Previously, I transcribed about 1,600 Oregon Covid-19 death records by hand and discovered the public facing presentation of deaths was misleading – it had time shifted past deaths into the present.

This made it look like January was the worst month for deaths, and that deaths were getting worse! OHA’s Director of Public Health made a statement that said we were experiencing the worst reported deaths so far. Except she was referring to the “reports”, which always lag, by a lot, and not the actual dates of death.

In reality, deaths peaked from in November to early December.

Yesterday, OHA acknowledged this by adding a new chart to their web site, as I had suggested to them. This new chart confirms that peak deaths occurred in late Nov/early Dec.

The data I analyzed has now unearthed another interesting pattern in the data.

  • Deaths actually peak roughly simultaneously with the peak in daily new cases.
  • A very good proxy for peak day of deaths is ICU bed usage.

This is contrary to the conventional wisdom that says deaths peak 3 to 4 weeks after the peak in new cases.

Conventional wisdom is we see a peak in positive test cases, then a peak in hospitalizations, then a peak in ventilators and then deaths peak sometime after that.

First you get sick, then if you get really sick you go to the hospital, and there, if you get really really sick you end up in the ICU – and if it goes downhill from there you end up on a ventilator and die sometime later.

The general perspective is there a 3 to 4 week delay from diagnosis to death. (And if we could see individual data, which we can’t, this may be true on an individual base – but we have no way to know.)

The data for Oregon, after correcting for the now acknowledged data reporting issue of time shifting deaths to the present, shows that deaths peak roughly at the same time as new cases.

Oregon Data

  • November 30, 2020 –> Peak day of hospitalizations
  • December 4, 2020 –> Peak day of positive test cases
  • December 9, 2020 –> Peak day of ICU beds in use (range 12/8 to 12/12)
  • December 9, 2020 –> Peak day of deaths, by actual date of death (range 11/25 to 12/9)
  • December 15, 2020 –> Peak day of ventilators in use (range 12/15 to 12/20)
  • December 15, 2020 and January 12, 2020 –> Peak days of reported deaths (range, 12/15 and 1/12, then 1/26, 1/5, 1/13, 1/14)

Look closely at the above list:

  • Peak days of death were coincident with peak hospitalizations, peak positive test cases and peak ICU – in fact, a near perfect match with ICU bed usage.
  • Peak days of death occurred before peak days of ventilator usage
  • Reports of deaths lag actual deaths by about 4 weeks

The conventional wisdom the deaths peak about 3 to 4 weeks after positive daily test cases is false.

Deaths are peaking almost simultaneously with peak cases.

There is not now any explanation for why this is true – however, figuring that out might offer important insights into the disease progression.

If this holds true across the country – and there is strong evidence to suggest it does – then the U.S. hit peak deaths somewhere from late December to early January.

The peak death numbers you see now at end of January are catching up with lagging death reports from weeks ago.

This is, actually, very good news, true.

CDC Chart showing peak daily new cases (updated Jan 28, 2021)

On a national level, peak ICU bed usage occurred during the first two months of January.

This suggests nearly half of the forecast additional 90,000 deaths over the next 4 weeks are, in fact, including people who already died in late December and January.

Furthermore, there was no holiday surge, there was no “surge upon surge” at all. Surges do not show up anywhere in any data set I have looked at. As noted previously, even California peaked before Christmas.

The above is based on limited data made available to the public. This shows how citizen data analysis can help in understanding this pandemic. But this work can only occur if public health makes data readily available to the public. Historically, such data has been difficult to obtain, and some, such as the State of California, deliberately hid their data from the public.


Last fall, Oregon discovered it had counted positive test cases incorrectly and thus had calculated the state’s positivity rate incorrectly. As the positive rate shot up to about 16%, the state criticized citizens for “not complying” and threatened us with more restrictions.

I saw the error in my chart but assumed I had done something wrong. At the same time, a Portland TV station spotted the error and determined what OHA had done wrong. The orange line should not have taken off to space as it had done.

OHA corrected the error – and corrected all positivity rates back to the start of the pandemic.

Overnight, our positivity rate was cut in half. The threats and criticism of our behavior went away and OHA thanked us for having done a good job in complying with guidance.

This too illustrates why it is essential that the pubic be allowed to audit government agency data.

2 thoughts on “Another interesting finding from my Oregon public health data analysis”

Comments are closed.