On the afternoon of Tuesday, March 24th, 2020, the Prime Minister of India, announced a 21-day nationwide lockdown, which was to be effective from the stroke of midnight of that very same day and was to be continued till the next twenty-one days.
Media link in Hindustan Times:
Media link in Outlook India:
After the first 21-day lock down, more back-to-back consecutive lockdowns were announced by the government.
As the government is stuck in a trade-off between saving lives and saving the economy, researchers at the Department of Management Studies, IIT Delhi have proposed a risk identification and mitigation framework, for the development and implication of an orderly state-by-state lockdown, instead of a complete nationwide lockdown. At specific time, there may not be a better alternative than a complete lockdown to flatten the curve and give the time to the state administration to prepare for the pandemic. While many of the states are still at a very critical stage, it is necessary to examine not only on the basis of total current cases of COVID-19 in states, but at other parameters as well, before planning for Lockdown 4.0. Such other parameters would typically gauge a states preparedness to address the aftermath of the pandemic as and when it spreads as well as the probability to get impacted in the long run.
We have developed a data science driven risk mitigation framework based on proxy measures for severity, likelihood and detectability of the pandemic for each of the state.
To develop their data driven framework, they considered certain factors or pillars upon which the magnitude of COVID-19 infestation in India strongly depends. The researchers considered data from multiple sources such as the count of domestic and foreign incoming travelers, area of the state, average household size, tested-positive rate, tests done per million population, population density, beds per thousand population and the number of testing centres.
The data for these respective fields were collected state-wise and then broadly classified into three main pillars: the ‘severity’, ‘likelihood’ and ‘detectability’ for the disease. From this, they calculated a ‘risk priority number’ for every state and union territory where at least one person has been detected positive for COVID-19. Based on the risk priority numbers of the states, the researchers performed an analysis with an unsupervised machine learning algorithm, dividing the country into 5 clusters of different risk levels.
The results of this grouping of the states are depicted through a heat map. The numbers assigned to each state in the figure depicts the seriousness of risk in their respective clusters. Example: ‘1’ depicting Madhya Pradesh is the riskiest state in the “High-risk cluster” which also has Uttar Pradesh and Rajasthan. Similarly, Maharashtra is the riskiest in the “Medium-high risk cluster”. Within each cluster, based on number of cases within a ward, the government can selectively define red zones, orange zones and green zones.
The findings of their work can be used to implement a step-by-step cluster-wise lockdown of the states, instead of locking down the entire nation at once, disrupting lives and livelihoods. Where the most-risky clusters should still extend lockdown, the other clusters can be progressively opened up for economic activities for a short span of time. For doing so a more holistic data driven approach should be taken and this should not only be driven by the absolute number of active cases.
This approach will not only let the economy die entirely but will give small and large businesses time to plan for contingencies. It would also give time to stranded people in other states to return back to their comfort zones, thus less expenditure by the government to bring them back later on government expenses. The framework can not only be used in real-time by substituting latest data for future lockdowns in India but can also be used on a state level to cluster districts and an international level to cluster countries.