Estimating Unreported Deaths Associated with COVID-19

Stear, Benjamin; Hernandez, Kyle; Manian, Vidya; Dutta, Som; Taylor, Deanne; Conley, Catharine

Since early April 2020, COVID-19 has been the third-largest cause of death in the US, becoming the largest single cause of death during the last two weeks of April. This pandemic has strained public health systems in many states, leading to extended delays in reporting deaths to the US Centers for Disease Control National Center for Health Statistics (NCHS), as well as rumors about the accuracy of reported data. To assess these concerns, we have adapted simple financial statistics to evaluate possible anomalies in mortality data and applied them to NCHS and US Centers for Disease Control Case Surveillance (CS) datasets. NCHS data released 21 October record almost 8000 more deaths from COVID-19 in March and April, US wide, than are captured in the CS daily provisional counts of COVID-19 deaths during those months. For five consecutive weeks during June, July, and August, the states Alabama, California, Florida, Georgia, and Texas each reported at least 40 more deaths per week to the NCHS than were captured in CS daily counts for the same period, with Florida and Texas reporting over 1000 and 4000 more COVID-19 deaths, respectively. In contrast, counties in multiple states attribute to COVID-19 fewer than 10% of deaths reported to the NCHS over the period February-September 2020, while simultaneously reporting over twice as many deaths from all causes compared to the same period in previous years. Metrics based on ratio analysis can provide useful information on the reliability of data used for inputs to epidemiological models.