My apologies for constantly bombarding you with numbers, and models, and projections (oh, my). But 1) I am an epidemiologist and it is my job, and 2) major media outlets are running with the story of a new COVID-19 projection that deserves your attention.
Released alongside a preprint publication yesterday, the Institute for Health Metrics and Evaluation at the University of Washington has developed a data visualization tool that projects national and state-by-state COVID-19 curves. By utilizing case, hospitalization, and death rate data from across the US, the tool projects when the peak of the curve will occur. Even with the social distancing measures of all 50 states factored into the model, it predicts that the US will reach the peak of hospital utilization and the peak number of daily COVID-19 deaths within two weeks, on April 15th.
US Projected COVID-19 Peak: April 15th
Based on the model at the peak of hospital resource use, the US as a whole may be short more than 50,000 hospital beds including a shortfall of more than 13,000 ICU beds.
The number of COVID-19 deaths per day in the US is estimated to reach 2,200 (with a possible range of 1,100 to 3,300) at peak on April 15th.
In total, the model predicts just over 81,000 COVID-19 deaths in the US by June 1st, with a possible range anywhere from 36,500 to 144,500 deaths.
To me, these models are interesting but they are still only models at the end of the day. They are projections of what may occur, but they are not necessarily predictive, and certainly not prescriptive, of what will occur in the US.
What I find more informative about these models are some of the state specific projections – not so much the numbers but rather the timing. Take my state, New Jersey, for example. It would seem we are ahead of the national curve, and not in a good way.
NJ Projected COVID-19 Peak: April 8th
The model predicts the peak of hospital resource use to occur in approximately one week, on April 8th.
Statewide, we will be very close to capacity in terms of overall beds, and almost at double the capacity of ICU beds.
The number of deaths per day peaks by April 9th at 90, with a range of 50-125 deaths per day.
By around May 1st, New Jersey total deaths hits its plateau (a full month ahead of the US as a whole).
Statewide, total deaths are projected to be between 1000-2600, with 1850 being the point estimate.
It would appear that New Jersey (and neighboring New York) hit peak a week earlier than the nation reaches its peak. It is certainly possible, though health and government officials around here have certainly been hoping to push the peak out at least a few more weeks longer than the seven days this model is giving us.
Just as states across the US have different curves and timing, it is probably that cities and communities within states also have different curves. These projections are based on state-wide data and are not specific to any one part of the state. Our city implemented social distancing measures about one week before there were state-wide mandates. Yet, our city is more densely populated and closer to the NYC epicenter than other parts of the state.
I wish to reiterate that these mathematical models are imperfect. They are based on current evidence, which is limited, and are not necessarily indicative of the actual situation. We already know that we are under-identifying and underestimating total cases. However, we should still do our best to prepare for a sooner-than-anticipated peak.
It is also unlikely that COVID-19 spread will simply level-off around May 1st in New Jersey or June 1st nationwide. As we begin to relax social distancing measures, the spread of COVID-19 can be expected to continue, but hopefully at a much lesser rate. In addition, a possible seasonal effect on COVID-19 transmission could bring us a resurgence in the winter months, especially if all social distancing measures have been removed.
Explore the Institute for Health Metrics COVID-19 Projection Tool
Note that in these visualizations, the dotted curves represent the single point estimates. The large swath of color surrounding those lines depicts the wide range of possibilities surrounding that estimate.