India COVID-19 SEIR Model Now Open-Sourced on GitHub

The iSPIRT India COVID-19 SEIR Model is now open-sourced on GitHub: https://github.com/seirforindia/seirdistrictmodel

Epidemiologists, Data Science & Modelling community — looking forward to your support to ​ex​plore, ad​opt, adapt, and en​hance this India COVID-19 modelling effort.

If you want to contribute or have any queries please reach out to c​ommunity@ispirt.in

India COVID-19 SEIR Model: A Technical Introduction

#FlattenTheCurve

When India locked down back in March, there was just one question on everyone’s mind: How can we flatten the curve? 

Chiseled for ~90 days, the India COVID-19 SEIR model, which went live mid June, 2020, is the digital public good to meet this purpose. It pulls publicly available data to predict infection trends across all of India’s districts over the next 15-30 days.

Creating the India COVID-19 SEIR Model

The India COVID-19 SEIR Model is developed on the foundations of SEIR, a mathematical model popular in the study of infectious diseases. This model labels each individual in the given population as either Susceptible, Exposed, Infectious, or Recovered. Typically the entire population starts from the Susceptible state and then moves from Exposed to Infectious to Recovered.  Each individual is assumed to be equally susceptible to the disease and equally likely to recover. The model assumes a closed population i.e. births and migrants are ignored. This implies that the number of individuals in the Susceptible bucket is highest at start of the simulation, and can only decrease thereafter.

The India COVID-19 SEIR Model is a SEIR Model with modifications that increase granularity. The crucial difference between the preexisting SEIR Model and the current one is this: the modified model categorises the population by granular districts and age. Given India’s large population of young people, this is critical. Moreover, this modification allows the model to be run on a district-level (e.g. Mumbai, Chennai, Bangalore), grouping populations with similar behavior patterns and environmental circumstances while also catering to district-specific on-ground interventions. Simulations run on the district-levels are then aggregated to provide metrics for the entire population i.e. India.

The model is optimised — a mathematical optimiser ensures that the prediction curves adapt to fresh data each day. So, the model self-corrects every night when the most recent statistics flow in.

What the data really means

Like any predictive model, the curves on the India COVID-19 SEIR Model will undergo constant flux. The absolute numbers are less important than the trend. The model’s power lies in its ability to predict early trends, identify potential hotspots in the next 15-30 days, and plan for resource surge, especially in terms of hospital beds, PPE, medical equipment, ventilators, and emergency supplies.

The model’s power lies in each one of us. Not sure how to get started? Check our blogs for Use Cases and use the India COVID-19 SEIR Model!

Pass the word around. #FlattenTheCurve.

About the contributors: The blog post is co-authored by our volunteers Nikhila Natarajan and Yashvi Jaju.

Bending India’s COVID19 Curve Through Science & Data-Led Models

Powered by data-led scientific rigor, the India COVID-19 SEIR Model delivers early infection trends for every district in India. The model is geared to help Indians from all walks of life plan life and work decisions around their region’s projected trends over the next 15-30 days. Hospitals can use the model to plan for a surge in demand for resources (beds, ICUs, ventilators); local and national level leaders across private and public sectors can use the model to decide how best to contain the spread of the disease and re-open safely. Epidemiologists can use the model to define how different behavioral and environmental factors affect disease transmission. We introduce 3 use cases in this blog post—the first in a series aimed at promoting scientific and modeling capability. 

Wherever the Coronavirus curve has bent to our will, it has happened on the back of behaviour changes based on data-led insights. Everywhere, simple shifts in behavior—staying at home, wearing masks, sanitizing hands—have been informed by predictive models that showed us the mirror to a dystopian future if we didn’t edit our lifestyles. As a digital public good for a billion Indians, the value of the India COVID-19 SEIR Model lies in its reach and widespread use. 

Until a vaccine is developed, we have to make sense of today’s numbers in the context of all our tomorrows. Individuals, policy makers—and everyone in between—can make smarter decisions if they know the evolving shape of the outbreak, and the India COVID-19 SEIR Model aims to do just that by enabling identification of potential trends and patterns in the next 15-30 days. 

The approach taken by the model provides flexibility and utilisation from both a view of trends as core model adoption / enhancement. We can all use it to bend India’s curve. That’s the ultimate use case, really — where the model tells us where it’s going and we, in turn, steer it in an entirely other direction. Models will change and that’s a good thing. It means we are responding. The power of models and data science in this particular moment is the ability to assist a very scientific approach to scenario planning during an ongoing pandemic. We can turn the course of this pandemic and transform what this model tells us, every 24 hours. We are already watching the shape-shifting in real time. It’s in your hands. Go on, try it. 

Use Cases:

About the contributors: The blog post is co-authored by our volunteers Nikhila Natarajan, Srikar V Cintalagiri, and Yashvi Jaju.

Close Bitnami banner
Bitnami