Coronavirus SARS-Cov2 (COVID-19)

[jayakris]

Earlier it looked like Karnataka had some infight and thigs might me getting messed up, but KAR has been quiet lately and not in the news. I am not sure if AP and Telengana are tracing and testing enough. Reports from the newspapers are all over the place, the numbers come at odd times, test numbers don't seem to be high enough, and politicians are talking too much. There are a lot of foreign-returnees in that state. I am a bit concerned.

I am not sure if we should stop being concerned about Punjab. There were some horror stories earlier, that thousands of Punjabis from abroad might have escaped airport screening etc. Things were quiet, but some +20 today to 99 cases now. Only 3 are from Tablighi. 6 in a family are positive after they were tested following the death of a lady 2 days ago. The cases that turn positive only 2 days after the death of a patient, shows that testing has just not been enough. Punjab has done only less than 3000 tests so far. Not great. They should be having easily at least 6K to 8K tests so far.

[prasen9]

China's number is possibly severely under-reported.

There is some research that shows that humidity is correlated positively and temperature negatively to the spread of the virus. Which is good news and bad news for India, specifically Kolkata. Article

[sameerph]

So, higher the humidity, the lower the chances of spread, right ? If yes, not proving the case in Mumbai so far which has high humidity level.

[depleter]

ICMR update

A total of 1,14,015 samples have been tested as on 07 April 2020, 9 PM IST. 4616 individuals have
been confirmed positive among suspected cases and contacts of known positive cases in India.
Today, on 07 April 2020, till 9 PM IST, 12,584 samples have been reported. Of these, 285 were
positive for SARS-CoV-2.

ICMR said the first batch of 500K anti body tests will get to them by tomorrow. As they have already given a go ahead for their use, wonder they will be used and when will those tests be added to total tests.

[jayakris]

Article

This paper is absolute bullshit. Done with no insightful look into the problem, by people (engineers) who have no background in epidemiology. That is sometimes okay, but the analysis and discussion with neural net models and regression, which will show some positive or negative correlation (the only two possibilities!) is just childish in this paper. And they did not wait till serious spread happened in many places in the world outside of China too. There have been a whole lot of horrendously bad papers that "ambulance-chaser grad students and professors" have written. I just disregard all of them. So many aspects of epidemiology like the effects of quarantine policies, police actions, testing delays (the biggest flaw in this paper), are not considered. This paper is like the BCG paper by the NY engineering professor's student group. At least that guy found clear correlation and his mistake was assuming causality. Here they have not even found clear-enough correlation and have assumed all kinds of causality.

My opinion is that the infection rate of this virus is correlated only to the physical activities done by people in various places, and the nature of the confined spaces where they were. Perhaps general immunity levels and vaccinations count too. That's it and nothing else. Certain activities are 100 times more infectious than certain other things. Nobody has really identified all of it, because insightful data preparation takes time - and the subject maybe touchy (I think this was the case with HIV too, at one time).

I have been looking left and right for somebody who has done a good enough analysis of this weird-behaving virus and find that there is so much BS research out there that you and I won't even think of doing if our students came to us, prasen! The fact is that insightful research on complex human systems just does not happen in 3 weeks time all that often. Actually I have had quite a few of my students who proposed all kinds of BS to do, again with an ambulance-chasing attitude to write something quickly and get citations. I had to nicely convince them that quality matters more than citation counts.

[jayakris]

ICMR said the first batch of 500K anti body tests will get to them by tomorrow. As they have already given a go ahead for their use, wonder they will be used and when will those tests be added to total tests.

I am curious too. Not sure how they are doing it. I thought the test locations were allowed to go ahead with Covid-19 testing after the viral load rapid tests. So, I assume the swabs from perhaps 20K or 30K people (assuming some 4 to 6% people showing viral infection of some sort) may have been taken already for Covid tests and the tests may be in progress? Should we expect some 3 or 4K Covid positives (or hopefully much fewer) from them in the next 3-4 days? No idea, really.

But it is good to see 12.6K tests today. It has been going up by about a thousand every day, it seems.

[jayakris]

Update: 5337 with +559 today, for a 11.7% growth. At least 100 above what I hoped for. That must be almost a final count, though we may hit 600 by end of day.

Delhi with 51, Telengana with 40 and AP with 11 (which I don't think is correct or final. Very inefficient reporting from AP/Telenagana)

My feeling is that we had only 250 or so Tablighi cases today. At least 300 are other cases with about 120+ of non-Tablighi cases from Mumbai/Maharashtra.

[Varma]

I am not sure if AP and Telengana are tracing and testing enough...

I can't vouch for the entire AP, but according my friends, my hometown (Eluru in West Godavari district) is doing a phenomenal job. When AP numbers spiked up overnight, Eluru went from 0 to 15 cases (all Tablighi returnees) setting a panic in town. Two (or may be three) areas were identified as hot-spots and they sealed these areas off completely with police monitoring day and night. One of these zones had 5 positives. My neighbor's son's in-laws live in that area. Although none of them went to Nizamuddin, they interviewed the in-laws (who live in the hot-spot), came and talked to my neighbor's son (since he is related) to make sure he did not go to Nizamuddin or met anyone who returned from there, and then came to the dad (dad and son live separately) and interviewed him too. That to me is going at least 6 or more levels deep in the chain which is a hell lot of work. It also depends on who is at helm, I suppose. Our current district collector is a very reputed young guy known for his sincerity and hard work. He is a native of the district who made news a decade or so ago for topping Civil services (he made it to the IPS the year before) despite coming from a very poor fishing community. He seems to be working relentlessly to keep things in check. I am sure there are several others doing similar job all over India. We owe a lot to these guys!

- Varma

[jayakris]

Great to know that, Verma. I had generally expected that things would be handled in AP and Telengana, but was seeing all these disconnected reports and got concerned. But like you say, in the end it is these IAS guys who are actually coordinating and doing things. Same in Kerala. The young collector in Pathanmthitta was so quick in grabbing a bunch of engineering students from the college there and setting up the tracking cell immediately on March 8th when the first cases from the west came in.

One thing that I have felt in recent years is that we have a lot of district collectors who are establishing their authority and getting things done - and most of them seem to be very smart. This is the case in Kerala too, though the ministers take the credit. Though the IAS guys in India were always good, I wonder something changed over the years to make them a bit more pro-active in things and not just guys who lived in Bungalows with British-style servants and pushed paperwork for politicians. Lately, that is not the impression I get. A whole heck of a lot of IAS guys are all very young too. It is good for India.

Update: 5341 at +563 seems to be out midnight count. The plus number is a bit higher than yesterday but lower than 2 days ago.

[arjun2761]

So, higher the humidity, the lower the chances of spread, right ? If yes, not proving the case in Mumbai so far which has high humidity level.

The article is likely not very good (and in any case only shows correlation) but I think the correlation it claims is the opposite. High humidity correlates positively to spread while high temperature correlates negatively to spread. That is cold and humid places will have more spread....

[arjun2761]

Here is another paper on the statistics of C19.

Not sure of the quality of the research, but my hunch agrees with its main point, i.e., the data we have is reflective of the testing and medical systems and is really quite some ways off from the actual numbers (of both infections and deaths). In countries where the testing and medical systems are better, the reported numbers is closer to actuals while the numbers in the developing countries is likely quite far from the actual numbers.

This is also enabled by the fact that this disease presents in many people without any real symptoms and often with flu like symptoms which in developing countries are treated at home or by doctors who aren't plugged into the reporting system. Likewise, a lot of the deaths are due to secondary causes of folks who are already sick/old, so that these may also not be reported as C19 deaths.

[jayakris]

Here is another paper on the statistics of C19.

Not sure of the quality of the research, but my hunch agrees with its main point, i.e., the data we have is reflective of the testing and medical systems and is really quite some ways off from the actual numbers (of both infections and deaths).

They are generally in the ball park. Based on data from Mar 17 and Mar 31st, they estimate how wrong the tested numbers were (though the testing rate stat is a poor variable in all this, when we look at the strategy of testing, that varied country-to-country). When I say they are in the ball park, I mean that their estimates are not an order of magnitude off. But they are up to 2 to 3 times the actual number or the actual numbers is 2 to 3 times their estimate (so a wide range, but better than the "system simulation" guys' estimates based on wild guesses on infection rates).

For India, they think there were 8,462 infections on March 17th. The actual number (excluding the Markaz fiasco that was just happening on the 16th-17th) is probably about 3500, out of which we may identified 1600 cases by March 31st, based on airport screening and observations, hospitalization and contact tracing, and perhaps an extra 400 after that, for a 2000 total. I am assuming that 1500 people never got into the observation net who were floating around by March 17. But we definitely did not have 7000 missed-infections floating around on March 17th to fit their 8462 estimate, because we would have seen so many more cases and deaths by now, 21 days later.

So, they are actually over-estimating our number on March 17th by a factor of about 2.5 (i.e., 8462 versus 3500 actual, and 2000 tested and found by now). If we had not done nothing like the lockdown after 22nd, and if Nizamuddin hadn't happened, we would similarly have been around 35K infections by March 31st, as opposed to their 83K estimate. They are figuring a 17.8% per day growth rate, to go from 8,462 to 83,250. With a lockdown which usually drops the infection growth rate by about 10 percent or so, my guess is that (keeping Markaz numbers out) we might have had a growth from 3,500 on Mar 17th to about 10K by March 31st at an 8% rate. But what we had the Markaz issue happen, and that infected some 2000 people right there around the 17th, and they got into the observation net only by March 31st. But since they were in lockdown too but had not been warned/educated by us like the airport observation group, they may have grown at a 13% rate (within mosques and smaller groups). I estimate about 11K infections in the worst case by Mar 31st for the Markaz group. So, my estimate for March 31st infections in India is worst-case 10K of "in-the-net" people, and 11K of Markaz-related people. A total of 21K for March 31st, which is 1/4th of their estimate. We had tested and found about 1600 of them by then.

In fact the numbers in Mumbai are probably a combination of untraceable contacts with Markaz folks, and untraceable contacts with people who were never in our observation net from airport screening.

I expect that the 21K people infected by March 31st may have grown at a rate of 6 or 7 percent in the last week, to about 30K or so by now, but we have only 5.5K confirmed. It could even be 40K by now. That mismatch of tested and actual numbers is natural. It is all due to the 10-14 day lag due to incubation period and testing delay. About 12-16 days lag in India.

In other words, the huge mismatch of confirmed-vs-real numbers, 140-vs-3500 on Mar 17th, is now down to 5500-vs-35000. If we keep the heat on, and continue to work hard, we can bring the mismatch to say 10K-vs-50K in a week. Then, say, to 23K-vs-60K in 2 weeks, if the actual infections growth rate is down to 3 percent in 2 weeks from fire-fighting. Then to 40K-vs-60K. Or this may take a month or two to happen depending on how we keep the unidentified portion of the infected pool from growing too much, by stopping fires as fast as we can.

That is, the mismatch due to the delay in detection and testing may stabilize to a 20K total or grow and drop to some value that seems okay. Or it may take longer and it may grow to a 40K mismatch. Say 70K-vs-110K. The mismatch can come down to 20K or something later too, and we may need to wait for whatever is acceptable. Like 100K identified but 125K infections out there, down from a 70K-vs-110K before that. That may require the identified pool to grow at around a 2 percent rate. Whenever the mismatch is acceptable to us, the problem is pretty much solved and we can move on with our lives. We continue to test, treat, and things will be fine. (I know this may sound a little complicated/weird). Then we may see 600K infections say by September, but we have tested and found 570K by then, and some 15K keeps coming in every weeks, and 15K are leaving the hospitals. No big deal.

The question is how well we will handle the 30 to 40K infections now, which are luckily in so many localities of 50 to 150 cases (except Mumbai) in or case. Probably some 50 areas to cordon-off, seal and fire-fight. 2 or 3 areas each in about 10-12 states (of which we may only know about 10-15 hot-spots now, the rest will be found in the next 2 weeks). Fire-fight and fire-fight... If we can do it well, as and when hot-spots emerge over the next two weeks, we will still have control by Apr 21st and keep it going for the coming weeks and months. Let us see what happens.

Or maybe the natural infection rate is so low that we just don't see 30K even by end of April. That will be great.

[prasen9]

I had to nicely convince them that quality matters more than citation counts.

And, I think quality will eventually lead to more citation counts and not have done analysis. Correlation is not causation. And this issue is very difficult to model without taking into account all the different factors: weather, genetics, when did people lockdown, density of the cities, whether you have idiots ruling the country, etc. BoJo was boasting that he shook the hands of everyone in the hospital who had COVID-19 and now he is in the ICU. Viruses do not get conned by idiots.

[gbelday]

Article

This paper is absolute bullshit. Done with no insightful look into the problem, by people (engineers) who have no background in epidemiology. That is sometimes okay, but the analysis and discussion with neural net models and regression, which will show some positive or negative correlation (the only two possibilities!) is just childish in this paper.

Hi all, some of you might find this interesting. It’s a one hour presentation by one of the professors, Anindya Ghose, from NYU Stern School of Business.
It’s on using technology, data and data science/predictive analytics to combat pandemics. It’s quite fascinating. You can get past the first couple of minutes and go directly to Prof. Ghose.

A video from YouTube usually appears here. Please contact an administrator.

A video from YouTube usually appears here. Please contact an administrator.

[prasen9]

So, higher the humidity, the lower the chances of spread, right ? If yes, not proving the case in Mumbai so far which has high humidity level.

The article is likely not very good (and in any case only shows correlation) but I think the correlation it claims is the opposite. High humidity correlates positively to spread while high temperature correlates negatively to spread. That is cold and humid places will have more spread....

I agree the paper is not very good with respect to what it is purporting to study. Very simplistic assumptions that do not take into account the complexity of the factors influencing and their impact. Having said that, it says the opposite (which Arjun explained).

Arjun's citation is a paper in the Lancet, which is a better-quality venue. However, only 1/3rd of the seminal papers in psychology could be validated. So, science does have a reproducibility problem. We need to be careful about drawing definite conclusions. I had include the paper because it said the opposite of what we believed to be true, i.e., humidity increases the likelihood of the virus being spread. Just to be clear, there is no conclusive evidence of any sorts to support that and it remains a hypothesis.