Covid-19 Modeling for Arizona Has Been a Complete Failure

Covid-19 Modeling for Arizona Has Been a Complete Failure

Just like the rest of the country, Arizona residents have since mid-March hunkered down, self-isolated and followed the advice of medical experts and our political leaders to stop the spread of Coronavirus. The rationale provided to us was that if extreme actions were not taken, hundreds of thousands of people in the Grand Canyon state would perish. Hospitals would be overrun. It might be as bad as the Spanish Flu of 1918.

How do we know Covid-19 might be this horrific? Whenever this question was asked, defenders of the shutdown immediately pointed to the various models proving their case. Two in particular were cited the most—IHME and CovidActNow. Both produced dire forecasts for Arizona and both recommended draconian mitigation strategies. Both have proven to be wildly wrong.

The IHME Model

Designed at the University of Washington, the IHME model has been the most frequently cited the last 3 weeks and was the foundation for the nationwide projections developed by the Trump Administration.

On April 1st, IHME predicted dark days for Arizona. IHME forecasted that by April 15th over 2,500 beds would be needed and that Arizona would be at its ICU and ventilator capacity. At our peak on April 27th, Arizona would have 4,000 hospitalized Covid-19 patients and a ICU shortage of nearly 100 beds. Over 1,300 would be dead by August 1st.

CovidActNow

Founded and developed by four volunteers with very questionable credentials, CovidActNow became a frequently utilized source for politicians and governmental entities early on during the pandemic. In Arizona, CovidActNow was the primary model used in March by the University of Arizona College of Public Health to develop their pandemic response recommendations, which were cited by the media and referenced by politicians throughout the state.

On April 1st CovidActNow had concluded that Governor Ducey’s shelter-at-home policy was far too lax and that a much stricter statewide lockdown was necessary:

Without a “strict” stay-at-home policy, CovidActNow declared that Arizona on April 15 would have nearly 1,500 hospitalized, a hospitalization peak of 40,000 by June, and 28,000 dead by mid-summer.

Models vs. Reality

It is embarrassing how poorly both of these models performed compared to reality. As of today (April 15), Arizona has fewer than 500 Covid-19 patients hospitalized and around 100 admitted to ICU, a fraction of the predicted amount. Ventilator use is in decline and the state has already reduced their federal ventilator request from 5,000 to 500.

Fatalities are lagging behind the IHME model and will never approach the laughable figure cited by CovidActNow. There has been no Coronavirus hospital crunch. In fact it has been just the opposite–healthcare workers have been furloughed due to all of the empty hospital beds.

The excuses to explain away the modeling errors have been coming as fast as the downward revisions being made to both forecasts. Defenders of the modeling claim that the lower projections only prove that the current policies are working and thus fewer deaths and hospitalizations have been the result. The flaw with this argument is that both models as of April 1 were based on the mitigation efforts that are in effect today. This argument is simply an attempt to move the goal posts and avoid any discussion about why the models missed so badly.

Others have suggested that it is not really the fault of the modeler’s–Covid-19 projections are difficult and that a lack of data and changing assumptions hampered their effectiveness. These are all valid points, except that none of these issues were ever brought up when the public was being sold on their reliability and used as the justification for a nationwide shutdown of the economy.

Some are even saying that it doesn’t really matter that the models were wrong and that everyone should just be thankful that it is not as bad as they thought. This argument is not only wrong, but offensive. Thousands of people have lost their jobs, their livelihoods, and some will lose their lives because of the actions taken based on these models. If the goal is to completely erode all public trust and credibility in our institutions, this is the quickest way to do it.

Moving Forward

Unfortunately, what has transpired up to this point cannot be undone. Mistakes were made, yet rather than dwelling on them we need to start working toward solutions that address our overreaction.

The top priority should be to reopen Arizona in a safe and healthy way.  Governor Ducey has announced that he is developing a plan to open up the economy; our hope is that it coincides with the expiration of the existing stay-at-home order on April 30th (if not sooner). 

Of course, this doesn’t mean that Arizona should not take any precautions or that there won’t be certain restrictions when the shutdown ends. The models were useless, but that doesn’t mean we don’t have data showing that Covid-19 is a real danger to certain segments of the population (predominantly seniors and those with chronic illness). The Club believes a solution exists somewhere between doing nothing and the current draconian shutdown.

The other lesson that should be learned is a healthy amount of skepticism when politicians start using forecast modeling as their justification for their radical policy prescriptions. The public might have been fooled this time by the so-called experts and their doomsday modeling, let’s just make sure it doesn’t happen again.