However, experts and heads of state continued to be relentless for dark weeks to come, noting that the revised model created by the Institute for Health Metrics and Evaluation (IHME) at the University of Washington is in conflict with many other models showing equipment shortages, more deaths and more deaths. projected peaks.
Some state leaders are also increasingly concerned about how the federal government is using IHME’s lower estimates to deny states’ increasingly desperate requests for equipment and assist them in their preparations. The marked differences between the IHME model and dozens of others created by states reveal the glaring lack of national models publicly provided by the White House or agencies such as the Centers for Disease Control and Prevention that local leaders can use in the planning or preparation.
“It’s not clear what the White House is doing on this front,” said Dylan George, who helped Obama’s White House develop models to guide its response to Ebola in 2014. “As a result, every state try to create their own models to anticipate their needs. And you have a model like the IHME which is adopted as a national guide. “
The danger of relying so heavily on a model is that the model may be wrong or overly optimistic.
“When you plan, you want to plan for the worst, not the average or the best,” said Natalie Dean, assistant professor of biostatistics at the University of Florida. “Because the risk is not proportional.”
This is how the striking models can differ.
Local district leaders say their model on April 3 estimates the epidemic in the nation’s capital will peak on June 28. The IHME model predicts that the peak will arrive in a few days, April 16. The district model predicts that hospitals will need 1,453 ventilators at the top. The IHME forecasts a need for only 107 people. The district uses the IHME model as the best scenario and the most severe model to prepare for a likely outbreak.
“Although we hope that our experience will follow a curve closer to the IHME model, we cannot use a single model for our preparation and risk being under-prepared. We continue to refine our models and assumptions and adapt them to the population and context of DC, “spokeswoman Alison Reeves said in an email.
In states more populated than the district, this vast gap in planning and modeling could mean a difference in life and death for tens of thousands of people.
LaQuandra S. Nesbitt, director of the district health department, explained how city leaders chose their model. Its name is CHIME and was created by researchers at the University of Pennsylvania.
“We felt that a model that determined that the district would have practically no medical needs was not indicative of what we expected would be our reality in the district and we thought that a model that did not overestimate the impact of social distancing in the United States “was the right one,” said Nesbitt.
In the two weeks since the release of the IHME model – the researchers announced revisions on Monday – some experts deemed it overly optimistic. But even critics are quick to note that in the absence of any tool offered by the federal government and no other model offering national estimates by state, the IHME could be a lifeline.
To coordinate their response, some states with little modeling resources or experts from the home state have used IHME forecasts that project peaks in mortality and the resources required. The White House partially relied on it to generate its estimate last week that the epidemic would kill 100,000 to 240,000 people across the country.
Most epidemiological models look at different populations that interact during an epidemic – people susceptible to infection, those who are infectious, and those who are already infected who die or recover.
Funded by the Bill & Melinda Gates Foundation, the IHME model takes an entirely different statistical approach, taking the trend curve for deaths from China and “fitting” this curve to new data on deaths in US cities and counties to predict what could follow.
For this reason, many experts considered the IHME to be too optimistic when it launched on March 26. Few states or cities in the United States are taking such drastic measures as those adopted in Wuhan, China – the cradle of the coronavirus pandemic – or even in northern Italy. the residents.
Another big difference between the IHME and other models is a fundamental assumption about the effectiveness of social distancing. IHME model creator Christopher Murray said that many state models assume that social distancing will only slow or reduce transmission to some extent. The IHME model, inspired by the example of Wuhan, assumes that policies such as social estrangement and home support orders can effectively reduce transmission to the point where an epidemic – at least in its first wave – is actually controlled by the authorities. .
More models, better prediction
When it comes to forecasting, statisticians have a favorite maxim: “All models are wrong, but some are useful. “
Using multiple models, experts are often able to better triangulate their predictions and assumptions. This is why meteorological experts often rely on several models rather than one to forecast storms, using a “set” or “suite” of models. Such sets are also the source of the hurricane uncertainty cone.
In Illinois, heads of state use four models – a version of the CHIME model and models from the University of Chicago, Northwestern University, and the University of Illinois at Urbana-Champaign. New York officials said they rely on at least four different models, including the IHME.
In North Carolina, state leaders use a “weather forecast” approach that combines multiple models – and instead of focusing on specific dates or numbers of beds or fans, they decided to predict the probability that the medical system is overwhelmed.
“What we focus on is less about” when is the peak time? »And« what is the exact height of the peak? “And more about the likelihood that demand for health care will outweigh supply,” said Kimberly Powers, epidemiologist at the University of North Carolina at Chapel Hill. “Are we going to need more than we have?” “
Their composite models predict a peak between mid and late May. As long as social distancing continues, they predict only a one in four chance of exceeding the capacity of acute care hospital beds. But if these orders were lifted after April, the chances of overloading hospital capacity would double.
Another concern of some experts is that the IHME model is used too much like an overweight crystal ball given its anticipated needs for ventilators and hospital beds and staff.
State leaders have said that Trump administration officials have used IHME numbers to push back and, in some cases, refuse requests for equipment and assistance. State authorities have cited emails and documents in which federal officials highlight the IHME projections as evidence that the state needs thousands of fewer fans and beds than the proposed model l ‘State. State officials spoke on condition of anonymity because they feared retaliation from the Trump administration that could lead to even fewer fans and less federal aid,
“If the federal government really makes these kinds of life and death decisions on one model and only on the lower end of that model, it’s scary,” said a state official.
Murray said he was well aware of the criticism of his model
But in the absence of any other state-by-state planning tool, he noted during a briefing with journalists on Monday, his model provides an indispensable public service – point even critics of the IHME model are often quick to highlight it.
Murray and his team have worked tirelessly since releasing their model to feed the new emerging data and refine its projections.
On Monday, they announced their most significant revisions to date – prompted by a large amount of new national and international data.
While their original model was based only on the Wuhan curve, the updated model now incorporates curves from seven regions of Italy and Spain where epidemics have also peaked.
The new version also found that deaths in some states – such as Florida, Virginia, Louisiana and West Virginia – could peak earlier than previous projections. But the death toll nationwide is expected to peak on April 16. The new model suggests that the number of acute care hospital beds needed for the spike could be reduced by almost half and the number of ventilators needed for the spike in increase could drop from 40,000 to 29,000. The model also suggested that the total number of deaths would be lower, with around 82,000 deaths from the first wave of infection, although the number could range from 49,000 to 136,000.
Murray agreed with others’ criticism that multiple models should be used.
“I couldn’t agree more,” said Murray. “What we have learned from 30 years of weather forecasting, even the Netflix forecast for movies … you make better forecasts when you use multiple models.” “