A small percentage of the people who get the virus need hospital care, but under many model projections, the state’s healthcare system will be pushed to the brink.
Estimates of the resources that will be needed, how long the pandemic will last and how many people will get infected have a wide range.
There’s less than a month’s worth of data and as much as a two-week lag between infection and when symptoms appear. On top of that, it takes up to another 10 days for test results to come back, making it difficult to accurately assess how many Virginians have the virus and the rate of spread.
But state officials need to model how the virus will spread in order to make plans. Hospital capacity can’t be scaled up overnight. Some states, including Washington, which has had a major outbreak, and Oregon, which has not, have already released modeling data.
“We don’t want to have the numbers we have in Virginia, but we do,” said Daniel Carey, Virginia Secretary of Health and Human Resource at a Monday press conference. “And we can help use that real data to land on a model that we feel comfortable predicting and planning on.”
State officials are using a model created by University of Virginia researchers with the Defense Reduction Agency to plan for the state’s healthcare needs. They want to release it soon, when they feel they have enough data, said Dr. Norman Oliver, Virginia’s Health Commissioner.
“As is typical for these pandemics, or an epidemic of any sort, it started off small and only recently has begun to really mount up. And we didn’t have enough cases to have the … requisite data for building a Virginia-specific model,” he said in an interview Tuesday. “Which is the reason why we haven’t done it. It’s not because we don’t want to. You just didn’t have the data.”
So far, state officials have also declined to provide the number of available hospital beds and ventilators throughout the state, two key metrics in assessing how prepared the commonwealth is for the coming surge.
To better understand how many people may become infected in the state and to assess how many beds and ventilators might be needed, The Virginian-Pilot examined several models used to estimate state infection and hospitalization rates, including one that the Virginia Commonwealth University Health System used to plan for their increased bed needs at the pandemic’s apex.
One of the models is optimistic the spread of the disease can be stopped, another is not.
Forecasting the coronavirus’ spread is a bit like predicting the weather, said Carl Bergstrom, a theoretical and evolutionary biologist at the University of Washington who studies the epidemiology of emerging infectious diseases.
“We can do a pretty good job three days out, but a month out we have no clue,” Bergstrom said. “We can make guesses … but the uncertainties multiply.”
Most scenarios in current models predict Virginia needs more hospital beds and likely more ventilators. State officials agree.
“How many more we need depends on the actions we all take right now. But we know we will need more,” Gov. Ralph Northam said this week.
It’s important to look at both optimistic and pessimistic scenarios because of how pandemics work, Bergstrom said.
“I think, kind of riding the middle line down between those is like rolling the ball along the edge of the fence, it's not gonna stay on there all that long,” he said. “You’re gonna fall off one way or the other.”
The model created by the Institute for Health Metrics and Evaluation at the University of Washington predicts that Virginia will have its highest total of deaths from COVID-19, the disease caused by the new coronavirus, in late May, with 58 deaths every day from May 19 through May 23. This projection changes daily, jumping from about 2,000 total deaths in Virginia on Tuesday night to more than 3,000 on Wednesday morning.
That is the more optimistic of the models. It also predicts that the pandemic will have subsided in Virginia by the end of July.
The White House on Tuesday touted this model while projecting that nationwide 100,000 to 240,000 people could die from COVID-19 in a best-case scenario if current social distancing guidelines are followed.
The IHME model, which projects roughly 94,000 deaths throughout the nation, is also based on significantly reducing contact, said Dr. Deborah Birx, who is helping coordinate the nation’s coronavirus response.
“I would definitely say the IHME model is the rosy picture,” said Oliver. “It assumes very strong social distancing as well as other protective measures.”
Bergstrom, who works at the University of Washington but was not involved in creating the model, says it appears to assume a level of social distancing achieved in Wuhan, China, which implemented extreme measures in January. It’s good to know what will happen in a best case scenario, but it doesn’t show the whole picture, he said.
“You can use that to set a lower bound. You could sort of say, ‘You know, no matter how little we’re worried about the issue, you have to at least do this much,’” Bergstrom said.
And even in that scenario, Virginia might not be prepared.
There are 18,500 hospital beds in Virginia. Of those, about 2,000 are for intensive care. The state has about 2,400 ventilators, including an emergency supply maintained by the state. That’s the total available for coronavirus patients and anyone else needing a bed or ventilator. Currently, the state is nowhere near peak resource use, with the health department reporting 208 total hospitalizations as of Wednesday morning.
The more optimistic IHME model estimates that the state’s need for resources will peak roughly on May 20, when coronavirus patients will need 6,154 beds, 918 ICU beds and 734 ventilators. But the model provides a range for each as well, projecting up to more than 19,000 beds, 2,900 ICU beds and nearly 2,400 ventilators needed on those days.
And if Virginia fails at social distancing, Bergstrom said we could be at or close to that peak need for much longer than the model predicts.
The model also assumes that the entire country will be on lockdown within the next seven days.
“That itself maybe is an optimistic assumption given the political realities,” Bergstrom said. “It's not clear whether our lockdown measures are going to work as well as Wuhan's because they don't strike me as being nearly as severe. So, if that is true, it may also be the case that we end up doing substantially worse than predicted by the model. … Worse than anything in the range.”
The Virginian-Pilot used The University of Pennsylvania Health System’s COVID-19 Hospital Impact Model for Epidemics (CHIME) for a second, more pessimistic analysis of the virus. (You can find The Pilot’s methodology at the end of this story).
CHIME does not predict death totals, but the model is being used by hospitals all over the country to assess how many beds they might need in the coming months. VCU Health used it when the hospital decided to ask for an additional 460 licensed beds to handle a surge of COVID-19 patients, said university spokeswoman Laura Rossacher.
The model assumes that eventually most of the population will get the disease. Still, with social distancing it is possible to stretch the rate of infection out long enough — over 18 months — to prevent the healthcare system in the state from being overwhelmed.
The Pilot’s use of CHIME, like other modeling efforts, shouldn’t be seen as a precise prediction. The modeling scenario relies on many assumptions when there is still a lot of uncertainty surrounding the pandemic. For example, it’s unclear to what degree warmer weather could limit infections, or if new treatments will prove effective.
The Pilot used two scenarios in CHIME, one assuming social distancing effectiveness of 25 percent, the other at 50 percent.
If residents in the state cut down on person-to-person contact by 25% below normal, the model predicts that more than 6,300 people would need ventilators and 11,500 could need ICU beds when the pandemic reaches its peak in the state. Those numbers are significantly above the state’s current capacity.
But if residents cut down on contact by 50%, less than 900 people would need ventilator support during peak times, a much more manageable amount for the state as officials work to ramp up capacity of the commonwealth’s healthcare system. It would also delay patient surges, buying more time for the state to increase testing and secure additional personal protective equipment for healthcare workers.
Both CHIME scenarios assume the coronavirus “is ripping through the population, everyone’s getting infected,” Bergstrom said. “And then the pandemic goes away because everyone has had it.”
Oliver said he’s looked at CHIME models that project the pandemic stretching out into the fall. In those cases, social distancing has decreased the number of people who have the virus at any one time but lengthened the period of time in which people get sick.
“Which sounds bad from the point of view of the individual community member,” he said, “but I think the point to make to them then is that it allows our healthcare system to actually deal with the problem as opposed to a huge surge in a short period of time.”
It’s unclear to what degree residents are following the practices that officials have been urging. Data from Unacast, which uses smartphone data to grade how each county and state have been practicing social distancing, showed that Virginians had traveled about 30% less than normal on March 26, the latest day available. The state has since issued a stay-at-home order and closed public beaches in an attempt to slow the spread of the virus.
How Virginia hospitals would handle a massive influx of patients depends on where you live.
Assuming a 25% reduction in social contact, each of Virginia’s hospital referral regions would struggle to cope with the coronavirus epidemic. How overwhelmed each would be depends on several factors, including the ratio of beds to people. That varies significantly by region, though some resources could be moved around the state.
In the more rural hospital region surrounding Lynchburg, for example, there are nearly 48 ICU beds per 100,000 adults. In the area around Arlington, there are about 23 such beds per 100,000 adults.
Assuming social contact is reduced by 25%, none of these hospitals would likely be pushed to their capacity by May 1. But two months later, some would be well over capacity, according to the CHIME model. Arlington, for example, would be at 190% capacity.
By Aug. 1, each of the regions would recede to about capacity again, with Lynchburg being just a tick over 100%.
At this point, it’s impossible to tell how the pandemic will play out. Time and social distancing will tell the tale.
“I do think there are kind of two different storylines. One storyline is suppression and whether it's 100,000 deaths or 200,000, or even 300,000, it’s in that range,” Bergstrom said. “And then the other storyline is, we lose control of it, and then we're in the millions, you know, a million or higher. And so we don't know which storyline we're on yet.”
Oliver said it’s better to be over prepared than under prepared.
“You’re planning for the big storm,” Oliver said. “Do you wind up doing too much because … it turns out to be the rosy picture of the IHME model and you’ve planned for the CHIME?”
The consequences for not planning for a big surge in cases are far worse, he said. More people would die.
“I don’t want to be the guy standing there explaining why we were sitting there looking at lots of plans for a huge surge and we didn’t do it,” Oliver said.
It’s important to remember, the models aren’t good at accounting how human behavior will change over time.
“These models all assume that people sort of do something and then what they do doesn't change according to what the disease is doing,” Bergstrom said. “But of course, as your friends start dying … we as a society adapt.”
Gary Harki, gary.harki@pilotonline.com
Peter Coutu, peter.coutu@pilotonline.com
Methodology
The University of Pennsylvania Health System’s COVID-19 Hospital Impact Model for Epidemics (CHIME) model was created to help hospitals and public health officials assess their needs in the coronavirus pandemic. It allows users to input more than 10 variables to model their hospital or region.The Virginian-Pilot modeled several state-wide scenarios to examine what might happen.
The degree of social distancing is one of the biggest variables in the scenario The Pilot ran through the CHIME Medicine model, which has to rely on a set of assumptions for how the coronavirus behaves. The Pilot models assumed 5% of total infections will result in hospitalization, 1.5% would end up in the ICU and .75% would need to be on ventilators.
People would stay in the hospital for seven days, the ICU for nine and on a ventilator for 10.
These estimates, at this point, appear to be on the low range for what studies have shown. For example, a study of more than 72,000 COVID-19 cases in mainland China estimated that 15% of total cases would end up in the hospital and 3% would need the ICU. Using those for the model would show an even greater strain on Virginia’s hospitals.
Under the CHIME model, about 110,000 people would be infected at its peak in Virginia if social contact is cut down by 25%. About 17,800 would be if contact is cut down by 50%.
The state has not provided the current number of open hospital beds, ICU beds and ventilators available in Virginia — only total capacity. The Pilot used 2018 resource estimates from the Harvard Global Health Institute, which are based on total capacity numbers that closely match the state’s figures.
Of the state’s roughly 16,500 hospital beds and nearly 2,000 ICU beds, roughly a third are available and about 60% are “potentially” available at any given time, according to this data. These figures are more generous than those used by the University of Washington’s Institute for Health Metrics and Evaluation study, which estimated Virginia had about 329 ICU beds available.
———
©2020 The Virginian-Pilot (Norfolk, Va.)
Visit The Virginian-Pilot (Norfolk, Va.) at pilotonline.com
Distributed by Tribune Content Agency, LLC.
