Covid-19 may soon peak in three major US cities. Here’s what that means

It’s not over yet.
It’s not over yet.
Image: REUTERS/Brendan Mcdermid
We may earn a commission from links on this page.

New York, Detroit, and New Orleans—three metropolitan areas with some of the highest number of Covid-19 cases in the US—may all be hitting their peak number of hospital beds needed within the next week.

“If mitigation in New York worked—and we believe it is working—the cases are going to start to go down, but the mortality will be a lag behind that because of the co-morbidities and other conditions,” said Deborah Birx, the coronavirus response coordinator for the White House Coronavirus Task Force, in a press briefing on Saturday.

Birx and her team base these predictions on epidemiological models—in particular, those created by the Institute for Health Metrics and Evaluation (IHME) based at the University of Washington’s School of Medicine.

When you hear of the pandemic “peaking” in these cities, what people are likely referring to is peak medical resource utilization. Many of these models, including the one from IHME, predict hospitalization rates, ICU admission rates, ventilator use, and deaths. In part, that’s because it’s simply easier to obtain that data—as opposed to, say, infection rates, which are impossible to know in the US since such little testing has been done.

Thanks to what IHME’s director calls “a massive infusion of new data,” the model’s creators can feel increasingly confident in the accuracy of those kinds of predictions, which were updated on Sunday.

But not every model agrees with IHME. In the absence of a single, definitive national epidemiological model, states and many other organizations have been left to create their own, so that public health officials can apply the right mitigation strategies and hospitals can prepare for how many patients they expect to receive.

Models can disagree based on the assumptions of their creators. At the beginning of an outbreak, epidemiologists use estimates for how many people might be susceptible to the disease, how long a person stays sick, how long it takes for an infected person to develop symptoms, how likely a person is to contract the disease if they come into contact with someone who has it, and how likely someone is to be hospitalized or die from it.

All that information is in very short supply when a disease is new, so researchers start with educated guesses. “In early stages you use a lot of assumptions, then you refine models over time using real data,” says Glen Mays, a public health expert at the Colorado School of Public Health at the University of Colorado.

For the first coronavirus models in the US, epidemiologists had to rely on data from other countries, which may have varied because of differences in socialization patterns or underlying health conditions. For example, it’s much more common for men in China to smoke than it is for men in the US, which may have affected estimates of hospitalization rates.

But now that more data is available, models can still vary widely. The IHME model, for example, predicts that Colorado has already hit its peak number of hospitalizations, while state-made models say the peak is yet to come.

“Models are approximations of reality and all predictions should be approached cautiously. There is a great deal of uncertainty in model predictions and it is important to communicate this uncertainty, as well as model limitations and assumptions, alongside any model projections,” says Eva A. Enns, an associate professor of health policy at the University of Minnesota School of Public Health.

Trump Birx Covid-19 models
Birx and Trump in front of a Covid-19 model at Saturday’s press conference
Image: REUTERS/Joshua Roberts

Faced with the choice between several conflicting models, public health officials tend to go with the more dire one. “All these models have a wide margin of uncertainty to predict date of peak. A wise thing to do if you’re a hospital administrator or a public official is to overshoot the mark with the understanding that the consequences of not having enough [hospital] capacity are really quite dire,” Mays says.

State leaders and experts have criticized the Trump administration for only relying on one model, and one that offers less dire predictions than others.

Still, no estimate is a guarantee. Things could still change—especially if residents interpret the news of an anticipated peak to mean that they can go back to their daily activities and stop their mitigation efforts. That is not the case, neither in these three cities nor anywhere else where hospitalization peaks may be farther in the future.

“As we noted previously, the trajectory of the pandemic will change—and dramatically for the worse—if people ease up on social distancing or relax with other precautions,” IHME’s Murray said in a press release. “Our projections are strengthened by the new downturns in more regions. This is evidence that social distancing is crucial. Our forecasts assume that social distancing remains in place until the end of May.”