Background
Coronavirus disease 2019 (COVID-19) is an infectious disease caused by severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) infection [
1]. It was first reported in December 2019 in Wuhan, China [
2], and the World Health Organization declared it a public health emergency of international concern on 30 January 2020 [
3]. South Korea became the third country to report an imported COVID-19 case on 19 January 2020 [
4]. Further spread of COVID-19 in the community was reported in mid-February, and the Korean Ministry of Health and Welfare declared the highest level of public health alert on 23 February 2020 [
5]. Since the first COVID-19 case was identified in South Korea, isolation of confirmed cases, contact tracing, extensive testing, and timely quarantine of all contacts have been conducted under the strategic guidelines for COVID-19 control from Korea Centers for Disease Control and Prevention (KCDC) [
6,
7]. Furthermore, combined public health measures including travel-related measures, case-based measures, and community measures were implemented across South Korea which helped control the first epidemic wave without a complete lockdown (Table S
1) [
5,
8]. Between 22 March and 19 April 2020, strict social distancing measures included recommendations to the public to stay at home and to delay or cancel social gatherings, as well as policies including closing schools and other public facilities, allowing greater flexibility in sick leave, and encouraging work-from-home and flexible working hours [
9,
10]. These strict social distancing measures were relaxed on 20 April 2020, because the daily reported number of cases was under 50 and the unknown origin of infection was less than 5% among total cases of investigation for the previous 2 weeks [
11].
Sustained increases in cases were observed as the strict social distancing measures were further relaxed by opening public facilities on 6 May 2020 when the first epidemic wave had already ended. Increase of the cases can be observed by the changes of public health efforts of extensive COVID-19 case finding and contact tracing [
12]. However, there is a lack of study to assess the changes of this active case finding strategy specifically.
To characterize the transmission dynamics of SARS-CoV-2 in South Korea, we identified the major cluster types of COVID-19 cases, and we estimated the time-varying effective reproductive number, serial interval distribution, age-specific infector-infectee matrices for the two COVID-19 epidemic waves. Furthermore, to assess public health efforts to find and trace cases in South Korea, we estimated the proportion of cases that were asymptomatic and the proportion of local cases with an unknown origin of infection at the time of first clinical assessment.
Discussion
In South Korea, the combined public health measures implemented early in the COVID-19 epidemic reduced the spread of SARS-CoV-2 [
5]. As mobility restriction has led to a substantial economic and social loss [
26], many countries including South Korea relaxed social distancing measures after the first epidemic wave despite concerns of a resurgence of COVID-19 cases. Our analysis of two different epidemic waves provided insight into the strategies to control COVID-19, particularly in countries that have relaxed their social distancing measures.
A large number of imported cases in the first epidemic wave challenged efforts to control COVID-19 in South Korea. In late March 2020, many Koreans in Europe and the United States traveled back to Korea due to the lockdowns. Public health authorities implemented mandatory laboratory screening at the port of entry and a 14-day mandatory self-quarantine program for travellers from Europe (22 March) and then all overseas travellers (1 April) [
27]. Local health authorities identified 555 imported cases and they were linked with 44 local cases during the first wave (Fig. S
1). Furthermore, there were 634 imported cases linked with 3 local cases (most from Asia and North America) identified during the second wave. To encourage the 14-day mandatory self-quarantine for travellers arriving in South Korea, the Korean government implemented strategies to support (by providing financial aid), monitor (by using mobile phone application), and sanction (by enhancing noncompliance penalty and subjecting to deportation) the individuals in self-quarantine [
27]. Therefore, the control of transmission from imported cases improved substantially in the second wave and the re-emergence was mainly driven by local transmission in the community.
We identified many clusters associated with religious activities where individuals spend prolonged periods of time in close proximity [
14]. These were the most frequent type of clusters reported in both first and second waves. Although the Korean government emphasized wearing face masks and keeping physical distance between persons during their meeting, transmission still occurred, possibly through the droplet-borne route or environmental contamination of shared surfaces [
28]. Resuming economic activities and school early in the second wave (COVID-19 relief funds started distribution to the public on 11 May and resuming school on 20 May) was likely associated with clustered outbreaks in workplace, shopping malls, and academic institutions, and this has likely contributed to the longer duration of
Rt around 1 more than a month.
We found that a large proportion of transmission occurred between individuals of the same age, and transmission between children was rare. However, school closure in the first wave may have contributed to reduced opportunities for transmissibility between children [
29]. After resuming school, the active daily-screening, monitoring, and following personal protective measures to the students may also reduce the risk of transmission between children in the second wave.
We identified that 16.6% of the local cases were asymptomatic at the time of presentation, which is similar to those in Shenzhen (20%) and Hong Kong (21%) in China [
30,
31]. The proportion of asymptomatic cases was lower in the first epidemic wave, particularly among those aged 20–39 years. This is likely because the case investigations were mainly symptom-based and required an epidemiological link in the earlier period. The emerging scientific evidence on the full spectrum of SARS-CoV-2 infections encouraged more testing on asymptomatic individuals. Young adults may have lower or delayed healthcare-seeking [
32], which may explain their lowest proportion of asymptomatic cases among all age groups. In the second epidemic wave, extensive contact tracing and screening for the large workplace and leisure-related clusters, probably allowed the detection of infected young adults who were still asymptomatic or pre-symptomatic. This suggests that active case finding and improving the awareness of the disease dynamics among young adults is crucial to reducing asymptomatic and pre-symptomatic transmission of SARS-CoV-2. The proportion of asymptomatic cases among 80s increased significantly from the first to second epidemic wave. This was likely affected due to rapid and massive screening of the elderly for SARS-CoV-2 in residential homes during the second epidemic waves.
In South Korea, all close contacts of laboratory-confirmed cases were quarantined, and public health authorities diligently traced the source of infection. The presence of unidentified cases indicates hidden and uncontrolled transmission in the community. Therefore, a low proportion of unlinked local cases is an indication of effective case finding [
33]. In our study, 20% of local cases were identified as unlinked, which was lower than the early phase of the epidemic (39%, 18 January – 2 March 2020) [
34], similar to that in Singapore (17%) [
33], but lower than that in Hong Kong (36%) [
35]. The proportions of unlinked local cases were low in the two epidemic waves and this indicates that extensive investigation was maintained in both waves.
A study in China demonstrated that serial interval distributions can be shortened by active case finding and enhanced public health measures [
22]. In our study, the mean serial interval was about 3 days, shorter than those reported in China (4 days) [
23], and a pooled estimate of 5 days [
36]. The implementation of rigorous public health measures, including registered mandatory digital applications (QR codes) in public places for contact tracing may contribute to earlier interrupt of the transmission of SARS-CoV-2. A negative number of serial intervals indicates the symptom onset in the infectee occurs prior to symptom onset in the infector. Based on the previous literature on the serial interval of COVID-19 [
25], large negative serial intervals in some of transmission pairs in our study should be interpreted with caution as there could be uncertainty of the direction of transmission or the infectee could have been infected by another unidentified infector.
A modelling study demonstrated that control measures, including contact tracing, testing, and self-isolation, would be less effective if asymptomatic infections are higher [
37]. Furthermore, a review study showed that restricting mass gathering was associated with a reduced incidence of COVID-19 [
38]. Our findings are consistent with these previous findings that continuous, strict social distancing measures and active seeking asymptomatic cases are critical to reducing the spread of SARS-CoV-2 in a community.
Recent studies in Hong Kong, China [
39], and South Korea [
5] demonstrated that social distancing measures were effective in controlling COVID-19. However, there have been few studies on the control of resurgences in transmission after the relaxation of social distancing measures. A simulation study also demonstrated that the second wave of infection would develop when contact tracing failed [
40]. Our findings of
Rt indicated that even when rigorous public health efforts were in place, relaxation of certain social distancing measures in the community, in our case further reopening of public facilities, may allow resurgence of COVID-19 within days. Further research on how social distancing and other public health measures should be relaxed is warranted. Furthermore, simulation studies based on empirical data with the counterfactual scenarios to predict a potential resurgence of COVID-19 would help public health authorities prepare for future outbreaks.
Our analysis has several limitations. First, we excluded the Daegu-Gyeongsanbuk region where the epidemic was mainly driven by large superspreading events in a religious group at the very beginning of the first epidemic wave. The outbreak occurred mainly in Daegu-Gyeongsanbuk region and well before the major control measures were implemented, and uncooperative attitude of the members of the religious group was reported during the epidemiological investigation [
14]; hence it did not reflect characteristics of typical community transmission in South Korea. Furthermore, the Korean government designated the Daegu-Gyeongsanbuk region as a special disaster zone and recommended travel with caution during the study period [
41]. Second, we have not included data after mid-August 2020 which reported a number of clusters from religious groups as publicly available data were limited. Third, recall bias could affect the description of symptom onset and the exposure period of the infectee. Fourth, asymptomatic cases who were not identified and the imperfect sensitivity of the RT-PCR test may affect the estimated transmissibility in our study. Fifth, local public health authorities may identify the infection source after publishing the case-information, which may overestimate the proportion of unlinked local cases. Sixth, due to data limitation, we have not considered the spatial transmission heterogeneity along with the temporal variations in this study. Lastly, the effects of seasonality on SARS-CoV-2 were not considered which may partly explain the change in transmissibility [
42].
Our study has several strengths including the estimation of daily
Rt of SARS-CoV-2 in South Korea from illness onset data, whereas previous studies [
6,
43,
44] estimated it from the daily confirmed cases, which might be subject to reporting bias [
45]. Furthermore, we estimated
Rt using the serial interval distributions, which is evaluated by constructing transmission pairs on this illness onset data. Whereas, the earlier estimates were based on the approximation of serial interval distributions, evaluated for different data and locations, mostly used the early finding of serial interval in China which did not even include the pre-symptomatic transmission [
46] and might have been different over time (here for two epidemics) [
22]. Third, we included all local clustered cases to estimate
Rt to better characterize the changes of transmissibility after relaxing social distancing measures which did not include earlier study [
5]. Finally, our study has an added value over our previous report [
5], providing a more detailed interpretation of the transmission dynamics by accounting for local clustered cases [
36], and included changes in transmissibility after relaxing social distancing measures in South Korea.
Publisher’s Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.