How many will die?

Maybe 2 million?

Maybe 250,000?

Just 70,000?

The computer-based models of the COVID-19 pandemic have been bouncing all over the place for weeks.

Now an apparent slowdown in new cases in many areas combined with the enormous economic damage wrought by business closures has prompted some states to move cautiously towards reopening their economies.

The White House has cited the steady decline in projections of the death toll in a model developed by IHME in urging governors to reopen their states. The US so far has 790,000 confirmed cases and some 40,000 deaths, with 2,000 new deaths daily. However, the rate of increase has flattened out.

In Arizona, the Institute of Health Metrics and Evaluation (IHME) model now predicts 70,000 deaths nationally, compared to the 250,000 predicted several weeks ago.

The model applied to Arizona says deaths likely peaked at 11 deaths per day on April 15. The model predicts a steady decline, with the last death coming on or about May 20 assuming the continuation of social distancing policies.

The projection says Arizona now has enough hospital beds, ventilators and ICU beds to weather the epidemic. Some hospitals are furloughing workers because the cancellation of elective surgeries and the decline in other hospital admissions has reduced patient counts far more than COVID-19 has added. The demand on hospitals for COVID-19 patients peaked on or about April 10 at 312 hospital beds, out of a statewide supply of 6,000. The demand for ICU beds peaked at 76 out of 508, according to the IHME model.

The model envisions 267 deaths in Arizona. Earlier this month, the same model predicted 1,000 deaths here.

Those projections make the gradual reopening of the economy in the next few weeks plausible.

But can we believe the projections?

Maybe not, say experts in infectious disease.

The growing debate about the IHME computer modeling has cast new light on the agonizing difficulty governor’s face when making life-and death decisions based on still incomplete information. The early crude projections were based on death rates that proved too high and an expectation that half of Americans would end up infected. The stay-at-home orders changed the dynamics of the infection, reducing the number of people exposed. In the meantime, disease experts discovered most people infected by the virus never have symptoms serious enough to prompt them to seek a test. Worse yet, many of the people who die of the effects of the virus don’t end up in the official death county. In New York, one study suggests that nearly a third of the COVID-19-related deaths don’t show up in the statistics on which the models are based.

The medical industry magazine STAT published a story about the doubts among epidemiologists concerning the reliability of the IHME model, driven in part by the great shifts in predictions from one week.

The IHME model suggests US deaths are already leveling off and will decline from about 2,000 per day in April to 10 or fewer by June.

The model differs significantly from previous models relied on by many epidemiologists, according to the STAT article.

Two types of conventional models generally also suggest we’re near the peak of new infections, but result in different total cases and deaths. A conventional SEIR model simply looks at the virus’ speed of spread and the vulnerable population. COVID-19 apparently spreads as easily as the flu in a population with no immunity. Other “agent-based models,” like the video game SimCity take into account more factors, like incubation times and more sophisticated estimates of social mixing.

IHME took a different approach by building models of the rise, leveling off and decline of the epidemic in Wuhan China and 19 cities in Italy and Spain. The model then matches the shape of the curve in new areas to those 20 cities.

The model has several potential weaknesses, say experts, including government responses. Wuhan China imposed drastic measures, locking down the population, people’s movements, imposing fines, mass testing, mandatory use of masks and frequently checking temperatures strictly enforced by an authoritarian government.

The scientists who developed the IHME model say they’re continuing to tweak it, but that overall, the buildup of new data has supported the model.

However, other health experts have raised doubts. Researchers from the Imperial College of London last week published a critique of the IHME projections in the Annals of Internal Medicine. The continued tweaking of the IHME model makes it unreliable, they argued. For instance, a projection of 81,000 actually means deaths could range from 38,000 to 162,000 – with the projection a midpoint in a range of possibilities.

Imperial College researcher Britta Jewell in the Annals paper wrote “this appearance of certainty is seductive when the world is desperate to know what lies ahead” and “rests on the likely incorrect assumption that effects of social distancing policies are the same everywhere.”

The IHME projections for deaths in the US last week went from 60,000 by August 4 to 68,000 – which means somewhere between 30,000 and 176,000. Numbers for individual cities have also bounced around from week to week.

More traditional models have also been yielding projections of steadily declining death tolls.

Researchers from the Massachusetts Institute of Technology three weeks ago predicted cases will level out this week at about 600,000. The US total is already at about 800,000 with almost 40,000 deaths, but new cases have leveled off.

A University of Washington model predicted a million cases by the end of the epidemic in early June, declining to nearly zero – providing social distancing policies continue, according to the STAT article by Helen Branswell and Sharon Begley.

Contact the writer at

Recommended for you

(0) comments

Welcome to the discussion.

Keep it Clean. Avoid obscene, hateful, vulgar, lewd, racist or sexually-oriented language.
Don't Threaten. Threats of harming another person will not be tolerated.
Be Truthful.
Be Nice. No name-calling, racism, sexism or any sort of -ism degrading to another person.
Be Proactive. Use the 'Report' link on each comment to let us know of abusive posts.
Share with Us. We'd love to hear eyewitness accounts, the history behind an article. Real names only!