On March 4, I had the great pleasure of giving a talk at UW – Superior on my research on cascading expert failure (ungated version here). You can find a video of the talk on my YouTube channel . Thanks to Dr Joshua K. Bedi for hosting and for the Wisconsin Institute for Citizenship and Civil Dialogue for sponsoring the event.
In the talk, I discuss how the decision early in the pandemic to reserve COVID-19 tests to just hospital cases led to cascading expert failure. I failed, however, to bring that story to its conclusion, so I rectify that here.
By restricting tests to hospitals, there was an upward bias in the results making COVID-19 appear far more deadly than it actually was. Those results were fed into models like the now-infamous Imperial College London model, leading to forecasts in the millions over a period of months. This, in turn, fed into the narrative of lockdowns. Even as more and more evidence emerged that the virus was not as deadly as once thought and that the lockdowns actually contributed to the spread, those bad policies persisted.
READER COMMENTS
steve
Mar 20 2025 at 12:44pm
I agree that was a mistake. There should have been a lot of surveillance testing going on. However, the fact is that we didnt have enough tests since we had very poor leadership at upper levels, people who largely had little public health experience but met political litmus tests. We needed a simple test, done quickly and lots of them. Several other countries achieved this. The US, OTOH, decided to try to produce a more complicated test that would also test for other viruses at the same time. That complication lead to failure and greatly delayed our getting tests.
Query- Did you read the Imperial College report. I did, several times. There are several things to note in it. First, they had many tables showing a very broad range of possible deaths (also ICU admissions) depending upon the R value and interventions taken. They didnt predict 500,000 UK deaths but rather a range from about 5,000 to 500,000. It was the people talking about the report that claimed 500,000.
Next, and really important, is that they were modeling based upon current conditions, still very early in the pandemic. People forget that the death rate for hospitalized patients dropped by about 30%-50% in the first 3-4 months as we figured out how to better manage patients. It also didnt allow for vaccines. TBF, we had never developed vaccines that fast before. Also, TBF, we had never imposed the kind of restrictions they posited and we had no idea how well people would comply with them.
The UK ended up with about 230,000 covid deaths, half of their maximum prediction. If we had not developed vaccines and if we had not improved inpatient management of covid, we would have likely exceeded their highest numbers. Anyway, looking at all of the other predictions that were made, theirs probably comes out the best especially given when it was made. There is a strong tendency to criticize the work of others who didnt have the availability of information in the past that we do now. This model by the Imperial College has been grossly misrepresented. If you want to go after anyone go after the people who misreported it.
Steve
Jon Murphy
Mar 20 2025 at 12:48pm
To your first point, agreed. And do not forget that FDA rules actually prohibited the US from importing tests.
To your second, I have read the ICL report. It’s a hot mess.
steve
Mar 20 2025 at 2:06pm
A hot mess? Had you ever read one of these before? It did what you hoped it would. It laid out the range of possible outcomes based upon what we knew at the time. It didnt actually make a specific prediction for the number of deaths. The most difficult part in my estimation was estimating compliance with any of the proposed mitigation measures. Since it had never been attempted before I thought they chose reasonable numbers. (In China, where compliance was enforced note that mitigation was very effective.) Note that they did follow up reports that changed as they acquired new and better info.
I suspect you would be unhappy that they did not include models of economic effects, but that’s not really their job or area of expertise.
Steve
Jon Murphy
Mar 20 2025 at 2:13pm
Precisely the problem. What was “known” at the time was nonsense. Indeed, it was known as nonsense at the time, at least by epidemiologists and statisticians. They should have known the data were biased but they didn’t. This they produced BA bad model.
steve
Mar 20 2025 at 10:12pm
That’s not true. I think you are engaging in retroactive bias. I was in the hospital caring for ICU pts. I followed the mortality rates in real time for our network, a group of about 20 major university ICUs we were in contact with and national/international numbers. When they published their paper mortality rates were still higher. They dropped pretty quickly over 3-4 months but their paper was published early in that 3-4 month period. We had not really established the R value. We had not definitely identified the mode fo transmission. Majorly, which you keep ignoring is that we had no way to predict compliance with mitigation efforts. Last of all we certainly didnt know long term effects as we were only a couple of months into covid when it was published.
Steve
Jon Murphy
Mar 21 2025 at 6:13am
Steve-
The problem is that using the tests to identify COVID only on hospitalized patients upward biased mortality figures. That led to models like the ICL’s being wildly inaccurate. That led to poor policies like lockdowns which ended up being ineffective.
If, instead, the tests were used to randomly test the population, we’d have gotten better data.
The ICL model treated the data as of it was random testing when it wasn’t. That’s a rookie mistake. And that led to the horrible policies that, in all likelihood, made the pandemic worse. Not to mention the social and economic costs, and undermining the credibility of the public health system.
TMC
Mar 20 2025 at 9:23pm
Good comment. Out side of this: “we had very poor leadership at upper levels, people who largely had little public health experience”. It was the most experienced who failed the hardest.
We have to give a lot of leeway at the beginning. It was an unknown. Eventually we figured things out and that’s when leadership got bad.
Craig
Mar 20 2025 at 12:44pm
Cool, I have been wanting to do a YT channel relating to history and travel but its hard to find time. I have often thought many should really try to reach out as much as possible through various channels and many don’t do YT. I listened to an excerpt I will listen to it in its entirety a bit later, but I can tell from that that you have presence. I don’t often subscribe to anything really but I saw you have one subscriber so how am I not giving you a subscribe there! Good look with that hope you can make an impact on the marketplace of ideas.
Comments are closed.