Data modeling may have been unfamiliar to most of us before the COVID-19 pandemic, but it’s now front-and-center in the national conversation.
Prominent in that conversation is a model from the Institute for Health Metrics and Evaluation (IHME), an independent population health research center at the University of Washington. The White House coronavirus task force referred to the IHME model — projecting numbers of COVID-19 deaths and hospital resources needed over time — in a presentation in late March, and it has remained in the news since.
“Modeling a pandemic is a bit like hurricane tracking,” says Professor of Health Metrics Sciences Ali Mokdad, “but instead of modeling for just one geographic region, you’re doing it for much of the world.”
That means sifting through information from around the world and making constant updates to forecast what could happen in the future. “With more data come better projections,” says Mokdad, who is also the UW’s chief strategy officer for Population Health.
Over the past several weeks, these projections have offered health-care systems and policymakers the tools to make informed decisions for managing the COVID-19 pandemic.
Data-driven policy
[pagelet id=37516]On March 31, Deborah Birx, coronavirus response coordinator for the White House Coronavirus Task Force, discusses data from the IHME’s Chris Murray model.
Though the Trump administration had initially hoped to reopen parts of the country as early as mid-April, its tone changed as stark data rolled in.
On March 31, the White House coronavirus task force presented sobering projections that influenced a decision to recommend states remain closed for much longer. That decision was based partly on a data model from IHME Director Christopher Murray and a team of IHME researchers.
The model, which would revise its projections as the pandemic unfolded, provides state, national and international COVID-19 data. By detailing the number of current cases and projecting when they’re expected to spike, the IHME model has been informing crucial policy decisions that, according to the Washington Post, may have saved thousands of lives.
“IHME’s new forecasts for a growing number of countries around the world demonstrate the wide range of responses policymakers and health officials have had to the pandemic,” said IHME Director Dr. Christopher Murray. “We aim to inform their decisions on how best to manage and mobilize for COVID-19.”
Models don’t predict the future — they help us prepare for it
As Washington state emerged as an early hot spot for the novel coronavirus, UW Medicine asked the IHME team for help predicting the numbers of resources — hospital beds, ICU beds and ventilators — it would need and how quickly it would need them. IHME leapt into action, and their work in this area then made it possible to provide state, national and international projections on the spread of the pandemic and the efficacy of containment measures.
Throughout the pandemic, and in response to critiques of the model, IHME has been open about the methodology used, the challenges of projecting an unprecedented event like COVID-19, and the importance of updating the model based on changing conditions — from policy changes to trends in individual behavior.
Changing behavior, changing data
Though earlier revised projections suggested there would be fewer deaths during the first wave of COVID-19 infections, on May 4 IHME once more made a sobering adjustment.
With some states lifting social distancing restrictions before their death rates had decreased, and with anonymized cell phone data showing that people were leaving home more, IHME’s model projected a death toll of more than 135,000 Americans by August — with a wider range of possible scenarios suggesting between 95,000 and 242,000 deaths. As new data becomes available, IHME continues to update these projections.
Murray said that the jump in projected deaths reflects “the effect of premature relaxation of restrictions” by state and local governments, as well as the fact that Americans were going out and interacting more as stay-at-home orders wore on. Including mobility information like the cell phone data adds a more traditional epidemiological model to the mix — a factor some earlier critiques had seen as a missing component in IHME’s approach.
The New York Times via the Seattle Times
What’s next?
With its early and aggressive response to the COVID-19 pandemic, Washington state successfully flattened the curve for the first wave of infections. But as the state and country gradually reopen, IHME is developing more robust data modeling to provide even better decision-making tools.
IHME’s updated model is a hybrid, combining both a statistical modeling approach and a disease transmission approach, leveraging the strengths of both types of models.
“The hybrid model allows us to better track changes to social distancing mandates and other drivers such as testing, contact tracing and temperature,” says Murray.
For the time being, without effective drugs or a vaccine, Mokdad says, “the virus is unfortunately in charge. This makes physical distancing a critical tool in the short term. The models can then help inform when some relaxation of distancing will be appropriate.”
Novel coronavirus (COVID-19) updates
- Novel coronavirus and COVID-19: UW facts and resources
- Recent news coverage: UW experts comment on COVID-19