Apparent reductions in COVID-19 Case Fatality Rates reflect changes in average age of those testing positive

Numbers of COVID-19 infections are rising in many countries, while death and hospitalisation rates remain low. Infection Fatality Rates (IFR) for individual year classes calculated from seroprevalence data and the ages of those dying in England show a very strong log-linear relationship with age, and allow us to predict fatality rates for those testing positive on each day based on their ages. Since the peak of the epidemic, reductions in the ages of cases account for an eight fold fall in fatality rates. Over the same period, increased testing intensity appears to have increased infections detected amongst the most vulnerable by a factor of at least five, and between 15 and 24 fold in the population as a whole. Together these two factors are sufficient to explain the large observed change in the ratio of deaths to reported cases. We can also use these methods to give a more precise early warning system of future increases in mortality rates than raw case numbers. Although case numbers are currently increasing markedly, a continuing reduction in numbers of older individuals being infected means that the predicted increase in mortality rates is much slower.