Big C said:
LunchTime said:
OaktownBear said:
LunchTime said:
OaktownBear said:
CalFan777 said:
The NE is doing worse than the South right now.
The last two words being key.
The rate of increase of cases in red states is substantially higher right now than blue states while the rate of testing is higher in blue states. Blue states have more cases because they got hit first, but red states are catching up.
I dont actually see the data to back that up. Maybe because I am ignoring testing and confirmed cases?
Either way, TBH, I would separate the country into Far Western states, and the West+East. Only Washington and California have had a proactive "good" response. Florida seems lucky so far (probably because tourists dont mingle with the residents as much as we think so the spread is slowed by people leaving for home).
Its interesting though. I think the West is doing well largely because the economy of Washington and California is so geared towards remote work anyway (compared to the rest of the country), so the decision to shut down has a smaller economic footprint. My work had us WFH two weeks before the shelter in place, as an example. GDC was canceled around that time, with most companies pulling out in late February.
https://fivethirtyeight.com/features/the-coronavirus-isnt-just-a-blue-state-problem/
That might be the dumbest Nate Silver article I have ever seen.
The methodology is poor, from using confirmed cases, to comparing established rates to ramping from no testing to testing is terrible.
Want a simple baseline? Nate silver shows Texas at the top, going from 352 to 1396 (+300%). Awesome. He also uses New York the same week going from 20.8k to 30.3k (78%)
But if you index to similar times, NY went from 327 on March 12 to 5699 a week later (+1700%).
Sorry, but Texas starting larger scale testing CLEARLY doesnt indicate it is growing faster than NY. It shows there is absolutely no value in number of confirmed cases. We know New York doesnt have a 17x increase per week. That's silly. But Nate wants to show Texas data as evidence that the rate is growing faster there?
So, let's make the argument that Texas does have a faster increase in infection. For arguments sake. That VERY clearly is not reflected in mortality. Texas mortality is increasing at a massively lower rate than NY. So either Texas has a much better healthcare system, a massive delta % of healthy people, climate is a critical factor, etc, but also that Texas' mortality is declining against growth of infected meaning they also have better treatments that they aren't sharing?
Personally I dont think there is a significant delta in healthcare, healthy people, or climate (given Italian rates). I also dont buy that infection rate in Texas is higher but Mortality is lower and declining. It just doesnt pass the sanity check.
Look, when we can do massively wide scale real testing we can see what the rate of increase is through that. Until hospitals reach capacity causing the mortality rate to increase, the rate of spread, 17 days ago, can only be realistically calculated through deaths.
I usually trust 528s analysis, because they are usually pretty good, but this is poor data, terrible analysis and very transparently so.
Isn't much of Texas' lower mortality rate (compared to NY) simply explained by the fact that more of their cases are recent and they haven't had time to die yet?
I know you rail against the usefulness of counting confirmed cases -- and certainly the reliability of that statistic depends on the testing that is occurring -- but when an area is still early on the curve, what else is there? First you see the number of cases rise, trailed by the number of deaths (also a potentially flawed number, to a lesser extent), followed later by the number of recovered cases (possibly flawed for a number of reasons).
The fact that these three numbers correlate rather consistently as different areas move across the curve tends to indicate that the number of confirmed cases says something, even though it needs to be taken with a grain of salt.
If I am missing something here, by all means explain it.
Quote:
Isn't much of Texas' lower mortality rate (compared to NY) simply explained by the fact that more of their cases are recent and they haven't had time to die yet?
Lets start with the baseline of math and science work. If we can agree on that we have a starting point. I am not interested in explaining to Trump how a cruise ship wont "double" our number of infected by doubling our confirmed cases.
That said, I am seriously not sure how to answer this, bit I will try:
Just because Javid Best started his 40 yard dash after
Tom Brady doesnt mean Tom was faster. There is a method for them to compare them by measuring from where they start, not where they are.
I have not said anything about number of people dead today. The numbers today are insignificant to what we will see in a week and two weeks.
FWIW, As of yesterday, 26 in Texas and 605 in NY. That is COMPLETELY irrelevant. What is relevant is how quickly they are growing, doubling. In a week, 605 total in NY will be the good times. When 600 are dying a day, it will be apparent why I am concerned with growth, not simple numbers.
For this, and
I posed the viz above, there is two issues:
1.
Exponential growth. Once you get out in front, it is VERY hard to bring it back in, the numbers accelerate very fast.
2. Starting location.
The first I fix by what I would call indexing. As is typical for a lot of groups watching in this pandemic, I started at 10 deaths.
Why deaths? GREAT QUESTION.
Testing has proven (again,
New York increased 1700% in one week, a completely irrelevant number) that this measure is subject to too many flaws. Flaws in who gets tested. Flaws in how many are tested. Flaws in how many tests are available. Flaws in how fast those tests are being completed ... they all create what I call bad data. what is bad data? Data that is not good, the early it starts the worse it is. It may be good in NY now, but when NY had 350 cases confirmed it was ABSOLUTELY not. And that is the point: Testing is very inconsistent. In South Korea its great. In NY two weeks ago, it sucked. In NY now its good. In Texas now its bad.
OTOH, Deaths, while not perfect (eg they are not counting that kid in LA who died as a corona victim, because he died of a co-morbidity) it is what it is. Testing doesnt change it. People who die get counted as having died.
And what do we know about disease?
1. It has a number of people affected (or infected) - we can measure this with testing if we have enough tests or deploy the tests methodically to collect data, not provide care. We could also test as many people as possible and after a significant threshold we can have a good answer. We are not doing either. Only South Korea has done that.
2. Transmission. We can get that answer the same way as the first.
3. Hospitalization. This is the first good data that we would get in a region that has a strong healthcare system and record keeping (and yes, the US does for this measure). People who need to be hospitalized WILL get hospitalized at a predictable rate. MAYBE it discounts 5 or 10 or 20%, but it is a consistent enough discount that you can measure growth from this number. Unfortunately, we dont know what percentage of people move from home care to hospital care, because the number of people who have it is unknown. The problem here is that hospitals dont readily share this data, and regionally it is difficult to find consistently.
4.
Acute care. This is where it starts getting to very good data. It is very clear across all countries, and all states, that the rate of people who need acute care who are hospitalized is similar when adjusting for demographics. For example, ICU admission in California will be ~20% than New York, based on demographics. We know that because of data from China, Italy, and Spain. Science, not conjecture.
5. Finally
deaths. Apparently in Italy and China, 49% of people who require ventilators to stay alive (ie those people laying face down in Italy) will die in 4 days, 5 days if they are moved to the ICU on arrival at the hospital.
Nearly every death is counted, and again, there is a discount, but the discount is very stable. There wont be a rush of people counted, and then a week of everyone staying home to die and being tossed in a lake.
Quote:
The fact that these three numbers correlate rather consistently as different areas move across the curve tends to indicate that the number of confirmed cases says something, even though it needs to be taken with a grain of salt.
Absolutely, unequivocally false.
For starters: because of the contained cruise ship debacle that provided an amazing test lab for this, we are very sure that the mortality rate is not 15% like it was in Italy BEFORE their hospitals were overrun not long ago, or 6% like France right now, or 3% like China a month ago. It is much closer to 1%. The country that showed this the most closely,
with the highest scale of testing is South Korea with 1-1.5% mortality depending on time lag of testing to death. So, are you relying on the fact that there is a completely unexplained delta in mortality across regions. A delta that cannot be explained by health care, demographics, time of crisis, resources, etc? I am not. That data is bad data. Incomplete at best, and completely useless comparing regions.
Quote:
If I am missing something here, by all means explain it.
So what does focusing on death get you? We know people dont die from COVID19 unless they have SARS-CoV-2. We can assume that for each region (ie California, Texas, etc) the mortality will stay pretty stable based on the local demographics. So, we KNOW that, until the hospitals are at full capacity, the mortality will be stable, irrespective of it being .01% like the flu or 30% like SARS-CoV-1. It doesnt REALLY matter now. Because if 30% of people die from it, consistently, the number of deaths will grow at the same rate as the number of infections. If .01% of people die from it, consistently, the number of deaths will grow at the same rate as the number of infections. It is basic, simple, math.
How do we know that is still true? Because despite all our best efforts, people who die are not dying slower. They die at a very predictable rate. Median 17 days from infection to death, for example. 9 days when the hospitals are overrun. How do they measure that? Because we know who dies. We dont know who has it, but we know who dies, and when they got symptoms and where they were likely to be infected. Hospitals in all developed nations, and most underdeveloped nations record this information.
The only place the mortality matters is if you need to slow the rate of growth in models because you run out of people to infect. Otherwise you just have to watch for median time to death changing and hospital capacity. Its not perfect but it is a hell of a lot cleaner than any of the alternative data outside South Korea.
So that all adds up to:
In comparing the states with 4 or 5 days since 10 deaths, it is pretty damn clear who has it the worst. It is very clear who will have it the worst. It is not a predictor of policy to determine when to shut down, I would do a completely different analysis for that. Predicting what states will be next, and how bad they will have it in two or three weeks is not possible with this methodology. BUT, Florida, Georgia, Texas, Louisiana, NY, Michigan, California, and Washington all have enough data to predict how terrible it will be for each. It will be worse for Texas than it is for Ca, but not nearly as bad as NY and that is just math, not conjecture.
Texas, with current conditions, would have to have an Iran style licking things event to change the math. I dont see that behavior, but maybe I missed it.
Finally, since running these numbers, expanding to each region that hits 10 deaths, the predictions have been very accurate. It is disturbing how accurate death growth rate continues until roughly 3 weeks after an effective goes into effect.
And to be honest, if 8 billion people contract something that doesnt kill them, I could not care less. I care about the outcomes. I care about how many people will seek help in local hospitals and die. Or die because the hospitals have no capacity. Not fairyland 6% or 15% or .5% mortality rates and face infection numbers.