There is no crystal ball to show the future of the coronavirus pandemic, but researchers across the country have been working quickly to come up with the next best thing: an accurate statistical model.
These are the now-familiar arcing graphs showing case numbers rising above hospital capacity – or, hopefully, not rising above capacity – and then dipping back down. But, because the coronavirus is so new, there’s no scientific consensus about how contagious it is and what percentage of people stricken by it will require hospitalization.
That makes the current models vary quite a bit and, in turn, means policymakers have to essentially pick their model when deciding how to respond to the spread of COVID-19, the disease caused by the coronavirus. You see this playing out now in disputes between the federal and state governments – where states are preparing hospitals for massive patient surges that the feds say won’t arrive.
“I mean, it could be you have shortages, and it could also be that you have some that have way overestimated the number of ventilators they need,” President Donald Trump said last weekend.
This uncertainty in modeling is particularly evident in Colorado. One model last week showed the state likely coming up thousands of hospital beds short of what it needs. Then, this week, that same model updated to say Colorado is already past its peak, even as the state continues to build up treatment capacity.
So, here are three models with an explanation about how they were constructed and what their limitations are.
The CU modelGov. Jared Polis and other state officials have been relying on a model built by researchers at the University of Colorado Anschutz Medical Campus. It’s also sort of a middle-ground model in this set.
The model predicts that there are currently 17,000 to 18,000 cases of COVID-19 in Colorado and that the state will be able to avoid overwhelming the hospital system only if Coloradans do a really good job of staying home.
“High levels of social distancing, sustained throughout April, can not only flatten the curve but bend the curve such that we will see a decline in cases and hospitalizations such that hospital capacity is not exceeded,” the CU team wrote in a memo released Monday.
In the model, bending the curve means reducing in-person social interactions by 70% or more from normal. If the state achieves only a 60% reduction in social interactions, the critical-care capacity of Colorado hospitals would become overwhelmed by mid-May.
How well the state complies with the stay-at-home order will also determine when the peak arrives. If there’s poor compliance, the peak would likely hit sooner, at a time when there aren’t enough hospital beds and ventilators to treat everyone. If there’s good compliance, though, the peak may not arrive until late-summer, when the state hopes to have thousands of more beds and ventilators to treat patients.
“This is why social distancing is so important,” Dr. Rachel Herlihy, the state epidemiologist, said in a call with reporters on Monday. “… We are social distancing to save lives. If we don’t take this seriously, we will continue to see an increase in cases and a tragic loss of lives.”
The CU team updated its model Sunday to take into account new data on the impact of the statewide order closing bars and restaurants – which Herlihy said showed that the rate of increase in new cases has slowed. But data on the impact of the full stay-at-home order won’t be available until later this week at the earliest. So Herlihy said there’s still uncertainty in the projection.
Dr. Jonathan Samet, the dean of the Colorado School of Public Health, which is helping produce the model, said researchers were recently able to refine their projections by adding data about Colorado’s age distribution. And they are also receiving patient-level information from the Colorado Department of Public Health and Environment to better understand how COVID-19 has spread across Colorado.
“This has been a very rapid two weeks of modeling efforts,” he said.
While state officials believe the CU model is the most thorough, it also has plenty of unknowns.
Samet said researchers are still trying to understand how infectious the coronavirus is, the differences in how it spreads in rural areas versus urban settings, and what it might look like when the state begins to lift social distancing orders.
This is a big unknown for all the models out there. A paper by researchers at Imperial College in London argued that the United States and other countries will likely see a second surge in patients in the late fall and winter – meaning governments will probably have to toggle between relaxing and increasing restrictions for months, with cases yo-yoing up and down as a result.
Right now, the CU team hasn’t fully projected what it would look like if Colorado lifts the current restrictions. Herlihy said the state is still trying to get a handle on the health impacts of the stay-at-home order before figuring out what turning it off would do.
But Kathryn Colborn, one of the CU researchers working on the model, said the team has some hints. In order for the state’s epidemic to peak, the number of people that every infected person passes the virus onto – a figure known in epidemiology as R0 – has to drop below 1. Otherwise, she said, the epidemic will continue to expand.
In running scenarios, Colborn said the CU team has found that social distancing has to stay high – in the 60% to 70% range – for R0 to go below 1. That means, if the state even just relaxes its social distancing rules, cases will likely climb back up.
“In a nutshell, we anticipate a secondary spike around June” if the state eases social distancing rules later this spring, Colborn wrote in an email.
The Washington modelPerhaps the most widely cited coronavirus model out there is one produced by the Institute for Health Metrics and Evaluation at the University of Washington.
This is the model Trump and federal officials most often cite in their press briefings. And it’s the one that caused Dr. Deborah Birx, the coordinator of the White House coronavirus task force, to include Colorado in a list of emerging hotspots.
“We are watching them because they are starting to go on the upside of the curve,” Birx said of Colorado and other states like Pennsylvania.
But then the IHME model updated on Sunday, and it now shows Colorado already past its peak – and likely to be well under the state’s hospital capacity.
In a blog post, IHME researchers acknowledged that Colorado’s projections have seen “fairly substantial changes” caused by increasing availability of data about deaths in the state and a better understanding about the course of outbreaks across the globe. The model originally looked to China to understand how COVID-19 cases rise and fall, but the latest version also takes into account the experiences of Italy and Spain.
But researchers in Colorado say the IHME model has incomplete data from the state and, therefore, isn’t as reliable as CU’s model.
“We have not hit the peak,” Herlihy said bluntly.
Herlihy said IHME doesn’t have access to the patient-specific data that CDPHE is providing to the CU modeling team. It also doesn’t have precise data on the impact of Colorado’s stay-at-home order.
The IHME model assumes that social-distancing orders will remain in place until June 1. But it also doesn’t predict what happens when those are lifted.
“The question of whether there will be a second wave of the epidemic will depend on what we do to avoid reintroducing COVID-19 into the population,” the IHME team wrote on its website. “By the end of the first wave of the epidemic, an estimated 97% of the population of the United States will still be susceptible to the disease and thus measures to avoid a second wave of the pandemic prior to vaccine availability will be necessary.”
The worst-case scenarioThere’s a third model that hasn’t received as much attention as the first two. It’s from a group of data scientists from around the country who organized under the name Covid Act Now.
“We built covidactnow.org to solve an urgent problem: If we try to fight COVID in the present we will lose,” the team wrote on its website. “We can only beat COVID by understanding what it will do in the future. Our leaders need a forecasting tool to better understand the future spread of COVID.”
It makes sense, then, that this model presents the bleakest picture for Colorado. Part of its purpose is to hammer home how serious this virus is.
If Colorado had taken no action, the model projects that the state would have seen 61,000 deaths from COVID-19, nearly twice the number the CU model projected. Even with social distancing measures in place, if compliance with those measures is poor, the model expects Colorado to exceed hospital capacity by early May and then surge up to nearly 32,000 hospitalizations by mid-June.
Asked about this model, CU’s Colborn said it’s a legitimate model built by a strong team of scientists. So, while the Covid Act Now model hopefully doesn’t show what will happen in the state, she said it’s still worth considering. This is how you build a deeper understanding of the disease and what’s coming.
“Multiple models and different approaches are fine with me,” Colborn wrote on Twitter. “They show just how uncertain we should be in any one prediction. I do not believe in one model to rule them all!”