Use of Modeling and Simulation for Understanding COVID-19 Dynamics

The novel COVID-19 disease has disrupted human life across the globe. Understanding its impact is a multi-faceted and complex systems problem. Mathematical modeling is a technique that allows capturing the complexity of the dynamics of such diseases using interpretable mathematical relations. These models can be combined with simulations to study the dynamics at micro and macro levels as well as experiment response mechanisms to minimize the disease impact. An effective Modeling and Simulation (M&S) solution provides a scientific platform to not only analyze the dynamics but also facilitate evidence-based decision making. A situational awareness and early warning system based on real data feeds would be ideal but in the absence of such information or where the mechanisms to collect such data are not available, M&S plays a vital role and is an effective replacement. With this motivation, we present a simulation framework that leverages several mathematical models to simulate the spread of diseases such as COVID-19 in urban environments and allows us to analyze the disease impact under different scenarios (e.g., lockdowns, self-isolation, and contact tracing).

THE SIMULATION FRAMEWORK

Our simulation framework considers the individual user mobility and infectiousness in an urban environment and makes use of several mathematical models to simulate the disease spread (Figure 1). We use these models to consider the real-life scenarios of ordinary citizens. For example, individuals would reside in a specific area and would mobilize for work, shopping and socialization. In this process they would be exposed to other individuals and the probability of exposure can then be computed based on how many people they would come in contact with and how many of those could be positive on average. In our work, the mobility of the individuals is represented by a mobility model (Figure 2), while the probability of disease contraction and transmission comes from an epidemic model (Figure 3). Finally, we also consider a recovery model (Figure 4) based on real data for Case Fatality Rates (CFR) that predicts whether a person recovers or deceases. Overall, we implement our framework using a spatial model that considers the different types of locations of an urban environment; a geographical depiction of the entities in our environment is shown in Figure 5.

Figure 1: The underlying models of our simulation framework

Figure 2: The Mobility model used to change the locations of individuals in our simulator.

Figure 5: The geographical visualization of the spatial model representing a real city with different locations.

WHAT-IF ANALYSES

Governments around the world have implemented different kinds of interventions to control the spread of COVID-19. Understanding how these interventions affect the disease spread plays a pivotal role in making the right decisions to reduce the impact. It is here that our framework becomes highly useful as it can provide those insights through what-if analysis. As a demonstration, we are going to use our framework to understand how self-isolation, mask-wearing, and contact tracing can help in reducing the spread of COVID-19:

1. Self-Isolation: In this scenario, the infected individuals who get symptomatic isolate and quarantine themselves either at their residences or by getting hospitalized until recovery.

2. Mask Wearing: We consider the impact of mask-wearing on the spread as prior studies have shown that individuals who wear a mask reduce the likelihood of transmitting the disease.
3. Contact Tracing: Here, we consider the scenario where government agencies can notify all individuals, who come in close contact with an infected individual, to isolate themselves for 14 days.

The results (Figure 6) obtained from our framework suggest that self-isolation can help in reducing the infections, but we need to have additional interventions. In that regard, we consider scenarios for mask-wearing and consider contact tracing under different compliance rates; the compliance rate is a parameter in our framework that specifies the percentage of the population that conforms to the isolation instructions. In Figure 6, we can observe that if individuals start wearing masks, the number of infections reduces significantly. Moreover, from Figure 7, we can see that if we have contact tracing enabled along with mask-wearing, we can minimize the number of infections and the deaths. Overall, our analysis shows that contact tracing and mask-wearing helps in controlling the spread if the compliance rate is high. However, if compliance lowers, the number of infections can increase. So, the population's compliance rate is a significant factor that government agencies need to consider in developing COVID-19 mitigation strategies.

Figure 6: Disease spread for the scenario of Self-Isolation and Mask wearing

Figure 7: Disease spread for the scenario of Contact Tracing

In conclusion, the last demonstration shows how our framework can allow us to perform such what-if analyses and help us better understand the COVID-19 spread dynamics that will ultimately support in making informed decisions for controlling the spread of such diseases. Additionally, we can use our framework to perform risk profiling for not just COVID-19, but for any other future hazard.

Acknowledgement: The development of this framework has been initially done as part of an early warning system for COVID-19 developed by Teradata GDCs led by Kamran Shafi. Currently, this framework is part of Teradata’s COVID360 initiative led by Christopher Jackson, where the aim is to help countries restart their economies in the post-COVID world.