Transmission
Our estimates of SARS-CoV-2 transmissibility in South Korea vary substantially over time (Fig. 2). We estimated the time-varying reproduction number Rt using previously published methods [38] for cases between March 30th and July 13th, since the relative contributions to Rt from increased testing and increased transmission cannot be disentangled early on in the epidemic and imported case numbers were not initially available. Using a 7-day sliding window and confirmed cases by date of report, we estimated that Rt initially dropped below 1 in early April before increasing above one twice during May, first up to a maximum of 1.94 (95% CrI, 1.64–2.27) and then again in late May up to 1.56 (95% CrI, 1.37–1.79) (Fig. 2). These increases in estimated Rt coincide with the growth of highly localised transmission clusters in the Seoul Metropolitan Region (Seoul, Incheon city, and Gyeonggi province). In early June, Rt dropped below 1 again where it remained until the end of our study on July 13th.
Interventions implemented by South Korea
Social distancing
Unlike other countries affected by COVID-19, South Korea has not implemented a national “lockdown”. However, similar measures have been applied in a short-term localised manner in high-incidence regions such as Daegu where residents were asked to refrain from leaving their homes for at least 2 weeks after the large cluster of cases associated with the Shincheonji religious group was reported [6] (Fig. 1a). KCDC reported that movement decreased nationally by 38% during February 24th to March 1st as compared to January 9th to 22nd before the first case had been detected in South Korea [46, 47]. This decrease in movement is corroborated by Park et al. who found that the daily traffic volume decreased by 80% and 50% in Daegu and Seoul, respectively, following the identification of the Shincheonji cluster [47]. Schools were closed on February 23rd nationally, and the country transitioned to remote learning until schools reopened gradually between May 20th and June 3rd [12, 14, 31, 48]. On March 22nd, in response to the ongoing emergence of local small clusters, imported cases, and a reported increase in population movement up to 28% below baseline [46], the Korean government implemented a stronger social distancing campaign nationally (Fig. 1a). People were asked to leave their houses only for daily necessities, healthcare, and commuting to work, and many community spaces were closed with the potential for non-compliant facilities to be handed administrative orders [49, 50]. On April 19th, low-contact outdoor facilities and churches could reopen. On May 6th, South Korea transitioned again with the intention of establishing long-term routine distancing [34, 51]. The KCDC criteria for sustaining the routine distancing policy include low daily case numbers, low incidence of detecting apparently sporadic cases, and detection of > 80% of new cases in individuals who are already in self-quarantine. These conditions were being met until the increase in cases in the Seoul Metropolitan Region led to the daily threshold being exceeded on May 28th [52]. In response to increased local transmission, enhanced epidemic control measures have been introduced in the region until daily case numbers are back below 10 [13, 53]. High-risk facilities including nightlife venues, karaoke rooms, and PC cafes are subject to gathering bans or limited operation with potential legal action against non-compliance. Residents are requested to not attend social gatherings or venues frequented by lots of people. Following further transmission in churches, additional infection control guidelines became mandatory on July 10th including instruction not to hold face-to-face meetings outside of regular services, to check symptoms upon entry, and not to partake in singing or food sharing during services [22].
Testing
South Korea has a population of 51 million [54]. By licencing private companies and clinics to conduct tests early in the epidemic, testing capacity was rapidly expanded from 3000/day on February 7th to 15,000–20,000/day with a turnaround time of 6–24 h by the end of March [7]. As of March 25th, there were 118 institutions capable of conducting tests for COVID-19 including KCDC, 4 National Quarantine Stations, 18 Research Institutes of Public Health and Environment (RIPHEs), and 95 private medical laboratories and hospitals [4]. Testing protocol in South Korea changed multiple times throughout the outbreak (Fig. 3 and Additional file 1: Table S1, Table S2) [35, 55,56,57,58]. Initially, only symptomatic individuals with a history of travelling in areas with known infections, or contact with confirmed cases were tested [59]. However, the policy changed on February 20th to include testing of symptomatic individuals based on the physicians’ judgement regardless of travel history and to test close contacts of confirmed cases regardless of symptoms if deemed necessary [60]. This coincided with the Shincheonji religious group cluster and explosive growth in cases. The average number of positive cases per test increased during this period (Fig. 4). Mass testing has also been used in high-risk facilities such as hospitals and care homes in high incidence areas since March 18th. In Daegu, a total of 33,610 high-risk individuals have been tested with 322 testing positive as of April 4th [51]. As case numbers decreased, the average number of cases per test decreased again until the mid-May increase in cases, which resulted in another increase in the average number of cases per test. We are unable to infer the delay from symptom onset to testing from our data set collated from the KCDC daily press releases. A study of the first 28 patients (identified from January 20th to February 10th) estimated the mean delay from symptom onset to diagnosis to be 5.2 days (range 0–16 days) [61]. We did not find data on delays to testing later in the outbreak.
Identification of clusters and contact tracing
A high proportion of cases have been epidemiologically linked throughout the outbreak. Case-based contact tracing in South Korea can be separated into (i) cluster investigation utilising targeted mass testing of hospitals and communities and (ii) identification and follow-up of individual cases and their contacts (see Additional file 1: Section 2. Contact tracing for more detailed protocol) [30, 62,63,64,65,66,67,68]. Clusters now account for 66% of South Korea’s cumulative cases nationally, while individual case-based contacts account for only 10% (Fig. 5a). The large cluster associated with the Shincheonji religious group accounts for 39% of cases nationally. Although focused in Daegu, the cluster included group members in most regions of the country and made considerable contributions to overall case numbers in Gyeongbuk, Gyeongnam, Gangwon, Ulsan, and Gwangju (see Figure S1 for the regional breakdown of case type by epidemiological link or origin). Later in the epidemic, after the March 25th, the Shincheonji cluster contributed very few new cases. Other smaller clusters make up the further 27% of cases (ranging in size from 2 to 196 people). Prior to May 8th, these clusters were mainly in workplaces, hospitals, and churches, with most clusters of > 100 cases occurring in hospitals, where mass testing protocols were in place [69]. The relative contribution of local transmission to new cases steadily decreased until 82% of new cases in the fortnight prior to May 8th were imported; however, the Itaewon nightclub cluster [35] brought the relative contribution of local clusters back up to 40% of new cases reported in the fortnight ending on of May 11th (Fig. 5b). Since then, the majority of reported cases have been linked to clusters in the Seoul Metropolitan Region, largely associated with entertainment venues, religious events, after school private academies, and a string of door-to-door promotional events [13, 53]. The contribution of local transmission clusters to overall fortnightly case incidence reached a peak of 75% in June before beginning to decline back down to around 50% at the end of the study period (Fig. 5b).
Isolation
Isolation of cases and self-quarantine of contacts has been a consistent feature of the South Korean response. As of July 13, there were 986 confirmed cases currently isolated (Fig. 6). Although we were unable to identify accessible data sources that reported the number of contacts isolated over the course of the outbreak, media outlets reported that approximately 30,000 people were in self-quarantine on March 6th [70], 27,000 on April 2nd [71], and 46,000 on April 7th [72]. In addition, a study of contact tracing in South Korea between January 20th and March 27th found that for the 5706 index cases with at least one contact, 59,036 contacts were identified (who according to KCDC policy would have been required to self-quarantine) [73]. Based on the severity and risk factors, confirmed cases were either isolated in a hospital, at home, or in a residential treatment centre (RTC). Cases were isolated in a hospital in their own rooms or grouped with patients with the same clinical characteristics when ventilation and structural barriers between patients were available. Cases were isolated in RTCs when symptoms were too severe for home isolation, but did not require hospitalisation, when home isolation was not possible because no individual room was available or because they live with a high-risk person [66]. Medical staff in RTCs and public health managers (for home isolation) monitored and recorded the patient’s symptoms twice per day, transferring them to medical facilities if necessary. Close contacts of confirmed cases were asked to self-quarantine for 14 days and monitored daily.
We were unable to estimate the delay to isolation of a confirmed case from the KCDC press release data. Another study, which fitted gamma distributions to case data of 211 confirmed cases, estimated the mean delay from symptom onset to isolation of cases as 4.3 days, before the red alert warning on February 23rd and 3.3 days after [74]. The same study estimated the mean delay from exposure to isolation as 7.2 days before the red alert and 6.5 days after.
Confirmed cases were required to test negative twice in 24 h before release from isolation [66]. Close contacts with no symptoms in the 14 days following contact were released from home quarantine. If close contacts were health care workers or caregivers, they were released pending a negative at the end of the quarantine period.