On Sept. 9, 2020, UC Berkeley School of Public Health researchers reported that probability analysis showed the U.S. may have experienced more than 6.4 million cases of COVID-19 by Apr. 18, 2020. The study was published in Nature Communications. That is nine time more than the number of confirmed cases in the same period, which was 721,245.

The first known case of COVID-19 in the US was confirmed on January 21, 2020. For the first few months of the pandemic, the US Centers for Disease Control recommended that testing be prioritized for patients in the hospital who presented moderate or severe symptoms. However, studies suggest that 30–70% of individuals who test positive for the virus present with mild symptoms or may have none at all.

Berkeley Public Health’s Benjamin-Chung, Sean Wu, and colleagues estimated the total number of SARS-CoV-2 infections in each US state from February 28 to April 18, 2020, using a probabilistic bias analysis to account for incomplete testing and less than 100% test accuracy. The authors estimate that there were 6,454,951 cases of SARS-CoV-2 infection (19 per 1,000 people). This estimate is about 9 times larger than the number of confirmed cases during the same period (2 per 1,000 people) and suggests that 89% of infections were undocumented. The majority of this difference (approximately 86%) was due to incomplete testing, with the remainder due to limited test accuracy.

The authors found that COVID-19 incidence was highest in the Northeast, Midwest, and the state of Louisiana when using confirmed case counts or the estimated number of infections. Underestimation of the number of cases was more common in Puerto Rico, California, and some southern states. In 33 states, the estimated number of infections was at least 10 times higher than the number of confirmed cases.

The authors note that their methodology does not incorporate a transmission model and so they are unable to make forecasts about the spread of the virus. However, they argue that their method provides a more realistic picture of infection burden at a given point in time.