Category: UK

COVID-19 India Mortality is Not Low …

Summary

The RED line for India is proof that the deaths in India due to COVID-19 are at par with global trends for other countries. Taking the day-to-day data for cumulative Deaths per 1000 InFections (measured real data), India is showing a trend that is higher than that for Germany, South Korea, Australia, Canada, China and the US. It is lower compared to France, UK, Spain and Italy. So, it would be wrong to conclude that India has lower deaths than other countries due to inbuilt immunity, BCG vaccine or geographic temperature.

Background

I have brilliant friends around the world who have been thinking about the impact of COVID-19 on India. Initially, we debated on the low number of inFections in India. Eventually we all agreed that the published data on inFections in India is incorrect, since the number of tests/million is very low. No dispute here.

However, questions started pouring in that if the detected inFections are low in India, the data should at least show up in the death statistics. The logic was that there is bound to be lot more undetected inFections in the country and some of these folks would get really sick and would have to eventually reach a hospital for care. Sadly, some of them would succumb and this should show up in the mortality data.

As on 11 Apr 2020, the published data from JHUM (John Hopkins University and Medicine) for India indicated 288 deaths. This is a rather low number for the second most populous country in the world.

We examined many theories floating around. It was speculated that the Indian people had fantastic immunity. Some speculated that Indian’s had resistance because of the BCG vaccine, given for tuberculosis¬†(TB), since it is part of the childhood immunization program. Others felt that it was because of the higher temperature in India, especially due to the onset of summer.

Realization

One needs to examine data carefully before drawing conclusions.

As on date (11 Apr 2020) it is tempting to compare the 288 deaths in India (1366 million people) with that in other countries. For example, Italy (61 million people) has 19,648 deaths. But, keep in mind that the exponential growth of inFections in India stared on 22 Mar 2020 (Day-1 for India), while in Italy it started on 26 Feb 2020 (Day-1 for Italy). We are off by a month, if we just look at the data based on the calendar. I have found that when inFections reach about 400 in any country, the exponential growth phase begins. So, taking that point as Day-1, we can transform all the country data to a common starting origin. We can then begin to compare for an equal time lapse.

In addition, we have to remove errors in sampling. An effective way to examine the data is to calculate the day-to-day metric of “measured deaths per 1000 inFections”. By doing this we can remove some of the errors in sampling. For example, if we claim that the number of people tested is not sufficient, then there is a sampling error in the inFection data. But, if we take the ratio of such inFection data and match it with measured deaths, the errors would mathematically cancel out (at least to a large extent). Implicit in this argument is the fact that measured deaths will arise from measured inFections.

Therefore, the graph at the very top, is quite stunning in its result. It visually shows that India is following the death trend for other countries. It is somewhere in the median position. Its trend is actually higher than that for Germany, South Korea, Australia, Canada, China and the US. It is lower compared to France, UK, Spain and Italy. Variations in the relative positions in the graph can be attributed to the accessibility and quality of health care in a given country.

The graph also reveals another startling fact. It is telling us that the death/inFection ratio can skyrocket all of a sudden. For example, France and India have similar datum values until Day-15. Then France suddenly shoots up. Why is this? I attribute it to the inability of the medical care to cope up with the critical cases.

India has data for 21 days during its exponential growth (22 Mar 2020 to 11 Apr 2020). I have computed the average for the first 21 days for all the countries that I have analyzed. India has an average ratio of 26.37 compared to 22.87, which is the average for all the countries. The graph below shows this visually.

COVID-19 India: 2.4 Million Infections by 25 Apr 2020 (Mathematical Prediction)

Summary – Predicted Infections in India …

22 Mar 2020
= Day-1
Start of rapid InFection growth -exponential curve.
Based on Italian growth model for InFections.
20 Apr 2020
= Day-30
InfLection point
Slope of growth curve begins to decrease.
Slowdown in InFection
04 May 2020
= Day-45
Start of assymptote (based on trends)
InFection new cases begin to rapidly decline.
20 May 2020
= Day-60
Expect a true asymptote. Rapid decline in new InFections.
Rate of new InFections tending to zero.
Very small number of new cases each day.
DayDateInfections (Predicted)
Day-122 Mar 2020
Day-1031 Mar 2020 100,196
Day-1505 Apr 2020 272,945
Day-2010 Apr 2020 635,209
Day-2515 Apr 20201,180,269
Day-3020 Apr 20201,824,431
Day-3525 Apr 20202,407,025

The above are Predictions are based on the Italian infection raw data model. For their validity the standards of Lockdown and Social Distancing in India should be at par with that in Italy. The above Predictions would reveal in the testing data for India, provided India does a minimum 1500+ Tests/Million of population. There are also people with infections who never get tested and as such the true infection could be higher than predicted values by a factor of 5 to 10. For example, in Italy the true infections are projected to be 10 times higher than the published tested data.

A previous article has discussed the modeling methodology based on the Population Ratio (PR), Model Factor (MF) and taking Italy as the bench mark reference point. It is also shown there that the model works for almost all the countries evaluated, but not for India. In my first article on COVID-19, I had described a model based on simple scaling with PR. In this article, this idea is further refined to make a prediction of infections from 22 Mar 2020 to 25 Apr 2020.

Published India Infection Data – Paints a Faulty Picture

The published India infection data from the John Hopkins University (JHU) data set is faulty. JHU sources its data from the Government of India portals. The problem with this data is that it is based on 32 Tests/Million people in India. Comparing with other developed countries, the quantum of these tests are so low, that it does not reflect a representative sample of the population. So, the existing data for India, at least from 22 Mar 2020 to 31 Mar 2020, is useless to start building a model for prediction.

CountryTests/Million
India32
Italy8,405
France1,508
Spain7,596
Germany5,812
China2,820
South Korea7,940
Australia9,670
UK2,120
USA3,377
Canada6,450

Italian Infection Model is a Good Benchmark

The infection raw data for Italy, based on my analysis in previous articles, is a good benchmark for predicting the infection trends in other countries. I have found that once the infections reach a value of 400, the trajectory follows an exponential curve. The corresponding date is taken as Day-1 and all countries can start at the same point on a common graph. The analysis has also shown that the exponential curve for Italy can be scaled with PR and MF to model the infection grown curves for other countries.

This graph contains the real day-to-day data of COVID-19 infections for several countries. The PR ratio is shown in the legend. The idea is that if PR is more than 1 the curve should fall below Italy (blue line) and if it is less than 1 it should fall below Italy. US and Germany show this trend (PR > 1). Canada and South Korea are way below (PR < 1). England and France are close to Italy (PR close to 1). China is an anomaly, since it is hovering close to Italy, though its PR is 23.68! Thats because China’s data is tainted. India’s data is also tainted, since it is falling way below the curve for Italy. In conclusion the Italian curve is a good bench mark.

It must be noted that to bank on this model, the remedial measures such as lockdown and quarantine norms must be similar in all the countries. Otherwise there will be variations. For this reason, Spain is higher than it should be for its PR value. It is likely that Spain’s lockdown was not as effective as that of Italy.

Predicting the Infections based on Italian Model

Based on the above reasoning, using the Italian curve, other countries can be modeled by multiplying the Italian infection raw data with the PR.

First, a polynomial was established to fit the Italian infection raw data.

Y(x) = -0.2274 x^4 + 15.521 x^3 – 217.81 x^2 + 1451.4 x – 1540

Here, Y(x) is the Predicted Infection on day ‘x’, where ‘x’ is the same as time ‘t’ in days, with x = 1 and incrementing by 1 for each day.

Multiplying this by PR the individual curves for other countries are obtained.

The graph is not detailed for most of the countries, since the scale is enlarged because of China and India. Removing these, we get a much clearer graph for the other countries.