Mortality from COVID-19 in 12 countries and 6 states of the United States

Patrick Brown; Prabhat Jha

is the (which was not peer-reviewed) The copyright holder for this preprint . https://doi.org/10.1101/2020.04.17.20069161 doi: medRxiv preprint Initial infectious disease models estimated total COVID-19 deaths in the USA of 1 to 2.2 million (roughly 40% to 80% over and above the usual annual total of 2.8 million deaths) if no action were taken. 23 These same projections suggested 510,000 COVID-19 deaths in the UK, roughly doubling the current annual total. The Institute for Health Metrics and Evaluation (IHME) model considers physical distancing through May 2020 and has median death estimates lower than ours (Supplementary Appendix). However, the IHME model has some implausible results like suggesting that Netherlands will have 16,000 deaths which is only slightly below their estimate for France and Spain, despite being at least three times larger in population. 22 Moreover, the IHME analyses assumes a symmetrical bell-shaped mortality curve, whereas a skewed curve (which we use) is observed in the actual data. 1-3 Our totals for the USA fall broadly within the range projected by the White House Task Force of 100,000 to 240,000 deaths, 24 and our Canadian estimates are consistent with infectious disease dynamic model results. 25 There is little doubt that some of GCAM forecasts are wrong as undercounts of COVID-19 and delayed reporting are common. [4] [5] [6] For example, the undercount in UK COVID-19 deaths might have been about 70% for some weeks. 4 For France, Italy, the Netherlands and the UK, we could find some record of undercounts and correcting for these substantially raised the total deaths in each country (Supplementary Appendix). However, systematic examination of undercounts across settings is needed to establish if the undercounts have accelerated during the epidemic, which would mean that we have underestimated the speed of the upward trajectory, or if they are more random, in which case our results would mostly be unaffected. Future iterations of GCAM will incorporate undercounts in as many countries as document these. Differences in . CC-BY-NC-ND 4.0 International license It is made available under a author/funder, who has granted medRxiv a license to display the preprint in perpetuity.