When will the Covid-19 Pandemic End ?

Levitt’s Metric on Indian Data

Bhaskaran Raman, Dept of CSE, IIT Bombay 13 July 2020





Opening Remarks, Disclaimers

●  What you will see is childishly simple (high school math) [and hence blindingly brilliant]

●  It is nothing new: it has been known since Mar/Apr

●  Michael Levitt proposed this metric based on his analysis
of Hubei data, and later for other countries’ data

●  What I have done: simply applied it on Indian data

○ And other countries, for comparison





Levitt’s Metric, Advantages

Metric: H(t) = X(t)/X(t-1), where
X(t) is cumulative count until day t, of (a) Covid deaths, or (b) Covid cases Advantages:

●  Simple to understand, compute: no complex differentials/integrals

●  Independent of population size: can compare across regions: city, state,
district, country

●  Robust to different countries/regions having different testing capacities

●  Robust to varying definitions of what is a Covid death

●  Robust to systematic under-counting (so long as it is time invariant)





Acknowledgment

Data sources: https://covid19india.org, https://worldometer.info, Google for misc information

Ramachandran Shankar (ISRC) for Chennai Twitter data Mahesh Tirumkudulu (ChE, IITB) for feedback and discussion

Shashikant Suryanarayanan (ME, IITB) for perspective, helping keep my sanity, and for feedback

    



Lets plot H(t) as a function of t



Plot of Hubei Data (Levitt’s paper, 14 Mar 2020)

Central observations:

●  H(t) vs t follows a line
of downward slope

●  Approaches 1 after
about 90-100 days

●  Note: H(t) = 1 ⇒ no
new deaths ⇒ pandemic over

Note: Levitt does not explain why this happens, nor do I : future work for modelers, virologists



 

Covid Deaths in the UK

Population = 67 million Covid deaths = 45K DPM = 670
H(t) = 0.001 ⇒ 45 deaths per day = 1/30th of normal non-Covid deaths per day

H(t) hovering between 1.001 and 1.004 in the last week

 

Covid Deaths in Sweden

Population = 10 million Covid deaths = 5.5K DPM = 550
H(t) = 0.001 ⇒ 6 deaths per day = 1/35th of normal non-Covid death rate

H(t) fallen to around this value already

Check: pandemic has “ended” around when the linear fit reaches y=1



 

Covid Deaths in Brazil

H(t) = 1.001 will be reached when x=125, i.e 2 weeks from now

Population = 210 million Covid deaths = 72K DPM = 340
Now lower than UK/Sweden: will likely rise to similar levels eventually

 

Covid Deaths in Italy

H(t) very low for a long time now

Population = 60 million Covid deaths = 35K DPM = 583

Similar to Sweden, UK

Exponential decay may be a better approx., but the overall point holds

Check: pandemic has “ended” around when the linear fit reaches y=1



 

Covid Deaths in Spain

H(t) very low for a long time now

Population = 47 million Covid deaths = 28K DPM = 595

Similar to Sweden, UK, Italy

Exponential decay may be a better approx., but the overall point holds

Check: pandemic has “ended” around when the linear fit reaches y=1



 

Covid Deaths in Germany

H(t) very low for a while now

Population = 83 million Covid deaths = 9K DPM = 110
Need to understand reason behind this…

Check: pandemic has “ended” around when the linear fit reaches y=1



 

Covid Deaths in India

H(t) = 1.001 will be reached when x=185, or another 2 1⁄2 months from now

Population = 1380 million
Covid deaths = 23.7K DPM = 17 so far

 

Covid Deaths in Mumbai

H(t) = 1.001 will be reached when x=83, or another 2 weeks from now

Population = 20 million Covid deaths = 5.3K Will likely reach ~6-6.5K overall; H(t)=0.001 will mean single digit deaths/day

DPM = 265 so far

 

Covid Deaths in Delhi

H(t) = 1.001 will be reached when x=132, or another 2 1⁄2 weeks from now

Population = 20 million Covid deaths = 3.4K Will likely reach ~4.5K overall; H(t)=0.001 will mean single digit deaths/day

DPM = 170 so far

 

Covid Deaths in Chennai (Twitter Data Source)

H(t) = 1.001 will be reached when x=95, or another 10 days from now (noisy data source?, need to recheck)

Population = 7 million Covid deaths = 1.2K

 

Covid Deaths in Chennai

H(t) = 1.001 will be reached when x=97, or about now (noisy data source, need to cleanup and recheck)

Population = 7 million Covid deaths = 1.2K

 

Covid Deaths in Chennai: Day 70 onwards

H(t) = 1.001 will be reached when x=128, or about 1 month from now

Population = 7 million Covid deaths = 1.2K

 

Covid Deaths in Maharashtra

H(t) = 1.001 will be reached when x=164, or another 2 months

 

Covid Deaths in Tamil Nadu

H(t) = 1.001 will be reached when x=276, or another 6 months (we are likely too early in the pandemic, to tell right)

 

Covid Deaths in Kerala

Numbers are too small to make any sense; pandemic pain postponed or pending or prevented?

 

Covid Deaths in Karnataka

Pandemic pain is only just beginning !

 

Covid Deaths in Gujarat

H(t) = 1.001 will be reached when x=114, or about now

Population = 63 million Covid deaths = 2.1K DPM = 33 so far

 

When will the Pandemic End?



2.5 weeks Ended

   

2 weeks

   

2 months

Beginning

Most of India: 2.5 months

2-4 weeks

     

From: https://en.wikipedia.org/wiki/File:India_-_administrative_map.png