Abstract
We analyzed COVID-19 data through May 6th, 2020 using a partially observed Markov process. Our method uses a hybrid deterministic and stochastic formalism that allows for time variable transmission rates and detection probabilities. The model was fit using iterated particle filtering to case count and death count time series from 55 countries. We found evidence for a shrinking epidemic in 30 of the 55 examined countries. Of those 30 countries, 27 have significant evidence for subcritical transmission rates, although the decline in new cases is relatively slow compared to the initial growth rates. Generally, the transmission rates in Europe were lower than in the Americas and Asia. This suggests that global scale social distancing efforts to slow the spread of COVID-19 are effective although they need to be strengthened in many regions and maintained in others to avoid further resurgence of COVID-19. The slow decline also suggests alternative strategies to control the virus are needed before social distancing efforts are partially relaxed.
Highlights
Since its initial outbreak in Wuhan, China in late 2019 and early 2020 [1], COVID-19 caused repeated rapid outbreaks across the global from February through April 2020
As social distancing induces high costs to both society and individuals, plans to relax social distancing are discussed. Changes in both the transmission rates and detection probabilities over time coupled with stochasticity due to reporting delays makes differentiating between truly subcritical dynamics and reduced transmission difficult
We model the spread of COVID-19 as a partially observed Markov process with real-valued states S, E, I, and R to describe the latent population dynamics, and integer-valued states C0, Y1, D0:3, and Y2 to model sampling into the data
Summary
Since its initial outbreak in Wuhan, China in late 2019 and early 2020 [1], COVID-19 caused repeated rapid outbreaks across the global from February through April 2020. The extremely rapid spread of COVID-19 in China [2] does not appear to be an anomaly: the disease has shown a short doubling time (2.4-3.6 days) outside of China as well [3]. As social distancing induces high costs to both society and individuals, plans to relax social distancing are discussed. Changes in both the transmission rates and detection probabilities over time coupled with stochasticity due to reporting delays makes differentiating between truly subcritical dynamics and reduced transmission difficult
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
