Evolving epidemiological characteristics of COVID-19 in Hong Kong, January to August 2020.


We retrieved the official case series and the Apple mobility data of Hong Kong in January-August 2020. The empirical CDs and SIs were fitted to theoretical distributions, and factors associated with their temporal changes were quantified in terms of percentage contribution (PC), which was the percentage change in the predicted outcome (from multivariable regression models) relative to a predefined comparator. Rt was estimated with the best fitted distribution for SIs.

To characterize the epidemiology of the first two epidemic waves of COVID-19 in Hong Kong by (i) estimating the CDs, SIs, effective reproductive number (Rt), and the proportion of asymptomatic cases; (ii) identifying factors associated with the temporal changes of the CDs and SIs; and (iii) depicting the COVID-19 transmission by age assortativity and types of social settings.

Not applicable.

The two epidemic waves were featured by imported cases and clusters of local cases respectively. Rt rose to peak at 2.39 (wave 1) and 3.04 (wave 2). The proportion of asymptomatic cases decreased from 36.4% (0-9 years) to 12.9% (≥80 years). Log-normal distribution best fitted the 1574 CDs (mean:5.18 days; standard deviation [SD]:3.04) and the 558 SIs (17 negative) (mean:4.74 days; SD:4.24). CDs decreased with involvement in a cluster (PC:10.08-20.73%) and case detection in the public healthcare sector (PC:27.56%; 95% confidence interval [CI]:22.52, 32.33). SIs decreased over time (6.70 days [wave 1] versus 4.35 days [wave 2]) and with the tertiary transmission or beyond (PC: -17.31% to -50.75%), but was lengthened by mobility (PC: 0.83%). Transmission within the same age band was high (18.1%); and households (69.9%), followed by social settings (20.3%), were two most common settings at which the transmission happened.

COVID-19 plagued the globe, with non-pharmaceutical interventions and multiple SARS-CoV-2 clusters hinting on its evolving epidemiology. Since the disease course is governed by important epidemiological parameters, including containment delays (CDs) (i.e. time between symptom onset and mandatory isolation) and serial intervals (SIs) (i.e. time between symptom onsets of infector-infectee pairs), understanding their temporal changes helps to guide interventions.

First, the factors associated with reduced CDs suggested government-enacted interventions (such as public mode of case detection) were useful to achieve outbreak control and should be further encouraged. Second, shorter SIs associated with the composite mobility index called for empirical surveys that disentangle the role of different contact dimensions in disease transmission. Third, the pre-symptomatic transmission and asymptomatic cases reminded the importance of remaining vigilant about COVID-19. Fourth, the time-varying epidemiological parameters suggest the need to incorporate their temporal variations when depicting the epidemic trajectory. Fifth, the high proportion of transmission events within the same age groups supported the ban on gatherings outside of households, and that in household pushed forward the need for residence-center preventive measures.

MIDAS Network Members

This site is registered on as a development site.