Johns Hopkins COVID-19 Tracker - Significant Update
My alma mater Johns Hopkins created a COVID-19 tracker toward the beginning of the pandemic that shows how many cases there are around the world. It breaks it down into "Confirmed, Deaths, Recovered, Active."
https://www.arcgis.com/apps/opsdashboard/index.html#/bda7594740fd40299423467b48e9ecf6
Up until today, it had only shown numbers by state, which is useful, but not as useful as it could be. I live in Southern Jersey, and while New Jersey has a lot of confirmed cases, most of them are in Northern Jersey.
That seems to have changed. It now breaks the numbers down by locality, which (1) shows just how much this virus has spread throughout the USA and (2) gives you a sense of what localities have been hit hardest and need the most support.
I say "sense," because we have no way to know how many actual cases there are.
Stay safe everybody. Let's hope we can slow this down sooner rather than later...
I watch this pretty much on a daily basis. Unless I totally missed it, they don't total the number of deaths in the US (they just show it by state), so I have to go to a different site for that. They do have total incidence for the US and other nations, which I find useful.
EDIT: another thing I just noticed is that they are downplaying and removed from the front page the number recovered which I think was wise because it was such a misleading (and probably wildly inaccurate) number.
Also, the death rate remains steady at 1.2% - been that way for a few days now.
Click on US on the left (red font) with total cases, it displays current deaths on the right in big white font. Currently 499 for what it's worth*.
*I'm highly skeptical of numbers from here on out because, as some of discussed over the weekend, the two largest population centers in the US have virtually stopped most testing at this point.
ah thanks. I see whats going on. It doesn't work on the mobile version for me, but does on the main website.
That 499 is close to the 504 number I get from another site. And I agree that the incidence rate is way low due to what you're saying.
Are you talking about this?
It seems like they're not testing because they simply don't have enough supplies to test.
Yes, from that article:
Naturally, limiting testing in America's two largest population centers would also likely lead to widespread under-counting of total cases.
Or, depending on your choice of news providers, this:
https://www.foxnews.com/us/la-county-containing-coronavirus-skip-testing-patients
A surge in coronavirus cases has Los Angeles County health officials telling doctors to give up on testing patients in the hope of containing the outbreak and instructing them to test patients only if a positive result could change how they would be treated, according to a new report.
I'm talking daily with an emergency room physician at Cedars-Sinai in LA who is telling me they have the tests, they just haven't been allowed to order them since Thursday night unless it's a nursing home patient potentially going back to their nursing facility.
The mortality rate is a misleading statistic since, at least in the US, we have never had enough test kits to determine those who had COVID-19 who are asymptomatic or mildly ill. In some places, the mortality rate has been skewed by inadequate resources. Also, with contagion levels where they are now, testing people without symptoms doesn't do much as community spread has taken off in a big way.
What is critical is the number of people requiring hospitalization, often for an extended period. If that number exceeds our ability to treat the seriously ill, due to shortages of HC personnel and/or facilities, we must triage people and people will die.
The mortality rate is somewhat misleading because its not a very accurate number in a vacuum, but the number does provide some relevant and accurate insight into just how deadly this virus thing is. Here's what we know about the mortality rate:
1. The number of deaths if fairly accurate
2. The number of incidence is way low due to pretty much what you said in your first paragraph
This means that the 1.2% mortality rate as of today is likely the worst case scenario, and in reality the mortality rate is much lower. To me that is important. If the actual mortality rate is, lets say 0.1% for example, then we might be able to ease some panic/hysteria and maybe, just maybe, we can take steps towards getting back to normal.
While I think 1.2% is probably high because we haven't come close to capturing what the incident rate is, there are two opposing forces on incident rate:
1) Inability to confirm all cases (lots of mild cases not being confirmed) leads to undercounting.
2) Our open cases are not yet concluded. So a bunch of them will end up in deaths and that means we're understating the death rate by using a case count that includes open cases rather than a closed case count leads to overcounting (i.e. until a case has recovered, we don't yet know that it hasn't resulted in death).
The death rate is also impacted by interventions available. Places with enough ventilators to treat everyone will do better than places that don't. Some places also appear to be using some drug combos that might be effective while others aren't.
With regards to #2, I see what you are saying but I don't think "a bunch" of the open cases end up in deaths. Most experts are saying the majority, the vast majority, will end in recoveries.
Another thing the number of deaths can tell you is how quickly the number if deaths are doubling. Here is an analysis that is being updated regularly.
Since New York has the highest number of deaths in the US and the number is doubling every two days, it is not hard to see why they are concerned.
where many are taking this process and data the wrong way...comparing NYC and New York state to say Michigan is apples to cherries...NYC alone has nearly the population of Michigan as a whole state.
Michigan-9.996 million (2018)
NYC 8.623 million (2017)
as one of the precautions is 6 feet of distance, don't tell me population density has zero affect on contagious spread and death toll...not to mention Michigan's highest cases are...you guessed it, in our highest population density centers, Oakland Wayne, and MacComb counties....yeah it's serious, but using any 'model' with each state following their own general guidelines and population situations will play a role.
Also, the death rate remains steady at 1.2%
Not to be a broken record here . . . but the case fatality rate should be based on cases with an outcome. You're making the assumption that every active case (and nearly all U.S. cases are active) will end in recovery. That's not realistic. We have something like 2000 patients nationwide in critical condition.
We just have to wait and see what happens to all of these cases. When we get to a point where we have like 20,000 or 30,000 closed cases, then we can look at the U.S. CFR. We don't have enough data right now.
I am making the assumption that MOST cases will end in recovery. That is alignment with all experts on this. The issue with trying to discern something meaningful from recoveries number is that there doesn't seem to be a unifying definition what that means.
Also, the fatality rate by itself may not say much, but the trend over the course of several weeks to months can be very meaningful. Because you have to assume that if you don't die from it, you'll recover (duh!). So unless we start seeing a meaningful uptick in the mortality rate and it remains a trend, we can safely make the assumption that we are in the right neighborhood with regards to the mortality rate without having to fully understand recoveries.
The case to fatality rate is 14% at the global level. That number is going to be extremely high as a lot of countries aren't able to test every person and are generally only testing the most severe. The US for example, hasn't reported any recoveries, so there is 100% case to fatality being reported in the US right now. However, whatever that rate ends up being, would generally apply to those without a determined outcome yet, so, even if we don't get another case, your estimate of 1.2% would actually increase as some of the cases that you're assuming will recover will end up dying.
The thing is that we just don't have enough data right now to make valid assumptions. There is a selection bias because it helps make decisions on a micro level, but it gives us bad data on the macro level. Once testing ramps up, numbers will stabilize and we'll be able to have this discussion, but right now, there are too many gaps in the data for anyone to be able to make a sensible argument about the disease.
Also, I don’t understand why CFD is the only stat people are using to justify overcorrection (or “hysteria” for some pundits).
Anecdotally, there are a lot of very, very sick people that aren’t dying from this, and even if that is a non-trivial amount, you’re taking about some massive strain on the healthcare system.
March 24th, 2020 at 11:24 AM ^
" I watch this pretty much on a daily basis."
you during 9/11:
Information is great. The problem remains with the data set-very small, very narrow. Unless and until larger more randomized test results are in we are pretty much flying by the seat of our pants. Here, working from a lowest common denominator position in a matter such of this may be used as a definition of "hell". I've got many family members working on the front line...they are damn good at what they do...wish they were working with better than "best guess" information. "We can do this hard thing".
Been watching the Johns Hopkins chart daily and started putting the numbers in an excel chart. The hope from the 20-30 doctors I have spoke with is that the US rate comes down to a final range between 0.5 - 0.8% as more people are tested and the curve is flattened. The US population is 330M+ people to Italy 60M+.
Fantastic! I've been looking for this data (particularly the US) and I couldn't seem find it, so I started doing this myself late last week. But you've been doing this longer. The consistent downward trend for the death rate is not only positive, but also not surprising.
Johns Hopkins shares the raw data. It may be more complicated than you care to deal with but it's out there. Look in the text under the charts, you'll see a few links for the raw data sets.
Thanks. I typically view it on my mobile device which seems to have less capabilities than the actual website.
Also in regards to the falling death rate and without trying to sounds like fear mongering, the death rate will drop drastically as cases are increasing so quickly. We're discovering new people early in the progression of the disease. Theoretically, if there were 0 new cases from here on out, the death rate would begin to climb as people in the population start to die.
Looks like COVID-19 approaching a death rate near to that of the normal flu, which is great news.
It’s far too early to say that. Only about 800 cases nationwide are considered closed.
That death rate is deceptive, though, considering that over 95% of U.S. cases are ongoing. Fortunately we’re not at the point where patients are dying right away due to a lack of ventilators. But some go on die after several days in the hospital. In South Korea their mortality rate has inched up as some of their critical cases finally died.
This effort in San Miguel County Colorado seems like it could be quite informative.
“San Miguel County in Colorado announced this week it plans to test everyone in the county for COVID-19. And they’ll be using a blood test rather than the usual nose-and-throat swabs.”
“County officials said the company is providing it “free of charge,” that the test is “not mandated.” The county is already testing first responders, healthcare providers and law enforcement before beginning “widespread public testing” early next week.“
Will be interesting to see how complete results they will get and what they will show.
That's really interesting. Eventually the constitutional lawyers will have something to say about measures like this.
I am no conspiracy theorist, but it is amazing to me how willing everybody is to give up their rights over something they really can't prove to themselves exists. Apparently a coordinated effort between China, the EU and US leaders could usher in the NWO. Maybe the tinfoil hat brigade will be right eventually!
/s obviously. Trying to lighten up a day of what is going to be a stressful week.
I read an article last week where a town in Italy ordered everyone to stay in and get tested. If you tested positive you were isolated further. I'm not sure how drastic it all was, but eventually everyone tested negative.
https://www.theguardian.com/world/2020/mar/18/scientists-say-mass-tests-in-italian-town-have-halted-covid-19
Read it, I'm sure I mis-remembered the details.
I was following the data for quite a while, but I found that once they combined the Chinese numbers with the "Other Areas" data, it stopped being quite as useful, because it became almost impossible to see how fast the number of cases was doubling outside of China. They also took off the logarithmic scale, which just pissed me off.
Regarding the local data, however, it is similar to what Bing.com/covid is now doing. It's helpful, but like all of this data, is just showing confirmed cases. I suspect the actual number of infections is actually much wider.
At the beginning, almost everyone - including epidemiologists - were assuming that the number of people who had been infected but not tested was the harbinger of millions of deaths in the U.S.. I sure thought so, but now I am less so. Many studies have indicated that this virus was spreading in the country for several weeks, likely as much as two months, before some of the first reported cases were confirmed.
That has some implications, almost all of them are positive. First, there are signals in the Healthweather.us data that indicate that a very large number of people (hundreds of thousands) started showing signs of flu-like illness in early March based on their core body temperatures. We know, from news reports and confirmed cases, that most of those patients were not showing up at hospitals. If they had been, we'd have had an even bigger crisis on our hands by now given the rate at which the virus was expected to double (3-4 days). We don't; which means that the majority of people who are infected, even if they have a fever, don't become sick enough to require treatment at a hospital.
Second, Dr. Fauci has been talking for weeks about asymptomatic or mildly symptomatic carriers of the virus who don't know they have it. It's why social distancing and shutdowns are essential right now, but also underscores the point that confirmed cases are only one part of the total number of cases. I've read some studies that between 40-80% might have gone unconfirmed and/or undiagnosed (it depends on the study and assumptions). So, while Osterholm and others are saying we might see 1.7M-1.8M deaths in the U.S., that can only be true if you aren't taking those cases into account.
Suffice to say that I have found the combination of the JHU and Healthweater data to be particularly insightful when used together. I'm not 100% positive we've turned the corner yet; but I don't think the worst case scenarios are most likely either.
What does KENPOM think?
Better yet, when does Brian do the write-up titled: Section where I predict the same thing Kenpom does.
U.S. favored by 6 over virus
This is exactly right. We need to test antibodies of the general population to see how many people have already had it, or simply to find out how immune the population is at this moment.
How many people have had or currently have it ... Interesting question. I've been reading the healthweather data every night as soon as its posted (partly to reassure myself; partly because I haven't had too many good nights' sleep).
Here's some back of the envelope math I did for Oakland County: On March 8, there was a 0.85% spike in the population with a flu-like illness (as measured by the Kinsa thermometers). That's 9,760 people. While it's true that not all were likely COVID-19, the rate of illness across the county had been following a pretty typical, post flu-season decline. And then, all hell broke loose.
If Oakland was typical of the surrounding counties (the Kinsa data suggests it is), then 30,000-50,000 people in the Metro Detroit area got sick almost literally overnight on Mar 7-8. That fact alone is staggering to contemplate.
But they didn't all go to the hospital (obviously, or we'd be worse off than Italy right now). It's also reasonable to assume that even if seasonal flu is a significant part of that data, then so is COVID-19.
So, getting back to your question ... There are two ways to answer it. First, The University of Washington Virology Lab (https://depts.washington.edu/labmed/covid19/) started processing COVID-19 tests on or about March 1, and in the last three weeks has processed 24,359 tests. The daily average of positive tests is 7.5% (there have been 1,832 cases confirmed in the lab). Using 7.5% as a proxy for the rate at which a patient with flu-like virus presents in a healthcare setting and is positive for COVID-19, and applying it to Oakland County, here's one estimate of the actual number of cases: (1,270,000) x (0.0595) x (0.075) = 5,667 infected people on March 18 (the date of the peak rate of flu-like infection in the county).
- 1,270,000 is Oakland County's approximate population
- 0.0595 is the fraction of the population estimated by Kinsa to have flu-like illness on March 18
- 0.075 is the ratio of positive tests from U-W to overall tests performed on probable patients
My gut tells me that may be low, but it's a number.
Second, the NY Times published an article that about 7M people traveled from Wuhan in the January timeframe, some of whom probably arrived in or traveld through Michigan. Using that as a starting point, assume Patient 0 was in Oakland County on or about Feb 1. Doubling the number of cases every 3 days yields just under 10,000 cases on March 8, the date that the 0.85% blip appears.
So, somewhere between 5,000 and 10,000 people had COVID-19 in Oakland County by mid-March. Again, it's just a number, but if many of those people have no to mild symptoms, it may help explain why we are in better shape than some other areas around the country.
Good post - good analysis.
I think your 0.85% number for "Oakland County residents infected on 8-March" is probably not far off. Maybe a bit high, but it aligns directionally with other some other back-of-the-envelope calcs I've done and seen.
One thing that I am fairly certain of: this virus was circulating around America by late January. At the latest.
Thanks, Nittany. Your comments are very much appreciated. They really are.
I've been posting this kind of analysis every morning on my FB page for friends and family; mostly as therapy for myself, but also to help them be a little more grounded. I just pray to God that it's helping at least some of them.
As far as your remark that this virus was in the U.S. by late January: there's almost no question about it. Given the numbers and timing of cases around Puget Sound, plus the incubation period for Patient 1, someone with the worst carry-on luggage ever arrived in Seattle no later than mid-January.
Fantastic analysis! +1 Informative!
That daily average of positive tests of 7.5% is a lot lower than I expected based on the "r-naught". If that number is even remotely accurate, then that would be great news in my opinion.
Thanks, WCB. I very much appreciate the kind words.
I was also very surprised by the U-Dub data. I had expected it to be much higher. What it implies, though, is that there are at least a few bugs running their course that present like COVID-19. I've had that on my mind for a while. Many people I know - and based on reading some of the comments in these pages - indicates that this is a very bad year for viral illnesses of all types. We just happened to draw the short straw this year with a novel coronavirus on top of everything else.
March 23rd, 2020 at 10:42 PM ^
Speaking of a few bugs, Washington had a Flu B spike trending down in January 2020 just as Flu A was ramping up for February. It's a cocktail.
Unfortunately I think we can pretty confidently say we haven’t turned the corner yet.
“Surgeon General Jerome Adams warned Monday that the coronavirus outbreak will worsen this week and said people across the country are not taking the threat seriously enough.”
as I posted above, number of deaths doubling every two days in New York, I wish there was evidence that we we were turning the corner.
The stories emerging from hospitals this week are going to be heartbreaking; I have no doubt of that. I also don't want to be misunderstood - "turning the corner" means that the rates of new cases, hospitalization, and death start to decline. That's all. It means that, at least for the time being, we have gone over the top of the wave and started back down. The number of people hospitalized, to Mad Hatter's point, will remain very, very high for quite a while; likely several weeks.
Yes, but the data does not suggest that the number of deaths are starting to decline.
Again, it's a lagging indicator. It will be several days to a few weeks before cases catch up with the Kinsa data. But, it must; otherwise, social distancing and stay at home orders are a waste of effort.
No matter how much data we collect we can't possibly know how many people actually contract the virus until we've had a chance to conduct controlled studies. This is just a scientific fact.
I thought about you today when I took my cat to the vet. They had a system where I had to call them from the car and then they came out (in protective gear) and got my cat. Didn't let me inside at all. When they were done they called me and returned him the same way. Did the payment over the phone by CC.
Maybe something your wife could do to keep open? Just a thought.
I appreciate that. We already are doing this, and praying it works. In fact the AVMA is basically requiring it of everyone now. If one of our staff come down with it or if people are reluctant to agree with necessary medical treatments, it could really hurt us.
EYB - Of course you are correct but that is very unlikely to happen soon. If you look at numbers in different states you’ll see a widely different % positive. For example, NY is running at near .075% while WA is at 4.75%. So a controlled test would have to be extremely wide spread.
https://covidtracking.com/data/
Overall the numbers are starting to look positive. The only number that will be reliable is the #deaths. For example, on 3/20 Michigan stopped reporting negative tests. WTF Whitmer? You go on national TV and complain that we need more DATA and then drop the ball in Michigan?
If anything, the death rate will be inflated as that is the number that WILL be reported.
PS - Shame on you Whitmer!
In comparing mortality, note that Washington’s outbreak started before New York’s, so their own curves aren’t parallel. In the long run I would predict that their rates will converge.
This University of Minnesota site has links to this tracker and a number of other trackers.
Here are some sobering numbers from the WHO:
It took 67 days to get to 100,000 cases.
It took 11 days to get to 200,000 cases.
It took 4 days to get to 300,000 cases.
Dr. Amesh Adalja, an infectious disease scholar from John Hopkins, estimates a 30% U.S. infection rate with a mortality of 0.6%. That's roughly 600,000 dead.