An Automatic Tool for Predicting the Influence of COVID-19-Infected Patients on Future Infections

The COVID-19 pandemic has rapidly spread across the globe, causing significant impact on various aspects of human life. To combat the spread of infectious diseases and reduce transmission rates, epidemiological surveys are crucial in identifying high-risk individuals. However, conducting these surveys manually can be time-consuming and burdensome for epidemiologists. Therefore, there is a need for an automatic tool to assist with epidemiological surveys.

In this study, researchers aimed to develop an automatic epidemiological survey tool that can predict the influence of COVID-19-infected patients on future infections. The researchers utilized a dataset that included information about interactions between confirmed cases, such as contact order, contact times, and movement routes, as well as individual properties like symptoms.

To incorporate this interaction information and individual properties, the researchers used graph neural networks (GNNs). They utilized two variants of GNNs, namely graph convolutional and graph attention networks, and compared their performance with traditional machine learning models. The results showed that the graph-based models outperformed the traditional models.

The researchers specifically focused on predicting the spreading of infections up to the 2nd, 3rd, and 4th order. They found that the graph-based models had higher performance, with improvements of 0.200, 0.269, and 0.190 for the area under the curve, respectively. This indicates that the contact information of an infected person is crucial in predicting their impact on future infections.

The findings of this study suggest that incorporating the relationships between an infected person and others can significantly improve the effectiveness of automatic epidemiological surveys. By considering interaction information and individual properties, the tool can accurately predict the spread of infections and identify high-risk individuals.

The COVID-19 pandemic has prompted extensive epidemiological surveys in Korea since February 2020. These surveys gather various information, including demographic data, contact history, time of contact, movement routes, and symptoms, from individuals who are infected or have been infected with COVID-19. The results of these surveys help classify the type of care for infected persons, whether it be at home or in a medical institution.

Furthermore, the aggregated data from epidemiological surveys can inform the government’s preventive measures and policies. The researchers in this study utilized the epidemiological survey data to create a relationship map between major infection clusters and Incheon City, Korea. This map allowed them to visualize the connections between infected individuals and the people they had contact with at different stages.

The objective of this study was to explore the development of an automatic epidemiological survey tool. The researchers proposed using graph neural network-based algorithms to predict whether specific confirmed cases would influence future infections. Graph neural networks are effective in capturing interaction information between confirmed cases and their properties.

The researchers utilized a COVID-19 infection network dataset from Incheon City, which included information about each person and their relationships. The goal was to predict whether an infected person would affect the future infection of others, even when future contact information was not accessible. The researchers aimed for realistic evaluation by assuming that contact information between current and future infected persons was not given.

In conclusion, this study developed an automatic tool using graph neural networks to predict the influence of COVID-19-infected patients on future infections. The results showed that incorporating contact information and individual properties significantly improved the accuracy of predictions. This tool can assist epidemiologists in conducting faster and more efficient epidemiological surveys, ultimately reducing the spread of infectious diseases.