Why our COVID-19 total death projections for the United States more than doubled
Published
May 20, 2020
At IHME, our guiding principle in projecting COVID-19 is to produce the best possible predictions given what we know today – and to continually improve these estimates to support further gains against COVID-19 tomorrow.
When we published our first COVID-19 projections for the United States on March 26, 2020, we used a statistical modeling approach – also known as curve-fitting – in which the available data and knowledge of the disease progression are described mathematically and then used to forecast probable future patterns. Using what information was available at the time about the possible effects of social distancing on the transmission of the virus and the daily deaths for each state in the US, we predicted that approximately 81,000 deaths in the US could occur from the virus (with an estimate range of 38,000 to 162,000 deaths) through August, when the initial wave of the pandemic was estimated to recede and containment strategies could begin to take the place of social distancing mandates. This model was developed as an effort to forecast when hospital resource needs in each state would be most intense rather than an effort to predict the end of the pandemic. Daily deaths due to COVID-19 provide a more direct link to the need for hospital resources than case counts – particularly given uneven testing patterns – which more typically underpin traditional models of disease transmission.
With many locations at or past their initial peak in daily deaths, on May 4, we released an adaptation of our model of the initial peak in deaths that links that model to our emerging understanding of disease transmission dynamics. This new hybrid approach between our initial statistical model and a more traditional disease transmission model will enable the exploration of changes in transmission intensity if – or as the data increasingly suggest, when – social distancing mandates are eased and/or human mobility patterns rise.
With this new modeling approach comes a revised prediction for the US, as of May 20: 143,357 cumulative COVID-19 deaths (with an estimate range of 115,378 to 207,364) could occur in the US through August.
“People are moving around, social distancing mandates are being lifted, and we’re now entering a new phase of the epidemic compared to when we started modeling COVID-19,” said Professor Emmanuela Gakidou, one of IHME’s lead COVID-19 modelers.
Prof. Gakidou says that the main reason why our COVID-19 modeling estimates nearly doubled on May 4 is that our initial assumption, which is that all states would continue strict social distancing through at least the end of May, is no longer true. Many locations are also past their initial peak daily death rates, and we’re now observing flatter and longer peaks in various countries and subnational locations.
Now that states are lifting social distancing mandates, we’re seeing a more prolonged epidemic than we had originally thought we would see, something that is reflected as we incorporate each new day of data into the model.
“That causes our estimates to go up by a lot,” said Prof. Gakidou.
We will continue to update our projections in the coming days and weeks to incorporate the world’s evolving evidence base on COVID-19.
“What’s very important to keep in mind is that as this epidemic evolves, so will the modeling strategies and the estimates, because we do the best to reflect reality on any given day with the information we have available” said Prof. Gakidou. “That information comes to us with very high frequency and speed, and we try to incorporate it all.”
Explore our projections for the US, Puerto Rico, Canada, all countries in the European Economic Area, Latin America, and some other locations with more than 50 reported COVID-19 deaths: covid19.healthdata.org/projections
