A Dire Prediction


By: Kyle Freese, PhD, MPH

Daily summary 4/01/2020

We learned today that the U.S. might experience at least 100,000 deaths from COVID-19, even with the current social isolation measures in place. Many epidemiologists’ and data scientists’ statistical models agree. Delaying widespread isolation measures can no longer be an option- the entire country must heed the advice of public health officials to prevent an unstoppable downward spiral.

Figure 1 shows the number of new, confirmed cases of COVID-19 per day for the U.S., Italy, and Singapore. The y-axis is presented on the log scale, so each horizontal hatch represents a 10-fold increase in the number of new, daily cases. These three countries were chosen to illustratively compare the U.S. with two, vastly different stories for how COVID-19 has progressed and controlled. Singapore on one end of the spectrum, which was known for aggressive interventions and Italy on the other, which was devastated by the outbreak. At its peak, Singapore had slightly more than 100 new cases per day; Italy had nearly 10,000; yesterday, the U.S. reported 26,400 new cases. Given, these are not adjusted for population, but in terms of crude numbers, they point to a troubling reality; if the trend continues, one can imagine the additional strain that will be placed on healthcare systems around the country.

Figure 1. Number of new, daily cases for March, 2020 in the U.S., Italy, and Singapore.

Case Fatality Rate Revisited

Figure 2: Case fatality rate by age category. Aggregate data from mainland China. n=44,672 cases

In order to precisely determine the case fatality rate of a pathogen, we must have precise ascertainment of the numerator (the number of deaths from the disease in a given time period) and the denominator (the number of cases of the disease in the same time period). The numerator is relatively straight forward. Supposing most of the known deaths occur in hospital settings and are confidently attributed to COVID-19, in most cases the numerator is [likely] fairly accurate. The denominator, however, is more difficult to ascertain. For illustration, let us take an artificial population of 10,000 individuals. In our sample population, 6 people have died over the past month in the hospital from complications from COVID-19. For simplicity, we will assume that no deaths due to COVID-19 went unreported or misdiagnosed. First, we know all the cases in the numerator are also in the denominator. In addition to these deaths, there were 95 confirmed cases of COVID-19 being treated in the hospital. If all the known cases in the population were accounted for, we would have a case fatality rate of 5.9% (6/101). But what about the cases that did not require hospitalization? There are few available data, but let us estimate that 10% of COVID-19 cases are serious enough to require hospitalization. Now we have an additional 909 cases of COVID-19 in the population that weren’t included in our original case fatality calculation. Accounting for these cases, the case fatality rate is 0.6% (6/1,010). Furthermore, we must assume that there are asymptomatic individuals within the community. For conservative estimates, let us add 2% to our estimated number of cases (new case total=1,030). Assuming 2% more cases, we arrive at our true estimate of case fatality rate in this population: 6/1,030=0.58%). This points to the importance of adequate testing in a community; without widespread testing across diverse populations, we cannot present these epidemiologic metrics with a high level of confidence.


Verity R, Okell LC, Dorigatti I, et al. Estimates of the severity of coronavirus disease 2010: a model-based analysis. The Lancet Infectious Diseases. March 30 2019; DOI: 10.1016/S1473-3099(20)30243-7.

For additional references, resources, or questions, please email Dr. Kyle Freese at