Researchers: Jianhong Wu, Nicola Bragazzi, Huaiping Zhu, Sanyi Tang

A biologically, epidemiologically and clinically informed model is a fundamental tool for scholars for capturing the biology of the COVID-19, identifying new targets and for public health decision- and policymakers for devising evidence-based measures and interventions. It is of crucial importance to provide accurate estimates, especially during the early phases of the COVID-19 outbreak. However, lack of knowledge of the severity of the outbreak, uncertainty and low data quality may negatively impact on the accuracy and precision of the estimations. Meta-analytical approaches that pool together various data sources provide more robust and accurate estimates to better- deal with uncertainty. Meta-analysis also helps understand sources of variability and heterogeneity among estimates, and quantifies the impact of initial assumptions, model and parameter choice on outcomes, especially when: a) a plethora of models and techniques for computing infectious agents-related metrics exists; b) there are discrepancies among such techniques; and c) no standardized methods exist to cope with contrasting results.

Building on collaborations initiated during the SARS crisis and enhanced through an IDRC funded bilateral project, we have developed partnerships with a Chinese mathematical modelling team at the Tianyuan (Mega) Mathematical Center in Northwest China (Sanyi Tang and Yanni Xiao), the China CDC (Yiming Shao, Chief Scientist), Epidemiological Survey Center of China Animal Health & Epidemiology Center (Youming Wang, Director) and Hubei University of Medicine (Yong Yu). These collaborations offer access to substantial epidemiological and hospital data from Hubei and Shaanxi Provinces, enabling us to form a COVID-19 database to support our modeling project. We will fund a team of disease modelling visiting graduate students from China, to compile, translate and interpret data from reliable Chinese sources to establish insights (model choice, parameterization, implementation, calibration and simulation) to support our modelling project for COVID-19.

Researchers: Julien Arino, Michael Li, Charmaine Dean

Once an epidemic breaks out in a given area, its transmission dynamics varies significantly based on the geographical locations of the initial infection and its connectivity with the rest of the region. To better understand the global spread of the disease, we propose to work on a range of hierarchical models to account for the numerous parameters and variables that affect the outcomes of our spatial transmission models. We will incorporate population density, economic and population mobility measures, based on stochastic simulations of SIR (susceptible-infectious-recovered) transmission models along with alternative transport parameters in terms of modal share as well as quarantine effects. The alternate transport parameters account for widely different scenarios. For instance, global spread is largely driven by air transportation channels, where ground transportation is the primary mode of conveyance in local vicinities or neighbouring countries. In such a case, alternative modes of movement will be embedded into our models to account for these differences, and capture the range of possible outcomes. Furthermore, different types of network models, that take into consideration fixed and random contacts, will be used, in conjunction with calibration of algorithms.

We will use data, accessed through our national and international collaborations (See Theme 1) with partners in China, to learn from their analyses and translate that learning to the Canadian context. For example, we will draw comparisons from data in sparsely populated regions in China, to emulate regions with similar population densities in Canada. In the event that there may be a transmission chain or chains in Canada, we intend to model the potential for spread to other locations, factoring in different situations and parameters. Furthermore, we will devise methods to integrate the contact tracing process into our transmission models. We aim to understand the lag from latency, to infection, to serious death, and the factors associated with the transmission rates between these states, including likely disparate spatial effects and potential mitigation factors with a view to understanding how to reduce death rates, and determine if sustained transmission is a consideration. We will draw on data from the initial reported cases in humans, as early as possible, ideally going back to the original zoonotic source, devising a number of stratified isolation and quarantine strategies, developing quantitative criteria (cases, transmission chains) for each type of intervention, ranging from the least invasive to the more severe and challenging in terms of social organizations.

Researchers: Ali Asgary, Jianhong Wu, Jane Heffernan, Zack McCarthy, Lydia Bourouiba, Adriano Solis

Public facilities, particularly health and transportation facilities like hospitals, long term care facilities, health clinics, subways, airports, airplanes, trains, and passenger ships are vulnerable and also play a major role in both transmission and containment of virus spread in the past and current COVID-19 outbreak. We will develop a multifunctional agent-based modelling and simulation platform to model, simulate and examine the extents that an index case can spread the virus in public facilities, focused on hospital, air, rail, and sea transportation settings, and examine the effectiveness of public health countermeasures. The agent-based modelling and simulation approach has emerged as a highly promising method for modelling and simulating disease outbreaks. Through this platform we will:

• Model the behaviour of multiple human and non-human agents; their dynamic and multilevel interactions and operations will be included, examined, and visualized;
• realistically simulate and evaluate the effect of public health responses to inform policy and decision making in emergency situations. Our access to and significant prior experience in the ADERSIM facility based at York will ensure robust outcomes of this aim;
• generate realistic synthetic outbreak data needed for equation based mathematical models and artificial intelligence models of the COVID-19 spread; and
• train more highly qualified personnel (HPQ) in Canada to develop this platform type to simulate and visualize outbreak trajectories in real settings, filling an unmet need.

We will use data from both our our Meta-Analysis (Theme 1), and several international sources, particularly hospitals in the infected areas in China, the Diamond Princess cruise ship, airports, subway and trains to calibrate our model. Our model will then be applied and tested on selected Canadian settings. We will turn the agent-based models into scenario-based visual simulations and will test them at ADERSIM for professional training and exercises of public health emergency management. Moreover, our research will yield guidelines for other public settings such as schools, shopping malls and sport facilities in Canada that can adopt these models to their own specific conditions.

Researchers: Junling Ma, Kumar Murty

Reduction of infectious contacts, contact tracing, quarantine measures, and school closures played critical roles in the control of emerging infectious disease such as SARS and the 2019 influenza pandemic and such measures are currently implemented in China, Canada, and other regions with confirmed cases. However, to evaluate their effectiveness for the Canadian public health settings, we need to answer the following questions:

• What is the generation time distribution of the virus?
• What is the scope of contact tracing, and quarantine needed to effectively control an outbreak?
• How does the ability of the virus to transmit asymptomatically impact the effectiveness of these control measures?
• How do the altered contact patterns caused by school closures impact the evolution of the epidemic?

We will evaluate the effectiveness of these control strategies by developing mathematical and computer simulation models based on contact networks that can faithfully replicate the disease spread in different regions. Although our agent-based modelling approach is a powerful tool for realistically modelling, simulating, and evaluation viral transmission and control measures, it provides little understanding of the mechanisms of the epidemic process due to its complexity, and thus cannot easily answer these questions. Contact networks, where nodes represent individuals and edges represent transmissible contacts between individuals, are powerful tools to model realistic community contacts. They quantify the effect of contacts on transmission, and thus allow us to answer the questions. The model will be coupled with a random mixing component to capture random mixings such as contacts in public events, taking public transportation, and public-health workers interacting with the general public. We will use data in China and other regions with confirmed cases, including a) daily confirmed cases, b) individually documented cases, and c) travel data from China to Canada, and between Canadian cities, to estimate the generation time distribution and calibrate our model. Our study will provide a tool to quantify the effectiveness of the contact-reducing control strategies, to evaluate their costs, and to provide guidelines to improve these control strategies that can be tested by our agent-based modelling framework.