LAUSR

Abstract COVID-19 testing strategies are primarily driven by medical need - focusing on people already hospitalized with significant symptoms or on people most at risk. However, such testing is highly biased because it fails to identify the extent to which COVID-19 is present in people with mild or no symptoms. If we wish to understand the true rate of COVID-19 infection and death, we need to take full account of the causal explanations for the resulting data to avoid highly misleading conclusions about infection and death rates. We describe how causal (Bayesian network) models can provide such explanations and the need to combine these with more random testing in order to achieve reliable data and predictions for the both policy makers and the public.