COVID-19 District Vulnerability Index
While India is aggresively testing, tracing and isolating, evidence from other countries has shown that senior citizens and citizens with existing conditions like cardiovascular diseases, diabetes, respiratory diseases, hypertension, etc are at the highest risk.
What does this mean to India?
India did a general lockdown and as it responds strongly, it is also rigorously monitoring cases, deaths, interventions and activities in it’s 718 districts.
It is time we leverage the massive public data available to us, learn from the evidence from other countries and identify the citizens who are at risk and separate them from the rest of population since a lot of them can be asymptomatic carriers of the COVID-19.
Therefore, I have attempted to make a COVID-19 District Vulnerability Index which factors six crucial parameters and can give policy-makers a sense on which are the districts which are the most vulnerable to COVID-19 and where should the state target it’s resources and accordingly build a data driven strategy.
Please note that I am not an epidemiologist, these are not predictions or projections but this is an index which is an interpretation of the public data and the weightages are based on my assumptions/evidence from other countries.
The map shown has some errors due to limitations on Tableau plus the data is a mix of 2011, 2015–16 and it might exclude some area where Census was not taken. The area marked in grey which is Ladakh is an integral territory of India which is not colored due to limitations in my shapefile.
1. Age Risk
This visualizes percentage of population in a district which is greater than 60 years of age.
Top 5 districts which are at the highest age risk are: Pattanamtitta (Kerala), Sindhudurg (Maharashtra), Kottayam (Kerala), Alappuzha (Kerala) and Ratnagiri (Maharashtra)
Bottom 5 districts which are at the lowest age risk and have a lot of young population are: Papum Pare (Arunachal Pradesh), Kurung Kumey (Arunachal Pradesh), Upper Subansiri (Arunachal Pradesh), East Kameng (Arunachal Pradesh) and Lower Subansiri (Arunachal Pradesh)
2. Hypertension Risk
This visualizes percentage of adults which have very high blood pressure.
Top 5 districts which are at the hypertension risk are: West Siang (Arunachal Pradesh), Tawang (Arunachal Pradesh), East Siang (Arunachal Pradesh), Anjaw (Arunachal Pradesh) and Mokokchung (Nagaland)
The bottom 5 districts which are at the lowest hypertension risk are: Mirzapur (Uttar Pradesh), Tehri Garhwal (Uttarakhand), Kottayam (Kerala), Lalitpur (Uttar Pradesh) and Bhind (Madhya Pradesh)
3. High blood sugar (Diabetes) Risk
This visualizes the percentage of adults with high blood sugar level i.e. which have hyperglycemia and Diabetes.
Top 5 districts which are at the highest high blood sugar risk are: Wayanad (Kerala), Kolkata (West Bengal), Guntur (Andhra Pradesh), Cuddapah (Andhra Pradesh) and Puri (Odisha)
The bottom 5 districts which are at the lowest high blood sugar risk are: Kargil (Jammu and Kashmir), Ramban (Jammu and Kashmir). Auraiya (Uttar Pradesh), Lahaul and Spiti (Himachal Pradesh) and Nandurbar (Maharashtra)
4. High BMI (Obesity) Risk
This visualizes the percentage of adults with high BMI i.e. which are overweight and obese.
Top 5 districts which are at the highest obesity risk are: Krishna (Andhra Pradesh), Guntur (Andhra Pradesh), Kolkata (West Bengal), Hyderabad (Telangana), Mahe (Puducherry)
The bottom 5 districts which are at the lowest obesity risk are: Simdega (Jharkhand), Dantewada (Chattisgarh), Narayanpur (Chhattisgarh), The Dangs (Gujarat) and Dindori (Madhya Pradesh)
5. ARI (Acute Respiratory Infections) Risk
This visualizes the percentage of children under 5 years with prevalent symptoms of Acute Respiratory Infections in the last 2 weeks preceeding the NFHS survey.
Evidence suggests children below 5 years are less vulnerable and this is an indication of high prevalent ARI infections. This might evolve and might be re-considered in the risk index.
The top 5 districts which are at the highest ARI risk are: Ramban (Jammu and Kashmir), Kishtwar (Jammu and Kashmir), South Garo Hills (Meghalaya), West Garo Hills (Meghalaya) and Kannauj (Uttar Pradesh)
The bottom 5 districts which are at the lowest obesity risk are: Sheopur (Madhya Pradesh), Longleng (Nagaland), Hailakandi (Assam), Karimganj (Assam), Bolangir (Odisha)
6. Current status factor
Along with considering the above risk factors, considering there are still a lot of factors which affect the spread and vulnerability, have considered a current status factor which gives us what is the situation today — accounting a lot of random events and other possible factors.
This visualizes the 4 types of districts according to the cases and spread.
District Vulnerability Index
The index which is attempted here factors in the above parameters, weightages are given appropriately (more weightage for co-morbidity and age) and we get a composite index number. This visualizes the District Vulnerability Index.
According to the index,
The top 5 districts which are highly vulnerable are: Coimbatore (Tamil Nadu), Nagapattinam (Tamil Nadu), Jammu (Jammu and Kashmir), Erode (Tamil Nadu) and Sangli (Maharashtra)
The bottom 5 districts which are least vulnerable according to the above index are: Datia (Madhya Pradesh), Ukhrul (Manipur), West Jaintia Hills (Meghalaya), Bhind (Madhya Pradesh) and Sirohi (Rajasthan)
This ofcourse is not an accurate projection or prediction but a measure and interpretation of risk according to our public data. It does not take into account tons of other parameters like travel, religious events, policy changes, etc and it may not reflect the reality on the ground today
Other important visualisations:
Here is another analysis which I found interesting. This visualizes % of households with any usual member covered by a health scheme or health insurance from NFHS 4 (2015–16). Things have changed drastically post Ayushman Bharat and I hope the insurance cover is changed but a worthwhile look at the data from 2015–16.
Districts like Bijapur (Karnataka), East Siang (Arunachal), Villupuram (Tamil Nadu), Dhamtari (Chattisgarh) and Srikakulam (Andhra Pradesh) have more than 82% households with any member covered in an insurance/scheme.
Baramula (Jammu and Kashmir), Senapati (Manipur) and Shopian (Jammu and Kashmir) have less than 1% households with any member covered in an insurance/scheme.
Population density also plays a big role in the spread of COVID-19. Along with Population Density, members per household also play a crucial role since following social distancing is very difficult.
Districts which are densely populated are also on the risk such as North East, Central and East Delhi, Chennai (Tamil Nadu), Kolkata (West Bengal) and Suburban Mumbai (Maharashtra)
Districts which are least dense are Deoria (Uttar Pradesh), Azamgarh (Uttar Pradesh), Upper Dibang Valley (Arunachal Pradesh), Lahaul and Spiti (Himachal Pradesh) and Leh (Ladakh)
Here, I have obtained the data on District-wise health infrastructure as on 31st March, 2017 from data.gov.in. Though the ideal measure should be health facilities per thousand population, due to some errors, I have not been able to analyse that.
Number of CHCs:
Districts like Alwar (Rajasthan), Purba Bardhaman (West Bengal), Jaipur (Rajasthan), South 24 Parganas (West Bengal) and Sikar (Rajasthan) have more than 30 CHCs in their respective districts.
Districts like Dharwad (Karnataka), Sheohar (Bihar), Arwal (Bihar), West District (Sikkim) and Saiha (Bihar) have 0 CHCs in their respective districts.
Number of PHCs
Districts like Belagavi (Karnataka), Tumakuru (Karnataka), Mysuru (Karnataka), Chennai (Tamil Nadu) and Hassan (Karnataka) have more than 130 PHCs in their respective districts.
While, districts like Yanam (Puducherry), New Delhi, Central Delhi, East Delhi and North Delhi have 0 PHCs in their respective districts. However, most districts of Delhi have more number of District level Hospitals and now the Mohalla clinics!
Number of Sub-Centres
Districts like South 24 Parganas (West Bengal) West Medinipur (West Bengal) Nagaur (Rajasthan), East Godavari (Andhra Pradesh) and Murshidabad (West Bengal) have more than 800 Sub-centres in their respective districts.
While, districts like Central Delhi, East Delhi, North Delhi, North East Delhi and North West Delhi have 0 sub-centres in the districts. Again as mentioned earlier, Delhi has more number of district hospitals and now the Mohalla clinics!
Interestingly, districts from Tamil Nadu i.e. Sivaganga, Tirunelveli, Pudukkottai, Thanjavur and Coimbatore have more than 12 sub-divisional hospitals in the respective districts
While, 313 districts in India have 0 sub-divisional hospitals. Sub-divisional hospitals form a crucial link between PHC/CHC and District Hospitals and are at a block level
Now that we have an estimation of the districts which might be vulnerable, how should we move ahead? I highlight a three point strategy which can be effective
- Identify the most vulnerable through targeted surveys and district level data collection: The Government has multiple online citizen channels it can leverage like Aarogya Setu and MyGov plus it has the district level administration which can effectively undertake a quick data collection excercise to identify people who are/have a. greater than 60–65 years of age and b. existing conditions like hypertension, diabetes, HIV, cancer, COPD, etc
- Separate the vulnerable from the rest of population: Once we have mapped the vulnerable population, post the general lockdown, we need to separate them from the rest of the population in shelter homes, hostels, facilities, etc. The fiscal burden of this needs to be calculated and carefully evaluated.
- Lift the lockdown with exemptions in hotspots and areas where the vulnerable population is kept: Let the rest of the population start working and develop herd immunity. When the cases start decreasing and the rest of the population have developed immunity, we can hopefully integrate them again with the vulnerable population.
Ofcourse these suggestions needs to be very carefully evaluated and considered but in my opinion, we should atleast know where the vulnerable population is and start getting ready to protect them first.
Would like to thank @Bhanu Prasad with his inputs and @Farhan Yusuf for sharing the NFHS data!
Open to other thoughts/ideas/critiques. Feel free to DM me or reach out on firstname.lastname@example.org
References: https://ourworldindata.org/coronavirus , Census 2011, NFHS 4 (2015–16), Open Data Government Platform (OGD)