Categories
Covid-times

India Covid deaths weekly projection – 11 Oct update

IT’S CLEAR NOW THAT INDIA PEAKED IN SEPTEMBER.
FROM 1.07 LAKH REPORTED DEATHS AS OF NOW, WE WOULD REACH AROUND 1.4 TO 1.6 LAKH BY YEAR END.
IT’S QUITE POSSIBLE THAT EVENTUALLY INDIA FLATS OUT AT MAX 2 LAKH TOTAL REPORTED DEATHS.


I have been looking at the India Covid figures since few months now. I update my projections every Sunday. This is how things look as of today.

The actual Covid death toll could be as high as twice the ‘reported’ figures, for various reasons documented here but what’s clear is that the avg daily death has been going down for three weeks now. This suggests India peaked by 4th week of September.

Graph plotted by Amrit Vatsa on 11 Oct 2020, based on publicly available data

The total reported Covid death toll for India right now is 1.07 lakh that basically translates to roughly 78 deaths per million.

In comparison, Brazil has over 700 per million dead, US & UK over 600 and Canada over 250.

One could try forecasting the future Covid deaths in India by simply using the existing rate of week-on-week growth / decline in deaths.

Chart created by Amrit Vatsa on 11 Oct 2020 from publicly available data

For three consecutive weeks, total weekly deaths in India has continued to decline.

For future, -8% (baseline), -12% (better case) and -4% (worse case) assumption for decline rate sounds good?

Below is how the forecast looks like, for those three scenarios.

Chart created by Amrit Vatsa on 04 Oct 2020 from publicly available data

The total death toll will reach somewhere around 1.4 to 1.6 lakh by 2020 end and it’s quite likely that Covid in India will be more or less over once around 2 lakh are dead. That will take many more months in 2021 given that growth is on decline.

Fingers crossed.

Let’s now try a slightly more nuanced (albeit indirect) approach to project future deaths that requires looking first at cases. Cases are important because even when you don’t die, just being infected seems to have its own issues.

From ‘brain fog’ to heart damage, COVID-19’s lingering problems alarm scientists

ScienceMag.Org

Like deaths, for cases too, if we look at the rate of week-on-week growth / decline, we can get some idea of how it’s probably going to unfold in the next few weeks.

5.05 lakh total positive cases were detected this week – that’s 11% lower than the total cases detected the week before (~5.7 lakh).

Let’s call this w-o-w rate of growth / decline in cases – ‘X’. X was -4% last week and -8% the week before (see the above chart). X=-11% this week. For my projection, I think I will assume a range of zero to -10% for X in the coming weeks.

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before (just a basic assumption). Do we have some idea of what %age of such cases die? We do actually.

6,574 Covid deaths were recorded this week, which is basically 1.2% of half of total cases from this week + half of total cases from last week.

Let’s call this %age Y; Y= 1.2%.

For the future, let’s assume a range of 1.1% to 1.3%?

So we can forecast now – I am going with the following 3 scenarios:

  • X=-5%, Y=1.2% (baseline)
  • X=-10%, Y=1.1% (optimistic: cases decline faster + lesser %ge of deaths)
  • X=0%, Y=1.3% (worse: cases don’t decline + higher death %age )

With the above assumptions, below chart shows the future cumulative death count.

This gives a similar estimate.

India’s reported Covid deaths would be around 1.4 to 1.6 lakh by year end.

It was interesting to see the New York Times report on the spread of Covid in rural India in pretty grim ways. This is their article from Friday (behind paywall). Number-wise things are actually getting better!

Before I end, below is a new Covid insight that you all should be aware of!

That’s it for this post. I’ll get back with updated projections next Sunday (19 Oct). Stay safe.

Categories
Covid-times

India Covid deaths weekly projection – 04 Oct update

REPORTED DEATH TOLL HAS CROSSED 1 LAKH BUT INDIA SEEMS TO BE APPROACHING THE PEAKING POINT NOW. ABOUT 50,000 MORE WILL DIE BY NOVEMBER AND AFTER THAT, WE WOULD MOSTLY FLAT OUT! THIS IS GOOD.


My big question every week (since May) has been, when will India cross 1 lakh total reported Covid deaths? It happened this week.

The actual Covid death toll could be as high as twice the ‘reported’ figures, for various reasons documented here.

On a global level, if you just look at total number of reported Covid deaths, you will find that India is at no. 3. But the moment you adjust for population (which makes more sense), you realize that India is in a much better position (the pink line; US is dark blue, Brazil green, UK light blue and Canada red).

Source – FT – Plot generated and screenshot taken on 04 Oct 2020

1 lakh total deaths for India translates to 73 deaths per million (Brazil is 900% higher than India’s per million deaths).

It will take many months for India to reach the kind of deaths per million figures that Brazil or US have already seen (if it ever does). After looking at data from the present and previous weeks, it doesn’t look like things will ever get that bad in India.

In any case, there are many countries doing better then India and there are many others that are doing worse (once you adjust for population and compare).

One could try forecasting the future Covid deaths in India by simply using the existing week-on-week growth in deaths.

Chart created by Amrit Vatsa on 04 Oct 2020 from publicly available data

For two consecutive weeks, total weekly deaths in India has continued to decline. This never happened before! And this is great news!

For future growth / decline estimate, -2% (baseline), -4% (better case) and +2% (worse case) assumption sounds good?

This is how the forecast looks like, for the following three scenarios.

Chart created by Amrit Vatsa on 04 Oct 2020 from publicly available data

It is quite possible now that India’s total death toll may not cross 2 lakh (145 deaths per million) this year. Both US and Brazil are already over 600 per million dead.

If what I am saying happens, chances are we will flat out below 200 deaths per million (similar to Canada that peaked after crossing 200 – if you scroll up and check the FT chart I put up).

Fingers crossed.

Let’s now try a slightly more nuanced (albeit indirect) approach to project future deaths that requires looking first at cases. Cases are important because even when you don’t die, just being infected seems to have its own issues.

From ‘brain fog’ to heart damage, COVID-19’s lingering problems alarm scientists

ScienceMag.Org

At a global level, when adjusted for population – total reported cases for India (pink in the below chart) are low when compared to the worst performers (US – dark blue, Brazil – green) but already higher than Canada (red) and could cross UK (light blue) soon.

Source – FT – Plot generated and screenshot taken on 04 Oct 2020

Anyway so like deaths, for cases too, if we look at the week-on-week growth, we can have some idea of how it’s probably going to grow / decline in the next few weeks.

~5.7 lakh total positive cases were detected this week, which is 4% lower than the total cases detected the week before (~6 lakh).

I am aware of the issues with low testing but I am not sure that’s the only explanation for decline in weekly cases.

I think India is now approaching the peaking point.

Let’s call this w-o-w growth in cases – ‘X’. X was -8% last week and +2% the week before (see the above chart). X=-4% this week. For my projection, I think I will assume a range of -2 to +2% for X in the coming weeks.

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before (just a basic assumption). Do we have some idea of what %age of such cases die? We do actually.

7,463 Covid deaths were recorded this week, which is basically 1.3% of half of total cases from this week + half of total cases from last week.

Let’s call this %age Y; Y= 1.3%.

For the future, let’s assume a range of 1.2% to 1.4%?

So we can forecast now – I am going with the following 3 scenarios:

  • X=0%, Y=1.3% (baseline)
  • X=-2%, Y=1.2% (optimistic: growth in cases declines + lesser %ge of deaths)
  • X=+2%, Y=1.4% (worse: cases grow + higher death %age )

With the above assumptions, below chart shows the future cumulative death count.

We will see around 50,000 more deaths by November, but once India crosses 2 lakh, the toll would not increase much beyond that.

Before I end, below is a new Covid insight that you all should be aware of!

That’s it for this post. I’ll get back with updated projections next Sunday (11 Oct). Stay safe.

Categories
Covid-times Gyaan

‘Invisible women’ in the Covid times

Thalidomide is a medication used to treat a number of cancers and skin conditions including complications of leprosy. The developers of the drug claimed that they “could not find a dose high enough to kill a rat” and so thalidomide was freely available since 1950s in stores as a mild over-the counter medication in many countries.In 1960, doctors began prescribing it to pregnant women who suffered from morning sickness.

It turned out that while the drug didn’t kill rats, it did affect foetal development. Before it was finally taken off the market in 1962, over 10,000 children had been born around the world with thalidomide-related disabilities! You can read a detailed story by NY Times.

Because of this scandal, the FDA (Food and Drug Administration) issued guidelines in 1977 excluding women of childbearing potential from drug trials!

But what happens when you exclude a certain group from clinical trial? Be it pregnant women or women altogether?

Over the last few weeks I have been reading a super insightful book – Invisible Women: Exposing Data Bias in a World Designed for Men by Caroline Criado Perez. Actually, I had started to read it long back (last year I think), but then got overwhelmed by the data / insights and had to take a break. I am glad I resumed. It is still overwhelming but one must read as much as one can.

Caroline essentially quotes examples after examples of how almost everything that is made / designed in the ‘man’-made world, miserably fails to take into account the specific needs of women. This reflects in design of roads to malls to phones to piano to drugs to vaccines.

Just because something works for men in a certain way does not mean it will necessarily work the same way for women. And that’s a problem when women are not represented adequately in any kind of user impact study.

In 2000 for example, the FDA had to force drug manufacturers to remove phenyl-propanolamine, a component of many over-the-counter medications, from all products because of a reported increased risk of bleeding into the brain or into tissue around the brain in women, but not in men.

To understand the difference between male and female bodies, at the most basic level you need to realize that women typically tend to have a higher body-fat percentage than men. This, along with the fact that blood-flow to fat tissue is greater in women, affects how women metabolise certain drugs. Also, male gut transit times are about half the length of women’s! This means women may need to wait for longer after eating before taking medications that need to be absorbed on an empty stomach.

When it comes to vaccine, it is well proven that women develop higher antibody responses and have more frequent and severe adverse reactions to vaccines.

The mechanisms leading to these differences can be:

  • hormonal (i.e. the different effects of testosterone, oestrogens or progesterone);
  • genetic (biological females have two X chromosomes while males have only one); or
  • related to differences in intestinal bacteria.

And yet, most phase 1 clinical trials – a) don’t bother to study sex specific results and b) don’t enroll women in adequate numbers in the first place! How many drugs that would work for women are being ruled out at phase 1 trials just because they don’t work in men? Nobody knows!

Sex matters even in animal trials. In a 2007 analysis of animal studies where rats of both sexes were identified, it was observed that in over half the studies, the drug-effect depended upon the sex of the animal! And yet, most animal tests don’t bother to sex-tag the results (if they at all get enough of male and female animals in the first place).

By the way sex and gender have different implications – and to those not clear on the difference between the two terms, the below figure is self explanatory.

Source

So how are we doing gender / sex wise in terms of analyzing Covid’s effect or vaccine development?

Most states are doing a bad job of reporting sex / gender aggregated data.

Bad quality of sex / gender aggregated data = a vaccine / drug that is designed mostly for men.

Source

In a still to be peer-reviewed study, researchers have found that only 416 of the 2,484 Covid-19 clinical trials mention sex / gender as a recruitment criterion on the ClinicalTrials.gov database. [Source]

During the time of ancient Greeks, the female body was seen as a ‘mutilated male’ body – ovaries were female testicles and didn’t have a name for themselves till the 17th century! For millennia, medicine has functioned on the assumption that male bodies can represent humanity as a whole. A 2008 analysis of a range of textbooks recommended by ‘twenty of the most prestigious universities in Europe, US and Canada’ revealed that across 16,000+ images, male bodies were used three times as often as female!

For things to change in the future, we all need to be at least aware of the implicit data bias that exists in every single aspect of our lives – before enough people can even begin to make noticeable noise about it. I can only hope that happens sooner than later.

If you found the insights in this blog fascinating and yet reading an entire book on this topic is a bit much, at least check out this Guardian article that has a lot more examples of data bias for women and how it affects them, even kills them.

Categories
Covid-times

India Covid deaths weekly projection – 27 Sep update

FOR THE FIRST TIME, THE TOTAL NEW CASES REPORTED IN A WEEK WERE LESS THAN THAT IN THE PREVIOUS WEEK!
HAS INDIA PEAKED ALREADY?
IN ANY CASE, INDIA WILL CROSS 1 LAKH REPORTED COVID DEATHS THIS WEEK (93K+ DEATHS PRESENTLY).


My big question every week (since May) is, when will India cross 1 lakh total reported Covid deaths? It will happen this week as we enter October.

Total cumulative Covid death toll as of yesterday (26 Sep) stands at 93k+ (actual figure could be as high as twice this value, for various reasons documented here).

On a global level, if you just look at total number of reported Covid deaths, you will find that India is at no. 3. But the moment you adjust for population (which makes more sense), you realize that India is in a much better position (the pink line; US is dark blue, Brazil green, UK light blue and Canada red).

93k total deaths means ~68 deaths per million (Brazil is almost ten times that figure). It will take many months for India to reach the kind of deaths per million figures that Brazil or US have already seen (if it does).

The reason I just compared India with literally the worst performers is not so that like Modi, I can claim it’s all good – I am just making sure you see things for how they are. There are many countries doing better then India and there are many others that are doing worse (once you adjust for population and compare).

One could try forecasting the future Covid deaths in India by simply using the existing week-on-week growth in deaths.

Chart created by Amrit Vatsa on 27 Sep 2020 from publicly available data

This is the first time that weekly death growth in India went significantly below zero percent!

Only once earlier, the growth had been negative (just shy of zero – in the 23-29 Aug week). For future average weekly growth estimate, 2% to 10% growth range sounds good?

This is how the forecast looks like, for the following three scenarios.

We will cross 1 lakh Covid deaths this week and then will touch 2 lakh by early December (2 lakh total deaths for India would be equivalent to 144 deaths per million; both US and Brazil are already over 590 per million dead). Even a relatively better performer Canada peaked only after crossing 200 deaths per million.

Alright, let’s now try a slightly more nuanced (albeit indirect) approach to project future deaths. This requires looking first at cases. Cases are important because even when you don’t die, just being infected seems to have its own issues.

From ‘brain fog’ to heart damage, COVID-19’s lingering problems alarm scientists

ScienceMag.Org

At a global level, when adjusted for population – total reported cases for India (pink in the below chart) are low when compared to the worst performers (US – dark blue, Brazil – green) but already higher than Canada (red) and will cross UK (light blue) soon.

Anyway so like deaths, for cases too, if we look at the week-on-week growth, we can have some idea of how it’s probably going to grow in the next few weeks.

~6 lakh total positive cases were detected this week, which is 8% lower than the total cases detected the week before (6.5 lakh)

I will not conclude that India has peaked already unless total cases continue to be negative week after week. This week could just have been an anomaly.

Let’s call this w-o-w growth in cases – ‘X’. X was 2% last week and 14% the week before (see the above chart). X=-8% this week. For my projection, I think I can assume a range of 2 to 10% for X in the coming weeks.

Let me also make sure you understand how growth works – when something grows at 10% every week, it means it will double in less than two months. But if it grows at 2%, it would take almost nine months for it to double!

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before (just a basic assumption). Do we have some idea of what %age of such cases die? We do actually.

7,760 Covid deaths were recorded this week, which is basically 1.2% of half of total cases from this week + half of total cases from last week.

Let’s call this %age Y; Y= 1.2%.

For the future, let’s assume a range of 1% to 1.2%?

So we can forecast now – I am going with the following 3 scenarios:

  • X=5%, Y=1.1% (baseline)
  • X=2%, Y=1% (optimistic: slower growth in cases + lesser %ge of deaths)
  • X=10%, Y=1.2% (worse: expecting faster growth in cases)

With the above assumptions, below chart shows the future cumulative death count.

What’s going to happen this week doesn’t change (crossing 1 lakh) but this indirect estimate for deaths tells us that only in the worse case scenario will we cross 2 lakh in 1st week of December.

India will cross 1 lakh total deaths this week and will probably touch 2 lakh in December.

Now, 1 lakh total deaths for India is basically equivalent to 72 deaths per million of the total population (currently we are at 68 per million Covid deaths).

To what extent would the death toll figures keep going up – before it flattens / peaks?

If we look at other countries, death toll for many started to flatten out only after anywhere between 400 to 600 per million of their population died!! Scary, I know!

If we assume that for India, the death toll flattens out even at say 200 deaths per million, that would be equivalent to ~3 lakh total deaths!

It’s difficult to imagine why India would see any less no. of deaths than that. Let’s look at some of our cities / small states.

Y axis represents weeks; 1= the week when the city / state first reached ~10 deaths per million

Pune for example, has crossed 800 deaths per million.

The only populous countries across the globe where death toll flattened at much lower levels (like say Japan and China) happened when they somehow didn’t let the total deaths cross even 5k (Japan for example didn’t even let it cross 1k). We clearly couldn’t control things to that extent in India (most countries haven’t). So now let’s just be hopeful that the total death cap estimate that I am guessing is on the conservative end – otherwise, we could lose even up to 5 lakh people (or 362 deaths per million)!

That’s it for this post. I’ll get back with updated projections next Sunday (04 Oct). Stay safe.

Categories
Covid-times

India Covid deaths weekly projection – 20 Sep update

85K+ HAVE REPORTEDLY DIED ALREADY.
WE WILL CROSS 1 LAKH DEATHS IN 2 MORE WEEKS AND AT THE PRESENT RATE, ANOTHER ONE LAKH WOULD BE DEAD BY NOV END.


My big question every week (since May) is, when will India cross 1 lakh total reported Covid deaths?

Total cumulative Covid death toll as of yesterday (19 Sep) stands at 85k+ (actual figure could be as high as twice this value, for various reasons documented here). Remember that this number was just over forty thousand in first week of August.

On a global level, if you just look at total number of reported Covid deaths, you will find that India is at no. 3. But the moment you adjust for population (which makes more sense), you realize that India is in a much better position (the pink line; US is blue, Brazil green).

85k total deaths means just over 60 deaths per million (Brazil is ten times that figure). It will take many months for India to reach the kind of deaths per million figures that Brazil or US have already seen. The reason I just compared India with literally the worst performers is not so that like Modi, I can claim it’s all good – I am just making sure you see things for how they are. There are many countries doing better then India and there are many others that are doing worse (once you adjust for population and compare).

One could try forecasting the future Covid deaths in India by simply using the existing week-on-week growth in deaths.

Yes there are many ups and downs in the weekly growth of Covid deaths but if one has to extrapolate, a 5% to 15% growth range seems to be a good guess?

This is how the forecast looks like, for the following three scenarios.

We will cross 1 lakh Covid deaths in like two weeks and then will touch 2 lakh sometime in November (2 lakh total deaths for India would be equivalent to 144 deaths per million; both US and Brazil are already over 550 per million dead).

There is another way to forecast future deaths, but let me take a quick light break first.

Alright, let’s now try a slightly more nuanced (albeit indirect) approach to project future deaths. This requires looking first at cases. Cases are important because even when you don’t die, just being infected seems to have its own issues.

From ‘brain fog’ to heart damage, COVID-19’s lingering problems alarm scientists

ScienceMag.Org

At a global level, when adjusted for population – total reported cases for India, as of now are much low. But they are growing – so would they remain low forever?

Anyway so like deaths, for cases too, if we look at the week-on-week growth, we can have some idea of how it’s probably going to grow in the next few weeks.

6.5 lakh total positive cases were detected this week, which is only 2% higher than the total cases detected the week before (6.4 lakh)

Let’s call this w-o-w growth in cases – ‘X’. X was 14% last week and 15% the week before (see the above chart). X=2% this week, possibly because of not enough increase in testing capacity?

Anyway, for my projection, I think I can assume a range of 5 to 15% for X in the coming weeks.

Let me also make sure you understand how growth works – when something grows at 10% every week, it means it will double in a little less than two months (7-8 weeks). But if it grows at 15%, it will double in just five weeks. On the slower end, if something grows at 5%, it would take almost 4 months for it to double!

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before (just a basic assumption). Do we have some idea of what %age of such cases die? We do actually.

8,147 Covid deaths were recorded this week, which is basically 1.3% of half of total cases from this week + half of total cases from last week.

Let’s call this %age Y; Y= 1.3%.

For the future, let’s assume a range of 1% to 1.3%?

So we can forecast now – I am going with the following 3 scenarios:

  • X=10%, Y=1.2% (baseline)
  • X=5%, Y=1.1% (optimistic: slower growth in cases + lesser %ge of deaths)
  • X=15%, Y=1.3% (worse: expecting faster growth in cases)

With the above assumptions, below chart shows the future cumulative death count.

The indirect method more or less gives a similar estimate as the direct death projection.

India will cross 1 lakh total deaths in two weeks and by Nov end, the figure is likely to cross 2 lakh (i.e. 1 lakh more deaths will happen just in Oct-Nov).

Now, 1 lakh total deaths for India is basically equivalent to 72 deaths per million of the total population (currently we are at 62 per million Covid deaths).

To what extent would the death toll figures keep going up – before it flattens / peaks?

If we look at other countries, death toll for many started to flatten out only after anywhere between 400 to 600 per million of their population died!! Scary, I know!

If we assume that for India, the death toll flattens out even at say 200 deaths per million, that would be equivalent to ~3 lakh total deaths!

It’s difficult to imagine why India would see any less no. of deaths than that. Let’s look at some of our cities.

Y axis represents weeks; 1= the week when the city first reached ~10 deaths per million

Pune for example, crossed 700 deaths per million last week itself.

The only populous countries across the globe where death toll flattened at much lower levels (like say Japan and China) happened when they somehow didn’t let the total deaths cross even 5k (Japan for example didn’t even let it cross 1k). We clearly couldn’t control things to that extent in India (most countries haven’t). So now let’s just be hopeful that the total death cap estimate that I am guessing is on the conservative end – otherwise, we could lose even up to 5 lakh people (or 362 deaths per million)!

That’s it for this post. I’ll get back with updated projections next Sunday (27 Sep). Stay safe.

Categories
Covid-times

India Covid deaths weekly projection – 13 Sep update

INDIA IS AT 77K+ REPORTED DEATHS (56 PER MILLION) AS OF NOW AND WILL CROSS 1 LAKH TOTAL DEATHS BY SEP END / OCT FIRST WEEK. PUNE IS LEADING THE GROWTH WITH 700+ PER MILLION DEAD ALREADY.


My big question every week (since May) is, when will India cross 1 lakh total reported Covid deaths? Total cumulative Covid death toll as of yesterday (12 Sep) stands at 77k+ (actual figure could be as high as twice this value, for various reasons documented here).

For the first time, daily reported deaths every single day in a week remained over 1,000. That’s almost six full capacity A320s crashing and killing everyone on board everyday.

One could try forecasting the future deaths by simply using the existing week-on-week growth in ‘total deaths in a week’.

Yes there are many ups and downs in the weekly growth of Covid deaths but if one has to extrapolate, a 5% to 15% growth range seems to be a good guess?

This is how the forecast looks like, for the three scenarios.

We will cross 1 lakh Covid deaths either by this month end or latest by first week of October.

Alright, let’s now try a slightly more nuanced (albeit indirect) approach to project cumulative deaths. This requires looking first at cases. Cases are important because even when you don’t die, just being infected could be troublesome.

From ‘brain fog’ to heart damage, COVID-19’s lingering problems alarm scientists

ScienceMag.Org

Let’s look at the week-on-week growth rate of ‘total cases in a week’.

6.4 lakh total positive cases were detected this week, which is 14% higher than the total cases detected the week before (5.6 lakh)

To make sure you understand how growth works, when something grows at 10% every week, that means it will double in a little less than two months (7-8 weeks). If it grows at 15%, it will double in just five weeks and if it grows as slow as 5%, it would take almost 4 months for it to double.

For my projection, I will assume a range of 5 to 15% (X) for week-on-week growth of cases. That’s my assumption #1.

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before. Do we have some idea of what %age (Y) of such cases die? We do actually.

7,891 Covid deaths were recorded this week that is basically 1.3% of half of total cases from this week + half of total cases from last week.

In other words, Y for this week is 1.3%. For the future, let’s assume a range from 1% to 1.3%? That’s my assumption #2.

Let’s forecast now…

Let me consider 3 scenarios:

  • X=10%, Y=1.2% (baseline)
  • X=5%, Y=1.1% (optimistic: slower growth in cases + lesser %ge of deaths)
  • X=15%, Y=1.3% (worse: expecting faster growth in cases)

With the above assumptions, below chart represents the future cumulative death count:

The indirect method more or less gives a similar estimate as the direct death projection.

India will cross 1 lakh total deaths by early October.

Now, 1 lakh total deaths for India is basically equivalent to 72 deaths per million of the total population (currently we are just over 56 per million Covid deaths).

To what extent would the death toll figures keep going up – before it flattens / peaks?

If we look at other countries, death toll for many started to flatten out only after anywhere between 400 to 600 per million of their population died!! Scary, I know!

Y axis = no. of days (all the countries are arranged in a way that starting point of 10 deaths per million is common to all)

If we assume that for India, the death toll flattens out even at say 200 deaths per million, that would be equivalent to ~3 lakh total deaths!

It’s difficult to imagine why India would see any less no. of deaths than that. Let’s look at some of our cities.

Y axis represents weeks; 1= the week when the city first reached ~10 deaths per million

Pune has already crossed 700 deaths per million and is still quite steep in terms of growth.

Delhi has somehow managed to grow much slower and doesn’t look like it will cross even 200 deaths per million. Mumbai may not be growing as fast as Pune, but it is almost at 400 per million deaths and far from peaking. So that’s the overall range we are looking at (to remind you again, at a country level we are at < 60 per million dead so far).

The only populous countries across the globe where death toll flattened at much lower levels (like say Japan and China) happened when they somehow didn’t let the total deaths cross even 5k (Japan for example didn’t even let it cross 1k). We clearly couldn’t control things to that extent in India (most countries haven’t). So now let’s just be hopeful that the total death cap estimate that I am guessing is on the conservative end – otherwise, we could lose even up to 5 lakh people (or 362 deaths per million)!

That’s it for this post. I’ll get back with updated projections next Sunday (20 Sep). Stay safe.

Categories
Covid-times

India Covid deaths weekly projection – 06 Sep update

WHILE THE TOTAL DEATH COUNT IS NOW AT 70K, THIRTY THOUSAND INDIANS DIED OF COVID IN JUST PAST ONE MONTH


My big question every week (since May) is, when will India cross 1 lakh total reported Covid deaths? Total cumulative Covid death toll as of yesterday (05 Sep) stands at ~70k (actual figure could be as high as twice this value, for various reasons documented here).

For the first time, avg daily deaths in a week crossed 1,000. That’s more than 5 full capacity A320s crashing and killing everyone on board everyday – for one full week.

Also notice above how just a month ago (01 Aug) deaths were less than 40k. So basically in just one month, 30k more died.

One could try forecasting the future deaths by simply using the existing week-on-week growth in ‘total deaths in a week’.

Yes there are many ups and downs in the weekly growth of Covid deaths but if one has to extrapolate, a 5% to 15% growth range seems to be a good guess?

This is how the forecast looks like, for the three scenarios.

We will cross 1 lakh Covid deaths either by this month end or latest by first week of October.

Alright, let’s now try a slightly more nuanced (albeit indirect) approach to project cumulative deaths. This requires looking first at cases.

For many weeks, I had maintained that deaths matter and cases don’t (because if you don’t die, you end up recovering). But someone pointed me this week to the growing evidence of long term effect on those who got infected but didn’t die.

From ‘brain fog’ to heart damage, COVID-19’s lingering problems alarm scientists

ScienceMag.Org

Till last month, total positive cases detected was about 2 million and now it’s over 4 million. So basically…

we detected as many new cases in just one month (2 million), as we had ever since the Pandemic hit India.

Let’s look at the week-on-week growth rate of ‘total cases in a week’.

5.6 lakh total positive cases were detected this week, which is 14.5% higher than the total cases detected the week before (4.9 lakh).

For my projection, I will assume a range of 7 to 15% (X) for week-on-week growth of cases. That’s my assumption #1.

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before. Do we have some idea of what %age (Y) of such cases die? We do actually.

7,011 Covid deaths were recorded this week that is basically 1.3% of half of total cases from this week + half of total cases from last week.

In other words, Y for this week is 1.3%. For the future, let’s assume a range from 1% to 1.3%? That’s my assumption #2.

Let’s forecast now…

Let me consider 3 scenarios:

  • X=10%, Y=1.2% (baseline)
  • X=7%, Y=1% (optimistic: slower growth in cases + lesser %ge of deaths)
  • X=15%, Y=1.3% (worse: expecting faster growth in cases)

With the above assumptions, below chart represents the future cumulative death count:

The indirect method more or less gives a similar estimate as the direct death projection except in one scenario (optimistic), we may cross 1 lakh deaths only by 10th October.

India will cross 1 lakh total deaths by early October.

Now, 1 lakh total deaths for India is basically equivalent to 72 deaths per million of the total population (currently we are just over 50 per million Covid deaths).

To what extent would the death toll figures keep going up – before it flattens / peaks?

If we look at other countries, death toll for many started to flatten out only after anywhere between 400 to 600 per million of their population died!! Scary, I know!

Y axis = no. of days (all the countries are arranged in a way that starting point of 10 deaths per million is common to all)

If we assume that for India, the death toll flattens out even at say 200 deaths per million, that would be equivalent to ~3 lakh total deaths!

It’s difficult to imagine why India would see any less deaths than that. The only populous countries across the globe where death toll flattened at much lower levels (like say Japan and China) happened when they somehow didn’t let the total deaths cross even 5k (Japan for example didn’t even let it cross 1k). We clearly couldn’t control things to that extent in India (most countries haven’t). So now let’s just be hopeful that the total death cap estimate that I am guessing is on the conservative end – otherwise, we could lose even up to 5 lakh people (or 362 deaths per million)!

That’s it for this post. I’ll get back with updated projections next Sunday (13 Sep). Stay safe.

Categories
Covid-times Gyaan

How much of the 23.9% GDP growth fall for India can be attributed to Covid?

Note: This article was updated on 04 Sep 2020 (few days after being first published)


Below is what the BJP twitter handle tweeted on 22 Aug.

They used IMF’s forecast from April, even when IMF already had a negative forecast out in June.

Anyway, so now we know what really happened.

My question is: how much of this drop is from Covid and how much of it has to do with actions from Govt. (typically lockdown). One may say that the action of Govt. is a response to Covid – so they are one and the same thing – but are they? Imagine two scenarios:

  1. Govt. doesn’t do much – the economy still goes down – this is purely because of Covid
  2. Govt. does something – here both factors are at play

So for India, how much of the fall is from Covid and how much is from what Govt. did about it+Covid? Can one attempt to find that out?

To begin with, I think it makes sense to first look at some other countries that were impacted by Covid too, the “Govt. response” being different for each country.

GDP growth fell for all of them – as you can see in the below graph that I put together.

Other than 2020Q2 for India, other figures from OECD

India is the orange line. From the countries that I have chosen, India’s last quarterly growth in 2019 was lower only than China (grey). This was pre-Covid. Also, by growth I mean, when compared to the same quarter the year before.

In the next quarter, China’s GDP growth took the sharpest fall (understandably). Other countries GDP growth fell too, but not as much. In fact, India was doing okay in Jan-Feb-March 2020.

And then Covid spread in India.

In the Apr-May-June quarter, India’s growth fell really really bad. The decline in GDP growth is only better than Peru, a country that is one of the worst hit by Covid (just scroll up and look at the per million deaths for Peru). Peru’s GDP growth fell down by almost 30%.

China is the only country whose growth returned to positive in this quarter. Now you tell me, what explains the orange line (India) falling this bad even when other countries were hit by Covid much worse? Or…

has India’s GDP growth fallen this bad BECAUSE we have not done as bad as most of these other countries in terms of Covid spread?

But if this were the case, what explains Peru? Its cases went up AND its economy fell too – both drastically.

Think about it.

How do you think a GoI’s propaganda channel is spinning the news? Just listen to Arnab describe this “global phenomena”. He literally quotes random numbers for countries like USA and Canada. Is he lying on national TV or am I missing something?

At one point, #GDPTruth was trending at no. 5 on Twitter. Most of them using this hashtag seemed to be paid propagandists whose only job looked like passing this off by calling it a “global phenomena”!

Is comparing India with US / Europe like comparing apples and oranges?

Responding to an earlier draft of this blog, some friends seemed to suggest so.

I agree. No two countries are the same. No two fruits are the same. Not even two apples are same for that matter! Comparison after all is just one way of looking at things – because looking at something in isolation is even more meaningless, no?

If you think about it, the figure of -23.9% itself is a comparison! It compares 2020 quarter 2 GDP with 2019 quarter 2 GDP. But 2020 is not like 2019, so why compare? Everything is apples and oranges – from Global Ease of Doing business rankings to GDP forecasts!

Some also responded to my first draft saying this was expected – we were in lockdown for most of the quarter in question – people were saving money (so drastic drop in demand) – factories couldn’t operate (so supply drop).

Nice. Let me quote Daniel Kahneman from Chapter 19 of his book Thinking Fast & Slow (the chapter is titled “the illusion of understanding”).

The mind that makes up narratives about the past is a sense-making organ. When an unpredicted event occurs, we immediately adjust our view of the world to accommodate the surprise.

Learning from surprises is a reasonable thing to do, but it can have some dangerous consequences.

A general limitation of the human mind is its imperfect ability to reconstruct past states of knowledge, or beliefs that have changed. Once you adopt a new view of the world (or any part of it), you immediately lost much of your ability to recall what you used to believe before your mind changed.

Daniel Kahneman, Thinking Fast and Slow

BJP now has a nice explainer video.

At least in this video (if you watch from 3.30) BJP admits that IMF changed its 1.1% GDP projection for India to -4.5% in June (unlike my first quoted tweet from 22 Aug where they conveniently ignored this).

That aside, I think the above video is a fair attempt to explain the GDP numbers and we will have to wait till Nov to see how much we recover in Jul-Aug-Sep.

The worst case scenario would be when GDP growth continues to be low and the cases / deaths continue to rise and get as bad as the other countries. Because then, paise bhi gaye or jaan bhi?

Categories
Covid-times

Making sense of Goa Covid figures – 01 Sep update

Summary – Goa is at 180+ reported Covid deaths presently. From now till October, ~450 more will die and then the overall growth in cases and deaths will likely slow down (but if Goa’s growth behaves like Pune, then God save us).


Some of you may be aware about my weekly India level projections. Now because I live in Goa, I think I will try to look at Goa figures too, if not weekly, then at least once a month (or may be twice if needed). This is the first such analysis.

Total positive cases* tested, stood at 16k+ on Saturday (29 August)

*not to be confused with active cases which are obviously lower, because people who don’t die, recover.

~2,700 positive cases were detected just last week (Sun 23 Aug – Sat 29 Aug), which was 13% higher than the total positive cases detected the week before (~2,400). When we track this week-on-week growth in total weekly positive cases for the past few weeks, we see that there is no clear pattern as such.

8 in this chart is the 23-29 Aug week, 7 = the week before that, and so on and so forth

If I have to make a guesstimate, I would go with a range of 15% to 35% weekly growth in positive cases for the next few weeks (with 25% as my baseline).

Total deaths stood at ~180 till this Saturday (29 Aug) which is 111 per million population (Goa has a population of 1.6 million).

38 people were reported dead from Covid just last week which was 1.5% of half of cases from last week and half of cases from the week before (let me call this the “death %age”)

This death %age was less than 1% eight week ago. It has gradually gone up but seems to be stable at around 1.4-1.5%

So if I consider three cases growth scenarios (A – 15%, B – 25% and C – 35% week-on-week), and use the 1.4% death %age approximation, below is how the future cumulative death toll will look like.

Goa is at close to 200 deaths now – in another three weeks, the figure will touch 300.

Till when will the figure keep rising?

To answer this, let’s divide the death numbers above by Goa’s population (1.6 million) and relook at the same chart.

We are at a little over 100 per million dead in Goa right now (and yes I know death figures are under-reported across the country and all that but I will explain towards the end why it is not so important if what we want to deduce is, when will the growth in cases and deaths end and things begin to stabilize, i.e. when we will peak).

In another three weeks, we will be around 200 per million dead.

Now let’s look at some other cities / states (deaths per million). In the below graph, I have arranged the starting point in a way that all of them begin from a similar time when they had their ~10 deaths per million deaths.

You probably can’t even see Goa in the above graph properly, but don’t worry, I will sort it out for you in a sec. I want you to look at Delhi – notice it sort of flattened at less than 150 deaths per million? Look at Mumbai and Pune now – Mumbai is above 400 and going strong and Pune is doing so much worse – it has crossed 600 and is still so steep (New York peaked after 1200 per million died btw). Also, as a context deaths per million hasn’t even touched double digit in Kerala (I realized this only when I was trying to plot all the graphs together).

Anyway, so to observe Goa more clearly, let’s take out Mumbai and Pune (for now let’s treat them like exceptions / outliers).

Compared to Delhi and Bangalore, Goa’s growth from ~10 deaths per million to 100 deaths per million has been slower.

But moving on, Goa can either behave like Delhi and start to flatten or it could behave like Bangalore and cross 150.

It doesn’t look like it’s going to behave like Delhi (the orange line). As you can see above, Delhi’s growth gradually started slowing from around 75 per million death onward and then started slowing a lot more once it crossed 100.

Goa on the other hand is as steep at 100+ as Delhi was at 50+

So going ahead, if Goa’s growth acts like Mumbai, then flattening could happen only when ~400 per million are dead (i.e. ~640 total deaths in absolute terms). That means over three times more the number of dead already – which if you scroll up and look at the Goa deaths per million projection, will happen some time in October.

Let me put this in perspective.

We are at ~180 deaths now. In Sept and October, ~460 more will die and then the overall growth in cases / death is likely to slow down.

But if Goa’s growth behaves like Pune, then God save us!

And now let me address the under-reporting of deaths bit.

Based on various reports available, in general under-reporting could be as much as half of all deaths (in general for India / various states). So when the reported deaths for Goa are 110, actual deaths could be 220 – nobody can really tell the exact number. And while this information changes the absolute number of deaths – it is irrelevant when one is looking at the growth rate (assuming under-reporting factor is constant week on week).

Even for the peaking analysis – when we are looking at other cities / states, given that all of them are under-reporting, a relative comparison nullifies the impact of under-reporting on our analysis.

Net, net – the under-reporting factor only tells us that the actual deaths are higher. It doesn’t impact when we will peak.

And before I end, let me address one question that some people throw.

Don’t cases matter too, than just the deaths?

There are reports, such as this for India and this for the world, that do indicate that even in the recovered patients, the long term health impact is not clear.

Source

I read these articles and I think that one should be aware of this, but for now it should not be a pressing concern for the general public. Recovery >>> death.

Guess, that’s all I got to say. Feel free to share this with your Goa friends and if you have inputs / questions, let me know.

Categories
Covid-times

India Covid deaths weekly projection – 30 Aug update

FOR THE FIRST TIME, THE TOTAL COVID DEATHS RECORD IN A WEEK (23-29 AUG), HAPPENED TO BE LESS THAN THAT RECORDED THE PREVIOUS WEEK.


My big question every week (since May) is, when will India cross 1 lakh total reported Covid deaths? Total cumulative Covid death toll as of yesterday stands at 62k+ (actual figure could be as high as twice this value, for various reasons documented here).

On an average 965 people reportedly died per day this week. This figure was the same in the previous week too – still as much as five fully loaded A320 airplanes crashing and killing every passenger every day.

One could try forecasting the future deaths by simply using the existing growth rate trend for weekly Covid deaths.

Yes there are many ups and down but if one has to extrapolate, a 5% to 15% week-on-week growth in total weekly deaths seems to be a good guess.

This is how the forecast looks like, for the three scenarios.

We are at 62k+ deaths (as of yesterday – 29 AUG). We will cross 1 lakh by September end.

Alright, let’s now try a slightly more nuanced (albeit indirect) approach to project cumulative deaths. Here we will first look at the week-on-week growth rate of total weekly cases (instead of weekly deaths).

4.9 lakh total positive cases were detected this week, which is ~9% higher than the total cases detected the week before (4.5 lakh).

And yes, we also set a new milestone this week in terms of largest single day recorded case anywhere in the world (but then it doesn’t mean much because what matters is [cases / population] – not the total cases – it will obviously be more for country with such a large population).

For my projection, I will assume a range of 7 to 15% (X) for week-on-week growth of total weekly positive cases. That’s my assumption #1.

Now in general, people who die of Covid in a given week, are either tested positive the same week, or the week before. Do we have some idea of what %age (Y) of such cases die? We do actually.

For the first time, the total deaths in a week is less than the total deaths in the previous week (albeit marginally as of now). 6,756 Covid deaths were recorded this week which is slightly less than 6,758 recorded last week.

Also, 6,756 is basically 1.4% of half of total cases from this week + half of total cases from last week.

In other words, Y for this week is 1.4%. For the future, let’s assume a range from 1% to 1.4%? That’s my assumption #2.

Let’s forecast now…

Let me consider 3 scenarios:

  • X=10%, Y=1.2% (baseline)
  • X=7%, Y=1% (optimistic: slower growth in cases + lesser %ge of deaths)
  • X=15%, Y=1.4% (worse: expecting faster growth in cases)

With the above assumptions, below chart represents the future cumulative death count:

The indirect method tells us, the growth could be slower (at least in one scenario) – and it might be first week of October by when we reach 1 lakh total deaths.

India will cross 1 lakh total deaths by Sep-end to Oct 1st week.

Now, 1 lakh total deaths for India is basically equivalent to 72 deaths per million of the total population (currently we are a little over 45).

To what extent would the death toll figures keep going up – before it flattens / peaks?

If we look at other countries, death toll for many started to flatten out only after anywhere between 400 to 600 per million of their population died!! Scary, I know!

Y axis = no. of days (all the countries are arranged in a way that starting point of 10 deaths per million is common to all)

If we assume that for India, the death toll flattens out even at say 200 deaths per million, that would be equivalent to ~3 lakh total deaths!

It’s difficult to imagine why India would see any less deaths than that. The only populous countries across the globe where death toll flattened at much lower levels (like say Japan and China) happened when they somehow didn’t let the total deaths cross even 5k (Japan for example didn’t even let it cross 1k). We clearly couldn’t control things to that extent in India (most countries haven’t). So now let’s just be hopeful that the total death cap estimate that I am guessing is on the conservative end – otherwise, we could lose even upto 5 lakh people (or 362 deaths per million)!

That’s it for this post. I’ll get back with updated projections next Sunday (06 Sep). Stay safe.