The next chart graphs the daily change in U.S. hospital census. This shows that the drop off at right, is not only steep but is 2-3 times steeper in the downwards direction than when cases were previously climbing. This is fantastic news!
When we look at the data in this way, we see that the peak in increasing hospitalizations occurred in late November/early December. This has critical ramifications for data interpretation …
Similarly, we can look at the first derivative of the 15-day moving average of daily new cases in the United States. As you can see, the change in daily new cases also basically peaked in mid-November.
We can see, from the above, that national new cases were also peaking co-incident with the peak in hospitalizations. While hard to tell from the chart, the peak in daily new cases (15-day smoothed) was Nov 15th (but note that smoothing shifted this to a later date).
The peak in early January is not so much a post holiday surge but is due to colleges and universities testing 100% of staff and both on and off-campus students (at some campuses). When you test a large number of asymptomatic individuals using diagnostic tests, you get a lost of false positives. In my state, these tests used high FP rapid tests and did not follow recommended guidelines to do a 2nd test on asymptomatic positives.
Surprisingly – to those not paying attention – peak deaths were likely to have occurred, on average, in December.
There was no post holiday “surge upon surge”. This is a myth that persists in spite of significant evidence to the contrary.
As I showed for my state – and the State’s public health department agreed with my analysis – their data presentation implied peak deaths occurred in January, but when their data was charted using “date of actual death” instead of “day the death report arrived in our office”, the actual peak of deaths was early December.
In fact, peak deaths occurred the same day as peak ICU utilization. And peak hospitalization day was very close to the peak in daily new cases.
This is contrary to the conventional wisdom that first you get sick, then you go the hospital, then you go to ICU, and then you die. This created a widely assumed meme that “reports of deaths” peak 3-4 weeks after peak daily new cases.
But this is not true. In my state, the median time from diagnosis to death is 14 days. Another epidemiological myth is shattered.
The chart above, combined with what we now know about my state’s own data strongly implies that peak national deaths -based on actual dates of death – likely occurred in December or perhaps early January.
The daily reports you are seeing now are weighted towards deaths that occurred some time ago, and then do to 7-14 day smoothing, are further time shifted into the future. This means we are now well past peak deaths. This is excellent news.
Related: The chart of daily census change shows again that the peaks came before Thanksgiving and long before Christmas holidays. Epidemiologists remain stuck on holiday “surges upon surges”. These claims are false.