Erratic IHME model changes again to project 8,000 fewer coronavirus deaths than previous week

The coronavirus projection model from the University of Washington's Institute of Health Metrics and Evaluation again revised the number of expected COVID-19 deaths to project 8,000 fewer overall U.S. deaths than it did the previous week.

The IHME model is possibly the most prominent of all available projection models, and likely has influence in state and federal government decisions about when to relax the social distancing policies, which have been deemed necessary to slow the spread of the virus, but that have had disastrous economic impact nationwide.

The new projection: On Friday, the IHME model projected a total of 60,308 U.S. COVID-19 deaths, with an estimate range of 34,063 to 140,381. On April 13, it projected 68,841 deaths and a range of 30,188 to 175,965.

The IHME model produced the projections that led President Donald Trump and the White House coronavirus response task force to announce that even if Americans did everything right, as many as 240,000 people would die of the coronavirus by the end of the summer.

The model has also overestimated the number of hospitalizations and the number of needed ventilators, and those projections have had political implications and also been economically damaging for health care providers that shut down all nonessential medical procedures to prepare for COVID-19 surges that, in some areas, never arrived.

What's the problem? The IHME model's method of producing projections has led to inaccuracy and volatility, both of which make it unhelpful for forming public health policy, according to STAT News. The model looks at the progression of the virus in other countries with earlier outbreaks, tries to determine where the U.S. fits on that bell curve progression, and produces projections from there.

The model assumes that the U.S. curve will mimic the curve in other countries, without accounting for all the variables such as differences in social distancing policy and different timing for implementing those policies. Because it is a statistical model that started with very little U.S. data, it can shift significantly every time new U.S. data is added to the model.

The IHME model has been inaccurate even when projecting the next day, STAT News reported, with next-day deaths falling outside the projected range 70% of the time.

"That the IHME model keeps changing is evidence of its lack of reliability as a predictive tool," said epidemiologist Ruth Etzioni of the Fred Hutchinson Cancer Center, according to STAT News. Etzioni has served on a search committee for IHME. "That it is being used for policy decisions and its results interpreted wrongly is a travesty unfolding before our eyes."

(H/T Washington Examiner)