Basic prediction methodology for covid-19: estimation and sensitivity considerations

Tom Britton

One of our main conclusions is that any prediction of the number of infected, using data from cumulative case fatalities, rely heavily on accurate knowledge of the infection fatality risk f . Without accurate knowledge of f , the prediction of the total case fatality once the epidemic outbreak is over, is equipped with a similar multiplicative uncertainty. It is hence of utmost importance to obtain more accurate estimates of f . The best way to obtain such estimates, now that seroprevalence tests are becoming available, is to conduct tests on a random sample of a community in which the outbreak is (close to) over. f is then simply estimated as the ratio of the total number of case fatalities and the estimated population number of people testing positive.