Coronavirus SARS-Cov2 (COVID-19)

[jayakris]

The whole social distancing thing is/was such a wrong message to spread.
Stay away from indoors or places with recirculated air should have been it from the start (and have everyone wear masks if you need to be in such a location).

Spot on. The messaging has all been horrible. That is the most frustrating part of it.

[jayakris]

ICMR update: 8,398,362 total... Sunday tests: 170,560 ... It better not be more than 16K tomorrow!

ICMR lab count update: +11 to 1047

Good start for Monday. -554 in the Unassigned category.

[jayakris]

How many tests in BLR?

Don't know, but newspapers seem to make general statements that the testing has not been much and one even said it has gone down, but I didn't see BLR testing numbers. My guess is that it is no more than 3000 to 5000 (at best). The state tests a lot, but BLR didn't need more than 2K earlier, with just 10 to 30 cases happening. I think they should have increased it at least 2 weeks earlier immediately after those ILI cases showed up and 11 deaths suddenly happened in the city. If they did that, we would've seen KAR test numbers going up by at least 5K by now. No indication of any of that. I think it is some 10K tests elsewhere and 4K max in the city. Hopefully it is 8K+6K and the positivity is not over 15 to 20%.

My guess was exactly correct, as per this newspaper article that just came out - Bengaluru’s positivity rate has gone up (Bangaloire Mirror)

I cut out the ending of the article title that was deliberately misleading ... They said positivity has gone up from 1.2% to 3.23%, which is incorrect. They are reporting the positivity on the cumulative total. That is some 3000+ in 90K+ tests so far. But the current rate is 3500 to 4500 persons tested a day, which actually means they got 800 cases from probably 4500 tests at about an 18% positivity on Sunday. They need to raise it to 15K, but I can almost guarantee that they cannot and will not. At best they will manage to do 6K or 7K when they are getting 1500+ next week. They need all hands on deck to set containment zones and prioritize testing. They don't have an Amit Shah bringing in the whole Central Govt machinery to tackle the virus with massive testing like at Delhi (but maybe they should ask him ) ...

It is just terrible. The only good thing is that I see a vibrant press from Bengaluru that is writing into the details of a lot of things (unlike in Hyderabad, where there is still no reporting of anything). It is good to see many people talking of the testing situation, starting from the minister Sudhakar. Let us hope that the silicon valley will not let it get out of control for too long.

[suresh]

The large group of Indian scientists at indscicov.in have just put out predictions for Chennai. They have been doing region specific models and they don't do ordinary SIR models. You can read details here: https://indscicov.in/for-scientists-hea ... indscisim/

Summary:

1. 7-13% of the population is infected. The total number of cases is underestimated by a factor of 10.
2. A peak in hospitalization numbers is expected in early August.

[suresh]

TN is the first biggie to report today. +3949 from just 30K tests (13.1% positivity).

[rajitghosh]

https://www.moneycontrol.com/news/busin ... 67941.html

Now JSW Steel in Bellary is the next company after Bajaj Auto to have an explosion of cases at its plant

[rajitghosh]

Even on a Monday the legends are coming to party. TN and Maharashtra show no signs of slowing. What is the effect of the lockdown in Chennai?

[suresh]

Even on a Monday the legends are coming to party. TN and Maharashtra show no signs of slowing. What is the effect of the lockdown in Chennai?

Yes, not a great day!

[rajitghosh]

And finally Gujarat and West Bengal also break out of the narrow band in which they were operating. What is going on? If a Monday brings such numbers what can we expect for the rest of the week?

[jayakris]

The large group of Indian scientists at indscicov.in have just put out predictions for Chennai. They have been doing region specific models and they don't do ordinary SIR models. You can read details here: https://indscicov.in/for-scientists-hea ... indscisim/

1. 7-13% of the population is infected. The total number of cases is underestimated by a factor of 10.
2. A peak in hospitalization numbers is expected in early August.

I really can't put much weight in what they are doing. I took a close look, and I don't think their numbers are that reliable at all. They are making the best effort possible, but still aren't using proper human mobility models (nobody is) and that is a huge problem. Not their fault. There isn't data. They are using a lot of international factors too, and seems to have totally missed modeling the contact-tracing "efficiency" that is causing a factor of 2 to 3 difference between cities' (and states) numbers. Just project with simple aggregate models and that will work better than trying to use dis-aggregate models (which are all inaccurate to start with, due to lack of data and thus poor assumptions). Plus they need to project at the urban zone level if they are going to do this kind of detailed modeling. The coupling assumptions on virus (and human mobility) could all go even more wrong otherwise.

And most importantly, the scientists need to SHUT THE HELL up about "10 times as many people infected". This is the same thing that scientists in the US, UK, and everywhere else say. That number they say has no meaning. Yeah, a lot of people have the virus. So what?

Khichdiwal's sidekick Sisodia parroted this in Delhi a couple of week ago. About 550K infected Delhi by end of June or something (and some ridiculous numbers on hospital bed requirements) and is already undergoing ridicule for that. People got really worried. They shouldn't do this.

ICMR's sero-survey is also showing a high fraction of infected people, so the scientists are not wrong or anything. The Kolkata sero-survey that just came out yesterday showed 15% infected among some 400 random samples, on antibody tests. RT-PCP may show some 5 or 10 positives among them too. But Kolkata does not have 600K cases that we need to worry about. They just don't. It is pointless to use that figure, because most of these "infected people" will simply never even show symptoms. To worry about them, would be idiocy, because our energis should be on managing the ones who may get into trouble soon.

Just focus on the number that need to be counted, for monitoring purposes with contact tracing. They are the ones who have received sufficient viral load that they may get into trouble. That seems to be typically about 20-40 times the deaths you are seeing. Buy the way, the death counts are not 10% of actual numbers. At most 1.5 to 2.0 times what the states report. These scientist-group's own death projections are actually matching what Chennai is seeing. That means the positives Chennai has detected is realistically the only number that need to be counted as "infected to the extent to be of concern" (in fact they needed to detect only half of what they have. Due to having spent energy on all the rest, they started really falling behind on tracing and containing. To have 90% people who are asymptomatic is just horrible over-testing).

The scientists need to stop scaring people, and do their job of telling the truth on what the public needs to know. Again, Indian scientists are better, in my view, than those elsewhere, though.

[jayakris]

And finally Gujarat and West Bengal also break out of the narrow band in which they were operating. What is going on? If a Monday brings such numbers what can we expect for the rest of the week?

Monday usually brings higher positivity, because the positives are first reported by the labs. It seems to take time for the labs to input data into the ICMR portal, so I have noticed the Monday positivity about 25% or so higher thus cases are only some 15% lower even if the tests are 30-40 percent lower on Sunday.

But it's weird that WB has been doing 9500 tests, almost precisely every day, for the last 40 days or so. They finally raised it to 10.5K on Saturday. Hopefully they will do 12K to 15K or so from now on, because the positivity is up to 6.5% now (from 3% a month ago).

[Sin Hombre]

The large group of Indian scientists at indscicov.in have just put out predictions for Chennai. They have been doing region specific models and they don't do ordinary SIR models. You can read details here: https://indscicov.in/for-scientists-hea ... indscisim/

Summary:

1. 7-13% of the population is infected. The total number of cases is underestimated by a factor of 10.
2. A peak in hospitalization numbers is expected in early August.

Like Jay, I don't trust these numbers at all.

It is very similar to the 30% number being parroted for Delhi.

The 95% confidence intervals for CFR is between 0.3% and 0.7% (this is from a study of studies).

Even at the lower end of the confidence interval, this would mean 20 million * 0.3 * 0.003 = 18000 deaths in Delhi.

For Chennai, that would be 7million * 0.1 * 0.003 = 2100 deaths.

There is no way we have gotten anywhere near those numbers, you cannot hide those many bodybags.

[jayakris]

^^^ I think there is serious confusion going on, about who should be counted as "infected". I did not start paying attention to it till the recent weeks, actually. And it is becoming more and more clear to me that only about 30 times (or 20 to 40 times) those who die need to be monitored - if you are tracing right, that is. And that is the actual number of infections we should count. Because only that many people are "sufficiently infected" (which I know, is a horribly naive characterization from a non-epidemiologist and non-doctor). That is because about 1/4th of those people will end up needing hospitalization (i.e, 5 to 10 times those who will eventually die)

So, if the deaths are properly correctly reported in a city and it is 200 people, then there should be around 6000 cases that have tested positive, and about 150K tests done for 45 positivity. That is doing it all about right.

The actual infections in a sero-survey of that city may say that there are 40 to 60K people infected. But it is totally counter-productive to pick up the newspaper boy who may have the virus to some level to turn an anti-body test or even RT-PCP test to turn positive, but if you start contact tracing every house where he dropped newspaper, you are wasting time. Because the boy himself may never get COVID-19 to even sneeze once. He will have SARS-Cov2 virus in him, but not COVID-19, the disease.

Chennai has 850 deaths. Let us say it was actually 1200 and only 850 got reported. Then I think Chennai should be having 1200*30 = 36K positives. But Chennai has 55K cases detected. 90 percent of the active cases at any time have been asymptomatic and overall, I believe only about 10K or 15K have actually ever felt the disease to any level. Chennai has for sure tested (over time) at least 75% more people than needed. Quite probably twice as many. That is my judgment.

As always, I may be wrong. These are just where logic is taking me now, looking at numbers. We try to learn as we go

[jayakris]

It is truly incredible what Amit Shah did in Delhi. ... They did 9.6K RTPCP and 6.5K Antigen tests and got just +2084 in Delhi. Positivity fell from 20+ to 12.9% in just about 10 days.

Wish Mumbai had done this when I suggested 100K tests in a week, some 2 months ago.

Looks like we won't be over 18.3K or so today. Good.

[Sin Hombre]

Jay, your mental model matches what the studies show.

200 deaths in 60k sero-cases in your example is 0.33% ; which is right what I used above.

Does that make sense?

The 95% confidence intervals are fairly accurate imo