International USR Summit 2022: Education and Action for a Sustainable Future

Predicting the spatiotemporal risk of COVID-19 is the key to formulating place-specific precise control measures. Most methods for predicting COVID-19 risk are based on confirmed/positive cases. However, there is a spatially variant delay of a few days from the symptom onset, when the cases’ infectiousness reaches the peak, to the case is tested positive. The prediction of symptom onset risk thus has a higher potential to support more timely anti-epidemic measures. This presentation introduces the applications of the lately developed extended Weight Kernel Density Estimation (E-WKDE) model for short-term COVID-19 spatiotemporal onset risk prediction. By flexibly incorporating the COVID-19 cases data, virus transmissibility, and mobility data in different granularities, the models have been used to predict the COVID-19 onset risk and evaluate the anti-epidemic measures in various countries, regions, and cities. Studies with the models (e.g., on the early Omicron break in South Africa) found that more strict travel restriction in the epidemic centre and looser epidemic control elsewhere can usually more effectively control the epidemic risk than nation- or region-wide medium control measures, which provides evidence for precise and strict control. The predictions have also been applied to decision-supports, such as identifying the epidemic risk levels in different secondary schools in Hong Kong, and informing the logistic companies in the Guangdong-Hong Kong-Macao Greater Bay Area the epidemic risk of their common destinations.