Heterogeneous length of stay of hosts movements and spatial epidemic spread

Infectious diseases outbreaks are often characterized by a spatial component induced by hosts’ distribution, mobility, and interactions. Spatial models that incorporate hosts’ movements are being used to describe these processes, to investigate the conditions for propagation, and to predict the spatial spread. Several assumptions are being considered to model hosts’ movements, ranging from permanent movements to daily commuting, where the time spent at destination is either infinite or assumes a homogeneous fixed value, respectively. Prompted by empirical evidence, here we introduce a general metapopulation approach to model the disease dynamics in a spatially structured population where the mobility process is characterized by a heterogeneous length of stay. We show that large fluctuations of the length of stay, as observed in reality, can have a significant impact on the threshold conditions for the global epidemic invasion, thus altering model predictions based on simple assumptions, and displaying important public health implications.

💡 Research Summary

The paper tackles a fundamental limitation in spatial epidemic modeling: the assumption that hosts either move permanently to a new location or commute with a fixed, short stay. Real‑world mobility data, however, reveal that the duration of stay at a destination is highly heterogeneous, ranging from a few hours to several months, depending on purpose, individual behavior, and socioeconomic factors. To capture this realism, the authors extend the classic metapopulation framework by introducing a stochastic “length of stay” variable for each movement event. They model the stay duration with probability distributions (e.g., exponential for moderate variability, Pareto for heavy‑tailed variability) and incorporate the first two moments of these distributions—mean and variance—into the disease transmission equations.

Using linear stability analysis, they derive a generalized invasion threshold that depends not only on the usual parameters (transmission rate, recovery rate, and movement flux) but also on the variance of the stay‑time distribution. The key analytical insight is that larger variance lowers the critical transmissibility needed for a global outbreak. In other words, when a small fraction of travelers stay for long periods while the majority have brief visits, the network of contacts becomes more “connected” in a temporal sense, facilitating the spread of infection even if the average stay is short.

The authors validate the theory with extensive stochastic simulations on synthetic networks. They compare three scenarios: (1) homogeneous fixed stay, (2) exponential stay distribution, and (3) Pareto (heavy‑tailed) stay distribution. Results show that the Pareto case produces the fastest epidemic growth, higher peak prevalence, and larger final attack rates than the other two. The exponential case yields intermediate outcomes, confirming that the shape of the stay‑time distribution, not just its mean, critically shapes epidemic dynamics.

Beyond the mathematical findings, the paper discusses public‑health implications. Traditional control measures—travel bans, border screenings, or commuter‑based quarantine—focus on reducing movement volume or targeting short‑duration commuters. The new results suggest that targeting long‑stay individuals (e.g., travelers on extended business trips, students, migrant workers) for pre‑departure testing, mandatory quarantine, or priority vaccination could be far more effective in curbing spread. Moreover, surveillance systems should identify “stay‑time hotspots” (places where people tend to linger) rather than only high‑traffic transit hubs. This shift could improve early detection and resource allocation during emerging outbreaks.

In summary, the study demonstrates that incorporating heterogeneous lengths of stay into metapopulation models fundamentally alters the epidemic invasion threshold and the predicted spread patterns. By moving beyond the simplistic permanent‑movement or fixed‑commuting assumptions, the work provides a more nuanced, empirically grounded framework for forecasting spatial disease dynamics and designing targeted, efficient intervention strategies.