IHME projections were always absurdly low. Was it proficiency, politics, or bureaucracy?
The University of Washington’s Institute for Health Metrics and Evaluation (IHME) mortality models have underestimated total projected mortality for weeks due to faulty assumptions underlying their methodology. They have now nearly doubled the death predictions to 134,000 and point to increased mobility due to reopening as the cause. In reality, their projections were always absurd, and reopening provides a pretext to update the numbers. Is this normal bureaucratic avoidance or politics of some sort?
By the numbers
The first IHME projections estimated were roughly 84,000 deaths, rose to 94,000, then were adjusted down until reaching a low point of 60,000 on April 17th. The estimated then ratcheted up to about 66,000 deaths on April 21st and then to over 72,000 deaths on April 29th. The May 4th prediction effectively doubled the prediction from the week before.
What’s going on here? In essence, the early models fitted a classic bell curve to the data, predicting a steep rise and an equally steep decline on the other side. In reality, Coronavirus deaths did not drop off steeply on the other side of the peak. Ultimately the IHME adjusted their models to account for this.
If you dive deeply into the explanation of the May 4th update, you find this:
Expanding the range of multi-Gaussian distribution weights for predicting epidemic peaks and shapes. Since our initial release, we have increased the number of multi-Gaussian distribution weights that inform our death model’s predictions for epidemic peaks and downward trends…This expansion now allows for longer epidemic peaks and tails, such that daily COVID-19 deaths are not predicted to fall as steeply as in previous releases.
In normal human-speak, they changed the parameters to skew the curve to the right and match what is actually happening.
Controlling the narrative
While the IHME has made appropriate changes to the model to reflect real world death trends already witnessed here and in other countries, that is not the narrative that the organization publicizes. Their May 4th press release states:
The revised projections reflect rising mobility in most US states as well as the easing of social distancing measures expected in 31 states by May 11, indicating that growing contacts among people will promote transmission of the coronavirus. Increases in testing and contact tracing, along with warming seasonal temperatures — factors that could help slow transmission — do not offset rising mobility, thereby fueling a significant increase in projected deaths.
The IHME is flatly stating in its press release that the doubling of their mortality due to changes in mobility and social distancing that are only now taking place. They are misleading the public by not acknowledging that “longer epidemic peaks and tails” were already self-evident weeks ago.
As quoted by CNN, Dr. Christopher Murray, director of the IHME, explicitly attributes the future increase in modeled deaths to mobility beginning to increase even before guidelines are relaxed. These would be future deaths; no acknowledgement or explanation is offered for the fact that previous versions had failed to predict the slope of the current death trend.
A quick Worldometer review of the mortality data of Italy, Spain, or even South Korea reveals right skewed curves with extended tails. All of the countries led the United States by weeks. By the middle of April it was easily evident in US data as well.
On April 16th, the day before the April 17th projection of 60,000 was released, there had already been 35,000 deaths, and the country was experiencing about 2,200 deaths per day. Did anyone take seriously the idea that only 25,000 more would perish by August? Likewise, on April 20th before the prediction of 66,000 deaths, there were already 43,000 deaths.
On April 28th, the day before the 72,000 mortality prediction was released, deaths were already over 49,000. Forget about Gaussian curves, epidemiology, or any other such trickery. By this point, the average junior high student could have made a more accurate prediction with graph paper an a strait-edge.
Kicking the can down the road
The Centers for Disease Control website lists a host of generally more sobering models. The IHME model, however, occupies a special place in our discourse. It has been referred to in presidential press conferences, and is widely cited in the news.
Why was the IHME so late to adjust its projections to reality, and why is it at best dissembling in its messaging? Three possible reasons, from least to most likely.
- Proficiency. Perhaps no one noticed the data deviating from the model when the sharp declines failed to materialize, though the the IHME itself noted that peaks were elongating at least as early as the April 17th update.
- Politics. One can construct scenarios where outside political influence impacted how the IHME communicated. It could have received pressure from the federal government to minimize projections, then when that was no longer feasible looked for the nearest available explanation. Read another way, the IHME could have been biding its time to for its own agenda, dropping giant leap in death numbers to influence the national debate on reopening.
- Bureaucracy. Proud, educated individuals don’t like to admit to mistakes. Organizations are made of people and magnify that tendency. Perhaps, with constant scrutiny in the press and daily White House press conferences, it was just easier to make small adjustments and incrementally raise the estimates as the mortality curve stubbornly refused to start its decline. Why cause a stir until you are really sure and have the numbers nailed down? Reopening, which will certainly lead to more mortality, presented a lifeline to deflect serious scrutiny.
Regardless of speculation, daily deaths were declining slowly long before the IHME doubled its prediction.
Brian E. Wish works as a quality engineer in the aerospace industry. He has spent 29 years active and reserve in the US Air Force, where he holds the rank of Colonel. He has a bachelor’s from the US Air Force Academy, a master’s from Bowie State, and a Ph.D. in Public and Urban Administration from UT Arlington. The opinions expressed here are his own.