Revisiting the IHME Model: Still Useless?

The University of Washington’s Institute for Health Metrics and Evaluation has become famous for producing a COVID-19 model that is widely referred to by policymakers and journalists. I have described the IHME model as useless as a guide to governors and other decision-makers, for several reasons.

First, because it runs out on August 4, by which time it assumes that fatalities will have been flat for some time. What happens after August 4? Is the epidemic over, or will it resume in the Fall? The model gives us no answer. Second, the model’s predictions of fatalities in individual states have fluctuated so wildly–e.g., declining by more than 90 percent in just four days–that no policymaker could properly rely on them. Third, while proclaiming that it “assumes social distancing,” it gives no guidance as to what will happen if a particular state reopens its economy–the key point on which governors need information.

The IHME model was updated a week ago; you can see the current information here. The most recent update didn’t change the fatality projection for the U.S. through August 4 by much. It sits at 143,357. What will happen after that is anyone’s guess, although the model implies that well before August 4 the disease has pretty much flat-lined. It projects fewer than 10,000 deaths between July 1 and August 4.

If you want to know what impact the current loosening on restrictions in various states will have on fatalities, you’ve come to the wrong place. IHME admits that its prior assumption that more liberal policies would lead to more fatalities was wrong:

With mobility rising throughout the US over the last several weeks, our team had expected to see large increases in reported COVID-19 cases and deaths in more recent days. After all, the time lag between heightened mobility and potential rise in COVID-19 infections is approximately two weeks. Yet such a surge has yet to materialize, suggesting that increases in human mobility alone may not fully capture risk of transmission.

This has been obvious in states like Florida, where liberal journalists wrote that the state was conducting an experiment in “human sacrifice” when Governor DeSantis relaxed his shutdown order.* Those predictions turned out to be entirely wrong, and Florida has been a shining example of how to deal with the COVID-19 epidemic.

The IHME team is now focusing on face mask use as possibly explaining why the disease hasn’t picked up in the wake of relaxed economic and social dictates. Good luck with that.

IHME has moved toward expressing COVID fatalities, by default, as a percentage of population rather than in raw numbers, although those are still available. Perhaps the most helpful item on the current IHME web site is this map, which shows COVID fatalities per 100,000 of population in each state. At the IHME web site you can hover over each state for information. But the map pretty much speaks for itself. Click to enlarge:

COVID-19 is a local or regional phenomenon. It has struck Northeastern states–some of them, anyway–hard, and has been a non-factor in most of the rest of the country. I believe there are a number of states where Wuhan fatalities are lower than those from a typical seasonal flu.

New York City and environs are obviously the center of the epidemic in the U.S. It still is not clear why that is, although Governor Cuomo’s mismanagement of the state’s nursing homes is undoubtedly a factor. Certainly densely-populated cities are likely to be hit harder than rural areas, but that doesn’t explain the low impact in San Francisco, or in the large cities of Texas, Florida and other states.

Also, you can see that in the Upper Midwest, Minnesota and Iowa perform poorly compared with other states. South Dakota, the only state that has refused to enter any sort of shutdown order and was viciously smeared by the Washington Post and other Democratic Party news outlets as a result, has one of the lowest fatality rates in the U.S. Not quite as low as Arkansas, however, which the virus seems to have skipped altogether. The fatality rate in New York is around 150 times that in Arkansas. The uniform solution insisted on by our national press makes, apparently, no sense.

Still, I have assumed that the virus will come to less-populated areas in time, and that eventually, case rates will be similar from one place to another. It looks increasingly like that assumption is wrong. In fact, there is evidence that the virus is dying out on its own.

What will happen in the coming months is anyone’s guess. To me, it is striking how little guidance America’s most-cited model even pretends to give.

* A reader reminds me that it was Georgia, not Florida, that the Atlantic accused of engaging in human sacrifice. Memory is fallible, but the point is the same: Georgia’s track record has been about as good as Florida’s. If liberals are looking for human sacrifice, they need look no farther than New York State and its feckless governor, or Minnesota, whose inept governor has allowed hundreds to die in nursing homes without taking any meaningful action.