🔬 Agent-Based Models: Simulating Epidemic Dynamics at the Granular Scale

Agent-Based Models (ABMs), also known as micro-simulation models, are essential when population heterogeneity—differences in movement behavior, contact patterns, susceptibility, or adherence to interventions—plays a central role in disease spread. Unlike classical compartmental models that assume homogeneous mixing, ABMs simulate infection dynamics at the level of individual agents.

---

📐 Compartmental Structure and Flow Explanation

ABMs consist of distinct, autonomous agents, each representing an individual with unique characteristics such as age, location, health status, mobility patterns, and behavioral tendencies. While disease progression may follow classical pathways such as Susceptible → Exposed → Infectious → Recovered, transitions occur through localized interactions governed by agent rules.

Agents
Represent individuals (similar to nodes in a network). Each agent carries attributes such as occupation, movement schedule, or vulnerability.

Interactions and Flow
Transmission occurs when susceptible agents come into contact with infectious agents. Proximity, environment, and mobility rules determine how interactions occur.

State Dynamics
Each agent’s epidemiological state changes according to probabilistic rules, reflecting exposure history and stochasticity in disease progression.

---

🧮 Mathematical Representation

ABMs are not defined by differential equations but by iterative update rules that govern agent behavior. At each time step, the state of each agent is updated based on interactions and stochastic events.

The state of agent i at time t+1 is:

sᵢ(t+1) = fᵢ(sᵢ(t), s₋ᵢ(t), ε(t))

Where:

• sᵢ(t) is the current epidemiological state of agent i.
• s₋ᵢ(t) represents the states of all interacting neighbors.
• ε(t) encompasses all probabilistic or random factors influencing transitions.

Through this update process, population-level epidemic curves emerge as the aggregate result of millions of micro-level interactions.

---

⚗ Parameter Definitions and Typical Ranges

ABMs require parameters describing both disease processes and demographic or behavioral complexity.

Parameter	Definition	Typical Range
R₀	Basic reproduction number	1.5 – 4.0
γ, κ	Recovery or progression rates	0.1 – 0.5 per day
γ (isolation effectiveness)	Fractional reduction in infectivity	0.9 (high effectiveness)
D_CT	Delay in contact tracing or isolation	0.25 – 0.5 days
P_CT	Fraction of contacts traced	0.5 – 1.0
s_age	Age-specific susceptibility multiplier	0.38 – 0.88 relative to adults

These parameters allow the model to capture both biological properties of the pathogen and the heterogeneity of the host population.

---

Applicability and Limitations

Use and Purpose

1. Modeling Heterogeneity
   Ideal when contact rates vary strongly across age, occupation, location, or mobility groups. ABMs allow detailed representation of networks and individual-level interactions.
2. Evaluating Granular Interventions
   Useful for policies requiring individual-level modeling, such as contact tracing algorithms, household quarantine, or location-specific restrictions.
3. Capturing Stochasticity
   Able to reflect inherent randomness in person-to-person transmission, especially important for early outbreak dynamics or small clusters.

Assumptions and Weaknesses

1. Computational Burden
   Simulating large numbers of autonomous agents with complex rules requires substantial computational resources.
2. Data Requirements
   ABMs need detailed input data such as realistic contact patterns, mobility maps, or demographic layers, which can be difficult to obtain or validate.
3. Interpretability Challenges
   Because ABMs are high-dimensional and rule-based, identifying causal mechanisms and parameter fitting can be challenging.

---

📚 Selected References

1. Ferguson, N. M., et al. Strategies for mitigating an influenza pandemic.
2. Peak, C. M., Childs, L. M., Grad, Y. H., & Buckee, C. O. Comparing nonpharmaceutical interventions for containing emerging epidemics.
3. Tang, L., Zhou, Y., Wang, L., et al. A review of multi-compartment infectious disease models.
4. Oshinubi, K., Chen, Y., Doerry, E., et al. A systematic review of spatial epidemiological modeling approaches applied during the COVID-19 pandemic.
5. Volz, E. SIR dynamics in random networks with heterogeneous connectivity.