Tracking the instantaneous effective reproduction number (Rt) of COVID-19 in real time can assess whether an intervention is effective in reducing infection transmission. However, this is difficult due to a lag of around 9 days between infection and case reporting. This inevitably delays timely public health interventions to mitigate the infection spread. Experts from HKUMed developed a framework that utilised digital proxies to model the transmission of infection in real time, enabling nowcast and short-term forecast of the epidemic.
Key takeaways from the study:
- Transport transactions on the Octopus card – a stored-value card system ubiquitously used by Hong Kongers for their daily public transport and small retail payments – were valid digital proxies for population mixing, given that the Rt estimates were highly correlated with the number of transactions for transport.
- The model could robustly estimate the number of cases in the community that were not yet picked up by surveillance due to the delay between infection and reporting.
- However, it was less accurate at generating short-term forecast on how the epidemic might evolve over time due to the presence of superspreading events (SSEs) that caused a sharp increase of infection numbers, and tended to result in the underestimation of cases.
- On the other hand, the incidence was overestimated when the prevalence of infection was very low.
Conventional methods of using social contact surveys to predict the epidemiology of infectious diseases can be difficult as they require real-time updates of population mixing on a daily basis. Integrating digital proxies that closely model after human mobility and activities into conventional models therefore offers an efficient solution, providing an accurate nowcast and short-term forecast of epidemics.
To read the original article published in Nature Communications, click here.
