A latent space model and Hotelling's T2 control chart to monitor the networks of Covid-19 symptoms

Abstract

In the COVID-19 coronavirus pandemic, potential patients that suffer from different symptoms can be diagnosed with COVID-19. At the early stages of the pandemic, patients were mainly diagnosed with fever and respiratory symptoms. Recently, patients with new symptoms, such as gastrointestinal or loss of senses, are also diagnosed with COVID-19. Monitoring these symptoms can help the healthcare system to be aware of new symptoms that can be related to the COVID-19 coronavirus. This article focuses on monitoring the behavior of COVID-19 symptoms over time. In this regard, a Latent space model (LSM) and a Generalized linear model (GLM) are introduced to model the networks of symptoms. We apply Hotelling's T2 control chart to the estimated parameters of the LSM and GLM, to identify significant changes and detect anomalies in the networks. The performance of the proposed methods is evaluated using simulation and calculating average run length (ARL). Then, dynamic networks are generated from a COVID-19 epidemic survey dataset.