How are the States handling this Virus so Far?

Oregon might have just have been fortunate to be where it is located. With CA and WA essentially imposing lockdowns and not having much in the way of population to your east and the ocean to your west you were more or less isolated to begin with. Then you implemented your own Stay at Home measures and that must have been enough to keep a lid on the virus getting out of hand up until now.

 

I dunno what 'actions' are being taken in WA, because where I live (bfe) nothing is being enforced. We have 15 infected (up 5 from a week ago) in my county and our first death just today. Seattle is still teaming, of course, and judging from the (lack of) traffic, is roughly 90% shut down. There isn't any less activity in our tiny town or the moderately sized town nearby, although about 1/3 of everyone is wearing some sort of mask (most of them look homemade, there has been zero on the shelves for weeks), which is a major improvement from last week.

 

The chart I looked at showed no deaths from Multnomah County. It may have been out of date. The virus is mostly in Salem but most of the infected came from nearby rural counties. In any case, the hospitals are far from at full capacity.

 

Necessity is the mother of invention. THANK YOU for posting this.

 

Fair enough. My nearest town has a population of about 90 and someone could die there and I wouldn’t hear about it for two weeks.

I don’t think anyone’s hospitals are full, are they? Even NY is seeing a higher discharge than admittance rate last I heard.

Glad Oregon is faring well. My wife has family there, some in the high risk demographic.

 

Bingo. The only way to come up with that data would be to crowd-source it. For this reason I've contacted the elections data team at the DailyKos. Yes, KOS is a very left-leaning website, but their election gurus are among the best in the world at getting data and getting it correct. And they were the very first to very successfully and correctly crowd-source the data for presidential elections by CD, since many states do not provide that data. This means that the data team already knows exactly what part of of what county belongs to a CD or not.

I'll let you know if they respond to me.

 

Many rural districts have no well equipped hospitals and the gravely ill would have to be taken to a nearby large city to get treatment. Which would skew the numbers. A more accurate comparison might be to use a ratio. Infections and deaths per capital. For instance Oregon and Idaho have similar numbers of cases, yet Oregon has three times as many people. Thus Idaho's infection rate is around three times that of Oregon.

Washington saw the first case yet has mostly contained it, as has Oregon. Mostly because they closed everything down and began enforcing social distancing early on.

 

Personally I think what's most important is not how many cases there are per capita- that's frequently a function of how many tests are given out in a given state. I think a more accurate figure would be how many -deaths- there are per capita. worldometers.info measures this as deaths per million. I would like to point out that deaths per capita isn't necessarily a reflection of the policies that states are using. There are frequently underlying factors such as air pollution that can make death tolls worse. Here are the states with the highest deaths per million, ranked by the highest:

 

I really like this suggestion. I am going to try and work on making that based on yesterday's cumulative total data.

So, I am thinking (Deaths Per Capita/Infections Per Capita) x 100 to get a percentage. This essentially tells us the percentage of confirmed infections that turn into a death and is normalized based on the population.

 

So here is what I came up with using your suggestion. I took all of the Per Capita Deaths for Each State and Divided it by the Per Capita Infections for Each State. I then made that number into a percentage. And I then ranked the States from largest to smallest. What I think this tells us is the percentage of infected persons who died in that particular state, normalized by population. I would ask folks with more statistical analysis background, like @Statistikhengst , @Derideo_Te , or @nobodyspecific to look this method and analysis over and see if they agree that this is a somewhat useful method for analyzing how a given region is handling the virus.

And I guess we define "handling" the virus as keeping infected from dying.

The first number for each region is the Cumulative Cases Per Millions and the second number is the Cumulative Deaths Per Millions, as of 04/09. I grouped them in bunches of 10 just so they could all fit on the same screen, but you should view it left to the right, top to bottom, to get from 1-56.

 

I'm leery of this methodology, primarily because it would inflate the "success" of treating people infected in states where more tests are done per capita. This is why I'm leery of the infections anywhere, not just in the U.S. I know that Iceland is one of the few countries that focused a fair amount of tests on people who had no symptoms at all- they found that 50% of people who were infected had 0 symptoms. I imagine those people were rather easy to "treat" .

 

I honestly do not have a statistical background. It is mostly a hobby I have picked up recently in following the coronavirus.

There is almost certainly a testing bias per state, which will throw off any calculation made when incorporating number of confirmed cases. But what you could perhaps do is if you want to utilize the confirmed cases, you could try to normalize the total positive cases by applying the total US positive rate across all states. This would be making the (incorrect) assumption that all states have approximately the same distribution of coronavirus cases. While almost certainly not true, it will probably be a bit closer to the truth than the mass disparity we currently have where a place like Connecticut is testing 29% positive, and a place like Idaho is around 10%.

This calculation is not too difficult. We have the total % positive off of covidtracking, so we just need to divide the nation's % positive rate by the state's % positive rate, and multiply the total cases by the result. For example,

Current % US Positive: 19.4%
Current % CN Positive: 29%
Current % ID Positive: 10.1%

Connecticut:
19.4 / 29 = .66, .66 * 10,538 = 7039

Idaho:
19.4 / 10.4 = 1.91, 1.91 * 1,396 = 2670

The result is to bring Connecticut's cases down and Idaho's up so that the total for each is in line with the current US average. This assumes Connecticut is being more targeted with their testing whereas Idaho is less so. You can of course use some other % you think is more appropriate instead of the current US average to achieve a similar affect. As to whether or not this is a useful calculation to incorporate or comparing cases / deaths per million is useful, I am not entirely sure.

I think deaths per million makes some sense as a comparison, although there is an inherent risk of small numbers artificially inflating things in smaller areas. Similarly, when the base numbers themselves are still very small (say less than 100), they can be highly influenced by breakouts in the wrong place, like nursing homes. And that can skew things significantly.

 

One additional thing I should mention here is that the results are not actually what my goal was to being with. If the assumption is that Idaho is overstating cases (due to lower % positive) and Connecticut is understating (due to high % positive), then you would want the result to be lower Idaho cases and higher Connecticut cases. That can be done by switching nation% / state% in the calculation to state% / nation%.

 

So I wanted to follow up on this thread with the newest information (4/12) and I added some more data. This is the latest that I have for this effort:


It's a lot of data to fit in one screenshot, but hopefully it is fairly self-explanatory. First off, this is ranked according to the metric that we previously discussed (again, thanks to @phoenyx for the idea). Again, that metric is to divide the Deaths Per Capita by the Cases Per Capita and turn that number into a Percentage of Confirmed Covid Cases which has resulted in Covid Deaths. And that data is normalized by the population.

So, for example, Arizona has seen 7.68% of their confirmed cases turn into deaths while Iowa has only seen 2.58% of their cases turn into deaths. And they have very similar case infection rates (493.5 vs 502.9). This does not automatically mean that Arizona has done a worse job because there are other demographic factors or timing factors that could explain the difference, but perhaps it gives us a better idea. They are also testing at similar clips as well (587 vs 557 per 100K of population), so it is unlikely that Arizona is just doing a better job of identifying their infected.

I have also added information, as per the suggestion from @nobodyspecific , regarding the testing conducted by each State. The 4th row under each State is the Positive % for each State and the 5th row under each State is the Tests Per 100K (note: The Cases and Deaths are still per million).

On this front, we can see a few points. As an initial matter, every State has tested at least 443 per 100K (Alabama at the low end) and up to 5.35K per 100K (New York). We can also see a wide disparity in terms of the positive testing percentage from 2.64% (Hawaii) all the way up to 48.8% for New Jersey. For a comparison, the US average for the last 7 days is 21.0%.

@Statistikhengst , @Derideo_Te , @LoneStarGal , @HereWeGoAgain

 

Very nice, thanks for updating this. Seeing states with high deaths per capita and also testing at % positive below 10% makes me wonder if they are simply testing the wrong part of the population. You'd think with that many deaths vs that many cases, they would have to just be missing who to test.

States with very high % positive and very high numbers of deaths would likewise lead me to believe that had they the capacity to test more, you would see a lot more cases but also a lot lower % positive.

One thing I do notice is I think you may have used New York City's population to calculate deaths per capita there. 1,000 seems like a magnitude 2x too high. The state population is ~20M, so 10K deaths should come out to ~500 deaths per capita.

Other than that, it is very surprising seeing the difference in deaths per capita in say a Nebraska vs an Arizona. Both appear to be testing about the same rate. Possibly it is the sheer low overall numbers from the smaller states that is skewing things. At only 12 deaths and some 300 cases, Nebraska could be skewed heavily based on where specifically the virus happens to hit.

 

Thank you for catching that one. I definitely google searched "Population of New York" and the response does not bother to even say New York City.

 

I was recently in South Africa and got a chance to check out some of the slums around Cape Town.

They don't give a **** about a flu with a 2% fatality rate. Every time they walk to the shops that's a 2% fatality rate.

Worries me because if you live in the slums the actual rational choice is to more or less go about life as normal.

 

Good work!

NJ was in the same situation as NYC which meant that they were focusing all of their resources on high density areas in the northern part and only testing those who were already presenting symptoms and those in contact with them. Then it appeared as though Philadelphia was turning into another hotspot which means that the southern part of NJ was being infected from there. NJ is a bedroom state for both NYC and Philly so they are getting it from both sides, unfortunately. On the plus side it looks as though Philly has it mostly under control so it won't be as bad as it was in the north.

 

ironically, the states that had super few case may be the worst off in the future outbreaks, the goal was to reduce the curve to a level that the medical system could handle, not to reduce it to basically nothing - as when we re-open, if the virus goes there, no one has gained immunity

now, the other possibility is that this came here in November or December and those states have already been infected and already had immunity - only addl testing could show this

then again, if we do get a vaccine, then the states with low cases will be fine

 

One week later and I would like to update this project and again, I would like to invite @nobodyspecific , @Statistikhengst , @LoneStarGal , @Derideo_Te , @HereWeGoAgain , @phoenyx , @FreshAir, @ImNotOliver to review and give their thoughts.

A couple of notes - the source for this data is covidtracking.com. The next point is that I have included a set of data for the five data points Cases Per Capita, Infections Per Capita, The Ratio Per Capita (this is what I used to sort the data for the screenshot you see below - again, the idea being that this ranks the States based on the percentage of confirmed cases that turn into confirmed deaths), the positive testing percentage, and the tests per 100K. Below that, I repeat the same five data points and give you the difference for that particular State compared to the data from one week ago.

Side note: On the Testing Positive % Difference and the Ratio Difference, the percentage that you see is the raw difference. So a -0.8% drop is essentially saying that it went from 1.8% to 1.0%.










There are a lot of ways to sort this data. For example, I can sort the data based on which state increased their testing per 100K the most (Rhode Island at +1357.521) to the least (Idaho at +113.34 and Puerto Rico at +82.31). Rhode Island is now testing the most on a per 100K basis (just edging New York at 3287.8 compared to NY's 3175.1). On the opposite end (not counting the territories) are Texas and Kansas at 636.6 and 622.6, respectively.

In terms of the Ratio from Confirmed Cases to Confirmed Deaths, the State which saw that number improve the best was North Dakota (dropping by -0.888%) and the States which did the worst were Michigan and Connecticut (increasing by 1.57% and 1.59%, respectively). Puerto Rico had a really bad week though and saw that Ratio jump by 4.06%. Interestingly, 46 of the 56 States and Territories saw this Ratio increase.

I also sorted by the Positive Testing % Difference to see how that number changed over the course of a week. For example, New Jersey - last week - had a 48.8% of all their tests conducted come back as positive. This week, it is 49.97%, which means it went higher. This means that the number of Covid-19 cases in New Jersey is easily 2.5x higher than the current numbers.

The biggest movers on that category are South Dakota, which saw its positive testing percentage increase by 5.02%, and Alabama, which saw its positive testing percentage drop by 6.08%.

One final note that I will make is that I feel really sad for Puerto Rico. It is getting wrecked by this virus, given that it's cases per capita increased by ~98 and it's deaths per capita increased by ~18.7, which represents a massive increase in its ratio compared to last week by more than 4% points. But it's American Citizens also getting very few new tests relative to every other States given than it has only increased its tests per 100K by 82.3, which is significantly lower than the lowest State.

 

Great work, thank you.

This metric is going to become a KPI in the near future since it is what is being used as the basis to reopen the economies of the states.

With that said I was just reading about what has happened in Singapore because they neglected to test their migrant worker population.

https://edition.cnn.com/2020/04/18/asia/singapore-coronavirus-response-intl-hnk/index.html

From being a shining example of what to do right Singapore is now scrambling to prevent this from becoming a serious problem. They will probably be able to contain it but it is going to be a heavy lift IMO.

The lesson is that FAILING to test EVERYONE means that another outbreak WILL occur during this Pandemic.

Testing needs to be universal and thorough and failure is not an option if the states want to reopen and NOT be FORCED to close again.

 

Are they manufacturing safety and health equipment (masks...) in State owned enterprises, or planning to do so. If not, why not?

Just staying home is NOT an option for too long. Nobody can say we will have a vaccine ready to kill it for good, or can say next fall we get something else and have to start all over.

Whether or not Humidity...geography..., has an effect, just crunching numbers may not tell us as much as people might think. Areas where the culture was to party, invite everyone over for dinner, and kiss the cheek...it's not just the response it's the culture that spread it first. Social distancing can change the culture, but not humidity...

 

If say for instance you correlated that with Red or Blue states, it won't say everything because the weather is different between Yankees and Southerners, I'm from Georgia and almost got knocked on the ground when I stepped off the plane in Samoa. But Samoa is like really not very populated. The problem still is major cities have more major problems, lots of people on subways and busses early on, it totally screws the data compared to more rural states that are hit last.

To truly test the data people would have to be working, with the virus, knowing it existed; the initial spread contaminates the data so as to be unusable as to response.

The response can only be measured by living around the lions and tigers and bears, not by staying home and never seeing one.

 

The weekly update to this project. First, the data - sorted according to the Case Infection (Per Capita) to Confirmed Death (Per Capita) Ratio:

The difference statistics are in comparison to the previous week. I am going to make another posts with some interesting findings.

 

So, here is a comment about the biggest movers and Shakers. For this, I am only going to discuss the States, District of Columbia and Puerto Rico.

In terms of the Cases PC, the five biggest movers were New Jersey, Rhode Island, Massachusetts, New York, and Connecticut. The range being +2664.4 to +2160.4. The five smallest movers were Vermont, Puerto Rico, Alaska, Montana, and Hawaii. The range being +62.3 to +20.4.

In terms of the Deaths PC, the five biggest movers were Connecticut, New Jersey, Massachusetts, New York, and DC. The range being +234.5 to +129.54. The five smallest movers were Oregon, Montana, Hawaii, Arkansas, and Alaska. The range being +4.05 to +0.00.

As for the Ratio, the five biggest movers were Minnesota, Connecticut, Puerto Rico, Wyoming, and DC. The range being +1.864% to +1.197%. The five smallest (or biggest improvements) were Nebraska, Wisconsin, Iowa, Arkansas, and Kansas. The range being from -0.236% to -1.258%.

As for Testing (Positive %) Difference, the five biggest movers were Nebraska, Delaware, Iowa, Kansas, and Louisiana. The range being +4.71% to +1.81%. The five smallest (or biggest improvements) were Alabama, California, New York, Michigan, and Georgia. The range being -1.86% to -3.86%. The fact that Georgia saw the biggest Positive % drop is a sign that Georgia is doing much better at finding its positive clients.

As for Tests Per 100K, the five biggest movers were Rhode Island, West Virginia, Massachusetts, Utah, and New York. The range was +1769.5 to +965.5. The five smallest were Idaho, Arizona, Wyoming, Puerto Rico, and Louisiana. The smallest increase is +172.5 down to +32.8. The fact that Louisiana has seen such a dramatic decline in testing is really hard to explain.