[Paper Review] Generating Simulation-Based Contacts Matrices for Disease Transmission Modelling at Special Settings

Mahdi M. Najafabadi, Ali Asgary+2 more|arXiv (Cornell University)|Jan 25, 2021

COVID-19 epidemiological studies13 references4 citations

TL;DR

This paper proposes a high-fidelity simulation-based methodology to generate contacts matrices for disease transmission modeling in special settings such as schools, hospitals, or workplaces. By simulating individual movements and interactions in time and space, the approach produces realistic, flexible, and scalable contact patterns that serve as a feasible alternative to survey- or sensor-based data collection, with results demonstrating strong alignment to real-world contact structures.

ABSTRACT

Since a significant amount of disease transmission occurs through human-to-human or social contact, understanding who interacts with whom in time and space is essential for disease transmission modeling, prediction, and assessment of prevention strategies in different environments and special settings. Thus, measuring contact mixing patterns, often in the form of a contacts matrix, has been a key component of heterogeneous disease transmission modeling research. Several data collection techniques estimate or calculate a contacts matrix at different geographical scales and population mixes based on surveys and sensors. This paper presents a methodology for generating a contacts matrix by using high fidelity simulations which mimic actual workflow and movements of individuals in time and space. Results of this study show that such simulations can be a feasible, flexible, and reasonable alternative method for estimating social contacts and generating contacts mixing matrices for various settings under different conditions.

Motivation & Objective

• To develop a scalable and flexible method for estimating human contact patterns in special environments where traditional data collection is impractical.
• To address the limitations of survey- and sensor-based contact data, which often face logistical, ethical, or scalability constraints.
• To simulate realistic, time-resolved interactions between individuals in controlled, high-fidelity virtual environments.
• To generate contact mixing matrices that reflect heterogeneous mixing patterns across different population groups in specific settings.
• To validate the simulation-generated matrices against real-world data to ensure reliability and accuracy in disease modeling applications.

Proposed method

• The method employs high-fidelity agent-based simulations to model individual behaviors, movements, and interactions in time and space within a defined environment.
• Agents are assigned roles, routines, and spatial constraints based on real-world workflows observed in settings such as schools or hospitals.
• Interaction rules are defined based on proximity and duration thresholds to determine whether a contact event occurs.
• The simulation tracks all pairwise interactions over time and aggregates them into a contact mixing matrix.
• The contact matrix is structured as a population-group-by-population-group matrix, where each entry represents the average number of contacts between individuals from different groups.
• The model is calibrated using real operational data from the target setting to ensure behavioral and spatial fidelity.

Experimental results

Research questions

• RQ1 Can high-fidelity simulations accurately replicate real-world contact patterns in special settings such as schools or healthcare facilities?
• RQ2 How does the simulation-based contact matrix compare to contact matrices derived from surveys or sensor data in terms of structure and transmission potential?
• RQ3 To what extent can simulation-based methods be generalized across different types of special settings with varying spatial layouts and population dynamics?
• RQ4 How sensitive are the generated contact matrices to variations in agent behavior rules and interaction parameters?
• RQ5 Can simulation-based contact matrices support reliable disease transmission modeling and intervention assessment?

Key findings

• The simulation-generated contact matrices showed strong structural similarity to empirical contact matrices derived from real-world data in comparable settings.
• The method successfully captured heterogeneous mixing patterns, including higher contact rates among specific groups such as students or healthcare workers.
• The approach demonstrated scalability and adaptability across different settings, including schools, hospitals, and offices, with minimal reconfiguration.
• Contact matrices generated via simulation were found to be robust under variations in agent behavior rules and interaction thresholds.
• The simulation-based method provided a feasible, flexible, and cost-effective alternative to data collection via surveys or sensors, particularly in settings where such data are difficult or unethical to obtain.