Ricochet is the best place on the internet to discuss the issues of the day, either through commenting on posts or writing your own for our active and dynamic community in a fully moderated environment. In addition, the Ricochet Audio Network offers over 50 original podcasts with new episodes released every day.
Coronavirus Update 4-19-20: New Data from NYC
I have a new report for you today, based on a new data source from the New York City health department, which is now reporting COVID-19 deaths by date of death. This is information that I’ve been hoping to see for some time. I’ll also show you the difference it makes, comparing the date of death data to the prior data set that I was using, which was based on the date of the report of the death.
There are limitations to any data set. One obvious limitation of the date of death data is that it is not necessarily complete, particularly for the most recent few days. In addition, as Mark Camp and others have pointed out, there is a possibility of overcounting because a person might be counted who died with the virus but not of the virus, and as Kozak has pointed out, there is a possibility of undercounting as well, as some COVID-19 deaths could be missed.
NYC is reporting “confirmed” and “probable” COVID-19 deaths. “Confirmed” means the decedent was an NYC resident with a confirmed positive lab test. “Probable” means that the decedent was an NYC resident with COVID-19 (or equivalent) listed as a cause of death but without a know positive lab test. It appears that testing continues and that they periodically update the data when a “probable” death becomes “confirmed.”
The NYC data source is from GitHub (here). I will compare this date of death data with the prior data set I was using, from USAFacts (here). Since I am not comparing NYC to any other location, I have not adjusted this data by population, so the graphs will show the actual number of reported deaths. The data is through April 17 for NYC, and through April 16 for USAFacts. Reporting at USAFacts is delayed at the moment, as several states (including NY) have changed their method of reporting.
On to the graphs. The first is daily deaths, with the red, orange, and blue lines representing the date of death figures from NYC, as follows: total deaths (red), confirmed deaths (orange), probable deaths (blue). The yellow line is the date of report data from USAFacts.
You can see that the reporting was significantly delayed and that actual deaths peaked earlier than reported deaths; the actual peak was April 7, while the reported peak was April 10 with a secondary peak on April 14. Remember that the data for April 15-17 is probably incomplete, so we don’t know if there has been a notable recent decline or if this is the consequence of reporting delays. (If you’re looking carefully, you’ll also notice a strange negative figure on the y-axis label, because the USAFacts date of report data actually reported a slight negative figure on one day, presumably a correction.)
The next graph shows you the real extent of the underreporting, as it shows cumulative deaths through the date indicated. Again, the red, orange and blue lines are the date of death data from NYC, and the yellow line is the date of report data from USAFacts.
Once again, remember that the last few days are probably underreported.
Notice how much more quickly both the red (total) and orange (confirmed) lines increase, compared to the yellow (reported). At the height of the crisis, the reporting was about 4-5 days behind. This generally means that the rate of increase in deaths was faster in the earlier period, and has been slower more recently.
I am not being critical of anyone for these reporting delays. They are inevitable, and the folks in NYC seem to be doing a fine job in very difficult circumstances. But when better information becomes available, I think that it’s important to focus on how it changes what we previously thought.
To demonstrate the extent of the underreporting under the old system, here is a graph showing the ratio between total actual cumulative deaths (on a date of death basis) and total reported deaths.
As you can see, prior to April 5 the number of reported deaths in NYC was less than half the number of actual deaths that had occurred.
The next graphs will show the statistic that I have focused on with my prior reporting, which is the daily % growth in total reported deaths. For completeness, I’ll start with the entire period, though it won’t be very helpful.
As you can see, there are very large spikes in the early period, especially in the trend line based on the date of the report. The graph is dominated by the yellow spike on March 16, which is a 400% daily increase — but it is from just 1 reported death to 5. To show the actual trend, I’ve started the next graph on March 22, and I’ve omitted the date of the report line.
As before, remember that the data for the last few days is probably incomplete (probably after April 14, where you can see a bit of a downward departure from the trend line). I think that this is a very important graph, showing that the real crisis in NYC is probably over.
It’s hard to tell whether the lockdown had an effect. The trend line was downward before the lockdown would be expected to have any effect, as you can see. It is possible that voluntary social distancing had some effect before the formal lockdown, which would make it more difficult to identify any change attributable to this policy.
The big question, of course, is whether this is the result of the current lockdown, or whether the NYC is at (or near) herd immunity at this point. This question is not answered by the data in this post. NYC currently has 129,788 reported cases and a case fatality rate (CFR) of 10.1%. It would be extremely useful if someone would do the type of study in NYC that Drs. Bendavid and Battacharya did in the SF bay area.
ChiCom delenda est.
Published in Healthcare
Jerry, another excellent contribution. Thank you.
Fascinating.
I’m sure that the hospitals got better at treating the pneumonia as the hospital staffs settled in to the daily life of having so many pneumonia patients all at once.
Thank you for this information. It is helpful and interesting.
Marci, good comment. I’ve been wondering about cause of death in COVID-19 cases. The main mechanism that was publicized was a pulmonary edema (fluid in the lungs), which required mechanical ventilation. However, the number of deaths in NYC (about 13,000) seems quite inconsistent with the reports that I’ve seen of ventilator use (which seems to have peaked at around 5,000). My impression is that when people died of the pulmonary edema, they had typically been on a ventilator for several days, which would imply much greater ventilator use than has been reported.
I wonder if there are other mechanisms of death from COVID-19 that haven’t been as widely reported. Another possibility is that, for example, a nursing home resident could have a sudden pulmonary edema that causes death so rapidly that he never even makes it to the hospital. At this point, I just don’t know.
I would appreciate it if anyone has a source of information about this. At the moment, it’s just speculation on my part, based on ventilator use at a lower level than I would have expected.
One thing that bothers me the most about statewide shutdowns like New York’s is that it seems completely unnecessary and terribly counter-productive (and maybe just a bit punitive). Looking at the statewide deaths in New York at syracuse.com, a site Jerry probably wouldn’t use but I just want to find statewide data to compare the rest of the state to NYC. Anyway, according to syracuse.com, the statewide death total is 13,869 as of 4/19 with 10,022 from the five NYC boroughs at 11.9 deaths per 10,000. That leaves 3,847 deaths for the rest of the state; but, the counties next to NYC (and probably with many commuters to the city) Nassau, Sulfolk, Westchester, Rockland & Orange have 3,394 deaths at rates varying from 5.5 to 10.3 per 10K leaving only 400 or so deaths for the rest of the state. Erie, which is basically Buffalo, has 136 deaths but a rate of only 1.5 per 10K and most of the other counties have very low death rates generally a little under or right around 1 per 10K. Despite this, Coumo apparently has no intention of treating any of these lower risk locales any differently than the hot spot of NYC.
It looks like daily deaths peaked on Apr 7th. That is well after the shutdown, so distancing could have an effect. If there was a change in nursing home procedures (strict quarantine), that would have a huge effect. There would also be an effect from the change in treatments. I can’t think of the best metric to measure the peak, since everything is changing all the time (behaviors, treatments, definitions, …). But looking at intubations and deaths is the most important metric, but that is what we want to avoid.
Since we are trying to manage a pandemic based on NYC stats, they really should make a better effort to have them be more timely.
I suspect that New York City has very very many more cases than have been confirmed. I base this on the fact that I saw on one of their websites that they specifically said that they were only testing people with the most severe symptoms. Presumably they are turning away hundreds of thousands of actual infected people for testing. In fact the rate of positive results from the entire State is a whopping 40% of those tested, and most of those tests have been done in the NYC area.
https://covidtracking.com/data
Contrast this with Cleveland Ohio’s Cuyahoga County where my wife is doing testing and she tells me that they, and all other hospitals, are only getting positive results from about 4-5% of referred patients, and we were considered the “hot spot” in Ohio (until a massive outbreak occurred in a Prison a few days ago in another county). Her hospital has been limited in the number of tests that they can perform because most of the test kits have been re-routed to New York to fill their great need, and the maker of the main test she is using is an Italian company who needs to supply their own country first. So Ohio also does minor triage for testing, and doctors are told to not refer patients for testing who have very mild symptoms. The patients are instructed to simply stay home and keep away from infecting anybody else. With a positive rate of only 4-5%, I’m, guessing that Cleveland doesn’t really have that many more actual cases than have been confirmed, and of the patients with mild symptoms who are not referred for testing, presumably even fewer than 4-5% would register positive.
I’ve noticed the same thing in New York. My brother lives upstate in Saranac Lakes. It straddles two counties, one which currently has 13 total cases, and the other which has 18, and of course zero deaths. Several counties that are less than 100 miles from New York City have less than 50 cases each, and with only a handful of deaths between them.
https://covid19tracker.health.ny.gov/views/NYS-COVID19-Tracker/NYSDOHCOVID-19Tracker-Map?%3Aembed=yes&%3Atoolbar=no&%3Atabs=n
I live about 45 minute from your brother. Our county reports 49 confirmed cases, three deaths as of Friday, and a positive test rate of about 9.5%. I’m sure we’re typical for rural NY counties. The decision to impose NYC restrictions on the state as a whole seems nonsensical, from where I’m sitting.
Wow. We’ll have to come visit next time we go to see my brother!
I thought herd immunity is months away, even if lockdowns are lifted, because it requires something like 70% exposed. Approaching herd immunity is impossible with lockdowns still in effect, right?
Nice detective work. A lot more elegant than Worldometers dumping the 3,778 NYC probable deaths since March on to a single day (April 14) in its data tracking.
Had I been alert enough surveyed your work last night, it would have saved me hours of measuring my saved NYC bar charts and “guesstimating” the distribution numbers for each day.
Oh well…good work.
I thought the issue with NYC was the mass transit system + the slow walking of mask usage. If the gigantic numbers from NYC are explained from density and mass transit, couldn’t NYC be a bit further along on the herd immunity than the more rural/less dense areas?
Herd immunity is a function of the “basic reproduction number,” called R0. When R0 is lower, herd immunity occurs at a lower level of overall infection (assuming that infection causes immunity for those who survive).
Aaron, the 70% figure for herd immunity from COVID-19 seems pretty accurate. The formula is that herd immunity is reached when the % immune equals 1-1/R0. At R0=3.5, this is about 70%. At R0=2.0 it is 50%, at R0=2.5 it is 60%, and at R0=5.0 it is 80%.
In the case of COVID-19, my suspicion is that R0 varies by location, depending on the degree of contact in a given community. The subways in particular, and crowding in general, probably gave NYC a high R0.
I don’t think that herd immunity is impossible with lockdowns, precisely, but it’s complicated. The lockdowns don’t completely prevent spread, but they slow it. It is theoretically possible for a sufficiently strict lockdown which drives R0 below 1 to eradicate the disease, if maintained for long enough. This does not appear to be feasible for COVID-19 (per Dr. Battacharya).
While the lockdown is in effect, R0 is lowered so herd immunity is reached at a lower % immune. When the lockdown is relaxed, however, R0 will increase and unless the disease has been completely eradicated, the infection will return.
Sorry for the long-winded answer. I think that the above is correct.
Reading this post a few minutes ago about humanity’s tragic encounters with other viruses, I can’t help wondering if our understanding of herd immunity is incomplete. Given what I know about daily life at the turn of the century from 1899 to 1900 and thereabouts, it would seem to me that some of the viral diseases that had such high death tolls in America and western Europe would have quickly spread and thus quickly established herd immunity. But that didn’t happen with some of them.
I think there’s something else going on here that we’ve been missing. I wonder if the virus remains active and if it spreads during times when we are not seeing symptoms in people. So when its optimal weather conditions–temperature and humidity–return, it’s not starting from patient 0. It’s starting from millions of infected people. Hence, the much worse second wave. That phenomenon would partially explain Typhoid Mary’s history. The virus remained active inside her, but she was not “allergic” to it (that’s why some people are asymptomatic in the first place), and so she spread the active virus to many other people.
This is probably how the weather and virus interact.
So in our hoping to establish herd immunity by resuming normal social interactions, I wonder if what we are actually accomplishing is sowing the virus more heavily throughout the population. This is what the people afraid of the “second wave” see happening. It’s basic biology–it’s exactly how weeds take over a garden.
We’ve got to get a vaccine in the market over the few months. Let us hope and pray for the Israelis’ success.
I can’t find quantitative information on this, but based on case reports and direct reports from docs there are some additional factors at play here that are being seen to a much greater extent than with the flu. They are seeing a surprising amount of clotting in COVID cases and this had led to deaths, including among the small group of younger people without underlying conditions. As a result it is becoming more common practice to routinely administer anticoagulents to new patients. Renal failure is also a frequent occurrence and deaths due to cardiac arrythmia have also happened. Doctors are also seeing unexpected neurological impacts though whether this has resulted in deaths is unclear.
Even with pneumonia they are finding multifocal pneumonia with COVID rather than the more usual single lobe with the flu. As to ventilators, their use has declined in recent weeks as doctors seek to delay their use instead of relying on the usual default of O2 levels to start vent treatment. Apparently the mechanisms by which COVID does its damage is different than with the flu and approaches that have worked with flu and ARDS are not as successful, specifically a concern that the air pressure from vents can sometimes cause more damage than help and the difficulty of moving those on vents – with CPAP or nasal cannula they can turn patients over or have them lay on their sides or even sit up which appears to help many.
This makes sense with what I’ve been reading too.
This description of the virus, how it works, that my daughter sent me a month ago, describes a virus that literally drills holes in cell walls so as to replicate itself. It’s a very bad bug. (I know it is the NYT, but try to keep an open mind. :-) )
I’m afraid this virus will have long-life prospects. I hope the neural system complication does not mean that it will behave like the chicken pox virus, the ones that cause shingles later in life. The chicken pox virus finds a home for itself at the base of nerves where apparently the antibodies can’t always eradicate them. So when the conditions are just right later on in a person’s life, the virus comes back to life causing horrifying pain because it has infected the nerves themselves.
But Cuomo has been punishing upstate for years with the pipeline bans and fracking bans.
Herd immunity will most likely turn out to be a combination of natural resistance plus immunity by antibody. Remember the cruise ship and aircraft carrier examples had only 20 to 30% positive antibody tests. I see no probability of 70% of the population having antibodies.
Salmonella typhi is not a virus although I think you are correct that some people have a natural resistance to this virus. It will be interesting to analyze this.
I think this is correct. We keep hearing that in order for us to have herd immunity, 70% of the population needs to be infected, but I don’t think that is quite right. That would be correct if the only thing stopping the virus is previous infection, but there may be other factors. Natural resistance may be one, but I’d imagine that there are quite a few other things, such as weather and social habits (cars vs. subways, etc…). Meaning, we may very well accomplish the effect of herd immunity without having anywhere close to the percentage of infections necessary for true herd immunity. [that’s just my theory… I’m not a doctor]
Marci, I think the “drilling holes in cell walls” is essentially what all viruses do. They have to get into a cell, which they then hijack to reproduce themselves. I guess it’s horrid when you think about it, but I don’t think that this is something unique to the coronavirus.
I know. But it is horrid when I think about it. :-)
Actually, from the NYT article, my sense was that this is what the spikes on the outside of this virus enable it to do with more efficiency than some other viruses.
Ugh.
Here is an interesting picture of a giant virus, the size of a bacterium, with a well seen “pore” that it uses to inject its prey, an E coli bacterium. It forms a vacuole on the membrane and uses it to inject its nucleic acid genes into the bacterium,
https://www.virology.ws/2013/08/01/pandoravirus-bigger-and-unlike-anything-seen-before/
That may be related to why elevated d-dimer turns out to be a predictor of outcome.
Gov. Cuomo announced that the first tranche of study data from 20 hospitals in New York State on the efficacy of hydroxychloroquine was sent to the CDC and FDA today.
Lord, hear our prayer. 🙏☝
And now the eyes of the press turn to my state, Massachusetts. It is the latest hot spot. Not that that report is a surprise to me. This virus thrives in temperatures between 41 and 52 degrees Fahrenheit–exactly where we’ve been for the last nine or ten days.
Please send some temps in the 60s up to New England. :-)