The Number of Deaths Attributed to SARS-CoV2 will not Peak in New York State Today
The most visible model for projecting deaths from SARS-CoV2 is that of The Institute for Health Metrics and Evaluation of the University of Washington (IMHE). Their model calls for deaths in New York State to peak today and decline from here on out. I don't think that's what's going to happen. Deaths are likely to stay this high for at least a week. Some days will be higher than today and some days will be lower.
I base this conclusion on the strong relationship between the number of active SARS-CoV2 cases and deaths. It's been disheartening to me to hear so many health care analysts declare that inconsistent testing methodologies makes useful analysis of these data impossible. Two things. . .
First, what data are we waiting for? The data we have on this pandemic is much better on a worldwide basis than the data we have on the flu. It's surprising, but the vast majority of deaths attributed to the flu are derived from secondary analyses; not from actual tests of the victims. This is the first pandemic in history where the data have been this extensive and widely available.
Second, before declaring data aren't useful in theory, one should check out whether they are useful in practice.
So, how to put them to use? Anyone whose death was attributed to SARS-CoV2 had to have been tested for it at some point. Unfortunately, it is not widely known at which point in time each individual was tested nor when the results became known. This is only part of the problem that has perplexed many.
In the US, New York State has conducted more tests than anywhere else. So, it is a good place to start. The number of deaths are volatile on a day-to-day basis. So, I tried to find a way to reliably predict the number of deaths that have occurred, on average, over the past three days. To get to the point quickly, the net number of new active cases discovered 8 days ago days fits the data the best. This makes a bit of sense, as the course of this illness can take weeks to resolve. You can see the relationship between these two factors in the chart just below:
As the number of newly identified cases has grown. So have the number of deaths. The problem with thinking deaths have peaked in New York is that the net number of new cases hasn't started falling yet. Those data are charted below.
The data of my first analysis above were through April 8. The number of active cases used to predict the number of deaths on that day are from March 31. There have been three days with higher new, active case counts since, which indicate more bad news is on the horizon.
Now, one could reasonably ask whether the new cases being identified were less severe than those that were being discovered 8 days ago. That one is hard to know. One thought is to see if fewer people are testing positive for the virus. That would provide some evidence fewer symptomatic people (those who tend to be the one's tested) actually have been infected. It doesn't really directly answer the question. It just would provide some evidence the virus had less hold among the population.
The data provide a little help here. Below is a chart that shows the percent of SARS-CoV2 tests administered in NY State that returned a positive result. While the figure has remained stubbornly above 40%, it is down from the high of 50% experienced on April 1 (ignoring the data anomaly observed on March 27 and 28).
So, it could be that New York's worst days are behind it. But 40%+ is still very high. The figure for the rest of the country for April 8 was 17%. So, unfortunately, the evidence points to there being a lot of SARS-CoV2 cases left to find in The Empire State.
I've performed this same analysis for the rest of the United States and the combined nations of Europe. The relationship holds true, as you can see from the charts below. I should note that on a state-by-state and country-by-country basis, the 8-day lag is not always the best fit.
This also makes sense since testing practices, cause of death assignment, and any number of other factors vary. The important thing here is that there is, perhaps, a better way to transparently anticipate when are where the peaks will come for this pandemic.
I'm working on that, and I hope to present some thoughts early next week.