November 22, 2020
NY Update. Sorry for the slow reporting — it was a combination of a busy week at work and hoping to get numbers that would give a clear sense of the trends in Westchester.
I think the numbers here might have flattened out a bit, rather than continuing the steep climb northward. At least, the upward surge is not as steep.
In terms of new cases, three weeks ago we had about 1000 cases in a week, but the following week those more than doubled, to 2100. But over the last seven days, we had 2376 new cases. In terms of daily new cases, on a rolling 7-day average, the chart looks like this:
So, still growing, but at what looks to be a somewhat reduced rate. We’ll need another week’s worth of data to put the last two weeks into better context, I think.
In terms of testing-positive percentage, a consistent story with the above but even more of a signal that the line is flattening:
The difference between the slope of the new cases chart and the testing-positive chart makes sense to me if you use the analogy of “casting a net” that a scientist interviewed for (I think) a NYT article I mentioned a few months ago. The testing-positive number is like casting a net into a pool — if it rises quickly, that is an indicator that the pool of cases into which you’re casting is teeming with more cases. When you then test more, you’re then going to find more of those cases. In particular, given that there are presumably many more cases with no visible symptoms than there are with symptoms, as you test more and more people, you find more of those asymptomatic people.
We’ve been testing more in Westchester over the past several weeks:
If the trend in performing more tests continues , we may still see higher numbers of new cases even if the testing-positive number continues to flatten or grow at a less steep rate. So you still may see media articles about our rising case numbers. But if the current trend of the testing-positive percentage continues to level off in the 4% range, I’m not as worried about more tests showing more cases.
Our other two more reliable measures of the severity of an outbreak are the lagging indicators of hospitalizations and fatalities. Hospitalizations in the mid-Hudson region, which includes several other counties besides just Westchester, have definitely been on the rise:
With respect to deaths, in September we were seeing one or two deaths per week. Last week, we had 15 fatalities in Westchester. Here’s a chart showing fatalities per week over the past two months:
Unfortunately, the fatality numbers are going to keep climbing, as that big surge two weeks ago rolls through. I would expect to see more like 25–30 deaths per week going forward, if we are finding over 2000 cases per week. More on that down below.
To get hyper-local for a moment, we have around 27 active, known cases in the Pelhams, according to the Westchester government:
There is, however, a new “yellow zone” in New Rochelle that comes right up to the Pelham border:
What’s frustrating about trying to interpret the state and county numbers is the multiple ways in which geography is parsed. The state provides detailed numbers by county for tests given, positive results and fatalities. It provides hospitalization numbers by region. The county provides data by municipality. No one, as far as I can tell, provides data in NY by zip code. But in defining clusters, the state says it goes beyond zip codes and other boundaries, which as far as I can tell must mean it is identifying cases by physical address, and then looking at cases based on proximity to each other? It’s not clear.
But what’s even more odd is that micro-clusters are being used to avoid shut-downs to larger regions that, under prior state guidance, would qualify for restrictions. And the Governor, at least, is using micro-clusters as a way to say that things are better overall than the numbers suggest.
So, for instance, Westchester has been at over 4% positivity on a 7-day rolling basis for two full weeks, and has had more than 30 cases per 100,000 for two weeks as well. The cluster zone guidance says an area goes into a yellow zone if the “[g]eographic area has 7-day rolling average positivity above 2.5% for 10 days AND Geographic area has 10 or more new daily cases per 100,000 residents on 7-day average.”
Yet instead of declaring the county a yellow zone, the state government first declared a cluster in Port Chester and said that the rolling average for the county as a whole was around 2.5%. But in a prior post I showed that if you removed the Port Chester numbers (at least, as best I could figure them based on the county map), that wasn’t correct. And now we’ve been over 4% for two weeks, and while the New Rochelle case numbers shown on the county map have gone up a bit over the past week, the increases are not nearly enough to indicate that all of Westchester’s increase over the past two to three weeks are caused by the cluster zones.
I see several problems here.
Problem #1 is simply transparency — the State should just be much more forthcoming about how they determine a micro-cluster.
Problem # 2 is that the lack of transparency leads to an inability to question what on the face of it seems to be weird decision making, and thus to mixed messages.
Don’t get me wrong, I don’t want the county shut down. But if government officials are telling us that the county-wide numbers are being skewed higher because of specific clusters, show us the data. Because when I look at the map, while there are some increases in specific municipalities that skew higher, there is definitely greater spread across the whole county.
Turning to New York State as a whole, we seem to be holding just under 3% in testing-positive percentage; here’s a chart of the last 14 days:
But, as shown in the 30-day chart, it did climb quite a bit over the past month before levelling off:
So, that’s presumably why hospitalizations state-wide continue to climb:
And this is also reflected in the number of deaths per day in the state:
Finally, I want to briefly turn to the national numbers. This is nothing you haven’t seen elsewhere, I’m sure, but the daily case counts continue to soar. This chart shows case counts on a daily basis since March 12:
We do continue to test more each day as well:
So, we’re getting close to testing 2 million people per day, and we’re seeing close to 200,000 new cases per day. Thus, we’re seeing a nationwide testing positivity rate of around 10%:
The take-away is that nationwide, we are going into the holiday season at a pretty high level across the US, so if there is significant travel and lots of get-togethers, there’s a real concern our numbers in early December are going to go even higher.
And here’s the grim reality of what that means. I’ve posed versions of the following chart before, which tries to show the lag-time in fatalities based on new cases. The key number needed to calculate this is the case fatality rate, which is hard to pinpoint because it changes depending on how much testing is being done and also changes over time depending on lots of factors (including how effective treatments become over time). Still, for these purposes we just need a very rough, ballpark figure so I have used a range of 1.4% to 1.9%. Just remember, this is not the actual overall case fatality rate for everyone infected with the virus — it is an estimated range of case fatality rate for those people who have actually been tested.
As you can see, you can look at the number of new cases nationwide in a two-week period and make some guesses about what the number of deaths may be in the two-week period that follows. So, the last two weeks in October we jumped up from about 700,000 new cases the prior two weeks to about 1 million new cases, and in the first two weeks in November we saw an increase from 11,322 death to 14,634 deaths.
We then saw a further jump to over 1.7 million new cases the first two weeks of November. Using a 1.4% fatality rate, that would imply over 23,000 deaths from November 15 to November 28. I’m going to guess the CFR continues to fall because we are doing more testing than ever, but even at 1.2% we’re talking 20,000 new deaths by next Saturday, for the current two-week period. As of yesterday — one week in — we had seen 10,000 new deaths, so since we’re now averaging close to 2000 deaths/day, it looks like we’ll unfortunately readily hit that mark.
What’s most disturbing about that is that we’re currently on track to hit over 2.3 million cases the current two-week period (we are halfway through and we are averaging 167,000 cases/day). So in the two-week period starting after Thanksgiving we are likely to see over 25,000 new deaths. That takes us back to where we were, nationwide, in April. The difference being that a pretty universal lockdown had gone into effect in March so the numbers started declining as April progressed, which may not happen here.
To put this in perspective, we’ve had about 250,000 deaths thus far since the beginning of the COVID outbreak. If the numbers hold we will be adding a tenth of that in a two week time period. We could easily pass the 300,000 threshold by the end of the year (and, excess death numbers suggest that the count is low anyway).
This exceeds the number of combat deaths of US troops in ANY war, full stop. It surpasses all other pandemic deaths in the US, save the 1918 “Spanish Flu” epidemic, which killed over 600,000 (but the US was much less populous then). I do get tired of seeing people still trying to downplay the severity of a disease killing 2000+ per day in this country alone.
We likely have a long, hard winter ahead of us, but as we head into this first holiday week, let’s count our blessings.
Stay safe, all.