Where is the peak for COVID-19 in NY? It can be foggy on the peak, but drop into this thread and let's analyze the terrain more clearly. There are some good signs in the data you probably haven't heard about. /1

First, let's look at the UW Model that the White House unveiled almost a week ago. WH said it was especially looking at the death counts predicted by this model, and it said it would peak on Apr 9. /2

Here's an overlay combining UW model deaths with NY data as reported by Cuomo at daily briefings. (Note there are slight differences: UW is using different source. Also, UW model updates daily, so 3/31 is first prediction for this updated data set.) /3

It predicts a peak of 855 deaths on 4/9. Note the lower bound of UW predicts a peak of 563 on 4/6 and an upper bound peak of 1107 on 4/10). This is how UW predicts everything else (admission rate, beds, etc). They shape their model off reported deaths. /4

This approach produces comical results. If you take their predictions for daily hosp admissions & beds needed, and overlay with actual results, you see what I mean. Publishing these predictions of resource use are scaring public officials (and public) to death. /5

Note: The predictions are off by a factor of 5X! Using the 5X beds req'd prediction, you get a peak of beds needed of 75K in 2 days (Apr 7)! Beds are cycled on avg every 12 days (common model assumption). /6

Using this same assumption of 12 day hospital stay avg, and adjusting by 21.6% (apx 5X lower), you get a more reasonable beds req'd prediction of just under 15K on 4/8. /7

Using this adjustment produces a prediction NY will peak on new hospital admissions at about 1452 *today*, give or take 500 for upper/lower bound ("Back off, man, I'm a scientist";) Note it was 1427 two days ago, and 1095 yesterday. /8

I understand why UW model starts with deaths and predicts resources. It's the only data that's readily accessible across recent epidemics (e.g. Wuhan, South Korea, Italy, etc.). But, some important assumptions in these models are proving wrong if you look at granular NY data /9

In fact, something interesting is going on with the actual beds required data. Beds required actual would be Cumulative admissions - cumulative discharges - cumulative deaths, no? Well, in a perfect world: we don't know how many deaths were never admitted to hospital. /10

But I have recorded most of that data from Cuomo briefings. Yesterday was the third day in a row that discharges out-paced new admissions (not counting deaths). I bet you hadn't heard that, even Cuomo didn't note it was 3rd day. Why is this significant? /11

As Cuomo notes, the most rapidly increasing number in his daily briefings as a pct day over day, are discharges. If this continues, it means people are getting better and being discharged more quickly than 12 day avg. Hmm. What might be causing that? /12

Look at the green column. It is a simple formula calculated directly from what Cuomo is reporting each day. He never combines these. And, to be fair, not every death starts with hosp admission. But surely the majority are. Did NY COVID-19 bed resources peak on 3/30 at 4,327? /13

Surely not, but it's a question you think some reporter at those briefings might ask. Of course, NYC hospitals are under heavy load being at the epicenter, but across the state Cuomo doesn't need 120K-140K beds, does he? /14

I don't think NY is at the peak yet. Because new infection growth rate needs to start falling off about .7%/day. It's at 10.72% yesterday, where it's been hanging out for 3-4 days. That needs to drop to 10% today and 9.3% tomorrow, etc., even as they keep testing more people /15

But I do think they are close, and if the discharge rate increases hold, then whether it comes today or 10 days from now, NY will be fine with their hospital beds. Let's keep praying. /16 END

PS, going through and updating my model numbers. But after this mornings report from Cuomo, the data modelers may be adjusting the numbers in turbo-drive now! LOL! Check out the discharge numbers vs hospitalized numbers!!