Coronavirus models: what the numbers show in Colorado about how we may ‘return to normal’

Masks, continued distancing, and stay-at-home for those 60 and older all look to be needed until either a vaccine or a highly effective therapy emerges.
April 27, 2020
Two young men demonstrate social distancing, a practice that the 'return to normal' models demonstrate is working in colorado
Models show that social distancing should be a big part of Coloradoan’s efforts to ‘return to normal.’ Photo: Getty Images.

A miracle therapy or surprisingly quick-to-market vaccine could change the coronavirus-response game entirely. But absent big surprises, Colorado residents can expect a long, slow return to the “normal” we knew before COVID-19. Along the way – and, probably, for many months – we will need to keep our distance from each other, wash our hands, disinfect surfaces, wear masks, and continue to urge older Coloradans and others at high-risk of serious complications to maintain substantial social distancing.

If we don’t do these things well, coronavirus cases will likely surge again, possibly swamping hospital and critical-care units and undoing what our weeks of staying at home have accomplished.

These are the conclusions of the latest report, released on April 20, from the state of Colorado’s COVID-19 Modeling Group, a team of experts from the Colorado School of Public Health, the CU School of Medicine, The CU Boulder Department of Applied Mathematics, CU Denver, and Colorado State University.

Jonathan Samet
Dr. Jonathan Samet

Models don’t make decisions; politicians and the people and organizations they govern do. Even the best model is only as accurate as the information and assumptions feeding it. But a good model can tell a story that helps to predict the future. That’s what this one does, and its projections are being used as a key basis for policy decisions affecting all of us in the state.

‘Balancing dilemma’

The Colorado COVID-19 Model Group’s April 20 report is an expansion and refinement of the team’s first model released on April 6. That earlier model considered how the statewide personal-distancing (also known as social distancing) mandates instituted on March 17 and 26 affected case numbers and particularly hospitalization and ICU demands. The model ran through several scenarios, varying the possible effect of personal distancing from 0% to 80% reductions in social interaction, the “contact rate” among people.

The new model uses the observed reduction in cases as an input to estimate just how much contact had been reduced: the closing of schools and many businesses, staying at home and other measures plunged Colorado’s average statewide contact rate by 75 to 78%, the team estimates. For its April 20 follow-up, the group modeled how various scenarios for opening up society might impact hospital- and critical-care bed needs.

Opening up depends on maintaining a delicate balance, says Dr. Jonathan Samet, dean of the Colorado School of Public Health and the leader of the Colorado COVID-19 Modeling Group.

“We have unemployment, the collapse of the economy, and other factors on the one hand, and human lives and the threat to health care workers and more on the other hand,” Samet said. “It’s the most incredible balancing dilemma that I think you could imagine at a societal level.”

The guiding principle in striking that balance has been straightforward: we must, in Colorado and beyond, collectively make sacrifices to keep coronavirus infection counts low enough that serious cases do not overwhelm hospitals and intensive care units where critical patients end up. If we don’t, “We will have doctors deciding if Patient A or Patient B gets the ventilator, which would be an impossible ethical situation,” Samet said.

Thanks to widespread adherence to the March 26 measures, we’ve been successful so far. But what happens when we relax those orders? That’s what the modeling team set out to understand.

The ‘return to normal’ model in Colorado

The modelers started with three baseline scenarios. These assume personal-distancing decreases of 45%, 55%, and 65% from the typical contact rate before the pandemic. Various mixtures of changes to personal-distancing measures (in terms of such things as businesses opening and children attending summer camps or not) will correspond to these different contact-rate reductions, but there remains great uncertainty about these correspondences.

“We’re learning as we go, here,” Samet says.

For all updates and to read more articles about the new coronavirus, please visit uchealth.org/covid19

But he does figure that the measures in place in Colorado between mid-March and the statewide stay-at-home order on March 26 brought about a roughly 65% reduction in the state’s contact rate. (Among those measures were the closing of schools and ski resorts and the cancellation of NBA and NHL games and other events.)

Assuming some ICU occupancy for non-COVID-19 patients and those who have elective surgeries, the team estimated there to be about 1,800 Colorado ICU beds available for a surge of coronavirus patients. Under the 65%-below-normal distancing regimen, the new model estimated that, without additional interventions, more than 3,000 patients could flood the state’s ICUs at the mid-November peak. The looser 55% scenario would invite peak ICU demand of more than five times current surge capacity in early September; the 45% scenario estimated peak ICU need at nearly nine times capacity in early August.

The modeling team then ran the same three distancing scenarios with the added assumption that adults 60 and older continue to observe the current, strict stay-at-home orders.

Those in that broad age group could consider such a scenario to be discriminatory. But according to UCHealth Today’s analysis of State of Colorado COVID-19 data, as of April 24, the 60-plus cohort had comprised 60% of the state’s serious cases and 90% of Colorado’s COVID-19 deaths. That cohort comprises just 20% of the state’s population. Compare that to people under 30, who had accounted for 4.5% of serious cases and 0.74% of deaths—despite comprising 39% of the state’s population. Coloradans 60 and older were 23 times more likely to end up sick enough to be hospitalized and 230 times more likely to succumb to the disease than people half their age or younger. These facts are reflected in Colorado Gov. Jared Polis’s “Safer at Home” executive order effective April 26, which includes individuals 65 and older among the “vulnerable individuals” instructed to continue to stay at home with few exceptions.

So it’s no surprise that the scenarios assuming ongoing stay-at-home living among those 60 and older projected less than half the peak ICU demand as the baseline scenarios. Even then, though, peak ICU demand would stay below the red line of 1,800 beds only if a low statewide contact rate of 65% below normal were maintained.

The modeling team separately considered the impact of widespread mask-wearing as well as heightened coronavirus testing and isolation.

Covid-19 modeling chart
Projected timing and magnitude of the peak number of ICU hospitalizations for interventions including high levels of social distancing by high-risk groups (scenario B), mask wearing in public (scenario C), aggressive case detection and containment (scenario D), and combinations thereof (scenarios E and F). Scenarios assume implementation on 4/27 and maintained indefinitely; the study assumes a coronavirus surge capacity of 1,800 ICU beds in Colorado. (Courtesy of the Colorado State COVID-19 Modeling Group)

The team found that, taken independently, 1) wearing masks and 2) improved testing and isolation kept the number of peak ICU beds below the threshold of 1,800 – but also only if the overall contact rate were held at 65% below normal.

Both masks and testing and tracing have their limitations. Masks only work if people wear them and do so correctly. For testing and tracing to be effective, Colorado needs far more testing and a much-enhanced ability to trace contacts. Both will happen over the next couple of months, Samet is confident. But COVID-19 is a tough disease to trace: it seems that contagiousness peaks about 17 hours before symptom onset. An estimated 44% of cases in China were seeded by those who spread the virus during the roughly three days from the onset of contagiousness and first symptoms. That finding jibes with research that found 43% of those who tested positive for COVID-19 in Iceland to have had no symptoms yet. In addition to hiring students and others to trace coronavirus contacts, technological aids such as the cell-phone-based system the Google-Apple collaboration aims to produce could prove to be vital.

By far the greatest benefits – ones that would save hundreds of lives in Colorado while lowering the peak case rates much earlier in time (otherwise, depending on the scenario, those peaks could come as late as December, the model showed) – involved the combination of maintaining social distancing for those 60 and older, wearing masks, and improving testing and contact tracing.

That combination was the only one that kept peak ICU demand below the red line of 1,800 beds in the 55%-reduction scenario.

No one knows what the difference between slashing our collective interactions 65% versus 55% might look like. Those 10 percentage points would translate into some degree of additional business, socializing, and getting out and about, assuming we decide the roughly 1,000 additional ICU admissions and perhaps 500 additional deaths at peak – another Colorado model estimate – are worth the price.

No numerical model can make such a decision; no human wants to. But decide we must.

Whatever balance we strike, it must respect frontline health care workers – that means staying below that ICU-surge red line.

Samet and his team will be the first to admit that their model isn’t perfect – no model is. But it reflects the team’s best estimate of how nature works, Samet says.

“We’re susceptible, we’re vulnerable, we get exposed, and we get infected,” he said. “We’re reflecting processes that we understand, so I think that the foundation is sound. Are all these assumptions perfect? Well, they can’t be. But they’re going to get better, and the model’s going to become more certain over time as we learn more. And it’s the best tool we have.”

About the author

Todd Neff has written hundreds of stories for University of Colorado Hospital and UCHealth. He covered science and the environment for the Daily Camera in Boulder, Colorado, and has taught narrative nonfiction at the University of Colorado, where he was a Ted Scripps Fellowship recipient in Environmental Journalism. He is author of “A Beard Cut Short,” a biography of a remarkable professor; “The Laser That’s Changing the World,” a history of lidar; and “From Jars to the Stars,” a history of Ball Aerospace.