Quantifying the effect of quarantine control in Covid-19 infectious spread using machine learning

Raj Dandekar; George Barbastathis

Owing to the relaxed quarantine and isolation policies in the US in its initial stages post the infection spread, our model converges to Q(t) ≈ 0.4 − 0.6 ( Figure 8b ) which is the smallest compared to other regions. Even though the effective R t is still greater than 1 as of April 1 2020 (Figure 8c ), its growth has started to show a decreasing trend and we expect the infection to start showing stagnation with R t < 1 by 20 April 2020 if the current US policies continue without change (Figures 9b, 10a ). This will be accompanied by a continuous ramp up of quarantine policies (Figure 9a ). At its peak, we forecast the infected count to reach approximately 600,000 before stagnation, again assuming no change in US policies. Our mixed model analysis for USA, employing Q(t) learnt from the models of Wuhan, Italy and South Korea in the USA model starting from 1 April 2020, reveals that stronger quarantine policies (Figure 10b ) might lead to an accelerated plateauing in the infected case count, as shown in Figure 10a , and subsequently smaller infected case count. On the other hand, in agreement with National Institute of Allergy and Infectious Diseases estimates (Miller 2020) , we forecast that relaxing or abandoning the quarantine policies gradually over the period of the next 17 days may well lead to ∼ 1 million infections without any stagnation in the infected case count (Figure 10a ) by mid April 2020.