COVID-19 in India: Does Density Really Matter?

As we try to understand the factors that contribute to the spread of COVID-19, references to Density are found most frequently. The clustering of people makes us vulnerable to infectious diseases. There is plenty of literature citing examples of New York London, Singapore, Tokyo etc. supporting this line of thinking. 

I was curious to see how density was playing out in India. We don’t have COVID-19 figures for urban and rural areas separately—they are for districts that typically include urban, suburban, and rural areas. I culled the 2011 census figures and was able to map the density for 592 districts, shown in the image below (left). Interestingly only 10 of these locations have a density of more than 25 people per hectare (or 2500 people per sqkm), shown in dark red on the map. 369 have less than 5 per hectare, shown in two shades of green on the map, and the rest lie in between. Predominantly, you can see, the map is green, except for the Indo-Gangetic plane which is densely packed and with a few other such pockets scattered around the country. The map on the right shows the distribution of incidence-frequency of COVID-19 cases across these locations. The reds show higher frequency than the greens; and yellow somewhere in between.

Believing the ‘density theory’, I was expecting to see some congruence between the two maps—hotspots (reds) and cold-spots (greens) in somewhat the same areas. Surprisingly this was not the case. You can see incidence hotspots are scattered all over, particularly over the west-central region.

 Just to confirm this visual finding, I ran a quick linear regression. The result was the same. No significant correlation between density and incidence. See below the scatter plot which shows R2 (R squared or Coefficient of Correlation) to be 0.13. For a significant correlation this value has to be as close to 1 as possible, the closer it is to 1 the stronger the correlation is.

This discovery was a bit of an anticlimax for me. Such correlations (between any two variables) can be used as predictive indicators.

While I don’t believe density doesn’t have a role to play, this didn’t come through loud and clear here. This happened for at least two reasons.

One, these districts are not uniformly populated—there are dense pockets within the city (such as Dharavi in Mumbai has 2000 people living per hectare while Mumbai's density is 64). Such pockets do not find expression within the COVID-19 data published by the government. At any rate, density generalised at the district level will almost always be less than that at urban agglomeration level. It would be interesting to visualise data at that finer level if it becomes available. Secondly, other factors also play a role in the spread, not just the density.

I wondered how we could devise another indicator that could predict the future spread. One simple possibility is an indicator that takes into account not only density but also the frequency of incidence. I multiplied the two and called the Proliferation Factor! See how it shows on the map below. 

Expectedly it mimics both density and incidence-frequency maps. The R2 value for this is almost three times that of the earlier one, though still not statistically significant. I strongly believe this could be a fair estimate of which districts the virus is heading to as India moves to reach the peak.

Currently, the incidence-frequency of COVID-19 cases across the top 30 locations in descending order is given in the table below:

Assuming consistency in policy and ground-situation the future scenario might emerge as shown in the following table

I am not presenting an academically tested theory here, just a statistical view, a hypothesis at best. What do you think about this? Please log in your comment below. I hope I am proven wrong here, and not only those mentioned here, but no other location gets much more affected, and the world is able to contain the uncontrolled spread of the virus soon.

While I am working on the next article please do send me your comments, suggestions, and questions/issues you’d like addressed here through ‘Comments’ below or Twitter (@UrbanVarshney).

Credits: I have taken COVID-19 data drawn from the Grain Mart India Website. Ishaan Varshney helped with extracting the data so I could use it. Thanks to both.