The models used to calculate the number of people who would need hospitalization were based on assumptions that didn’t prove out.

Early data from the U.S. Centers for Disease Control and Prevention suggested that for every person who died of COVID-19, more than 11 would be hospitalized. But that ratio was far too high and decreased markedly over time, said Dr. Christopher J.L. Murray, director of the Institute for Health Metrics and Evaluation at the University of Washington. IHME’s earliest models on hospitalizations were based on that CDC data and predicted that many states would quickly run out of hospital beds.

A subsequent model, released in early April, assumed about seven hospitalizations per death, reducing the predicted surge. Currently, Murray said, the ratio is about four hospital admissions per death.

“Initially what was happening and probably what we saw in the CDC data is doctors were admitting anybody they thought had COVID,” Murray said. “With time they started admitting only very sick people who needed oxygen or more aggressive care like mechanical ventilation.”