Cases and Clusters of SARS-CoV-2 in Hong Kong
As of 28 April 2020, there have been a total of 1,037 laboratory-confirmed cases of SARS-CoV-2 infection in Hong Kong, and one probable case based on clinical and epidemiological features. The first case of SARS-CoV-2 infection in Hong Kong was confirmed on January 23, 2020, imported from Hubei, China (Onset January 18, 2020. The first local cases without known travel history were confirmed on February 4, 2020 among a family cluster with symptom onsets ranging from January 22, 2020 to February 4, 2020 indicating potentially undocumented community transmission had occurred in mid-January 2020 prior to the first travel restrictions implemented from January 25, 2020 (Figure 1). Stable numbers of sporadic and cluster cases were detected until early March when a substantial increase in the number of imported cases was observed (Figure 1). Cases continued to rise until a total ban on non-resident entry and mandatory 14-day quarantine for all arrivals was implemented. A rise in the number of local cases including clusters initiated by local or imported cases saw the implementation of various physical distancing measures such as restrictions on gatherings and bar closures to prevent community transmission (Figure 1).
Overall the majority (51.9%; 539/1,038) of SARS-CoV-2 infections in Hong Kong have been associated with at least one of 135 known clusters. The median cluster size was two and the largest involved 106 local cases. The remaining 38.9% (210/539) of cluster cases solely involved imported cases where no onward local transmission could be identified. Most clusters comprised only imported cases (82/135; 60.7%), followed by clusters initiated by an imported case (30/135; 22.2%) and clusters initiated by a local case (23/135; 17.0%). Among the 499 sporadic cases not linked to any cluster, 90.0% of infections were acquired overseas (449/499), while the remaining 50 (10.0%) had no history of travel and could not be linked to any other case either as a source or subsequent generation. Of cluster cases, 225 (41.7%; 225/539) belonged to clusters initiated by another local case, compared to 19.3% (104/539) of cluster cases which belonged to clusters initiated by an imported case. The composition of clusters is described further in Supplementary Table 1.
Transmission Pairs & Individual variation in SARS-CoV-2 transmission
From the 539 infections that occurred in clusters, all 210 solely imported cases (210/539; 38.9%) were excluded from subsequent analysis due to uncertainties concerning transmission within each cluster whilst overseas. Within the remaining 53 clusters initiated by a local or imported infection, 245 (245/329; 74.5%) could be linked resulting in 171 unique infector-infectee transmission pairs with 94 unique infectors. The age of distribution of infectors and infectees was not significantly different (Supplementary Figure 1A) however a positive trend by age can be seen in the transmission pair matrix (Supplementary Figure 1B).
Figure 2A shows the empirical serial distribution between all infector-infectee pairs and fitted normal and lognormal distributions. The median serial interval was 4 days (IQR 3 – 8 days), and the mean of the fitted normal distribution was 5.6 days (standard deviation 4.3 days). We observed at least eight instances of likely pre-symptomatic transmission where symptom onset of the infectee preceded that of the infector by one day (N=2) or occurred on the same day (N=6). Thirty-four unique infectors (35/94; 37.2%) were linked to two or more secondary cases, and the largest number of individual secondary cases was 11. From the empirical offspring distribution and fitted negative binomial distribution shown in Figure 2B, we estimated an observed reproductive number (R) of 0.58 (95% CI: 0.45 – 0.71) and dispersion parameter (k) of 0.45 (95% CI: 0.31 – 0.76). Given a superspreading threshold of 6-8 secondary cases, we directly observed 2-4 SSEs where the source was known. Likelihood analysis based on final cluster sizes increased the estimate of R to 0.75 (95% CI: 0.6 – 0.96) and decreased estimates of k to 0.37 (95% CI: 0.16 – 1.16). Based on these estimates we determined a coefficient of variation of 2.5 and inferred that approximately 20% of SARS-CoV-2 infections are responsible for 80% of all transmission events in Hong Kong (Table 1). Epoch analyses by wave (before or after March 1, 2020) showed similar estimates of R and k (Supplementary Table 2) from their respective offspring distributions (Supplementary Figure 2). A peak however in the number of index cases per day could be seen in wave two (Supplementary Figure 3). Parameter estimates and distributions for all fitted distributions are shown in Supplementary Tables 3 & 4, and Supplementary Figures 4 & 5.
Chains of SARS-CoV-2 transmission in Hong Kong
The largest local cluster was 106 cases and was traced back to multiple social exposures among a collection of bars across Hong Kong (Figure 3A). Evidence suggested this “bar and band” cluster originated in Lan Kwai Fong among a few staff and customers before being spread to additional venues by a number of musicians, however the source and chains of transmission between many bar cases could not be determined from epidemiological data. The earliest reported exposure was reported on March 7th among two customers who later presented with symptoms on March 11th (Supplementary Figure 6A). The earliest symptom onset was among two staff members on March 10th & 11th, however neither case was confirmed until March 24th and 25th and neither reported their history of exposure at the bar during this time. Given the serial intervals between these four cases and later cases (Supplementary Figure 6A), it is possible that an undetected source may have initially infected these staff and customers on March 7, before being spread to the musicians. The earliest onset among the musicians was on March 17, with most subsequently affected customers reporting exposures between March 17 and March 20 (Supplementary Figure 6C), which is suggestive of at least one or more probably SSEs within this exposure period. The bars were voluntarily closed from March 23 for cleaning prior to mandated closures from April 3. Of the 73 primary bar cases, 39 customers, 20 staff, and 14 musicians were infected, while the remaining 33 infections were secondary, tertiary or quaternary contacts (Supplementary Figure 6A). This single outbreak accounts for 10.2% (106/1,038) of all cases in Hong Kong regardless of the source, but 30.4% of all local cases acquired in Hong Kong (106/349). Evidence of asymptomatic transmission was determined in a single case (a staff member at the bar) who was placed in quarantine before their wife subsequently presented with symptoms and tested positive. Supplementary Figure 6B shows the age distribution of all cases associated with this cluster.
Figure 3B describes a cluster of 21 cases linked to a wedding. Ten cases resulted from a previous social exposure (SSE), of which four cases subsequently attended the wedding. Individual transmission pairs within the wedding could not be determined, however there were at least seven secondary infections, and therefore another potential SSE, and an additional two tertiary cases among family members of the wedding guests. A final potential SSE of undetermined origin was associated with religious activities at a local temple and resulted in 11 primary cases and 18 recorded cases total (Figure 3C). Cases reported multiple exposures over a number of days such that a single point source exposure was unlikely. Six secondary cases were linked via family exposures. The last case who worked at the temple was also infected however remained asymptomatic. It is unknown if this case was the source of the temple exposures or was infected by the undetermined source/s.
Figure 3D shows all other local cluster cases and chains of SARS-CoV-2 transmission in Hong Kong. In total, 51 subsequent cases were quarantined. Among terminal transmission cases, 26.9% (45/167) ended in quarantine (excluding five quarantined bars cases whose terminal transmission position was ambiguous; Figure 4A; Supplementary Table 5). The odds that a terminal case ended in quarantine was 16.97 (95% CI: 2.27 – 126.69), while the probability that a case was terminal and quarantined was 21.0% (45/214) compared to 57.0% (122/214) for cases who were terminal but not quarantined (Supplementary Table 5). Transmission within families was the most frequent event among all recognised transmission pairs (96/171; 56.1%) followed by transmission in social settings (55/171; 32.2%) and at work (20/171; 11.1%). However, there was very strong evidence that the number of individual secondary cases resulting from social exposures was significantly higher compared to work or family exposures (μ = 2.89 vs. 1.36; p < 0.001, negative binomial regression model).
Relationship between delayed case isolation and potential COVID-19 cluster size
Figure 4A shows the median individual delay from symptom onset-to-confirmation/isolation among SARS-CoV-2 infections in Hong Kong by cluster membership on the basis of size. We found that the delay from symptom onset to confirmation among cluster cases did not predict the overall cluster size although if excluding the two largest clusters there was a slight increase in onset-to-isolation delays in some of the larger clusters (Figure 4A). Delay from symptom-onset to isolation among infectors similarly was not associated with the number of secondary cases or the overall size of each cluster (p = 0.955, linear regression; Figure 4B).