Mapping Healthcare Units in India

I think India is still a sleeping giant and we might encounter a large number of cases in the next few weeks. (Better be prepared than late)

Given how India is placed in the world (counting for 1/5th of the world population), it will be helpful to map the number of people to the nearest healthcare facilities available.

What do I mean by healthcare facilities data?

  • Hospital Coordinates
  • Number of beds
  • Number of Isolation wards
  • Number of doctors
  • Number of nurses
  • Number of ICUs
  • etc

I searched for a while & came across this dataset from This has all the required columns but values are empty, we would need something very similar to this but more complete.

OSM worked on a similar initiative & it is detailed here. (need to evaluate the output of the process)

In India, the population is mostly clustered in big cities & as we proceed towards lockdown, people start moving back to their roots, this will increase the chances of spread and 2nd/3rd tier cities have to be better prepared for this. There is some research by on mapping population movement & connectivity using cellphone data.

What needs to be done?

  • Populate the healthcare sheet using crowdsource or other innovative approaches (Satellite Imagery of hospitals from geo coordinates & then estimate the capacity based on the size of the building)
  • Understanding the current population distribution & how they are migrating because of the COVID-19 pandemic
  • Helping govt. authorities to better plan out the strategy in their areas based on the curated datasets

Possible next steps

  • Look at the population distribution data in India. Sources: CIESIN, WorldPop
  • Understand the population movement using the techniques mentioned above
  • Populate the sheet provided by NIN

I would like to hear how this can be achieved effectively form this community.


Hi @srmsoumya,

Great idea and much needed to map the healthcare facilities and get crucial data about care capacity (especially ICU-level care) where this info doesn’t exist or can’t be readily accessed.

Have you looked at:

For population distribution and demographic estimates, this may be another good source (it’s CIESIN’s work refined with ML by Facebook):

We’re working on exactly this mapping of healthcare facilities and understanding of their capacity to care for COVID19 patients over time in the USA at the moment. While some of the data will be unique to USA, I think much of our thinking and methodology may be relevant to your similar work in India: Mapping US health system capacity (ICU care in particular) for COVID19 surge preparedness

Thanks @daveluo. I see you have done some great ground work here. It will be good to replicate this in Indian context, the bottleneck might be availability of the datasets.

Also, great to see Rob in the gitter chat (big fan).

1 Like

Hi @srmsoumya,

Regarding population movement, people who work in IT companies form one of the largest demographics of people who will move to their hometowns. One (definitely not trivial) way to get this information is to obtain:

  1. A list of companies in each city who implemented Work from Home,
  2. Approximate number of employees who are working from home and
  3. somehow get those employees to anonymously post the location of the hometown where they’re heading.

This can help in alerting the govt authorities of the places where the employees are heading towards.

Hi @srmsoumya,

I have been also thinking that there are couple of major things that we need to start looking into on a immediate basis and on a long term basis. I have just scribbled my thoughts here.

On the immediate basis:

1. 	Coronavirus affected areas:

	Maintaining a database mapping each new patient on a location: district, state, city, co-ordinate
	Mapping the location's population distribution
	Regular updates on if people are following the bans
	Creating touch points and local networks for faster reaching out to local affected regions
	Flights in-out of affected areas

2. Enabling healthcare:

	Creating a healthcare units mapping (more detailed, better):
		Hospital Name
		Hospital Coordinates/Other location details
		# of Beds / Isolation wards
		# of ICUs / Ventilators
	Both already occupied / available
	Plan for times when healthcare starts to get excessive pressure
		More ventillators (Spain prototype)
		Private isolation wards
		Masks / protection suits
	Maintaining a database for health workers

3. Increasing virus testing

	Labs location for testing
	Number of samples being brought in on a daily basis and their demographics
	Testing kits availability, location
	Alternate testing methods for parallel testing
		molecular recognition (PCR): 24-48 hours response time
		serology testing: Blood sample (80% accurate)
		AI based model based on X-rays for faster testing

Long term:

Upcoming recession
Help for people losing jobs / pay
Enabling online work

I started looking into the first part about affected areas yesterday and could gather somethings (link below): link

Useful Links:

  1. Slack channel for working on CORD-19 dataset.
  2. COVID-19 analytics for various countries (we can build one for India).
  3. COVID-19 India Dataset on Kaggle

Hi @sg1791 @svdesai

I have reached out to a few friends who work with the government to get some data on the hospital capacity in India.

To get started, we can start looking at the existing data & build on top of it. Shall we create a channel for communication & Github repository to start pulling the data together?

Hi @srmsoumya,

Absolutely, let’s work together. How about slack?

yes, I agree. Works better than any other channel and good if we scale and get more people on board.

I have created a slack group:

This is a great initiative!

I’ve summarized some of the data around the healthcare capacity in India, trying to understand at which point it will be overburdened by COVID-19, and how likely that scenario is. In the references, I have also linked to the National Health Profile of India, which provides information about hospitals in more details. Hope this helps!

@aakashns also consider looking at the data here, see if this can be of any help.