! (41) . It is defined as the number of new infections produced by a single infection in a completely susceptible population. An epidemic has the potential to expand when ! is greater than 1, and it contracts when ! is below 1. ! is often expressed as the product of 3 parameters: (1) the probability that the virus is transmitted when a susceptible individual comes into contact with someone who is infected (transmissibility, noted ); (2) the average rate at which susceptible and infected individuals come into contact (noted ̅ ), and (3) the amount of time infected individuals remain contagious, i.e., they are able to transmit the virus (noted ). We thus have:

Controlling a viral outbreak then requires reducing one or several of those parameters, so that ! drops below 1 in a sustained manner.

SARS-CoV-2 can be transmitted before the appearance, or in the absence, of symptoms (42) (43) (44) (48) , and wearing facial masks to reduce the risk of infection from droplets and other aerosols (49) . Other interventions to reduce the transmissibility of SARS-CoV-2 may emphasize increasing ventilation of indoor space to limit transmission via aerosols (50) .

. CC-BY-NC-ND 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) Reducing the rate of contact between susceptible and individuals ( ̅ ), on the other hand, requires encouraging or mandating that individuals limit their social interactions. Often, these injunctions target infected or exposed individuals: after a positive COVID-19 test, a presumptive diagnosis or a potential exposure to SARS-CoV-2, patients are asked to isolate so that they do not transmit the virus to others in their household, community or networks (51) . In some circumstances, health authorities may also take wider measures to reduce the extent of contact within a population. They may temporarily prohibit attendance of certain places or events where individuals socialize and thus possibly transmit SARS-CoV-2. Since the beginning of the pandemic, many countries have even imposed periods of "lockdown" or "stay-at-home" orders, during which individuals are only allowed to leave their homes for restricted reasons and at very limited times.

In African countries, as in other areas of the world, governments and health stakeholders have promoted a mix of behavioral changes, which aim to reduce and ̅ simultaneously.

Investigations of the determinants of epidemic dynamics in African countries have however focused on changes in the rate of contact between infected and susceptible individuals ( ̅ ).

Several mathematical models have projected the effects of lockdown measures on epidemic trajectories (52, 53) . Two empirical studies have relied on time-use surveys to document changes in patterns of contact in African settings (54, 55) . They used these data to build detailed age-specific contact matrices. In comparison with pre-pandemic data, they have shown that lockdowns reduced the extent of social interactions in target populations and prompted a decline in estimates of the reproduction number of the epidemic. Other mathematical models have explored the potential impact on epidemic trajectories in African countries of i) selfisolation of suspected COVID-19 cases, ii) contact tracing, and iii) "shielding" members of the most at-risk age groups (i.e., elderly people) (20, 53, 56) . Several surveys have documented the . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint prevalence of protective behaviors targeting , i.e., the transmissibility of SARS-CoV-2, in African countries (32, 33, 54, (57) (58) (59) . However, they have seldom investigated whether the scale on which these behaviors are implemented might have changed over time.

Malawi is a low-income country located in southeastern Africa, with a population of approximately 18 million. It had an estimated life expectancy of 63.7 years in 2019 (60) . Since the 1980's, it is affected by a large HIV epidemic (61, 62) . In 2015-16, more than 1 in 10 adults had HIV, and among those only 76.8% were aware of their HIV status (63) . Prior to the COVID-19 pandemic, Malawi had made significant progress towards the achievement of various healthrelated goals, particularly those related to HIV treatment (64, 65) and to the improvement childhood survival (66) . In recent years, the incidence of non-communicable diseases linked to unhealthy behaviors has increased in urban and rural areas (67) (68) (69) .

Initial projections of the potential impact of the COVID-19 pandemic in Malawi suggested that 1 in 5 Malawians might become infected with SARS-CoV-2 in 2020, resulting in more than 60,000 hospitalizations and close to 2,000 deaths (15) . Malawi was one of the last countries in the world to record a case of COVID-19. This occurred on April 2 nd 2020, when infection was confirmed by RT-PCR in a traveler recently returned to the country ( Figure 1 ). Sporadic clusters of COVID-19 cases were then detected in urban centers for several weeks. In late May, additional importations of COVID-19 cases occurred among migrants returning primarily from South Africa, i.e., the African country with the largest documented outbreak. The recorded incidence of SARS-CoV-2 then increased sharply in Malawi in June and July, before starting to . CC-BY-NC-ND 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. Laboratory-based surveillance systems similar to those used throughout the world were activated in the country (70) . The capacity to detect SARS-CoV-2 by RT-PCR was established at the end of March at the National reference laboratory in Lilongwe, and at teaching and research institutions in Blantyre, in the southern region of the country (71) . Additional laboratory sites with capacity to test for SARS-CoV-2 were then added throughout the country between May and July . As of the end of December 2020, Malawi had carried out more than 85,000 tests, . CC-BY-NC-ND 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) Tests have also been conducted as part of contact tracing (72), i.e., efforts through which health workers get in touch with people who have come in contact with a confirmed case of COVID-19, notify them of their exposure, and encourage them to isolate in an attempt to break transmission chains. Between the beginning of the outbreak and the end of 2020, more than 10,000 contacts of COVID-19 cases have been listed and followed-up in the country.

In Malawi, the MoH and other stakeholders (e.g., organizations of the UN system, Non-Governmental Organizations) have sought to reduce the transmissibility of the novel coronavirus using various channels. Protective behaviors such as increased hand washing, coughing or sneezing in one's elbow, and physical distancing were thus rapidly promoted. Early in the outbreak, the MoH also emphasized the use of facial masks to prevent the transmission of SARS-CoV-2. Medical masks and respirators were recommended in healthcare settings, and among people who care for patients with COVID-19, whereas the general public was advised to use cloth masks "in settings where social distancing is not possible and where there is widespread community transmission" (73) . The recommendations were accompanied by guidelines intended for local tailors about the making of cloth masks (e.g., number of fabric layers). To respond to the increased spread of SARS-CoV-2 in June and July (see figure 1 ), the Government made mask wearing mandatory when in public in early August. The measure was . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint accompanied by a fine of 10,000 Malawian Kwachas (i.e., approximately 12 USD) for noncompliers.

Concurrently, several measures were taken to reduce contact between susceptible and infected individuals in the population. As part of the national disaster declaration at the end of March, schools and universities were closed throughout the country, and attendance of public gatherings was limited to 100 people. A national lockdown was then announced on April 18 th 

Malawi, we analyzed panel data collected during the first 6 months of the pandemic in the country ("COVID-19 panel" thereafter). This panel was initiated in April 2020, roughly a month after the Government of Malawi declared COVID-19 a "national disaster", and less than 3 . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint weeks after the first cases of COVID-19 detected in Lilongwe, the capital city. The COVID-19 panel now includes 4 rounds of data collection conducted approximately 6 weeks apart and ending in mid-November 2020. During that time, the recorded incidence of the SARS-CoV-2 virus varied greatly, with peaks of close to 200 cases per day in July vs. fewer than 5 cases per day on average in October and November (figure 1). The first two rounds of data collection also took place during a time when campaigning for the presidential election of June 23 rd was under way (75) .

Respondents in the COVID-19 panel had previously participated in a study of the measurement of adult mortality. This "pre-COVID-19 study" was completed before the introduction of SARS-CoV-2 in Malawi. It was nested within the activities of a health and demographic surveillance system (HDSS) located in Karonga district, in northern Malawi (76, 77) . This is an area bordering Lake Malawi to the east and located within 1.5-hour drive of the Tanzania border. Its economy is predominantly organized around fishing, subsistence farming (e.g., rice, cassava) and smallscale retail activities.

The last round of interviews for the pre-COVID-19 study focused on the health and family relations of respondents. It occurred between December 2019 and March 2020. At that time, we sought to obtain the mobile phone numbers of 1) a random sample of current residents of several HDSS population clusters located close to the lakeshore community of Chilumba ("current residents", thereafter), 2) a random sample of former HDSS residents who had migrated throughout Malawi ("former residents", thereafter), and 3) the siblings of these current and former HDSS residents ("referred siblings", thereafter). When the spread of SARS-CoV-2 accelerated throughout the world (including in other parts of Africa) in March 2020, we decided to initiate the COVID-19 panel: we contacted all participants of the pre-COVID-19 study for . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint whom a phone number was available, and invited them to participate in a series of COVIDrelated interviews.

In total, we sought to obtain the phone numbers of 1,036 participants in the pre- Due to its sampling frame, and to selective participation in a survey conducted by mobile phone, the sample of the COVID-19 panel is not representative of the population of Malawi, or Karonga district. We thus do not use these data to estimate the prevalence of various symptoms or behaviors in the country at various points during the pandemic. Instead, we use multiple measurements available for each panel participant to a) understand the potential scale of underreporting of COVID-19 in data derived from RT-PCR tests, and b) assess the pace at which study participants adopted protective behaviors against SARS-CoV-2.

. CC-BY-NC-ND 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. 

In each round of the COVID-19 panel, we asked respondents whether they had experienced, during the previous month, a number of symptoms previously reported among COVID-19 cases in clinical studies (11, 80, 81) . These symptoms included, coughing, shortness of breath, chest pain, fever, headaches or fatigue. We only included loss of smell/taste (82, 83) in the list of symptoms elicited in rounds 2, 3 and 4 of the COVID-19 panel. Starting in round 3, we asked respondents who reported any symptom possibly related to COVID-19 whether they self-isolated upon experiencing these symptoms.

We also asked respondents whether they had ever been tested for SARS-CoV-2. In round 4, we added a description of the sample collection procedure in our question about SARS-CoV-2 testing (i.e., that a long swab was inserted in the patient's nostrils). This description was introduced to help prevent confusion with other COVID-related screening procedures (e.g., temperature checks) widely in Malawi implemented since the start of the pandemic.

. CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

In each round of data collection, we investigated the behaviors respondents used to reduce the transmissibility of SARS-CoV-2 and/or limit their contacts in the month before the survey. First, to explore changes in ̅ , i.e., the rate of contact between infected and susceptible individuals, we inquired about respondents' living arrangements. We asked respondents how many people resided in their households, and how many rooms their house included. This provided an indicator of the potential for intra-household exposure to/transmission of SARS-CoV-2. Second, to measure contacts outside of the household, we asked respondents whether they had attended various places and events. We drew a list of places where people commonly interact and socialize in Malawi, which included markets, churches and mosques, neighbors or relatives'

houses, workplaces, football games or funerals. Then, we asked respondents if they had attended each of these places/events in the past 7 days before the interview.

We also asked respondents to list the behaviors they had used to reduce the spread of SARS-CoV-2 in the past month. This question was adapted from instruments previously administered in COVID-related surveys conducted in high-income countries (84) . During pre-testing and piloting activities, we drew a list of possible answers to this question. This list included behaviors that reduce , e.g., use of facial masks, physical distancing or handwashing, as well as behaviors that reduce ̅ , e.g., avoiding crowded areas. Interviewers did not read this list to respondents. Instead, they let them spontaneously recall their behaviors, and coded their answer using the list of potential behaviors. If a behavior was not included in the list, they coded it as "other" and specified the respondent's answer in a follow-up question. Multiple answers were allowed, and after each answer, interviewers were instructed to probe further in a nonspecific manner, by asking respondents whether there was anything else that they did to prevent the spread of SARS-CoV-2.

. CC-BY-NC-ND 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) In rounds 3 and 4, we asked respondents who reported wearing a facial mask in the past month how often they wore one during that timeframe. Possible answers were "always", "often", "sometimes" and "rarely". We also asked them to list the reasons why they wore a mask.

Interviewers let respondents answer spontaneously and they coded their answer using a list of possible reasons. This list included, for example, "self-protection", "to protect others", "to meet a requirement (e.g., to enter a building)" or "to avoid a fine". Multiple answers were allowed. After each reason mentioned by a respondent, interviewers were instructed to probe further by asking whether there was an additional reason.

Data analysis: we first described the socio-economic characteristics of panel participants who completed all 4 rounds of data collection, including their reported gender, age, marital status, occupation and religion. Second, we analyzed trends in their experience of COVID-related symptoms over time. We classified participants in 3 groups according to their reported experience of symptoms during the course of the study: those without any reported symptoms, those who developed a cough, and those who developed one or more other symptoms previously associated with COVID-19. We then assessed the level of testing for SARS-CoV-2 in our panel sample, and we explored whether testing rates varied across these 3 groups. To measure testing rates, we only considered SARS-CoV-2 testing histories reported during round 4, because we introduced a more precisely worded question during that round. Low testing rates among those with new onset of various COVID-related symptoms constitute an indication of the extent of under-reporting of the SARS-CoV-2 burden in this sample.

Third, we described trends in behaviors associated with social contacts in our panel sample. To assess the potential for intra-household contact, we reported the ratio of the number of household residents to the number of rooms in the house. Then, we tabulated reports of attendance of places and events in the past week collected during each round. We measured . CC-BY-NC-ND 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. Rural participants relied more frequently on boreholes as their primary water source (55.1% vs.

4.8% in urban areas) and they engaged in activities related to agriculture and fishing at a much higher rate (46.0% vs. 8.6% in urban areas). Crowded households (i.e., with more than 3 people per room) were slightly more common in rural areas. In addition, <1% of respondents reported using residential strategies (e.g., moving to a new house or flat, appendix A2) to reduce the spread of SARS-CoV-2. place of residence, and b) the type of COVID-related symptom experienced. In particular, participants who experienced a cough were more likely to report self-isolating than participants who reported other symptoms (e.g., fever, headaches). Self-isolation was especially high during round 3 among urban participants who had recently been coughing (i.e., 70%).

Some panel participants also reported "avoiding crowded areas" or "avoiding going out in general" to reduce the spread of SARS-CoV-2 (appendix A2). However, the proportion of respondents using these strategies to reduce contact declined after round 1. Very few respondents also reported reducing their mobility to prevent the spread of SARS-CoV-2 . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint (appendix A2). In round 1, >10% of panel participants in urban areas reported "avoiding visibly sick people" to prevent the spread of SARS-CoV-2, but in subsequent rounds, this proportion declined to <2%. Virtually none of the panel participants reported "avoiding healthcare facilities" to reduce their risk of acquiring or transmitting SARS-CoV-2.

various behaviors that modify the transmissibility of SARS-CoV-2. In each round, more than 90% of participants reported washing their hands more often (figure 8). In round 1, the next most commonly reported behaviors affecting transmissibility were the use of hand sanitizer among urban respondents and covering up coughs/sneezes among rural residents. In subsequent rounds however, panel participants reported adopting other strategies as their main preventive behaviors. For example, whereas the use of physical distancing was reported by approximately 10% of participants in round 1, it increased to close to 60% in round 2 among participants in urban areas, and to similar levels in round 3 among participants in rural areas.

The use of physical distancing then stopped increasing in rural areas in round 4, and even declined among urban participants.

In all rounds, the use of facial masks in the month prior to the survey was more widespread among participants in urban areas than in rural areas (figure 8). Reported mask use did not increase markedly between rounds 1 (4.4%) and 2 (7.1%) among participants in rural areas, whereas the proportion of urban participants who reported using a mask almost doubled during the same timeframe (20.3% in round 1 vs. 38.3% in round 2). The proportions of participants reporting use of facial masks in the month prior to the survey then increased particularly sharply . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint in subsequent rounds: in round 3, 89.6% of participants in urban areas, and 67.7% in rural areas, reported using a mask in the past month.

The proportion of participants who reported using a mask at any point in the past month remained approximately constant in rounds 3 & 4 of the panel study. However, the frequency of mask use declined markedly between round 3 and 4 ( figure 9 ). In rural areas, the proportions of mask users who reported wearing masks either "always" or "very often" declined from 71.2% to 39.7%, whereas in urban areas, corresponding figures were 78.9% and 44.7%.

The reasons for wearing masks reported by participants also varied between rounds 3 and 4. In round 3, self-protection and protection of others were the most commonly reported reasons for mask use. In round 4, on the other hand, more than 50% of participants also reported wearing a mask due to requirements, either to enter a building or to receive a service (appendix A3). In rural areas, the proportion of mask users who reported wearing a mask to comply with a requirement nearly doubled between rounds 3 and 4. Fewer than 1% of respondents reported wearing a mask because they worried about fines imposed by the Government, because of social pressure or because they were sick or presented symptoms.

In this study, we found that the magnitude of the SARS-CoV-2 outbreak in the country is likely larger than reflected in case-based surveillance datasets that rely on RT-PCR diagnostics.

When we compared participants' self-reports of a) symptoms often associated with COVID-19 (e.g., cough, chest pain) and b) their history of RT-PCR testing for SARS-CoV-2, we found that . CC-BY-NC-ND 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. estimates of the proportion of participants tested for SARS-CoV-2. We also did not investigate repeat testing for SARS-CoV-2 among panel participants, nor did we assess whether testing coverage varied over time. Based on these data, and also owing to the selectivity of our study sample (see below), we could not produce an estimate of the true number of infections with SARS-CoV-2 in Malawi. Nonetheless, our findings corroborate results from a serosurvey conducted in Blantyre (87) and several similar surveys conducted in other African settings (12, 88, 89) . In these studies, the high prevalence of IgG antibodies against SARS-CoV-2 implied a number of cases (much) larger than recorded by surveillance systems relying on RT-PCR testing.

Whereas our data indicated that daily case counts based on RT-PCR diagnostics might underestimate the scale of the outbreak, they also suggested that such laboratory-based . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint surveillance systems might adequately capture epidemic trajectories. In particular, the panel data we collected on symptoms associated with COVID-19 indicated that an epidemic peak likely occurred in urban centers in June/July (figure 3), similar to trends detected in daily reports of PCR-confirmed cases (figure 1). At that time, the proportion of urban participants who reported symptoms commonly seen among COVID-19 cases (e.g., cough, chest pain) nearly doubled relative to levels observed in earlier months (i.e., round 1 in April/May). In subsequent rounds (i.e., rounds 3 and 4), this proportion rapidly returned to '"normal" levels. In contrast, the proportion of panel participants residing in rural areas and who reported a cough or another symptom, declined steadily from round to round, consistent with seasonal patterns of respiratory infections (90) .

We found that several protective behaviors were adopted widely in our study sample during the first few months of the COVID-19 pandemic in Malawi. These included in particular behaviors that reduce the transmissibility of the novel coronavirus. More than half of panel participants reported adopting the recommended practice of physical distancing between rounds 1 & 3.

Mask use also increased very sharply during that time frame, with more than 8 out of 10 urban participants reporting that they had worn a facial mask outside of their household in the past month ( figure 8) . Comparatively, behaviors that affect contacts between infected and susceptible individuals were more stable in our panel sample. Indicators of intra-household contact remained largely constant throughout the 4 panel rounds, and there were few changes in attendance of places and events where people socialize (and thus possibly spread SARS-CoV-2) during the course of the study Our data on behavioral responses to the COVID-19 pandemic also indicate that panel participants have recently reduced their protective behaviors. For example, urban respondents are increasingly celebrating weddings (figure 6), a social occasion that had led to large SARS-. CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

CoV-2 clusters in other contexts (91) . Most respondents have also reduced the consistency of their mask use: whereas in round 3, more than two thirds of mask users reported "always" wearing a mask outside of their household, this proportion dropped to <25% in round 4. Mask use appears increasingly sustained by requirements to wear masks to access various buildings and services, rather than by a motivation to protect others. These trends might enhance the likelihood of local transmission of SARS-CoV-2 in the event of new imported cases and/or the emergence of more transmissible viral strains (92) .

These findings about the adoption of protective behaviors suffer from several limitations. First, respondents might have over-reported their use of recommended protective measures. For example, a recent study found that, in a sub-county of Kenya, self-reported mask use greatly exceeded the levels of mask use recorded by independent observers (93) . In our study, the extent of such social desirability bias in self-reports of protective behaviors might have varied over time: incentives to report compliance with mask use recommendations might have been higher after the MoH had mandated the use of masks in public and imposed fines on noncompliers. Since this mandate was adopted when confirmed cases were near their peak (i.e., early August), this might have created a spurious temporal association between increasing reported mask use and declining case/symptoms. In our data however, the reported frequency of mask use declined sharply in round 4, at a time when the mandate to wear masks in public was still in place.

Second, our data on patterns of contact also lacked a proper baseline. Our first panel measurement took place in April/May, i.e., a few weeks after the Government declared COVID-19 a "national disaster", closed schools and universities, and imposed limitations on the size of public gatherings. Reductions in attendance of social events (e.g., weddings) have likely occurred before our first data collection round. Such reductions in contact might have helped . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint delay the start of the outbreak and reduce its magnitude. However, they did not fully prevent the local transmission of SARS-CoV-2: incidence increased sharply in June/July when restrictions on public gatherings were still in place. Our data on self-isolation are also limited because they was only collected in rounds 3 and 4. They indicated limited compliance with self-isolation guidelines, except among residents of urban areas who presented with a cough at the peak of the outbreak ( figure 7) . This factor might have contributed to slowing down the first wave of the pandemic in the country.

Third, there are several sampling limitations. Despite high enrollment and follow-up rates (figure 2), the sample size available for panel analyses was limited. We thus could not investigate the correlates of observed trends in symptoms and behaviors (e.g., gender differences, differences by occupation). Because it relied on mobile phones to conduct interviews, our sample was also selective. We thus reached more educated potential participants at a higher rate than potential participants with no schooling, or only primary education (78) . If patterns of behavioral changes differed by educational levels, then our study results might be affected. Furthermore, the study . CC-BY-NC-ND 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) . CC-BY-NC-ND 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) . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

Notes: The experience of symptoms was ascertained over the course of the panel study, whereas the experience of testing for SARS-CoV-2 was reported in round 4. COVID-related symptoms elicited during the surveys included fever, chest pains, shortness of breath, fatigue, shivering, headache, muscle pain, runny nose, sore throat, cough and wheezing. 

Notes: the data plotted in this graph were collected from a series of questions asking respondents whether they had attended the places listed on the y-axis in the past 7 days prior to the survey. The point estimates are labeled by their round number on the plot. The places that appear on the y-axis are ordered according to their attendance in urban areas in round 1. 

Notes: the question about self-isolation was only applicable for respondents who reported experiencing one or more COVID-related symptoms in the month prior to the survey. This question was introduced in round 3 of the panel survey. In round 3, the sample size for these analyses was n = 206. In round 4, the sample size was n = 192. Estimates of the standard errors (used to calculate confidence intervals) were adjusted for the clustering of observations within families. CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

Notes: the data plotted in this graph were collected from a question asking respondents what they had done in the past month to prevent the spread of SARS-CoV-2. Respondents were not provided with potential answers. However, after each response, interviewers were instructed to probe further, by asking respondents: "is there anything else that you have done?". In this graph, the behaviors listed on the y-axis are those that reduce the transmissibility of the coronavirus. Respondents also listed other behaviors, that reduce the contact between infected and susceptible individuals (appendix A3). In each round, <1% of panel respondents reported not doing anything to prevent the spread of SARS-CoV-2. The point estimates are labeled by their round number on the plot. The behaviors that appear on the y-axis are ordered according to their prevalence in urban areas in round 1. . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint FIGURE 9: FREQUENCY OF MASK USAGE Notes: in rounds 3 & 4, we asked respondents who reported that they had used a mask to prevent the spread of SARS-CoV-2, how often they had done so. These data were not collected in rounds 1 & 2. In round 3, the sample size for these analyses was n = 498. In round 4, the sample size was n = 496. Data collection round Proportion of Respondents who reported using a face mask in past month Always Often Sometimes Rarely . CC-BY-NC-ND 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) . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

Notes: the data plotted in this graph were collected from a question asking respondents what they had done in the past month to prevent the spread of SARS-CoV-2. Respondents were not provided with potential answers. However, after each response, interviewers were instructed to probe further, by asking respondents: "is there anything else that you have done?". In this graph, the behaviors listed on the y-axis are those that reduce contact between infected and susceptible individuals. In each round, <1% of panel respondents reported not doing anything to prevent the spread of SARS-CoV-2. The point estimates are labeled by their round number on the plot. The behaviors that appear on the y-axis are listed according to their prevalence in urban areas in round 1. 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

Notes: the data plotted in this graph were collected from a question asking respondents who reported having used a mask in the past month, why they had done so. Respondents were not provided with potential answers. However, after each response, interviewers were instructed to probe further, by asking respondents: "was there another reason?". The point estimates are labeled by their round number on the plot. The reasons that appear on the y-axis are listed according to their prevalence in urban areas in round 1. To protect others

To protect self % reporting reason for using facial mask . CC-BY-NC-ND 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 February 23, 2021. ; https://doi.org/10.1101/2021.02.21.21251597 doi: medRxiv preprint

HOUSEHOLD OCCUPANCY RATIO Notes: the y-axis is plotted on a logarithmic scale

The COVID-19 epidemic

COVID-19: a fast evolving pandemic

The SARS epidemic in mainland China: bringing together all epidemiological data

Interhuman transmissibility of Middle East respiratory syndrome coronavirus: estimation of pandemic risk. The Lancet

Weekly epidemiological update -29

All-cause excess mortality observed by age group and regions in the first wave of the COVID-19 pandemic in England

Estimation of Excess Deaths Associated With the COVID-19 Pandemic in the United States

Monitoring life expectancy levels during the COVID-19 pandemic: Example of the unequal impact of the first wave on Spanish regions

Exploring excess of deaths in the context of covid pandemic in selected countries of Latin America

Epidemiology and transmission dynamics of COVID-19 in two Indian states

Epidemiological characteristics of coronavirus disease 2019 (COVID-19) patients in IRAN: A single center study

The puzzle of the COVID-19 pandemic in Africa

Africa in the Path of Covid-19

Looming threat of COVID-19 infection in Africa: act collectively, and fast. The Lancet

The potential effects of widespread community transmission of SARS-CoV-2 infection in the World Health Organization African Region: a predictive model

How Africa has tackled covid-19

COVID-19 in Africa: Dampening the storm? Science

Preparedness and vulnerability of African countries against importations of COVID-19: a modelling study. The Lancet

Urbanization and income inequality in Sub-Saharan Africa. Sustainable Cities and Society

The relatively young and rural population may limit the spread and severity of COVID-19 in Africa: a modelling study

Why is There Low Morbidity and Mortality of COVID-19 in Africa?

Genetic Adaptation and Neandertal Admixture Shaped the Immune System of Human Populations

Considering BCG vaccination to reduce the impact of COVID-19. The Lancet

Demographic science aids in understanding the spread and fatality rates of COVID-19

Besides population age structure, health and other demographic factors can contribute to understanding the COVID-19 burden

Life under lockdown: Illustrating tradeoffs in South Africa's response to COVID-19. World Development

COVID-19, and emerging infectious disease: lessons learned and future preparedness. Current Opinion in Ophthalmology

Covid-19, Ebola, and HIV -Leveraging Lessons to Maximize Impact

GeneXpert for the diagnosis of COVID-19 in LMICs. The Lancet Global Health

A crosscountry database of COVID-19 testing. Sci Data

A scandal of invisibility: making everyone count by counting everyone

COVID-19 Related Knowledge, Attitude and Practices in a Southeastern City in Nigeria: A Cross-Sectional Survey

Social Science Research Network

Attitudes and Practices Towards COVID-19: An Epidemiological Survey in North-Central Nigeria

COVID-19 related knowledge, attitudes, practices and needs of households in informal settlements in

Knowledge level and factors influencing prevention of COVID-19 pandemic among residents of Dessie and Kombolcha City administrations

Attitude, and Practice Towards COVID-19 Among People Living with HIV/AIDS in Kigali, Rwanda

Monitoring COVID-19 Impacts on Households in Malawi: Findings from the First Round of the High-Frequency Phone Survey

World Bank Group

Monitoring COVID-19 Impacts on Households in Malawi). Available from

Estimates of the Potential Impact of the COVID-19 Pandemic on Sexual and Reproductive Health In Low-and Middle-Income Countries

Protective behaviours and secondary harms from non-pharmaceutical interventions during the COVID-19 epidemic in South Africa: a multisite prospective longitudinal study. medRxiv

Falling living standards during the COVID-19 crisis: Quantitative evidence from nine developing countries

Infectious Diseases of Humans: Dynamics and Control

Presumed Asymptomatic Carrier Transmission of COVID-19

Asymptomatic and Presymptomatic Infectors: Hidden Sources of Coronavirus Disease 2019 (COVID-19)

The implications of silent transmission for the control of COVID-19 outbreaks

Drug treatments for covid-19: living systematic review and network meta-analysis

Improving handwashing habits and household air quality in Africa after COVID-19. The Lancet Global Health

Viewpoint -Handwashing and COVID-19: Simple, right there…? World Development

Scientific and ethical basis for social-distancing interventions against COVID-19

Face masks for the public during the covid-19 crisis

How can airborne transmission of COVID-19 indoors be minimised? Environment International

Feasibility of controlling COVID-19 outbreaks by isolation of cases and contacts. The Lancet Global Health

Modeling the transmission dynamics of the COVID-19 Pandemic in South Africa

Response strategies for COVID-19 epidemics in African settings: a mathematical modelling study. medRxiv

The impact of COVID-19 control measures on social contacts and transmission in Kenyan informal settlements

Impact of social distancing regulations and epidemic risk perception on social contact and SARS-CoV-2 transmission potential in rural South Africa: analysis of repeated cross-sectional surveys. medRxiv

Glèlè Kakaï R. Predicting COVID-19 spread in the face of control measures in West Africa

Level and Determinants of Adherence to COVID-19 Preventive Measures in the First Stage of the Outbreak in Uganda

Adherence towards COVID-19 mitigation measures and its associated factors among Gondar City residents: A community-based cross-sectional study in Northwest Ethiopia

Limiting spread of COVID-19 in Ghana: Compliance audit of selected transportation stations in the Greater Accra region of Ghana

Trends and measurement of HIV prevalence in northern Malawi

The development of the HIV epidemic in Karonga District

Malawi Population-Based HIV Impact Assessment (MPHIA) 2015-2016: Final Report

Ministry of Health

Adult mortality and probable cause of death in rural northern Malawi in the era of HIV treatment

Population-level reduction in adult mortality after extension of free anti-retroviral therapy provision into rural areas in northern Malawi

Malawi and Millennium Development Goal 4: a Countdown to 2015 country case study. The Lancet Global Health

Hypertension and diabetes in Africa: design and implementation of a large population-based study of burden and risk factors in rural and urban Malawi

Investigating associations between rural-to-urban migration and cardiometabolic disease in Malawi: a population-level study

Prevalence of impaired renal function among rural and urban populations: findings of a cross-sectional study in Malawi

Addressing COVID-19 in Malawi. The Pan African Medical Journal

Challenges of COVID-19 testing in low-middle income countries (LMICs): the case of Malawi

Covid-19 contact tracing: a briefing

Public Health Institute of Malawi. National COVID-19 Preparedness and Response Plan

Public Health Institute of Malawi

National-COVID-19-Preparedness-and-Response-Plan_Version1_April%202020%201.pdf/detail 74. Speech by President Mutharika on 21-day Covid-19 lockdown | United Nations in Malawi

Malawi at the Crossroads: Does the Fear of Contracting COVID-19 Affect the Propensity to Vote?

Profile: the Karonga Health and Demographic Surveillance System

Evaluation of a villageinformant driven demographic surveillance system in Karonga, Northern Malawi. Demographic Research

Knowledge and Behaviors Related to the COVID-19 pandemic in Malawi. medRxiv

The ABCs of Covid-19 prevention in Malawi: Authority, benefits, and costs of compliance. World Development

A Comparative Study on the Clinical Features of Coronavirus 2019 (COVID-19) Pneumonia With Other Pneumonias

Characteristics of and Important Lessons From the Coronavirus Disease 2019 (COVID-19) Outbreak in China: Summary of a Report of 72 314 Cases From the Chinese Center for Disease Control and Prevention

A New Symptom of COVID-19: Loss of Taste and Smell

The Loss of Smell and Taste in the COVID-19 Outbreak: a Tale of Many Countries

Perceptions and behavioural responses of the general public during the COVID-19 pandemic: A crosssectional survey of UK Adults

Malawi Population and Housing Census

COVID-19 in LMICs: The Need to Place Stigma Front and Centre to Its Response

High SARS-CoV-2 seroprevalence in health care workers but relatively low numbers of deaths in urban Malawi

Seroprevalence of COVID-19 in Niger State. medRxiv

Seroprevalence of anti-SARS-CoV-2 IgG antibodies in Kenyan blood donors

Seasonal influenza epidemiology in sub-Saharan Africa: a systematic review

Large Outbreak of Coronavirus Disease among Wedding Attendees

Sixteen novel lineages of SARS-CoV-2 in South Africa

Self-Reported Mask Wearing Greatly Exceeds Directly Observed Use: Urgent Need for Policy Intervention in Kenya. medRxiv