May 14, 2020

Covid-19 models are finally reaching consensus. (That won't last, experts say.)

Daily Briefing

    After weeks of disparate projections, models estimating the number of deaths related to the novel coronavirus are beginning to reach a consensus on the projected U.S. death toll, but experts say one factor could force researchers to alter these projections in the coming weeks.

    Monday: Your 30-minute global Covid-19 weekly update

    Coronavirus models reach consensus on number of projected deaths

    Last month, models on the new coronavirus had reported very different projections for the number of likely U.S. deaths tied to the virus, but recent updates show the models are largely coming into alignment on the country's projected coronavirus death toll.

    For instance, a New York Times analysis looked at past and present projections from five of the most widely cited models:

    • The University of Washington's Institute for Health Metrics and Evaluation (IHME) model, which is frequently cited by the White House;
    • The Imperial College London's model;
    • Columbia University's model;
    • The University of Texas at Austin's model; and
    • The independent data scientist Youyang Gu's model.

    The Times found that while some of the models, including those by Imperial College London and Columbia University, often predicted above average death tolls when compared with the other models, the five models on average are now expecting 31,000 to 42,000 more Covid-19-related deaths through mid-June, resulting in about 120,000 deaths total nationwide.

    Nicholas Reich, a biostatistician at University of Massachusetts Amherst who also developed a method to merge the results of eight Covid-19 models, came to a similar projection of 110,000 cumulative U.S.  deaths tied to the new coronavirus by June 6. According to Reich and his team, among the eight models they reviewed, the difference between the model with the lowest number of projected deaths and the model with the highest number of projected deaths has decreased by about 50% in two weeks, from a gap of 36,000 deaths to 17,000 deaths.

    While Reich noted that's still a large difference, the gap is significantly smaller than it was just a few weeks ago.

    Scientists are now working with new data and have been able to better assess how state stay-at-home orders are affected the epidemic, which has made the virus' real-world effects slightly easier to predict.

    Easing social distancing measures could cause projections to change—again 

    But some experts say researchers might be forced to drastically alter the models' forecasts as policymakers in more than 30 states are moving to ease social distancing measures and reopen nonessential businesses, moves that could lead to increased transmission of the new coronavirus.

    University of Washington's IHME saw its model's death projections increase after considering "key drivers of viral transmission like … easing of distancing policies" in its analysis, according to researchers. For example, a revision to the model last week increased its projection to 135,000 deaths by Aug. 4—almost double its projection on April 29.

    Another revision on Tuesday caused that number to reach more than 147,000, up almost 10,000 from the last projection on Sunday, according to researchers, who cited easing social distancing requirements as the reason for the spike.

    Some experts said the IHME revisions reinforced warnings from Anthony Fauci, director of the National Institute of Allergy and Infectious Diseases, who this week warned that easing social distancing measures prematurely would cause more outbreaks of the virus.

    However, Lauren Ancel Meyers, a professor of integrative biology at the University of Texas, said the impact of easing social distancing measures will likely show up in the projections within three to four weeks. "There's a lot more uncertainty because the system is changing under our feet," she said (Bui et al., "The Upshot," New York Times, 5/12; Aizenman/McMinn, "Shots," NPR, 5/13; Gorman, Reuters, 5/12).

    Have a Question?

    x

    Ask our experts a question on any topic in health care by visiting our member portal, AskAdvisory.

    X
    Cookies help us improve your website experience. By using our website, you agree to our use of cookies.