Coronavirus Model Now Estimates Fewer U.S. Deaths

The total number of coronavirus deaths in the U.S. may be lower than previous estimates, according to the Institute for Health Metrics and Evaluation at the University of Washington School of Medicine.

The new assessment predicts approximately 81,766 deaths, with a range between 49,431 and 136,401. That is a downward revision of the estimates of the higher end. The institute also projects deaths from the virus will peak earlier than it estimated before, with the likely peak on April 16, with a projected 3,130 deaths nationwide on that day, according to IHME.

The change is significant because IHME's estimates are followed closely by the White House and its coronavirus task force. The revision is the latest of a flurry of widely varying predictions released in recent weeks as statisticians try to provide up-to-the minute guidance to researchers, policymakers and hospital administrators who are trying to blunt the impact of the epidemic. It was driven largely by the experiences that have been seen so far, especially in other countries where there are early indications of a slowing of the trendline of cases and deaths. The institute said it reflects the success of so-called social distancing and warned that if those efforts are curtailed too soon, the more positive trendline might not hold.

Estimates of projected deaths, for instance, have ranged from 2.2 million to 49,431, at the lower end of the new analysis.The institute's updated analysis also looks at COVID-19's impact on hospitals, and it scales back the need for acute-care beds, intensive-care beds and ventilators needed to care for the sickest patients from its prior analysis.

"We started doing this work because our hospital system asked us to help plan for ICU beds and obtain ventilators," Murray said Monday.

Hospitals in hard-hit areas will need approximately 25,000 ventilators over the next several weeks. The institute projects a shortage of approximately 16,323-intensive-care beds.

Plotting deaths at the state level can provide vital clues to the trajectory of the epidemic, indicating whether deaths are likely to soar, as they did in New York or mount much more slowly, as they did in Washington.

Murray said Monday the revised forecasts reflect "a massive infusion of new data," incorporating a total of 16,000 hospital admissions versus just 500 that formed the basis of the institute's March 26 assessment.

Experts say the new calculations also do a better job of factoring in the uncertainty of state reports on an epidemic that is evolving at blinding speed, overtaxing the resources available to track its progress and treat patients.

The data include detailed information on COVID-19 health services provided in a number of states, including New York, Massachusetts, Georgia, Colorado, Pennsylvania, Florida and California.

In addition, estimation of the likely peak of the epidemic in each state has been strengthened by evidence that the epidemic peaked after social distancing measures were adopted in seven more locations internationally. Those seven locations are Madrid and Castile-La Mancha, Spain; Tuscany, Italy; Emilia-Romagna, Italy; Liguria, Italy; Piedmont, Italy; and Lombardy, Italy.

"As we obtain more data and more precise data, the forecasts we at IHME created have become more accurate," Murray said in a statement. "And these projections are vital to health planners, policymakers, and anyone else associated with caring for those affected by and infected with the coronavirus."

Experts note that the analysis draws heavily on data from seven states – California, New York and Florida among them – with a lot of variation among them. The variations are likely due to differences in patient mix and hospital care, among other things. The differences between California and New York are striking, with California reporting one death for every 10.6 COVID-19 hospital admissions, compared with one death for every 4.3 admissions in New York.

"I appreciate the work they've put into this, and it's encouraging if correct," says Jeph Herrin, associate professor of medicine at the Yale Center for Outcomes Research and Evaluation. "(But) these are large states, so their different numbers likely reflect different health systems and different demographics, not just random noise. It would be good to know how they reflected those differences in national estimates."

Herrin also noted that it's difficult to assess how IHME factored in the impact of different state social distancing policies. "Until they release the technical details," he says, "it's not clear how they incorporated the impact of different state social distancing policies; it would be good to know what weights they used, and whether they give the same weight to different states that are being more or less rigorous in enforcing their policies."