Overview
We developed a basecase deterministic compartmental model to simulate COVID-19 epidemic under implementation of interventions and gathering restrictions in Guangdong Province, China from 28 December 2019 to 31 August 2020. We then developed hypothetical scenarios of holding a large-scale international event (event hereafter) involving participants from Guangdong, other provinces in Mainland China, and overseas, and other Guangdong residents not participating the event. Sensitivity analyses were performed to analyze the epidemic impacts of key model parameters.
Study area
Guangdong, a province with 113.46 million population in 2018,[11] has reported 1641 confirmed cases including 8 deaths by 30 June 2020.[12] Imported cases, defined as COVID-19 patients whose infection originated outside Guangdong, accounted for around 76% of confirmed cases.[12] Contact tracing, testing, confirmed case isolation, and quarantine were implemented throughout the epidemic period. COVID-19 epidemic in Guangdong has been under control with less than 2 daily new confirmed cases since 21 March 2020.
Basecase model structure and assumptions
This is an open model with individuals entering and leaving the province (Figure 1, Appendix p.2). Besides quarantining close contacts, health quarantine of inbound travelers from listed origins have been imposed upon travelers’ arrival, and polymerase chain reaction (PCR) testing for SARS-CoV-2 (95% sensitivity) performed. We assumed all symptomatic travelers testing positive were directed to the hospital compartment, while those testing negative or positive without symptoms were directed to the quarantine compartment. After 14-days’ quarantine, individuals diagnosed would be directed to the hospital compartment, while the rest flowed back to the susceptible compartment. The quarantine arrangement applied to travelers who were from a) Hubei Province, China, between 14 February 2020 and 23 March 2020; b) overseas since 1 March 2020; and c) Hong Kong and Macau, both Special Administrative Regions (SAR) of China, as from 27 March 2020. A net proportion of travelers, who were susceptible and not fulfilling quarantine criteria at the time of arrival, entered the susceptible compartment. Infected individuals not fulfilling quarantine criteria at the time of arrival (i.e. not in the list of designated countries) entered either the pre-infectious compartment or symptomatic compartment.
Data Source
Model inputs included reproduction number (R0), biological parameters (latent period, time to recovery in asymptomatic infections, time from hospitalization and intensive care units (ICU) to recovery in symptomatic infections, mortality rate of COVID-19, asymptomatic proportion), demographic parameters (permanent resident population in 2018 in Guangdong, travelers staying overnight in 2018 in Guangdong), policy and healthcare system parameters (contact tracing, quarantine, and testing), and event parameters (eTable 1). Other interventions such as school closure, group gathering restrictions, and business and premises restrictions are absorbed in background and reflected by the proportion of contact rate reduction comparing with pre-epidemic level. Model parameter values or ranges were derived from provincial yearbooks, local government reports, scientific literature and assumptions.[13-18] We fitted model predictions over time to daily number of confirmed cases in the province by 29 March 2020, using mle function under negative log likelihood in R (stats4 package). In this calibration process, we simultaneously varied the proportion of asymptomatic infection and proportion of reduction of contact rate comparing with pre-epidemic level. Modelling results were validated with the daily number of confirmed cases between 30 March and 15 May 2020. IRB approval and a waiver of consent were obtained from Dermatology Hospital of Southern Medical University, China.
Event scenarios
The setting of the event was the biannual Canton Fair, which was organized virtually in 2020 because of the COVID-19 epidemic. Event scenario was developed on the assumption that this Fair was held on-site in a large conference venue in Guangzhou between 1 and 14 June 2020. In the event, we assumed the contact rate among participants doubled with reference to that in the community. Assuming only half of the overseas participants (100,000) in the past two years have joined the event, and the number of local participants remained the same as in the past for Guangdong (38,000) and other provinces of China (100,000), six scenarios were developed to examine the impacts of different pre-event interventions of quarantine and testing (assumptions in Appendix p.2).
- None of participants were quarantined before the event, unless they were contact traced;
- All participants travelling from overseas were quarantined for 14 days before the event;
- All participants (regardless of origins) were quarantined before the event;
- Scenario 2, and all Mainland participants were tested before attending the event;
- Scenario 1, and all participants were tested before attending the event;
- Scenario 1, and all participants were tested before and on day 7 following opening of the event
Model outcomes
The main modelling outcomes were the cumulative number and proportion change of cumulative new infections generated locally (regardless of symptom presentation) above basecase scenario by 31 July 2020, to allow enough time for demonstrating the impact on incidence caused by the event. The secondary outcome was the cumulative number of confirmed cases.
Sensitivity analyses
To account for variability of R0, contact rate, and proportion of asymptomatic infections over time, we have performed 5000 simulations in the basecase model (Appendix p.18). We performed one-way sensitivity analysis around the key parameters in the following scenarios: contact rate reduction from baseline ranged between 0% and 90% since 1 May 2020, prevalence of SARS-CoV-2 in overseas participants (0.00005-0.001) and other provinces’ participants (0.00001-0.005), proportion of all infected individuals in Guangdong who attended the event (0.001-0.1), and testing coverage (0%-100%) for participants in scenarios 4-6. We also varied the contact rate at the event from doubled, tripled and ten-fold with reference to that in the community. To examine the impact of duration with very low contact rate (10%) through intense interventions at epidemic peak (9 February 2020) and the change of contact rate after relaxing the interventions, two-way sensitivity analysis was performed.