Andrew Kerr and Chuck Ross, DCNF
- The Institute for Health Metrics and Evaluation (IHME) sharply reduced its coronavirus projection model on Sunday evening.
- The IHME model has produced some wild swings on its state-by-state projections since its first iteration on March 26.
- IHME Director Dr. Christopher Murray said Monday the encouraging numbers shouldn’t be seen as an excuse to ease up on social distancing measures.
A coronavirus projection model that has informed the White House’s response to the pandemic sharply reduced its estimations on Sunday night after encouraging data from Italy and Spain suggests that peak spread of the virus may arrive in the United States sooner than previously expected.
The Institute for Health Metrics and Evaluation (IHME) said in its April 1 projection that 262,029 hospital beds would be needed on the peak date of the coronavirus’s spread in the United States but on Sunday evening it cut that figure down by 58 percent to 140,823.
The institute’s projected number of American deaths by the start of August also dropped by 12 percent from 93,531 to 81,766.
“As we obtain more data and more precise data, the forecasts we at IHME created have become more accurate,” said the director of the IHME, Dr. Christopher Murray, in a press release Monday.
Murray told reporters during a conference call Monday that the reductions in the IHME’s model are largely driven by new figures from Italy and Spain showing that peak daily deaths arrived sooner than previously predicted, which suggests that social distancing policies have been effective in containing the spread of the virus.
Italy reported 525 coronavirus deaths on Sunday, down from a peak of 919 on March 27. Spain recorded 694 deaths on Sunday, down from 961 fatalities at its apex on April 2.
Murray pointed to publicly released cell phone mobility data showing that mobility started to decline in Washington, one of the first states to be hit by the outbreak, long before government stay-at-home mandates went into effect.
However, Murray cautioned that the promising trend shouldn’t be seen as an excuse by policymakers to ease up on social distancing measures.
“If you ease up prematurely, the epidemic can rebound right back to the level we’re at right now in a matter of weeks,” Murray said. “The potential for a rebound is enormous if we let up on social distancing.”
Murray said the time to begin the debate on when to ease up on stay-at-home orders and transition to a strategy of contact tracing and case-by-case quarantining shouldn’t begin until “after June.”
“Even then, it will require states have the funding and capacity to execute on a testing program with contact tracing,” Murray said.
IHME model: accurate in some areas and lacking in others
In terms of estimated nationwide death figures, the IHME has been largely on point. For example, the model’s first iteration on March 26 estimated that there would be 1,406 deaths nationwide on April 4.
The actual coronavirus death toll in the U.S. on April 4 was 1,330, according to Worldometer.
However, the IHME’s model has produced some wild swings on its state-by-state projections. Murray acknowledged that the model’s state-level projections could use improvements in a conference call with reporters on Monday.
“I’m sure we’ll be wrong in some states but we are working on steadily improving the state-specific forecasts,” Murray said. “But at the national level we think they’re going to be accurate, but they do require keeping social distancing in place until at least the end of May.”
Researchers revised expected fatalities drastically downward for states like California, Texas, and Mississippi on Sunday evening. States like New York, New Jersey, and Massachusetts were projected to have far more fatalities than initially projected.
Much of the difference is reflected by how far into the future the model projects each state to reach its peak number of daily fatalities, and how high the peak will reach.
For Texas, the earlier model predicted 102 deaths occurring on May 4. The revised model predicts that the outbreak will subside sooner, and with fewer deaths at the peak. It now predicts 72 deaths at a peak occurring on April 20.
Overall, the model now predicts 2,025 deaths compared to 4,150 under the previous projection.
California is projected to see an even steeper decline in fatalities, from 6,109 down to 1,783.
The revised model projects that California will reach its peak number of fatalities, 70, on April 17. The previous model predicted 148 deaths on its peak date of April 25.
Whereas the earlier model projected that fatalities would fall to 24 by June 1, the revision projects that the state will reach a similar number of deaths — 23 — a full month earlier.
The model revised the expected number of deaths in Mississippi from 2,292 to 237, a nearly 90% decrease in deaths for the state. Researchers initially projected 92 deaths at a peak on April 23, but now expect eight deaths at a high point on April 19.
The earlier model predicted 21 deaths for Mississippi on Monday, and projected that the state would have 105 fatalities so far in the pandemic. The Mississippi Department of Health reported eight deaths on Monday, raising the overall total to 51.
The model significantly increased fatalities expected in New York and New Jersey, the epicenters of the pandemic in the U.S.
The model projects 9,690 fatalities in New Jersey, nearly five times higher than the 2,096 predicted in the earlier version.
The massive increase in New Jersey’s numbers represents a spike in the model’s prediction for the number of peak deaths. The previous model predicted a peak of 104 deaths on April 10. The new projection predicts a peak of 584 deaths on April 16.
One data point suggests that the upwardly revised model is not entirely accurate. It predicted 207 deaths on Sunday, but New Jersey Gov. Phil Murphy reported that there were only 86 fatalities.
The model predicts a 52% jump in expected deaths in New York, from 10,243 to 15,618. A peak of 878 deaths was projected on April 9 in the initial model, versus a peak of 547 deaths on April 8 in the revised version. The revised model predicted 713 deaths on Sunday, though Gov. Andrew Cuomo said at a press conference Monday that 599 deaths were recorded. He also expressed optimism that deaths have peaked and begun to flatten out.
The model also drastically increased projections for Connecticut and Massachusetts, though initial data suggests that the revision overestimated the fatalities.
The model upped its forecast for Connecticut from 378 to 5,474 and Massachusetts from 1,782 to 8,254 fatalities.
For Massachusetts, the new model projects a peak of 373 deaths on April 18, a change from the 81 deaths predicted at a peak on April 16.
For Sunday, the model predicted 78 deaths in the Bay State. The actual number of deaths for the day was far lower, 15.
The revised model also lists the Massachusetts death tally at 350 through Sunday. As of Monday afternoon, the state had reported only 260 fatalities.
The revised model projects a far steeper peak for Connecticut, the domicile for many residents who work in New York City. The state was expected to reach a peak of 16 deaths by April 8. Now, the model predicts 192 deaths at a peak on April 22. As of Monday, the state had reached 206 deaths.