Predicting COVID-19 positivity and hospitalization with multi-scale graph neural networks

Abstract

The pandemic of COVID-19 is undoubtedly one of the biggest challenges for modern healthcare. In order to analyze the spatio-temporal aspects of the spread of COVID-19, technology has helped us to track, identify and store information regarding positivity and hospitalization, across different levels of municipal entities. In this work, we present a method for predicting the number of positive and hospitalized cases via a novel multi-scale graph neural network, integrating information from fine-scale geographical zones of a few thousand inhabitants. By leveraging population mobility data and other features, the model utilizes message passing to model interaction between areas. Our proposed model manages to outperform baselines and deep learning models, presenting low errors in both prediction tasks. We specifically point out the importance of our contribution in predicting hospitalization since hospitals became critical infrastructure during the pandemic. To the best of our knowledge, this is the first work to exploit high-resolution spatio-temporal data in a multi-scale manner, incorporating additional knowledge, such as vaccination rates and population mobility data. We believe that our method may improve future estimations of positivity and hospitalization, which is crucial for healthcare planning.

Advances in the Development of Representation Learning and Its Innovations against COVID-19

“…Li et al [161] proposed a prediction model that combines the lioness optimization algorithm with the graph convolutional network, which can capture spatiotemporal information from feature data to achieve accurate predictions of COVID-19 case numbers. Skianis et al [162] developed a multi-scale graph model utilizing demographic data, medical facilities, socio-economic indicators, and other related information to improve the prediction accuracy of COVID-19 positive cases and hospitalizations. The model is capable of automatically learning the interrelationships between different features, thereby enhancing its predictive capability.…”

Evidence-driven spatiotemporal COVID-19 hospitalization prediction with Ising dynamics

“…Pandemic progression patterns have spatial similarity depending on the underlying population, economic, geographic, and mobility factors in various locations. Recent works suggest that utilizing these spatial similarities can guide the model to better extract progression patterns and make more accurate predictions 4,27,42,43 . In this work, instead of pre-defining a connected location graph with fixed edges, we aim to learn the adaptative connectivity between locations.…”

“…To further improve the spatiotemporal prediction performance, a significant line of research focuses on extracting and utilizing spatial dependencies. Graph neural networks (GNN) 4,27,42,43 and metapopulation analysis models [44][45][46] have achieved remarkable success in solving this issue. In these works, counties and states are modeled as nodes in the graph, and the edges are defined using geographical and sociological similarities or mobility scores.…”