Modelers at the University of Washington are starting to tackle one of the questions on everyone’s mind as the arc of the novel coronavirus pandemic appears to be flattening in places like Washington state: When will it be safe to begin easing up on the restrictions keeping the virus in check?
According to the UW’s Institute for Health Metrics and Evaluation (IHME), Washington could cross that threshold the week of May 18. Estimates for other states range from as early as May 4 to as late as the end of June, based on the local status of the epidemic.
The latest projections are an outgrowth of IHME’s ongoing work to forecast the scope of the epidemic nationwide — in every state — and in several other countries. They come as the group’s work, which has been influential from the White House to state houses coast-to-coast, is attracting criticism from some disease experts.
In a media briefing Friday, IHME director Dr. Chris Murray cautioned that the potential “opening dates” represent a first stab and are likely to change as more information comes in from individual states. Among the key variables are whether deaths are likely to drop off sharply once they peak, or whether — as seems to be happening in New York — they will plateau and decrease slowly.
“That will make a difference to when it will be safe to relax social distancing,” Murray said.
Another key factor will be how prepared states are to quickly diagnose and isolate newly infected people and everyone they have come in contact with. States that beef up their health departments and expand testing capacity might be able to start opening up sooner, Murray said.
The dates represent the modelers’ best estimate of when the daily new infection rate in each state will drop below one per 1 million people.
They picked that cutoff based on estimates that only a small percentage of actual infections are currently diagnosed, as well as health departments’ ability to detect and isolate new cases.
This week, health officers from King, Snohomish and Pierce counties all stressed that the number of new infections must be very low before restrictions are lifted, or their staffs will be quickly overwhelmed — as they were when the epidemic first began exploding across the state.
They also emphasized the need to be ready with widespread testing and adequate hospital space and equipment to deal with a possible upswing.
In Washington, where the UW group’s modeling suggests hospitalizations and deaths should have already peaked, many indicators are positive, but not yet definitive, Murray said.
According to the Washington Department of Health, the number of new cases averaged 210 per day during the week of April 6. That is equivalent to 27 new infections per million people.
IHME’s modeling has been criticized for being both too optimistic and too pessimistic. Earlier estimates that as many as 90,000 American could die have been scaled back to about 60,000, while the swings in some individual states have been even more extreme. Health officials in Massachusetts have said they are not convinced by new IHME forecasts that nearly quadrupled the state’s potential death toll, based on lower assumptions about the effectiveness of social distancing.
In Washington, IHME initially estimated as many as 1,400 people could die from COVID-19, the disease caused by the new coronavirus. The model is being updated every three days as new data comes in, and the most recent estimate puts Washington’s death toll at about 855.
Ruth Etzioni, a biostatistician at Fred Hutchinson Cancer Research Center, said she is not critical of the model itself, but how it has been interpreted and applied.
Now that the death estimates have been lowered, some officials and politicians are citing that as proof the U.S. response has been effective, when in reality the early projections were highly uncertain and the change mainly reflects the new data and understanding.
“It’s completely putting too much credibility on those early numbers, and that’s a terrible way to use the modeling results,” Etzioni said.
Murray pointed out that all of its U.S. death estimates have fallen within the uncertainty bounds of the original forecast, which ranged from about 35,000 to about 150,000.
One of the most recent changes in the model was based on mobility data from cellphones, which showed that people have curtailed their movements more than expected in places where fewer restrictions have been imposed, including some Southern states.
As a result, the death estimates in those states have been lowered, Murray said.
But he also said no one should rely solely on IHME projections as they decide when and how to ease life back toward normal.
“If I was a governor of a state, I would certainly not make a decision based just on our model.”