Thank you for these analyses. A definite public service and worthy of study. I do want to share a few things about state data:
After accessing a recent CDC report through the link below, I noticed some alarming inconsistencies in the COVID-19 data reported by Florida. One graph (Figure 1) showed statewide hospital and urgent care discharge diagnosis of influenza (rather than the previously reported COVID 19 as a chief complaint.) At Week 12 there was approximately, 3% discharge of influenza, with the rate plummeting to near 0% for Weeks 14-19. A dramatic decline; a reflection of a sudden cure? There's no reporting of COVID-19, and there's an apparent sense that all influenza has disappeared.
But, a later graph (Figure 4) showed approximately 375 deaths at Week 12 from influenza, supposedly exclusive of COVID-19, plummeting to about 250 deaths in Weeks 15-17. These data are saying that influenza which can include COVID-19 (but not chiefly COVID-19) has plummeted to 0% at discharge, while there are 250 deaths from influenza that are not associated to COVID-19.
So, how can the near 0% discharge coincide with 250 deaths? No deaths at the hospital? And the near lack of deaths is for patients not chiefly with COVID-19? But Figure 4 says there were approximately 250 deaths from influenza without COVID-19. Further, we know that COVID-19 is five to ten times (at least) more lethal than the typical influenza. So, there are inconsistencies between the two graphs, and the graphs are inconsistent with what we know about death rates from the typical influenza vs. death rates from COVID-19. It's likely that the more predominant COVID-19 involved deaths are more likely being counted in the Figure 4 accounting of data, while being ignored in the Figure 1 accounting of data.
Further, when I clicked on Georgia in the United States map in the CDC link, the message was that I had no access to this "public information". Another state with questionable counting is Alabama. The Washington Post has reported that in Alabama , if a patient with COVID-19 dies of cardiac arrest, then the patient is not counted as a COVID-19 death. . Question: If someone is hit by a drunk driver and has a heart attack on the way to the hospital in the ambulance, is that death due to the drunk driver. According to the logic of Alabama, and likely Florida, and other states, these deaths are not due to the drunk driver. See how that logic flies in court when there's a trial of the drunk driver.
Yes, there's a deliberate, dishonest undercount, which is endangering everyone. Further, it's giving many a false sense of security, and given the near incitement by the White House is causing divisiveness rather than cohesiveness in attacking this pandemic.
See link below.
Sincerely,
Mark Y. Czarnolewski, Ph.D. (retired)
11231 Columbia Pike
Silver Spring, 20901.
I got this link for the CDC from Linkedin
Click on:
https://bit.ly/2ViFflZ
Go to : Coronavirus Disease 2019 (COVID-19) Go to U.S. map and click on Florida
Updated May 15, 2020 Key Updates for Week 19, ending May 9, 2020
ILI Activity Levels; Influenza-Like Illness (ILI) Activity Level Indicator Determined by Data Reported to ILINet; 2019-20 Influenza Season Week 19 ending May 09, 2020
------------------------------
Mark Czarnolewski
------------------------------
Original Message:
Sent: 06-01-2020 21:16
From: Robert Agnew
Subject: A "cubic model" for COVID-19 deaths
Update. Using my curve-fitting methodology, daily US Covid case and death trends appear to have bottomed. New York is headed down after a huge surge. But other big states are headed up, California in particular. Bob
------------------------------
Robert Agnew
Original Message:
Sent: 05-28-2020 03:10
From: Jonathan Siegel
Subject: A "cubic model" for COVID-19 deaths
Appreciate this.
An alternative explanation for declining relative death rates might be changes in selection bias. Previously, limited testing capacity meant only people with relatively severe symptoms got selected for testing. But with increased testing capacity, people with milder symptoms are getting tested, along with more asymptomatic people. And perhaps also more false positives. This hypothesis - changes in the set of people selected for testing, not necessarily changes in care - might also explain a decreasing relative death rate.
------------------------------
Jonathan Siegel
Director Clinical Statistics
Original Message:
Sent: 05-27-2020 22:11
From: Robert Agnew
Subject: A "cubic model" for COVID-19 deaths
I found this thread interesting so I took a shot at curve-fitting using R time-series functions. Data is from the New York Times github site and runs through May 26 for Covid cases and deaths. Since this is indeed curve-fitting, I limited myself to 14-day forecasts. Results show some decline in daily cases and a more marked decline in daily deaths. This could make some sense in that hospitals may be doing a progressively better job of keeping people alive. I plan to keep running this script on updated data to see how the trends and forecasts evolve.
------------------------------
Robert Agnew
Original Message:
Sent: 05-05-2020 07:42
From: Jonathan Siegel
Subject: A "cubic model" for COVID-19 deaths
The Washington Post reports that the Trump Administration has rejected a Johns Hopkins based model of the Covid-19 epidemic predicting deaths rising. The article reports that in doing so, the White House is relying in part on a "cubic model" prepared by a White House economic advisor showing deaths sharply decreasing in May.
Nate Silver attempted to reproduce the "cubic model" by putting current death figures into a spreadsheet and fitting a cubic regression. It shows deaths hitting zero on approximately May 15. (After that, being a cubic regression, the curve shows deaths becoming negative.)
Perhaps all this is speculative. It isn't clear what the Washington Post's source meant by a "cubic model", or if Nate Silver's reproduction is accurate. We don't yet know the facts.
But it is nonetheless a matter of concern. If this is accurate, I would respectfully suggest that the ASA, if wants to be a public interest professional association, it should be far, far more concerned about the use of junk science that gives results the audience wants to hear in major public policy decisions with life-and-death public health implications than it has been about such matters as what hour economic results get released or where a government department is located.
If one is allowed to select from the whole gamut of possible curves, one can always find one that fits the data, appears plausible to a casual observer, and extrapolates to the future one wants to see.
This opinion, like all of mine here, is mine only and does not represent my employer.
https://www.google.com/amp/s/www.washingtonpost.com/health/government-report-predicts-covid-19-cases-will-reach-200000-a-day-by-june-1/2020/05/04/02fe743e-8e27-11ea-a9c0-73b93422d691_story.html%3foutputType=amp
https://mobile.twitter.com/natesilver538/status/1257476755574718470
------------------------------
Jonathan Siegel
Director Clinical Statistics
------------------------------