Efficient and Accurate Simulation of Disease Outbreaks on Changing Networks

Greg Howard
5th June, 2025

Efficient and Accurate Simulation of Disease Outbreaks on Changing Networks

The High-Acceptance Sampling (HAS) algorithm achieves efficient and exact simulation of adaptive networks by bypassing the frequent network updates required by the Stochastic Sampling Algorithm (SSA) to jump directly to infection events (a), utilizing derived edge existence probabilities (b) and a tight upper bound to leap over rejection steps (c).

Image adapted from: Gubela et al. / CC BY (Source)

Key Findings

  • Researchers from Freie Universität Berlin, Max Planck, Robert Koch Institute, and Johns Hopkins developed a new simulation method for epidemics on networks where interactions change over time
  • The new high-acceptance sampling (HAS) algorithm produces exact predictions of disease spread much faster than traditional methods
  • HAS can model how people adjust their behavior during outbreaks, helping to quickly test various intervention strategies
In recent years, scientists have increasingly relied on computer simulations to understand how diseases spread through populations, especially when individuals’ interactions change over time. Traditional simulation methods, such as Gillespie’s algorithm, have been very useful, but they often struggle when applied to large and complex social networks with time‐varying connections. A new study from research teams at Freie Universität Berlin, Max-Planck Institute for Molecular Genetics, Robert-Koch Institute, and Johns Hopkins University[1] addresses this challenge by introducing a novel simulation method called high-acceptance sampling (HAS). HAS is a rejection-based stochastic sampling algorithm designed to simulate disease spread on adaptive networks. In simple terms, it is a method for generating possible scenarios of an epidemic, where the pattern of interactions among individuals changes over time and may even be influenced by the epidemic itself. The algorithm is “exact,” meaning that its simulation results perfectly match the predictions derived from theoretical models, and it can be multiple orders of magnitude faster than the well-known Gillespie’s algorithm in many cases. Speed is important here because faster simulations allow researchers and public health officials to quickly explore different “what-if” scenarios during an outbreak. The study tackles the problem of simulating infectious disease dynamics when the network of contacts—how people interact—is not constant. In many epidemics, both social behavior and the disease spread influence one another. For instance, when people become aware of a disease outbreak, they might change their behavior—reducing social interactions or seeking medical attention—which in turn can affect how widely and quickly the disease spreads. This interplay between behavior and infection creates a complex system where standard methods often fall short because they assume that interactions are independent of the disease process. HAS is designed to work in these adaptive settings without relying on simplifying assumptions about distinct time scales. In many models, researchers assume that changes in the contact network occur either much faster or much slower than the disease transmission process. However, this assumption often does not hold in real-world scenarios. By contrast, the HAS algorithm works well regardless of whether the contacts among individuals change faster, slower, or at a similar pace as the disease spreads. This versatility makes HAS particularly well suited to exploring contemporary problems in epidemiology, such as those seen in recent outbreaks. One significant application demonstrated in the study is in modeling virtual epidemics similar to Mpox and COVID. The researchers specifically looked at how diagnosis-driven and incidence-driven behavioral changes influence the spread of infection. Diagnosis-driven changes occur when individuals modify their behavior as soon as they are diagnosed or learn of a case, while incidence-driven changes occur in response to the overall number of cases reported in a community. Both types of adaptive behavior are crucial elements in understanding why some outbreaks slow down more rapidly than expected. For example, earlier research on the 2022 global Mpox outbreak among men who have sex with men (MSM)[2] found that a spontaneous reduction in high-risk sexual behavior played an important role in decreasing the number of cases. The new study employs HAS to simulate similar scenarios, allowing for detailed exploration of how behavioral changes can alter the course of an epidemic. The methodological innovation of HAS lies in its use of rejection-based sampling to streamline calculations without sacrificing accuracy. In a standard rejection-based method, potential simulation steps are repeatedly proposed and then “rejected” if they do not comply with certain probabilistic criteria. What distinguishes HAS is its high acceptance probability, meaning that most proposed steps are used, which greatly enhances the efficiency of the simulation. The researchers show mathematically that this algorithm produces exact results, which builds confidence in its predictions, and they compare its computational performance directly with Gillespie’s algorithm. While the benefits of HAS are most evident in cases where the disease and contact dynamics are interdependent, the study demonstrates that HAS remains robust under a wide variety of conditions. Not only does HAS provide a faster and equally accurate tool for modeling epidemic spread on adaptive networks, its flexibility also extends to complex scenarios where contact dynamics do not follow simple patterns. Many infectious diseases are influenced by the behavior of the population, such as how frequently people interact in high-risk settings like social clubs, households, or community gatherings. In the context of the previous Mpox outbreak[2], the major transmission route was found to be within the sexual contact networks, with comparatively minor contributions from other settings. The HAS algorithm is capable of incorporating these complex, real-world details, offering a framework where adaptive responses and behavioral intervention strategies can be rigorously tested in silico. This work represents an important step forward in computational epidemiology by addressing a critical gap in the ability to simulate the coupled dynamics of human behavior and disease spread. Virtual epidemic scenarios that include adaptive behavior—such as changes in socializing habits in response to diagnosis or rising case numbers—are more reflective of actual outbreaks. Thus, tools like HAS can support public health authorities in designing interventions, evaluating the likely impact of policies such as contact tracing and targeted immunization, and ultimately improving response strategies during emergent epidemics. In summary, the novel HAS algorithm offers an exact, efficient, and versatile simulation tool for modeling infectious diseases on adaptive networks. Its capability to accurately capture and rapidly compute the effects of dynamic social behaviors makes it particularly promising for studying modern epidemics. By successfully bridging computational challenges and real-world complexities, HAS stands to contribute significantly to our understanding of epidemic processes, complementing earlier insights such as those from the Mpox outbreak analysis[2] and paving the way for more effective disease control strategies.

MedicineHealth

References

Main Study

1) Efficient and accurate simulation of infectious diseases on adaptive networks

Published 2nd June, 2025

https://doi.org/10.1371/journal.pcsy.0000049

Related Studies

2) The decline of the 2022 Italian mpox epidemic: Role of behavior changes and control strategies.

https://doi.org/10.1038/s41467-024-46590-4


Related Articles

An unhandled error has occurred. Reload 🗙