I know that the reported numbers from this week have a lot of people saying "I told you so", but from a modeling perspective, the rapid rise in cases+deaths in the US is fairly unprecedented.
I follow data for 70+ countries and the US is truly in a league of its own.
I follow data for 70+ countries and the US is truly in a league of its own.
I am estimating that the number of new infections right now (275,000 per day) is exceeding the level we've seen in March/April.
That means that 1 in every 90 Americans (1.1%) is currently infected (including asymptomatic individuals).
This may still be an underestimate.
That means that 1 in every 90 Americans (1.1%) is currently infected (including asymptomatic individuals).
This may still be an underestimate.
Our infections estimate matches the curve of the nationwide hospitalization data (after accounting for a 2-week lag).
Just 3 weeks ago, there was a lot of disagreement on the future trajectory, even among the more accurate models.
This is intrinsically a difficult problem and forecasts are probabilistic rather than perfect. I encourage looking at confidence intervals.
This is intrinsically a difficult problem and forecasts are probabilistic rather than perfect. I encourage looking at confidence intervals.
Our model only trains on daily reported deaths to minimize overfitting. But that has the downside of being slow to react to large spikes in cases.
As more data comes in, it should be able to adjust accordingly. I'm working to better incorporate the recent trends.
As more data comes in, it should be able to adjust accordingly. I'm working to better incorporate the recent trends.
The only other country that shares a similar trajectory is Iran. They've seen their daily deaths increase 3x since they reopened back in April.
If the US follows a similar trajectory, we'd be looking at 1,500 deaths/day by end of August (albeit unlikely for now).
If the US follows a similar trajectory, we'd be looking at 1,500 deaths/day by end of August (albeit unlikely for now).
Just for the record, here's how our forecast from 2 weeks ago has fared so far.
I included IHME's projections as well, because why not?
I included IHME's projections as well, because why not?
