New Model Predicts Hospital-Capacity Strain From COVID-19 Pandemic

By Will Boggs MD

April 15, 2020

NEW YORK (Reuters Health) - A new model can be used to predict the strain on hospital capacity created by the COVID-19 pandemic, researchers say.

"A key takeaway for me was that despite all the well-intended focus on how we're going to run out of intensive-care unit (ICU) beds and ventilators, our model predicts that we are far more likely to run out of regular hospital beds," Dr. Scott D. Halpern of Perelman School of Medicine at the University of Pennsylvania, in Philadelphia, told Reuters Health by email. "This is important as health systems contemplate converting regular beds to ICU beds."

Dr. Halpern and colleagues developed the COVID-19 Hospital Impact Model for Epidemics (CHIME) in order to forecast the course of COVID-19 in their region and to project COVID-19-related demand for ward beds, ICU beds and ventilators.

The base-case model projected a peak of 2,671 hospitalized patients with COVID-19, requiring 545 ICU beds and 189 ventilators. At the peak of new clinical cases presenting to the health system, the model projected 386 hospitalizations per day, of which 61 required an ICU bed and 19 required invasive mechanical ventilation.

In the worst-case scenario (based on a doubling time of two days), the model predicted peak simultaneous demand of 9,188 ward beds, 1,742 ICU beds, and 577 ventilators, yielding 99.6% probability of exceeding current capacity for hospital ward beds, 40.0% probability of exceeding ICU bed capacity and 30.0% probability of exceeding ventilator capacity during the epidemic.

The mean expected time to exceed current ventilator capacity was 58 days. But it ranged widely, from 17 days in the worst case to never in the best case, the researchers report in Annals of Internal Medicine.

According to the model, the expected time to remain above current capacity was 37 days for hospital beds, 33 days for ICU beds, and five days for ventilators. Depending on different scenarios, however, there was nearly a seven-fold difference in the expected peak simultaneous demand for ventilators.

"While physicians and all other clinicians, and indeed all members of the acute-care workforce, including all those who disinfect the rooms and deliver the food, focus on the patients in front of them, we should also recognize our roles in advocating and lobbying for strict and durable physical distancing policies," Dr. Halpern said. "The differences between the best- and worst-case scenarios shown by our model equate to enormous differences in lives lost, and those differences are largely due to our distancing efforts."

"The analyses in the manuscript are not intended to generalize to other geographic areas or health systems and certainly should not be interpreted that way," he said. "However, a key virtue of the CHIME system is how easily it can be used by others to generate estimates that do represent local circumstances."

Co-author Dr. Gary E. Weissman, also at Perelman School of Medicine, told Reuters Health by email, "One important limitation of CHIME, like all SIR (susceptible, infected, removed) and related models, is that the predictions are valid only for a few weeks at a time. As social distancing policies, testing and quarantine efforts, and closure of public events and spaces change over time, the dynamics of the epidemic will also change."

"Thus, CHIME, like IHME (Institute for Health Metrics and Evaluation) and other epidemic models, is meant to be updated iteratively as more empiric data on the disease course become available," he said. "This is important for two reasons: first, unlike predictions about the weather, what we do as citizens and policy makers really does alter the course of the future. So if the predictions are wrong, it's worth understanding if it's because of the model or because of policies and behaviors in response to the model."

"Second," he said, "long-term forecasts are unreliable and CHIME, or any epidemic model, shouldn't be used right now to identify a specific time to 'reopen' the economy or end quarantine policies. Such approaches misunderstand the limits and uses of epidemic modeling."

"To have a successful plan to lift social distancing policies, we would need much more empiric evidence that the epidemic has petered out (which we don't have right now), in addition to a robust response for testing and quarantining to prevent a second wave," Dr. Weissman said.

The authors have made the CHIME model and interface available for public use at http://penn-chime.phl.io.

SOURCE: https://bit.ly/3cfYu6k Annals of Internal Medicine, online April 7, 2020.

Comments

3090D553-9492-4563-8681-AD288FA52ACE

processing....