key: cord-333693-z2ni79al
authors: Wu, Lin; Wang, Lizhe; Li, Nan; Sun, Tao; Qian, Tangwen; Jiang, Yu; Wang, Fei; Xu, Yongjun
title: Modeling the COVID-19 Outbreak in China through Multi-source Information Fusion
date: 2020-08-06
journal: Innovation
DOI: 10.1016/j.xinn.2020.100033
sha: 
doc_id: 333693
cord_uid: z2ni79al

Modeling the outbreak of a novel epidemic, such as coronavirus disease 2019 (COVID-19), is crucial for estimating its dynamics, predicting future spread and evaluating the effects of different interventions. However, there are three issues that make this modeling a challenging task: uncertainty in data, roughness in models, and complexity in programming. We addressed these issues by presenting an interactive individual-based simulator, which is capable of modeling an epidemic through multi-source information fusion.

and reported. This gap includes the incubation period and any delay in medical visit, diagnosis, or reporting. To make matters worse, modelling may be influenced by authors' prejudice, interest relationships or preconceived ideas. Therefore, we argue that scientific researches should combine multiple sources of data worldwide rather than be based on a single source, and they should treat estimates from global researchers as elastic constraints imposed on models.

The second challenge is roughness in popular models, which is caused by oversimplification. It introduces significant errors and makes it impossible to reduce uncertainty by combining different types of information. The most common epidemic dynamics models are compartment models such as SEIR (Susceptible, Exposed, Infectious and Removed) and SIR (Susceptible, Infectious and Removed), which have been adopted widely in the simulation of COVID-19. The state vector of each person in a compartment is simplified as homogeneous and Markovian (memoryless), and transitions among compartments are modeled by differential equations with fixed parameters such as incubation rate, transmission rate and recovery rate. However, oversimplified models are not capable of incorporating multi-type uncertain information like clinical courses, viral shedding, subclinical transmission, infections, confirmations, deaths, or interventions, so they cannot reduce uncertainty by multi-source information fusion.

The last but far from the smallest challenge is the complexity of programming. This is a problem for researchers and reviewers who do not have a background in computer science. When it comes to individual-based models (IBMs), implementation is impossible without rich experience in object oriented programming (OOP), which compartmentalizes data into objects and describes object contents and behavior through J o u r n a l P r e -p r o o f the declaration of classes. Therefore, scientists have begun to call for sharing model codes so that the results of papers can be replicated and evaluated. However, sharing model codes is not enough given that there are many programming languages and it is no small task to install and configure corresponding development environments and run publicly shared codes.

To tackle the three challenges of modelling epidemic dynamics, we have developed an interactive simulator for individual-based models in this paper. This is described in detail in supplemental materials. In contrast to compartment models, individual-based models represent each individual via an independent set of specific characteristics that may change over time. This feature allows a more realistic and informative analysis of an epidemic. It is capable of interactively modeling parameter ranges as probability distributions, heterogeneity as independent objects and randomness of transmission as stochastic processes through a terminal or webpage without coding. The output of this model consists of daily values of infected cases, confirmed cases, recovered cases, deaths, and effective reproduction numbers. We can fit input parameters and output results with reported and inferred data from multiple sources. 

Supplemental Information includes supplemental materials, methods and results.

The authors declare no competing interests.

Clinical features of patients infected with 2019 novel coronavirus in Wuhan

Estimating the potential total number of novel Coronavirus cases in Wuhan City

Early Transmission Dynamics in Wuhan, China, of Novel Coronavirus-Infected Pneumonia

Nowcasting and forecasting the potential domestic and international spread of the 2019-nCoV outbreak originating in Wuhan, China: a modelling study

Estimating infection prevalence in Wuhan City from repatriation flights

Estimates of the severity of coronavirus disease 2019: a model-based analysis

Estimating the burden of SARS-CoV-2 in France

Changes in contact patterns shape the dynamics of the COVID-19 outbreak in China. Science

Estimating the asymptomatic proportion of coronavirus disease 2019 (COVID-19) cases on board the Diamond Princess cruise ship

Temporal dynamics in viral shedding and transmissibility of COVID-19