http://www.drroyspencer.com/2020/04/correcting-recent-u-s-weekly-death-statistics-for-incomplete-reporting/
I am seeing an increasing number of people on social media pointing to the weekly CDC death statistics which show a unusually low number of total deaths for this time of year, when one would expect the number to be increasing from COVID-19. But what most people don’t realize is that this is an artifact of the late arrival of death certificate data as gathered by the National Center for Health Statistics (NCHS).
This first came to my attention as a tweet by some researchers who were using the CDC weekly death data in a research paper pointing out the downturn in deaths in early 2020 and had to retract the paper because of the incomplete data problem. A disclaimer at the CDC website points out the incomplete nature of recent data. While they say that the new totals could be adjusted either upward or downward, it appears that the adjustments are almost always upward (i.e. recent data have a low bias in reported deaths).
As a first attempt to possibly correct for this under-reporting problem, I downloaded the data two weeks in a row (approximately March 30 and April 5, 2020) to examine how the recent data changes as new death certificate data are obtained. I realize this is only one week’s worth of changes, and each week would provide additional statistics. But the basic methodology could be applied with additional weeks of data added.