An interactive platform of district-level COVID risk projections for India
The second-wave of Covid-19 caught most of India — both governments and citizens — thoroughly unprepared. Several commentators have pointed out that lack of preparation, slack amongst government and citizens in pandemic precaution, and a complacent attitude toward evidence exacerbated the tragedy.
The second wave may be waning in Mumbai, and test positivity rates may be declining in Delhi, but this wave is far from over. Can we do better in districts that have not been through the worst yet? Can we buy some time? Can we allocate resources (oxygen, PPE, funds, medical supplies) based on careful calculation of which district is likely to suffer the most next? The crisis has already hit. It’s definitely too late to wish it away. There is still time, however, to make sure the ongoing response is grounded in as much evidence as possible.
In a conversation with Milan Vaishnav on “The Grand Tamasha”, Anup Malani likens the pandemic to a war, except the enemy is a virus and not another country. As in war, our response to the enemy should be based on intelligence of where the enemy is likely to attack next. To this end, a team of researchers and data scientists have collaborated to create an interactive platform illustrating real-time COVID risk projections at the district level. The map below illustrates the pandemic growth rate at the district level. What is the pandemic growth rate? Read on.
The map above illustrates the variation in the Rt: the COVID reproduction rate. For a given infectious disease, the reproductive rate (Rt) is the number of additional cases in a community that a single person creates. When Rt is less than 1, the number of new cases per day will be falling over time. Conversely, when the reproductive number exceeds 1, the number of people infected grows exponentially.
The graph above plots the predicted reproduction rate for two adjacent districts — Buxar (Bihar) and Ghazipur (Uttar Pradesh) — over time. Close attention to the variation in the reproduction rate over time and across space can give us a clearer sense of the geographic trajectory of the second wave of the pandemic. An adaptive pandemic response policy approach would use the estimated Rt at the local level and make district-wise decisions about lockdown policies and allocation of medical resources.
As of last week, Mumbai, Thane, Raipur, and South Goa were among the 56 districts with a reproduction rate below 1. On the other end of the spectrum, the graph below highlights the ten districts with the highest predicted reproduction rate calculated using available data. The analysis suggests that risk of COVID-19 remains particularly high in districts in Karnataka, Assam, Orissa, and Haryana.
A caveat is that the district-wise calculations on the portal are derived from official reports. While this is an imperfect solution, it is quick. It could also serve as an incentive for districts to test more if aid is linked to district-level projected COVID risk.
The underlying dataset of the portal is open-access and has information on total cases, deaths, estimated reproductive rate, total clinics and hospitals at the district level. Our hope is that residents of high-risk district will adjust behavior if their area has a precariously increasing reproduction rate over time. Even better if aid and medical support that many organizations are mobilizing at an impressive pace could be allocated based on district-level evidence. District-level bureaucrats can incorporate this additional information in planning their pandemic response (most of us have read about the striking example of what the District Collector of Nandurbar was able to achieve to prepare against the second wave). Finally, central and state governments could tailor their pandemic response given the obvious paucity of resources and time based on district-level risk estimates.
Overall, knowing where the virus will strike next can help save lives — by guiding behavior change, local public health measures, and allocation of scarce resources.
For details on the methods for estimation of the predicted reproduction rate, please get in touch on the e-mail provided on the portal.
— Anup Malani, Satej Soman, Sabareesh Ramachandran, Ruchir Agarwal, Sam Asher, Tobias Lunt, Paul Novosad, and Aditi Bhowmick