What Happens When the Coronavirus Models are Wrong?
As we hunker down under social distancing and stay-at-home orders, the Chinese coronavirus continues to extract a human toll in actual illness, as well as under-reported socio-economic costs. Many businesses are closed, people are not working, not earning an income, unable to socialize with friends and family.
The human tragedy that follows the virus will be horrific. Drug addiction, alcoholism, mental health issues, domestic violence, homelessness, and suicides may extract a toll far worse than the virus. Yet these costs are given scant attention, leaving the focus on ICU beds, hydroxychloroquine, and death counts.
When will the restrictions be lifted? How long until life returns to some semblance of normal? The answers will come from state governors and the president, but how do they know? They have advisors, like the basketball player and scarf queen, who use models, which are simply educated guesses based on certain data and assumptions.
Models aren’t a bad thing, but they have their limits. They follow the expression, “garbage in, garbage out.” Models are used to forecast everything from the weather to stock prices.
For example, models are used to predict hurricanes. Recall the spaghetti line plots preceding every hurricane, each squiggly line based on a particular hurricane model. Some veer into the Gulf of Mexico, others hit South Florida, and others head north to the Carolinas or out to sea. At most, only one line will be correct, but each is based on a model.
Other models predict future climate. Some have predicted horrific storms, others predicted melting polar ice caps and coastal cities under water. These models haven’t panned out either.
Models used to predict the course of the coronavirus are based on data and assumptions, providing estimates of cases, hospitalizations, resource need, and deaths. How have these models performed?
The IHME model is considered the gold standard. In mid-March, without social distancing, they predicted 2.2 million American deaths. By early April they reduced their death projection to 100,000 to 240,000 assuming social distancing measures in place. Their April 17 update now projects 60,308 deaths, 3% of their original prediction.
What changed? Social distancing was in already in place when the death predictions dropped by a factor of four. For perspective, 61,000 Americans died in the 2017-18 flu season.
On April 5, Colorado was preparing for a worst-case scenario, according to the Denver Post, predicting a potential shortage of 10,000 ventilators. Plans were being made to ration ventilators, prioritizing patients based on age, medical status, and other factors. This is called euphemistically the “crisis standard of care,” but instead is a blueprint for care rationing. It’s certainly good to be prepared for the worst possible scenario, yet how did the models lead to ventilators needing to be rationed?
On that same day, April 5, according to the IHME model, Colorado was projected to need 88 ventilators with 554 ICU beds available in the state, less than 20 percent capacity, and less than 1 percent of the above prediction. Note that virtually all ICU beds come with a ventilator, making the two parameters interchangeable.
Two models, one predicting the hurricane hitting New Orleans, the other hitting Long Island. Which model should be believed? Or if the hurricane makes landfall in Boca Raton, both models were wildly incorrect.
The CDC had its own models predicting gloom and doom. In mid-March they projected 160 million to 214 million infected and 200,000 to 1.7 million deaths. Did that leave President Trump any choice but to hit the off switch on the U.S. economy? What if the models were wrong?
YouTube screen grab
Here in Denver, the Denver Convention Center is being converted into a field hospital but is unlikely to be needed by the time construction is complete. The models said this was necessary. These models also spurred Seattle to create a field hospital at the CenturyLink Center where the Seahawks play, but closed it down after not seeing a single patient.
Another factor influencing models is the death count from this coronavirus. Projected deaths determine projected need, as in the ventilators mentioned above. How accurate are the death counts?
Task force member Dr. Deborah Brix, on April 7, said the U.S. government is classifying the death of any patient who tested positive for coronavirus as a coronavirus death, regardless of any underlying health conditions that genuinely killed them. If one has a heart attack, and happens to test positive for the virus, he or she will be classed as a coronavirus death. Garbage in, garbage out.
Going further, New York City is including in their death counts, “people who had never tested positive for the virus but were presumed to have died of it.” Data by presumption? Models based on presumption?
Climate scientists “presume” temperatures are rising due to human activity and, voila, we have a consensus telling us that global warming is real and an existential threat to civilization, although not as bad as the Orange Man.
New York City created multiple field hospitals to accommodate the model that predicted the surge in patients. The Javits Convention Center field hospital, with a capacity of 2,500 beds had 340 patients this week, about 15 percent of capacity. The Billie Jean King Tennis Center, home of the US. Open, is now home to a 350-bed field hospital, expecting to transfer 30 patients there, perhaps doubling that number next week if needed, but well below the capacity predicted by the models.
The USS Comfort, with a 1,000 bed capacity has treated 130 patients, 60 already discharged. Assistant Secretary of Defense for health affairs Thomas McCaffery said last week, “I wouldn’t say the states got it wrong.” Yet they did, in a big way, because they followed models that got it wrong.
Remember how New York Governor Andrew Cuomo demanded President Trump send NY 40,000 ventilators? This number was presumably based on the same models predicting the need for multiple field hospitals. Now New York is sending the ventilators they have to other states as New York doesn’t need them.
The models are why the country has been shut down for the past month. Leaders, from the president to governors and local officials, are presented with predictions from models that their advisors, like the basketball player and scarf queen, say are credible and accurate. Leaders then have no choice but to act accordingly, hitting the 'off' switch to a $20 trillion economy. This plays out not only here in the U.S. but around the world.
In a moment of candor, Dr. Anthony Fauci admitted, “We can’t really rely on coronavirus models.” Yet his advice to the president and the task force is based on what? A Ouija board, tarot cards, or his vaunted models?
The president and his team are making decisions affecting most Americans, sending businesses into bankruptcy, adding trillions of dollars to the national debt, sending millions to the unemployment office. Other politicians want to destroy the economy through schemes like the Green New Deal over climate models and predictions.
Make no mistake, many of these decisions are also political, way beyond the realm of science and models, but are superficially based on models. Money and power corrupt good science. The silent war continues.
But the coronavirus decisions are based largely on models. What happens when the models are wrong?