Revealing latent factors of temporal networks for mesoscale intervention in epidemic spread

Laetitia Gauvin, André Panisson, Alain Barrat, Ciro CattutoarXiv 2015 · arXiv

Introduction

Epidemic control strategies typically oscillate between two extremes: micro-interventions, such as vaccinating specific high-degree individuals, and macro-interventions, such as city-wide lockdowns or school closures. While micro-interventions are often data-intensive and difficult to implement in real-time, macro-interventions impose significant socio-economic burdens. This research introduces a "mesoscale" approach, identifying intermediate social structures—such as lunchroom mixing or classroom cohorts—that drive disease transmission within temporal networks.

Methodology: Tensor Decomposition of Social Interactions

The researchers model temporal networks as 3-mode tensors (Node × Node × Time). To extract meaningful patterns from high-resolution interaction data, they employ Non-negative Tensor Factorization (NTF). This technique decomposes the complex network into a sum of rank-1 tensors, each representing a "mesostructure": a specific group of individuals interacting during a defined time window.

Visualization of the 14 mesostructures identified via NTF, showing node membership and temporal activity patterns.

Because this decomposition is additive, it allows for "surgical" interventions. By removing a single component and reassembling the network, the researchers can simulate how an epidemic would spread in the absence of that specific social structure.

Data and Validation

The framework was tested using a high-resolution dataset of face-to-face interactions among 231 children and 10 teachers in a primary school. Interactions were recorded every 20 seconds using wearable sensors. The NTF method successfully identified 14 distinct components. Crucially, these mathematical constructs aligned with real-world metadata: most components corresponded to individual classrooms, while others captured cross-class mixing events like lunch breaks.

Key Insights: The Danger of Temporal Bridges

A primary finding is that the most intuitive targets—dense clusters like classrooms—are not always the most critical for epidemic spread. Instead, temporally-localized mixing patterns (e.g., lunch breaks) often act as "bridges" between otherwise isolated groups.

The study utilized Susceptible-Infected (SI) and Susceptible-Infected-Recovered (SIR) models to rank these structures. The results showed that removing components associated with lunch breaks (specifically components 12 and 14) slowed the epidemic significantly more than removing classroom-based structures, despite the latter often having higher total interaction "weight."

Delay ratio τ_r for the removal of each component. Higher values indicate a greater delay in the epidemic spread.

Results and Practical Implications

The simulations demonstrated that targeted removal of these "bridge" mesostructures could reduce the final epidemic size by approximately 50% in certain transmission regimes. This suggests a highly actionable strategy for public health: rather than closing an entire institution, administrators could implement targeted "mesoscale" changes, such as staggering lunch breaks or modifying shared space schedules, to achieve substantial mitigation with minimal disruption.

Conclusion

By bridging the gap between individual-level tracking and population-level lockdowns, this tensor-based framework offers a powerful tool for epidemic management. While the current study is limited by a two-day observation window and high data requirements, it paves the way for more nuanced, less intrusive intervention strategies in complex social environments.

Summary updated

Abstract

The customary perspective to reason about epidemic mitigation in temporal networks hinges on the identification of nodes with specific features or network roles. The ensuing individual-based control strategies, however, are difficult to carry out in practice and ignore important correlations between topological and temporal patterns. Here we adopt a mesoscopic perspective and present a principled framework to identify collective features at multiple scales and rank their importance for epidemic spread. We use tensor decomposition techniques to build an additive representation of a temporal network in terms of mesostructures, such as cohesive clusters and temporally-localized mixing patterns. This representation allows to determine the impact of individual mesostructures on epidemic spread and to assess the effect of targeted interventions that remove chosen structures. We illustrate this approach using high-resolution social network data on face-to-face interactions in a school and show that our method affords the design of effective mesoscale interventions.
BibTeX
@misc{gauvin2015revealing,
  author = {Laetitia Gauvin and André Panisson and Alain Barrat and Ciro Cattuto},
  title = {{Revealing latent factors of temporal networks for mesoscale intervention in epidemic spread}},
  year = {2015},
  eprint = {1501.02758},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/1501.02758},
}
Figure