We Should Not Have Trusted ‘Expert’ Epidemiological Models
Neil Ferguson and his Imperial College COVID-19 Response Team colleagues scared the world silly in mid-March. They announced they ran a sophisticated epidemiological model which said doom was on the way. Of course they adjusted the model as soon as the facts contradicted it, as modelers do. But their first forecast shaped the extreme global response.
They said that if we did nothing to stop the coronavirus, disaster would result. “In total,” they wrote, in an unmitigated epidemic, we would predict approximately 510,000 deaths in GB [Great Britain] and 2.2 million in the U.S., not accounting for the potential negative effects of health systems being overwhelmed on mortality.”
If health systems were swamped, and we did nothing, the totals would soar even higher. To avoid this apocalypse the authors recommended two possible strategies.
“Mitigation” or the More Extreme Option: “Suppression”
They were: “(a) mitigation, which focuses on slowing but not necessarily stopping epidemic spread – reducing peak healthcare demand while protecting those most at risk of severe disease from infection, and (b) suppression, which aims to reverse epidemic growth, reducing case numbers to low levels and maintaining that situation indefinitely.”
Relying on their model, they insisted that “optimal mitigation policies … might reduce peak healthcare demand by 2/3 and deaths by half. However, the resulting mitigated epidemic would still likely result in hundreds of thousands of deaths and health systems … being overwhelmed many times over.”
By mitigation they meant such things as:
- “combining home isolation of suspect cases,
- home quarantine of those living in the same household as suspect cases,
- and social distancing of the elderly and others at most risk of severe disease.”
The team went on to suggest suppression rather than mitigation.
Suppression meant a combination of:
- “social distancing of the entire population,
- home isolation of cases and household quarantine of their family members,”
- “school and university closures,”
- and other similar policies, many of which were adopted.
Authorities would need to enforce suppression, they said, for “potentially 18 months or more,” until a vaccine could be discovered.
Their Death-Toll Predictions for Mitigation vs. Suppression
What if society just mitigated? It would still be a catastrophe. “In the most effective mitigation strategy examined, which leads to a single, relatively short epidemic,” they wrote, “[and] even if all patients were able to be treated, we predict there would still be in the order of 250,000 deaths in GB [Great Britain], and 1.1-1.2 million in the U.S.”
Okay. What about suppression? If authorities did so quickly, deaths in Great Britain could be anywhere from 5,600 to 48,000, depending on how the disease was transmitted and the use of ICUs. They said numbers would be proportional for the U.S., but did not quote any figures.
Reported deaths at the moment in the whole United Kingdom are 11,329. Since that’s on the low end of the range with suppression, does that mean model was right? No. That’s because we should have doubted the model from the start. Let me explain.
Hideously Complex Models for Predicting Human Behavior
The model itself is from a class of tools used by Imperial College and the Institute for Health Metrics and Evaluation in the U.S. They use math and statistics to represent how humans behave, the supply and use of hospital resources, how diseases are transmitted, how deadly they are, and so forth. These models are hideously complex. But then so is the subject, or subjects, being modeled.
In practice, it’s so hard to predict how people will behave that no one can claim to have mastered it. If somebody has, think of the killing he could make in the stock market. Yet these models claim to foretell what a country of 66.7 million souls would do in a novel pandemic under a range of posited government actions.
Besides the expertise of the model creators, what evidence was there to believe the Imperial College model in the first place?
It’s Shocking Anybody Swallowed These Astounding Numbers
Again, their model predicted that if government did nothing, about 510,000 would die in Great Britain and 2.2 million in the U.S. These are staggering numbers. There are about 66.7 million people in the U.K. and 328.2 million in the U.S.
The model said that deaths per day per 100,000 population would peak at about 21 in Great Britain (around June 1st) and 17 in the U.S. (around June 20th).
These figures translate into about 14,000 deaths per day in Great Britain and 56,000 per day in the U.S. at the peak. Per day! These numbers are astounding. I’m shocked that anyone swallowed them.
There is, of course, no way to measure if the model would have gotten these morbid predictions right. After all, both countries suppressed and mitigated in various ways. How much is, of course, subject to debate.
The forecast for a business-as-usual scenario was so far removed from experience, though, that anyone not invested in the model should have doubted it. Here’s why.
An Absurd Prediction
The Spanish flu of 1918 was a horrific event. It befell a world fresh from a global war, one with poor medical care, aspirin poisoning, shortages of every kind. Between 17 and 58 million were killed worldwide.
The CDC estimated that about 675,000 Americans died, when the population in the U.S. was about 106 million. This makes 637 per 100,000 dead of Spanish flu in the U.S.
Imperial college predicted 670 per 100,000 would die of the coronavirus.
Now, when the COVID-19 Response team constructed their model on 16 March, there were only “6,470 deaths confirmed worldwide” and 97 in the U.S. Yet, somehow, even with these low figures and modern medicine, the team predicted the coronavirus would be deadlier than the Spanish flu.
This is idiotic. The model should have been strapped to a ventilator, not used to shape global policy. This is one of many lessons we’ll need to learn from this episode, if we don’t want to see it happen again.
William M. Briggs is a senior contributor to The Stream, author of Uncertainty, blogger at wmbriggs.com, philosopher and itinerant scientist. He earned his Ph.D. from Cornell University in statistics. He studies the philosophy of science, the use and misuses of uncertainty, the corruption of science, and the uselessness of most predictions.