Using random testing to manage a safe exit from the COVID-19 lockdown

Markus Mueller; Peter Derlet; Christopher Mudry; Gabriel Aeppli

a decrease of the infected population. The smaller the current growth rate k 1 , the longer the time to detect it above the noise inherent to the finite sampling. How long would it take to detect that a release of restrictive measures has resulted in a nearly unmitigated growth rate of the order of k 1 = 0.23 (which corresponds to doubling every 3 days)? Even with a moderate number of r = 15 000 per day, we find that within only ∆t 1 ≈ 3 − 4 days such a strong growth will emerge above the noise level, such that countermeasures can be taken (see Fig. 6 ). During this short time, the damage remains limited. The infection numbers will have risen by a multiplicative factor between 2 and 3. This degree of control must be compared to a situation where no information on the current growth rate is available, and where the first effects of a new policy are seen in the increased number of symptomatic, sick people only 10-14 days later. Over this time span, with a growth rate of k 1 = 0.23, the infection numbers will have grown by a factor of 10-30 before one realizes eventually that an intervention must be made.