Modelling COVID-19-III: endemic spread in India

TL;DR

This paper introduces a novel spatio-temporal gravity model to analyze the endemic spread of COVID-19 in India, leveraging extraneous covariates like air traffic to improve modeling accuracy at various spatial scales.

Contribution

It proposes a new gravity model within a generalized linear model framework that incorporates extraneous covariates, outperforming existing models in capturing COVID-19 endemic behavior in India.

Findings

The gravity model provides consistent estimators.

It outperforms traditional models on Indian COVID-19 data.

The model effectively captures local mobility and social interactions.

Abstract

A disease in a given population is termed endemic when it exhibits a steady prevalence. We address the pertinent question as to what extent COVID-19 has turned endemic in India. There are several existing models for studying endemic behaviour, such as the extensions of the traditional temporal SIR model or the spatio-temporal endemic-epidemic model of Held et al. (2005) and its extensions. We propose a "spatio-temporal Gravity model" in a state of the art generalised linear model set up that can be deployed at various spatial resolutions. In absence of routine and quality covariates in the context of COVID-19 at finer spatial scales, we make use of extraneous covariates like air-traffic passenger count that enables us to capture the local mobility and social interactions effectively. This makes the proposed model different from the existing models. The proposed gravity model not only produces consistent estimators, but also outperforms the other models when applied to Indian COVID-19 data.