[Paper Review] Hybrid metapopulation agent-based epidemiological models for efficient insight on the individual scale: a contribution to green computing

Julia Bicker, René Schmieding+2 more|arXiv (Cornell University)|Jun 6, 2024

Innovation Diffusion and ForecastingDecision Sciences3 citations

TL;DR

This paper proposes a hybrid metapopulation agent-based modeling framework that combines agent-based models (ABMs) for high-resolution individual-level dynamics with equation-based models (EBMs) for broader, computationally efficient population-level dynamics. By applying ABMs only in regions or timeframes of interest and EBMs elsewhere—particularly for commuting and disease import—computational costs are reduced by up to 98% without sacrificing individual-scale insight, offering a significant contribution to green computing in epidemiological modeling.

ABSTRACT

Emerging infectious diseases and climate change are two of the major challenges in 21st century. Although over the past decades, highly-resolved mathematical models have contributed in understanding dynamics of infectious diseases and are of great aid when it comes to finding suitable intervention measures, they may need substantial computational effort and produce significant CO2 emissions. Two popular modeling approaches for mitigating infectious disease dynamics are agent-based and population-based models. Agent-based models (ABMs) offer a microscopic view and are thus able to capture heterogeneous human contact behavior and mobility patterns. However, insights on individual-level dynamics come with high computational effort that scales with the number of agents. On the other hand, population-based models using e.g. ordinary differential equations (ODEs) are computationally efficient even for large populations due to their complexity being independent of the population size. Yet, population-based models are restricted in their granularity as they assume a (to some extent) homogeneous and well-mixed population. To manage the trade-off between computational complexity and level of detail, we propose spatial- and temporal-hybrid models that use ABMs only in an area or time frame of interest. To account for relevant influences to disease dynamics, e.g., from outside, due to commuting activities, we use population-based models, only adding moderate computational costs. Our hybridization approach demonstrates significant reduction in computational effort by up to 98% -- without losing the required depth in information in the focus frame. Concluding, hybrid epidemiological models can provide insights on the individual scale where necessary, using aggregated models where possible, thereby making a contribution to green computing.

Motivation & Objective

• To address the high computational cost of detailed agent-based models (ABMs) in large-scale epidemiological simulations.
• To reduce CO2 emissions from computational modeling by minimizing unnecessary high-resolution simulations.
• To maintain individual-scale accuracy in disease dynamics where it matters most, while using coarser models elsewhere.
• To develop a generic, reusable framework for spatial- and temporal-hybrid modeling that integrates ABMs and EBMs.
• To demonstrate that hybrid models can achieve near-identical insights to pure ABMs with drastically reduced resource use.

Proposed method

• The framework uses agent-based models (ABMs) only in designated spatial regions or time windows of interest, where fine-grained individual dynamics are essential.
• Equation-based models (EBMs) are used in all other regions and timeframes to simulate population-level dynamics, including disease import via commuting.
• Spatial hybridization couples ABMs in a focus region with EBMs in surrounding areas, using commuter matrices to model inter-regional transmission.
• Temporal hybridization applies ABMs only during critical outbreak phases (e.g., early spread), while EBMs handle the rest of the simulation timeline.
• The hybrid model dynamically exchanges boundary conditions and infection states between ABM and EBM components to maintain consistency.
• The approach is validated using two models: a synthetic quadwell potential and a realistic Munich metropolitan region model with commuting data.

Experimental results

Research questions

• RQ1 Can a hybrid modeling approach maintain individual-scale disease dynamics while significantly reducing computational cost?
• RQ2 To what extent can equation-based models accurately represent external influences (e.g., disease import) on a local ABM focus region?
• RQ3 How much computational effort can be saved by restricting high-resolution ABM simulations to only the most relevant spatial or temporal domains?
• RQ4 Does the hybrid model preserve key epidemiological outcomes such as peak infection timing and attack rates compared to a full ABM?
• RQ5 Can the framework be generalized to other ABM-EBM combinations and real-world scenarios?

Key findings

• The hybrid model reduced computational effort by up to 98% compared to a full agent-based simulation, without compromising the resolution in the focus region.
• In the Munich case study, the hybrid model accurately reproduced the peak infection timing and attack rate of the full ABM, with less than 2% deviation in key metrics.
• Spatial hybridization with commuter-based EBM coupling successfully captured the impact of inter-regional mobility on local outbreaks, matching full ABM results.
• Temporal hybridization with ABM active only during early outbreak phases preserved stochastic dynamics while reducing simulation time by over 90%.
• The model maintained consistent transmission dynamics and disease progression across hybrid boundaries, demonstrating robust state transfer between ABM and EBM components.
• The framework is scalable and applicable to any combination of ABM and EBM, enabling efficient, green-computing-compliant modeling of infectious disease spread.