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

Markus Mueller; Peter Derlet; Christopher Mudry; Gabriel Aeppli

We point out before proceeding further that this is a contribution from physicists that makes simplifying assumptions inconsistent with details of medical and epidemiological reality to obtain some key estimates and Dynamics of the pandemic with and without a feedback and control scheme in place, as measured by the fraction i of currently infected people (logarithmic scale). After the limit of the health system, i c , has been reached, a lockdown brings i down again. The exponential rate of decrease is expected to be very slow, unless extreme measures are imposed. The release of measures upon a reboot is likely to re-induce exponential growth, but with a rate difficult to predict. Three possible outcomes are shown in blue curves in the scenario without testing feedback, where the effect of the new measures becomes visible only after a delay of 10-14 days. In the worst case, i grows by a multiplicative factor of order 20 before the growth is detected. A reboot can thus be risked only once i ≤ i * * ≡ i c /20, implying a very long time in lockdown after the initial peak. Due to the long delay until policy changes show observable effects, the fluctuations of i will be large. Random testing (the red curve) has a major advantage. It measures i instantaneously and detects its growth rate within few days, whereby the higher the testing rate the faster the detection. Policy adjustments can thus be made much faster, with smaller oscillations of i. A safe reboot is then possible much earlier, at the level of i ≤ i * ≈ i c /4. illustrate the basic principles of feedback and control as applied to the current pandemics. When reduced to practice, special attention will need to be paid to all aspects of the testing methodology, from the underlying molecular engineering paradigm (e.g., PCR) and associated cost/performance trade offs, to population sample selection consistent with societal norms and statistical needs, and safe (i.e., not risking further infections) operation of testing sites. Furthermore, in preparation for the day when more is known about the immune response to COVID-19 and possible vaccines, we plan to revise our models for feedback derived from a reliable immunoassay with well-specified performance parameters, such as lag times with respect to infection.